Trade- offs in childrens time allocation : Mixed support for embodied - - PDF document

1

Trade-offs in children’s time allocation: Mixed support for embodied capital models of the demographic transition in Tanzania

Sophie Hedges1*, Rebecca Sear1, Jim Todd1,2, Mark Urassa2, & David W. Lawson3 1 Department of Population Health, London School of Hygiene and Tropical Medicine, London, UK 2 National Institute for Medical Research, Mwanza, Tanzania 3 Department of Anthropology, University of California, Santa Barbara, USA *Corresponding author. sophie.hedges@lshtm.ac.uk

ABSTRACT

Children’s time allocation is fundamental to embodied capital models of the demographic

• transition. Yet few studies have directly investigated the impact of market integration on

children’s time allocation in contemporary rural populations undergoing socioeconomic ‘modernization’. We present a study of children’s time use in two communities in Mwanza, Tanzania, representing the extremes of a local rural-urban gradient. Consistent with embodied capital theory, market integration increases investment in education, reduces children’s work, and is associated with lower opportunity costs to schooling. However, these patterns apply primarily to boys, with herding work relatively incompatible with schooling. For girls, schooling is more readily combined with domestic chores. Furthermore, contrary to predictions, the strongest time allocation trade-offs are not between school and work, but between school and leisure time, particularly for girls. Mixed support for embodied capital models may partially explain why fertility decline has stalled in many low-income countries, despite education uptake. Higher opportunity costs to boy’s education in herding communities, may account for recent, unexpected trends of higher school enrolment for girls.

2 Finally, we advocate that international development programs consider the wellbeing implications of reduced leisure time triggered by market integration, particularly for girls maintaining a ‘double-shift’ of school and domestic work.

KEYWORDS

Education; children’s work; time allocation; embodied capital; market integration

3

• 1. INTRODUCTION

1 Children’s lives and adult expectations of childhood have changed dramatically during the 2 past half-century (Bledsoe et al. 1999; Zelizer 1985). The ideal ‘modern’ childhood is work- 3 free, care-free, and spent primarily in formal education (Abebe 2007; Ansell 2005; 4 Nieuwenhuys 1996). These aspirations are increasingly realized worldwide; 93% of primary- 5 school-aged children are estimated to be enrolled (UNESCO 2015) and children’s involvement 6 in work has rapidly declined (ILO 2013). ‘Embodied capital’ models of the demographic 7 transition conceptualize these trends as a functional response to changing perceived returns 8 to parental investment. Under these models, market integration incentivizes parents to invest 9 in education because it enhances adult income. However, schooling is costly, both directly 10 and through opportunity costs arising from the re-allocation of children’s time away from 11 productive work. This favors an emphasis on child ‘quality’ over quantity, leading to lower 12 fertility rates (Becker 1960; Kaplan 1996). A considerable body of research on fertility 13 patterns, much of it carried out by evolutionary demographers and anthropologists, supports 14 key tenets of this perspective (Lawson and Borgerhoff Mulder 2016; Sear et al. 2016). But 15 little research has directly examined the impact of market integration on parental decisions 16 to educate children, and the anticipated trade-offs between education and child work 17 (Hedges et al. 2016). 18 Classic economic and embodied capital perspectives on the demographic transition are 19 primarily informed by historical European patterns of socioeconomic and demographic 20 change, rather than contemporary transitions occurring in low-income countries. Such 21 accounts are vulnerable to the fallacy of the ‘developmental paradigm’; Thornton (2001)’s 22 term for the assumption that societal change is linear and universal, with societies differing 23

4

• nly in their position along the same developmental trajectory. Thornton describes how this

24 model has conflated cause and effect, and led to the widely accepted view that changes from 25 ‘traditional’ family structures to ‘modern’ ones drive socioeconomic development (Thornton 26 2001). Thus, low fertility, gender equality and youth autonomy are framed as prerequisites 27 for economic development, while ‘traditional’ family structures are viewed as impediments 28 to ‘progress’. Development programs therefore promote family planning, female education, 29 and reduced work during childhood, as the key to both individual and societal wealth and 30

• wellbeing. Yet contemporary economic and demographic transitions differ in several

31 important respects, not least the role of external agencies in changing both the real and 32 perceived costs and benefits of education and high fertility. The objective of this paper is to 33 assess the applicability of embodied capital theory to patterns of child time allocation in a 34 contemporary low-income, high-fertility setting in rural Tanzania. 35 Anthropology has a strong tradition of time allocation scholarship, emphasizing context- 36 dependency in the costs and benefits of alternative behavioral strategies. While many 37 anthropologists have collected detailed data on children’s time allocation (Bird and Bliege 38 Bird 2002; Bliege Bird and Bird 2002; Gurven, Kaplan, and Gutierrez 2006; Kramer 2002; Lancy 39 2012; Turke 1988), few studies have been conducted in settings where school attendance is 40 the norm (but cf. Mattison and Neill 2013). Relevant research in these settings has 41 predominantly been carried out by development economists concerned with minimizing 42 harmful ‘child labor’. However, this research most often defines ‘work’ as activities done to 43 generate cash income. Work generally done by women and children, including household 44 chores and childcare, is frequently overlooked in time use surveys (Esquivel et al. 2008). This 45

5 devalues household work, yet these duties may be time- and energy-consuming, essential to 46 household functioning, and disruptive of schooling (Ilahi 2000). 47 We present a novel study of children’s time allocation in two communities in northwestern 48 Tanzania, representing the extremes of a local rural-urban gradient. Departing from much of 49 the prior literature on this topic, we take a holistic perspective on children’s time allocation 50 throughout a complete day, including contributions to domestic and farm work, and leisure 51

• time. In the next section we describe in more detail the embodied capital framework on

52 childhood and the demographic transition. We draw into this discussion the influence of the 53 developmental paradigm in shaping policy and interventions surrounding children’s work and 54 schooling in contemporary rural, low-income settings. We then elaborate on gender as a 55 determinant of parental investment and children’s time allocation, and highlight how the 56 value placed on market work by post-transition societies has led to gaps in our knowledge 57 about how girls’ time allocation may change with development. These sections lead to five 58 hypotheses derived from embodied capital theory regarding the impacts of market 59 integration and gender on (i) school enrollment, (ii) patterns of child work, and (iii) the trade- 60

• ffs between these activities.

61

1.1 EMBODIED CAPITAL AND MARKET INTEGRATION

62 Embodied capital theory assumes that children’s time is allocated to different activities based 63

• n perceived returns, whether that be immediate production, or long-term investment in

64 embodied capital (Gurven and Kaplan 2006). Embodied capital is defined as the skills, 65 knowledge, experience, physical growth and strength acquired during childhood and 66 adolescence, which increase adult social and reproductive success. Time allocation is 67 predicted to favor activities that improve long-term social and reproductive success, but there 68

6 may be trade-offs between activities with short-term returns, and activities with long-term 69 returns which are not immediately productive. Returns are determined by local 70 socioecological context, household factors, and individual-level factors such as child gender 71 (Bock 2002b; Gurven and Kaplan 2006). Parents invest preferentially in the child who will 72 benefit most, and expectations about children’s time allocation vary with the benefits of 73 different children’s work. With transitions to skills-based, competitive labor markets, school 74 replaces traditional forms of learning as the main way of acquiring embodied capital, with 75 corresponding reductions in children’s time allocation to immediate production. With 76 children less able to offset their own costs through productive work, parents limit fertility; 77 investing more in fewer children. These ‘quantity-quality trade-offs’ are hypothesized to have 78 driven the global decline in fertility over the past two centuries (Becker 1960; Kaplan 1996; 79 Kaplan, Bock, and Hooper 2015). From this perspective, parental investment strategies may 80 not be fitness-maximizing in all contexts, but represent an evolved response to perceived 81 costs and benefits of investment rooted in adaptive strategies (Kaplan 1996). Thus, high 82 socioeconomic benefits to ‘quality’-focused investment strategies reinforce modern low 83 fertility norms, but fertility limitation is nevertheless maladaptive because individuals with 84 few children leave fewer long-term descendants (Goodman, Koupil, and Lawson 2012; Kaplan 85 et al. 1995). 86 Support for this model primarily focuses on studies of fertility rather than parental investment 87 in education (Lawson and Borgerhoff Mulder 2016; Sear et al. 2016), using data from historical 88 European demographic trends and variation in fertility within modern affluent populations 89 (Goodman, Koupil, and Lawson 2012; Kaplan et al. 1995), rather than contemporary low- 90 income populations undergoing market integration (but see Gibson and Sear 2010; Gibson 91

7 and Lawson 2011; Hedges et al. 2016; Mattison and Neill 2013; Neill 2011). The paucity of 92 anthropological studies directly investigating the impact of modernization has recently been 93 highlighted (Mattison and Sear 2016; Sear et al. 2016). More studies of the impact of market 94 integration could provide important opportunities to both test existing theoretical 95 frameworks, and to consider the wider assumptions and practices of the international 96 development sector (Gibson and Lawson 2015). 97 98

1.2 THE ‘DEVELOPMENTAL PARADIGM’

99 The primacy of the historical European demographic transition in influencing demographic 100 transition theories is problematic because the process of modernization in contemporary 101 populations is distinct (Thornton 2001). In the central narrative of both the embodied capital 102 framework and the developmental paradigm, the benefits of education are clear, but this is 103 not necessarily the case for individuals within rural populations undergoing transition. In 104 many contexts, education quality is so poor that children leave school without even basic 105 literacy and numeracy skills, with no prospect of finding the kind of employment promised by 106 development programs (Pritchett 2013). Furthermore, families may remain reliant on 107 subsistence livelihoods and labor from young household members. 108 International global organizations, such as the United Nations (UN) and the International 109 Labor Organization (ILO) aim to promote education, particularly for girls, and eliminate child 110

• labor. Yet these policies frequently ignore the complexities of investment decisions,

111 stigmatizing children’s work and promoting unrealistic expectations about the returns to 112 education (Bourdillon et al. 2015). Indeed, the ‘right to education’ has now assumed strong 113

8 moral and ideological connotations. While child work is framed as exploitative, harmful, and 114 ‘backwards’, education is portrayed as the pathway to both societal and individual 115 development and poverty alleviation (Abebe 2007; Ansell 2005; Mukudi 2002; Nieuwenhuys 116 1996). 117 By contrast, anthropologists studying childhood emphasize the diverse forms of work and 118 learning seen across societies, and the importance of children’s production in supporting their 119 families, and developing skills and self-esteem (Bourdillon, Levison, and Myers 2010; Kramer 120 2005; Lancy 2012; Nieuwenhuys 1993). Several studies have demonstrated the 121 complementary nature of many children’s activities, with learning occurring predominantly 122 through play and ‘on-the-job’ training rather than direct instruction, with more time spent in 123 productive work as skill and strength improve (Bock 2002a; Lancy 2010). Anthropologists are 124 thus well-positioned to counter the prevailing ethnocentric view of childhood, giving a more 125 neutral perspective on the costs and benefits of education and work for children and their 126 parents (Bock 1999; Bourdillon et al. 2015; Lancy 2012). 127 There are also concerns that the widely-assumed trade-off between children’s work and 128 school attendance is exaggerated in current theory and policy discourse. Although at an 129 aggregate level it is true that school enrolment is associated with reduced children’s work, 130 few studies have demonstrated a direct trade-off between time spent in work versus 131 education (Beegle et al. 2008; Pörtner 2016). There is surprisingly little evidence that 132 interventions promoting schooling reduce children’s time spent working (Gibbons, Huebler, 133 and Loaiza 2005; Ravallion and Wodon 2000). In reality, work and school may often be 134 complementary, with, for example, children earning money for school expenses through part- 135 time work (Nieuwenhuys 1993). Additionally, given the emphasis on work and school as the 136

9

• nly important axes of children’s lives in policy-orientated research, there is little data on how

137 else children spend their time in leisure or personal activities (Bacolod and Ranjan 2008; 138 Biggeri et al. 2003). More holistic studies of children’s time use that include domestic chores, 139 farm labor and also leisure time are required. 140

1.2 GENDER, SCHOOLING AND CHILDREN’S WORK

141 Globally, girls are less likely to be in school than boys (United Nations 2015a). In recent 142 decades, development programs have increasingly focused on promoting female education 143 (United Nations 2015a; United Nations 2015b). Female education is considered to be 144 particularly beneficial through its association with other key development indicators, 145 including lower fertility, maternal and child health, and greater household wealth (Bedasso 146 2008). Inequality in education is usually attributed to broader patriarchal norms favoring 147 investment in boys. Development programs are thought to lead to gender equality in 148 education by lessening the financial pressures that foster biased investment, and changing 149 attitudes by highlighting the benefits of educating girls, and stigmatizing early marriage and 150 childbearing. 151 Evolutionary models of parental investment view biased investment as parents optimizing 152 returns to investment across children, allocating resources preferentially to the gender which 153 will give the highest returns to that investment (Borgerhoff Mulder 1998; Sear 2011). One 154 study showed that in a patrilineal context, educational investment is male-biased, while in a 155 matrilineal setting it is female-biased, and that these biases increase with household wealth, 156 probably because wealthier parents have more certainty over the returns to their investment 157 (Gibson and Sear 2010). Another study investigated education in a contemporary 158

10 transitioning context and found evidence for female-biased investment, due to greater 159 perceived benefits of school for girls’ future employment (Neill 2011). 160 Embodied capital models focus on education as an economic investment, assuming that 161 market integration incentivizes investment in education based on the expected returns in the 162 adult labor market. In a patrilocal, patrilineal context, such as we study here, sons remain 163 nearby as adults, meaning parents may anticipate greater returns to educating boys. 164 Furthermore, men typically earn higher wages than women in Tanzania and are more likely 165 to have a job requiring formal education (FAO 2014). However, education has consequences 166 beyond purely economic outcomes. Education has been proposed to improve maternal and 167 child health by increasing expenditure on health and improving knowledge of health-seeking 168 behaviors (Bedasso 2008). Educated women may have greater social status, and previous 169 studies have shown that this may give them the opportunity to marry higher status men, and 170 that they attract higher bridewealth (Ashraf et al. 2015; Cronk 1989). Furthermore, 171 development programs may alter perceptions of the costs and benefits by offering subsidies 172 for girls’ education but not boys’ (Mburu 2016). 173 Assessing the costs and benefits of education for each gender is additionally complicated by 174 differing opportunity costs of boys’ and girls’ work. Typically, girls do more domestic chores 175 and childcare, while boys are more involved in work outside the household (Murdock and 176 Provost 1973). There is mixed evidence as to whether girls’ work or boys’ work has higher 177

• pportunity costs. Many anthropological studies have highlighted the importance of girls’

178 childcare in underwriting the costs of high fertility in pre-transition societies (Bereczkei and 179 Dunbar 2002; Kramer 2002; Turke 1988). Other studies emphasize the importance of boys’ 180 labor in contributing to household subsistence (Cain 1977). However, in post-transition 181

11 societies, market-based work, typically performed by men, is most valued, while unpaid 182 household and care work, generally performed by women and children, is not valued 183 financially or socially, and is often overlooked in studies of work and time allocation (Esquivel 184 et al. 2008). Thus it is likely that previous studies of children’s work and education have 185 underestimated the amount of work done by children, particularly girls (Assaad, Levison, and 186 Zibani 2010; Ilahi 2000; Levison, Moe, and Knaul 2001). A lack of information on how 187 household work impacts schooling is likely to impede efforts to promote education, 188 particularly female education (UNICEF 2016). There have also been few studies of the 189 differential compatibility of schooling and work for boys and girls. 190

1.3 HYPOTHESES

191 We test five predictions derived from embodied capital theory, using residence in a 192 neighboring village and town as a proxy for market integration. While predictions are drawn 193 from embodied capital models, our analyses are also exploratory, given the unpredictability 194

• f returns to investment in a transitioning context, and the limitations of the models outlined

195

• above. Taking a holistic approach and investigating all time uses, and disaggregating analyses

196 by gender, we model trade-offs in time allocation in a more nuanced way than previous 197 studies. 198 Our first two hypotheses concern parental decisions to enroll children in school. In the town, 199 we anticipate higher returns to investment in skills acquired through school, due to the 200 greater potential for formal employment (Kaplan and Bock 2001; Mattison and Neill 2013). 201 Our first prediction therefore is that (1) market integration will be associated with greater 202 school enrolment. As discussed above, predictions about differential investment by gender 203 are complex. National and regional data, and historical trends within this area 204

12 (Supplementary Material (SM), Figure S3) indicate higher enrolment rates for boys, suggesting 205 greater perceived returns to male education. This leads to our second prediction, that (2) girls 206 will be less likely to be enrolled in school than boys, but this gender difference will reduce with 207 market integration. 208 Our third and fourth hypotheses concern children’s work. Agricultural and particularly 209 pastoralist livelihoods are associated with high labor demands, traditionally met partly 210 through children’s labor (Kramer 2002; Sellen 2003; Sieff 1997). Market integration is 211 associated with lowered reliance on agriculture and reduced livestock ownership, and so is 212 expected to be associated with lower returns to children’s agricultural work. Additionally, 213 better access to water and smaller household size (i.e. fewer household members) in the town 214 is expected to reduce the returns to children’s household chores. This leads to our third 215 hypothesis, that (3) market integration will be associated with less work overall, particularly 216 farm work. 217 Gendered division of labor is observed across societies, and children are socialized to fulfil 218 these gendered roles as adults. According to traditional Sukuma labor division (the primary 219 ethnic group in our study population), farm work and cattle herding are boys’ tasks, while 220 household chores are girls’ tasks (Varkevisser 1973). Market integration leads to a reduction 221 in farm work, and so it is expected that gender differences in work will be reduced in the town 222 compared to the village. Our fourth prediction is therefore that (4) boys will do more farm 223 work, and girls will do more household chores, and gender differences in work patterns will 224 reduce with market integration. 225 Finally, we examine the trade-off between work and education at the heart of embodied 226 capital models of child time allocation. As time is a limited resource, school attendance is 227

13 expected to reduce time spent in other activities. Furthermore, as the returns to children’s 228 work are expected to be lower, and the returns to school attendance higher in the town, the 229

• pportunity costs of school are expected to be lower. Thus, we hypothesize that (5) there will

230 be a trade-off between work and education, but that market integration will reduce the 231 magnitude of this trade-off. 232

• 2. SOCIAL CONTEXT

233 Field research was conducted in the Mwanza region of north-western Tanzania. Primary 234 school enrolment in Tanzania increased dramatically following the universal education 235 movement in the 1970s, but stagnated and even declined in the 1980s (Burke & Beegle 2004). 236 Less than 60% of boys and girls progress to secondary school, and there are concerns over the 237 low quality of schooling available (Hivos/Twaweza 2014). A large proportion of the population 238 is still involved in subsistence agropastoralism, with children also working on household farms 239 (ILO 2013; USDoL 2013). Under-5 mortality has declined substantially over the past decade, 240 but fertility decline has stalled in recent years (Kishamawe et al. 2015; PRB 2016). The most 241 recent Demographic and Health Survey shows that the Lake Zone, where Mwanza is located, 242 has particularly high fertility, with 6.4 children per woman on average (MoHCDGEC 2016). 243 Thus, the study area provides an example of a context in which social, economic and 244 demographic transitions are occurring, and in which the trade-offs between education and 245 work are likely to be complex. 246 The main ethnic group in this area is the Sukuma, the largest ethnic group in Tanzania, 247 representing about 17% of the nation (Malipula 2015). Most people speak Swahili and 248

• Sukuma. Traditionally, the Sukuma lived in large, dispersed homesteads and kept large herds

249

14

• f cattle, but cattle keeping is declining, as land holdings decrease in size, and the purchase

250

• f consumer goods becomes more important as an indicator of wealth (Wijsen and Tanner

251 2002). Most households farm, and many engage in petty trade of products such as milk, rice, 252 tomatoes, and fish (Kishamawe et al., 2015; own data not shown). The Sukuma are 253 traditionally polygynous, but polygyny is becoming less common with the adoption of 254 Christianity and the influence of education (Wijsen and Tanner 2002). In the town, 92% of 255 households identified as Christian, 5% as Muslim, and 2% as having traditional or no religious 256

• beliefs. In the village, 57% of households identified as Christian, while 42% reported

257 traditional or no religious beliefs. Religion and education are linked, both being associated 258 with urbanization, wealth, and ‘modern’ ways of life (Wijsen and Tanner 2002). Churches 259 sponsor schools and special tuition classes for local children, and Muslim children often 260 attend madrasa classes, which provide religious instruction and teach Arabic, in addition to 261

• school. Some children in this study attended madrasa but primarily on Saturdays; none

262 mentioned it as an activity during interviews. 263 The Tanzanian education system includes seven years of primary, four years of lower 264 secondary, and two years of upper secondary education. Education is compulsory between 265 seven and fourteen, but many children start school late. Parents do not pay fees for 266 government primary schools, but bear costs such as uniforms, books, and exam entry fees 267 (UNESCO 2011). Recently, secondary school fees have been abolished, though this legislation 268 came into effect after this study took place, and it is unclear how schools will now be funded, 269 given they already face a serious deficit (GEM Report 2015). Tanzania has ratified 270 International Labor Organization legislation on child labor, including a minimum age of 14 for 271

15 paid employment, and programs to reduce the most harmful forms of labor, for example in 272 mining (USDoL 2013). 273 The quality of education in Tanzanian government schools is generally acknowledged to be 274 poor, with overcrowded classes, limited teaching supplies, teacher absenteeism, and 275 frequent use of corporal punishment (UNESCO 2011). Learning outcomes are devastatingly 276 low; for the 2012 Standard 7 exam, which students must pass to progress to secondary school, 277 there was a meaningful pass rate of only 6%. Many children leave school unable to read, write, 278

• r do basic arithmetic (Hivos/Twaweza 2014). Swahili is the language of instruction in primary

279 school, changing to English in secondary school, representing a learning barrier for many 280 children, who may only speak the language of their ethnic group at home (Mwinsheikhe 281 2009). In the study area, private schools are most desirable, but beyond the means of many 282 families. 283 For the purposes of this study, we are interested in children and young adults aged 7 to 19, 284 the ages of school education in Tanzania. The study area chosen is encompassed by the Magu 285 Health and Demographic Surveillance Site (HDSS), covering everyone resident in the seven 286 villages of Kisesa ward, located approximately 20km east of Mwanza city on the main road 287 between Mwanza and Kenya. The site was established in 1994 to measure mortality, fertility 288 and mobility in the general population, and so provides a representative example of a rural 289 Tanzanian community (Kishamawe et al. 2015). Our data come from two out of the seven 290 villages to represent the two ends of the gradient from rural (Welamasonga) to relatively 291 urban (Kisesa) (Figure 1). While this is a narrow geographic focus, it provides an effective case 292 study that we believe translates more broadly. Kisesa is now more accurately described as a 293 town, and is situated on the main road close to several large businesses, such as a Coca-Cola 294

16 distribution center and a Mwatex textile factory. It has a busy market center, a health clinic 295 and dispensary, three government primary schools, and numerous private schools. 296 Welamasonga is the village in the HDSS furthest from Mwanza and retains the more 297 traditional wide dispersion of households. There is a small village center with two or three 298 shops, a dispensary, and two government primary schools. 299 [Figure 1 here] 300

• 3. METHODS

301

3.1 DATA COLLECTION

302 The HDSS provided a sampling frame of all households with members aged 7-19, from which 303 550 households were randomly sampled. Households are self-defined in the HDSS as “a group 304

• f people living together in the same compound and who regularly eat together from the

305 same pot” (Kishamawe et al., 2015). Two local research assistants carried out interviews in 306 Swahili or Sukuma depending on respondent preference. Fifty-eight households had moved 307

• ut of the study area, seven refused to be interviewed, twenty-three no longer had resident

308 eligible children, and six more were unknown to the facilitators, giving a final sample of 456 309

• households. Household surveys were carried out using Google Nexus 7 tablets with Open Data

310 Kit (ODK) Collect software (Brunette et al. 2013). 311 The sampling frame provided a list of the expected members of the household, together with 312 their gender and age. We first checked current household membership against this roster, 313 amending it accordingly, as well as collecting information about adult members’ education 314 and occupation, and the household head’s marital status, ethnicity, and religion. A series of 315 questions was asked about household assets, land and livestock ownership, and business 316

17

• involvement. This was followed by food security questions, based on the Food and Agriculture

317 Organization (FAO)’s Household Food Insecurity and Access Scale (HFIAS; Coates, Swindale, & 318 Bilinsky, 2007). This index asks questions about food security during the past month, including 319 anxiety about food supply, limiting food quality and reducing food quantity, and the 320 frequency of these experiences. For each child aged 7-19, an additional survey was answered 321 by their parent/guardian, collecting information on the child’s education and work history. 322 The primary outcome for this study is children’s time allocation on the previous school-day. 323 Studies of children’s time allocation typically rely on proxy respondents, generally the mother. 324 However, the International Labor Organization (ILO) recommends using child respondents 325 where possible (ILO 2004). A study in a similar Tanzanian context interviewed both children 326 and parents, allowing for a comparison of estimates, and found that parents consistently 327 underestimated the time spent by their children working. The difference between parents 328 and children shrank with the age of the child, suggesting that younger children may 329 exaggerate time spent working (Janzen 2015). We chose to interview children themselves 330 (Figure 2), allowing assistance from an older sibling or parent if the respondent was unsure of 331 exact times. 332 [Figure 2 here] 333 After the household and child surveys, the whereabouts of all eligible children were 334 established and any children present were interviewed. Other children were followed up at a 335 later date where possible. 1,278 children were followed up out of a total of 1,387 eligible 336 children (92.1%). The majority of those not followed up were away at boarding school (3.8% 337

• f total sample) or travelling (2.6%). During the time allocation interview, children were

338 shown a diagram representing the day, and were asked to remember everything they did on 339

18 the previous weekday, from when they woke up until they went to sleep. A diagram was 340 shaded to indicate the time and duration of the activities (an example of the time diagram is 341 shown in SM; Figure S1). 342

3.2 OUTCOME VARIABLES

343 We constructed three binary educational outcomes: schooled, where 0 indicates the child has 344 never been to school and 1 indicates the child has been enrolled in school; enrolled, where 0 345 indicates the child is not currently enrolled in school and 1 indicates the child is currently 346 enrolled in school; and progressed, for those aged 14 to 19 only, where 1 indicates the child 347 has progressed to secondary school. A value of ‘0’ includes children still enrolled in primary 348 school, who have therefore been delayed in their progression through school, and those who 349 have finished primary school; thus a value of ‘1’ is an indicator of ‘ideal’ educational 350 attainment for older children. Additionally, the highest grade attained or the current school 351 grade was used to calculate the hazard of dropping out of school in particular grades. 352 Activities from the time allocation interview were coded into one of five categories; leisure / 353 personal (hereafter referred to as ‘leisure’), education, household chores, farm work or 354 herding (hereafter referred to as ‘farm work’), and market work. Table 1 lists the activities 355 mentioned by children and the final code assigned. Total time spent in each activity category 356 was calculated and divided by the total number of hours covered by the interview (5am to 357 10pm; 17 hours) to give the proportion of time spent in each activity category. 358 [Table 1 here] 359

3.3 DATA ANALYSIS

360

19 Logistic regression models investigated the effect of town residence (our proxy for market 361 integration) and gender where the outcome of interest was schooled, enrolled, or progressed. 362 The clustering of children (Level 1, n=1,367) within households (Level 2, n= 456) was 363 accounted for using mixed effect models, including a random effect for household in schooled 364 and enrolled models, using the Stata command ‘xtmelogit’. Progressed models did not include 365 a random effect because the clusters are more sparsely populated, which may overestimate 366 fixed and random effects (Clarke 2008). All models adjusted for child age and food security as 367 a proxy for household wealth. An interaction between gender and residence was included to 368 investigate whether gender differences were reduced in the town. 369 As an alternative method of testing predictions regarding school enrollment, the hazard of 370 dropping out of school in each grade was investigated using discrete-time event history 371

• analysis. This method has the advantage of including more information, as the time to event

372 (here, dropping out of school) can be analyzed, rather than simply whether or not the child 373 was still enrolled. The results were consistent with other outcomes, however, and so only the 374 results of the simpler logistic regression models are presented; results from the event history 375 analysis are presented in SM (Figure S2). Linear regression was used to test predictions about 376 the effect of town residence and gender on children’s work patterns. Each activity had a 377 separate regression model, and the outcome was the amount of time, in hours, spent in the 378

• activity. The outcomes were therefore time spent in leisure, household chores, farm work,

379 and market work, as well as overall time spent in productive work (total hours spent in chores, 380 farm, and market work). An interaction between residence and gender was included to 381 investigate whether gender differences differ between the town and village. Analyses were 382 stratified by school attendance (attended on the previous school day or not) and age group 383

20 (7-13 and 14-19), as work patterns change with age, and differ between those who attended 384 and did not attend school. Models were adjusted for age and household food security, and 385 school enrolment for those who did not attend school. 386 Time uses are automatically negatively correlated, because time spent in one activity reduces 387 the time available for other activities. Fractional multinomial logistic regression models were 388 used to account for this automatic correlation, and investigate the trade-off between 389 education and other activities (Buis 2017). The outcome variables are the proportions of time 390 spent in education, leisure activities, household chores, farm work, and market work. For each 391 child, these proportions add up to 1. These models were stratified by age group, and adjusted 392 for age, gender, household food security and school enrolment. Models give predicted 393 proportions of time in the five categories, subsequently converted back into hours. All 394 analyses were carried out in Stata version 14. 395

• 4. RESULTS

396

4.1 DESCRIPTIVE STATISTICS

397 Table 2 shows descriptive statistics for sampled households. Town households are smaller, 398 and more likely to have an educated household head. They are less likely to own land, grow 399 crops, or own cattle; more likely to have a formal business or member with a salary; and have 400 greater access to public services such as electricity and water in the town. These data provide 401 support for our assumption that town residence is a proxy for market integration. Food 402 security is similar across the village and town, suggesting that land and livestock are important 403 for household wealth in both locations. Household food security is used as a proxy for 404 household wealth in our analyses. We believe this is an effective measure of household 405

21 wealth in the context of a food insecure population and avoids comparability issues when 406 using alternative metrics than cannot easily be compared across livelihood types (e.g. land- 407

• r livestock-based, compared to income-based). Previous research in northern Tanzania

408 confirms close associations between food security and alternative wealth metrics health 409 (Lawson et al. 2014). 410 The vast majority (95%) of children had at some point been enrolled in school; only 77 out of 411 1,387 children had never attended school, primarily because they were still considered too 412

• young. In the village, 79% of girls and 71% of boys were currently enrolled, while in the town,

413 84% of girls and 87% of boys were enrolled. Among those aged 14 to 19, just over half had 414 progressed to secondary school; 33% of girls and 29% of boys in the village had progressed, 415 and 76% of girls and 68% of boys in the town had progressed. For the 263 children who had 416 previously been enrolled in school but were not currently enrolled, the mean years of 417 schooling was 5.9 in the village, and 8.1 in the town, indicating that on average children in the 418 village do not quite complete primary school (7 years of education). Of the 1,278 children 419 followed up for a time allocation interview, 1,032 were enrolled and 907 had attended school. 420 Therefore 371 children had not attended school, of whom 246 were not enrolled and 125 had 421 missed school. There was no significant difference in missing school between the village and 422 town. 423 [Table 2 here] 424 Figure 3 shows children’s time allocation by age, stratified by gender and location. Children 425 spend about half their time in leisure activities. Leisure time decreases with age, while time 426 spent in education and work increases with age; except among the oldest children who spend 427 little time in education and more in either work (village) or leisure (town), depending on 428

22

• location. Farm work is predominantly done by village boys and older village girls. Girls do

429 more household chores than boys in both town and village. Time spent in market work by any 430 children is minimal. 431 [Figure 3 here] 432

4.2 Education patterns

433 Table 3 shows results from our logistic regression models of the three binary education 434

• utcomes: schooled, enrolled and progressed. These results support prediction (1): town

435 residence is associated with higher odds of enrolment (both ever-enrolled and currently- 436 enrolled), and progression to secondary school. Our second prediction (2), stated that boys 437 will be more highly educated than girls, but that this difference will reduce with market 438

• integration. Contrary to prediction (2), although consistent with our prior research in an

439 agropastoralist area of northeastern Tanzania (Hedges et al. 2016), girls have higher odds of 440 enrolment than boys, though there is no gender difference in progression to secondary 441

• school. There are interactions between residence and gender, with gender differences being

442 reduced in the town (though this is only marginally significant for ever-enrolled). In the village 443 therefore, boys are less likely to be in school than girls, but in the town, the overall level of 444 educational investment increases, and the gap between boys and girls is reduced. 445 [Table 3 here] 446

4.3 Work patterns

447 Figure 4 presents predicted hours spent in chores, farm work, leisure, and overall productive 448 work (chores + farm work + market work) (full model output shown in Table S1). Results for 449

23 market work are not presented given the negligible amount of time spent in this activity. We 450 first discuss work patterns for children who did not attend school, before considering children 451 who did attend school. Results are presented stratified by children of primary and secondary 452 age groups, as work patterns change with age. 453 Prediction (3) stated that market integration would be associated with less work, particularly 454 farm work. Among children who did not attend school, village/town differences are clear and 455 in line with our predictions. Figure 4a and Table S1a show that those in the town do less 456 productive work, and that this difference is statistically significant. Prediction (4) stated that 457 boys would do more farm work and girls more household chores; but that these differences 458 would reduce with market integration. Our results show that gender differences in the type 459

• f work done are substantial, and in the expected direction, but that gender differences in

460 the amount of work done are more complicated. Among 7–13-year-olds, there are no 461 significant gender differences in work and leisure in the town, but village boys do marginally 462 more work and have marginally less leisure time than village girls. Among 14–19-year-olds 463 however, girls do approximately four hours more chores than boys in both the town and the 464

• village. This translates to 14–19-year-old boys in the town doing much less productive work,

465 and having much more leisure time, than girls. Gender differences are therefore actually 466 exacerbated with market integration, with 14–19-year-old girls doing around 4 hours more 467 productive work than boys in the town, but no significant gender difference in the village. 468 Among children who attended school, gender and village/town differences are much smaller. 469 Consistent with prediction (3), those in the town do slightly less productive work; a difference 470 which is significant among 14–19-year-olds. Those in the town also have significantly less 471 leisure time than those in the village, because they spend more time in education (results 472

24 shown in SM). In line with prediction (4), some gender differences are statistically significant, 473 in that girls do more household chores than boys, and this difference is reduced in the town. 474 In the village, boys do more farm work than girls, while neither boys nor girls do much farm 475 work in the town. This leads to girls doing more productive work and having less leisure time 476

• verall in the town, particularly among 14–19-year-olds.

477 These results suggest that the lower enrolment rates seen for boys in the village may be due 478 to their time spent in farm work. They also suggest that there may not be a straightforward 479 trade-off between work and school, because girls do similar amounts of, if not more, 480 productive work than boys, and yet are not less likely to be enrolled. In the next section, we 481 directly estimate the trade-offs in time allocation between work, leisure, and school. 482 [Figures 4a and 4b here] 483

4.4 Trade-offs between work and school

484 Figure 5 shows the predicted difference in time allocation, in hours, between school attenders 485 and non-attenders (full output given in Table S2). This gives us an indication of the opportunity 486 costs of schooling, as it shows which activities are reduced to allocate time to education. In 487 Figure 5, confidence intervals that cross 0 indicate a non-significant difference between 488 school attenders and non-attenders. For example, school attendance had negligible impacts 489

• n market work for both genders. As to be expected, school attendance substantially

490 increases time in education, particularly among older children, who allocate 9-11 hours a day 491 to education. Which activities are reduced to make space for schooling is dependent on 492 gender and location. 493

25 Among 7–13-year-olds, school attendance primarily reduces leisure time, by up to seven 494 hours a day. Village boys are the outliers in this, in that school leads to a relatively small 495 reduction in leisure time, but a larger reduction in farm work of around four hours a day. 496 Village girls also have a small reduction in both chores and farm work with school attendance, 497 while for both girls and boys in the town, only chores are reduced. These results imply that 498 the opportunity costs of schooling are highest for village boys, while there are surprisingly 499 small trade-offs between work and education for girls or town boys. 500 Among 14–19-year-olds, the effect of school attendance on reducing work is greater for girls. 501 School attendance reduces time spent in household chores by approximately five hours for 502 girls in the town. For village girls, school attendance reduces household chore time by around 503 three hours, and farm work by around two hours. Village boys see trade-offs between 504 education and farm work in this age group, as in the younger age group, with school 505 attendance decreasing farm work by around five hours. Town boys, in this case, are the 506

• utlier, as they only show small trade-offs between work and education. Instead, school

507 attendance reduces leisure time for town boys by nearly eight hours. Thus, at older ages, the 508

• pportunity costs of school attendance increase for both town and village girls, to a level

509 similar to village boys, but are negligible for boys living in a more market-integrated setting. 510 In summary, prediction (5), that work would trade-off against education but that this trade- 511

• ff would decrease with market integration, is only partially supported. We do find trade-offs

512 between work and education, particularly for older children, but a substantial amount of the 513 time that children spent in education comes from leisure time, rather than work. Further, 514 there are gender differences in this trade-off. Market integration impacts boys’ time 515 allocation to a greater degree than girls’. In the less market-integrated setting, boys’ work has 516

26 very high opportunity costs, and this appears to impact their enrolment. However, in the 517 more market-integrated setting, boys do not reallocate this time to other work activities, 518 lowering the opportunity costs of boys’ schooling. Girls’ work patterns on the other hand 519 show much smaller differences with market integration, with the opportunity costs of older 520 girls’ time being quite high in both the town and village. 521 [Figure 5 here] 522

• 5. Discussion

523 Our results provide mixed support for embodied capital models of the demographic transition 524 and challenge prevailing assumptions of the ‘developmental paradigm’. Consistent with these 525 perspectives, we present evidence from two communities in Tanzania that market integration 526 increases investment in education, reduces farm work, and is associated with lower 527

• pportunity costs to schooling. However, contrary to expectations common to both

528 frameworks, the strongest trade-offs in time allocation are not between school and work, but 529 between school and leisure time. Furthermore, the classic narrative of embodied capital 530 theory applies primarily to boys; male-dominated farm work is relatively incompatible with 531 schooling, while female household chores are more readily combined with school. These 532 findings have important theoretical and applied implications for our understanding of 533 economic and demographic ‘modernization’ and its profound impacts on childhood 534 experience. 535

5.1 Mixed support for embodied capital models

536 Proponents of both embodied capital theory and the development paradigm have typically 537 emphasized the role of economic returns to education. Some economic studies suggest these 538

27 returns outweigh opportunity costs, reinforcing the idea that education is a rational and 539 economically-motivated choice (Ansell 2005). However, in this context, which we believe is 540 broadly representative of many contemporary low-income settings throughout East Africa 541 and beyond, the opportunity costs of educating children appear modest. Schooling primarily 542 trades-off with leisure rather than work time, especially in the relatively market-integrated 543 setting where livelihood shifts have reduced demand for farm work, and particularly for girls 544 who combine school work and domestic chores. This implies that the returns to education do 545 not need to be high, since there are only modest opportunity costs to be compensated, in 546 turn suggesting that parental decisions to educate children may be driven by alternative 547 forces not captured by the embodied capital framework. It also indicates that declines in work 548 may precede enrolment increases, rather than the other way around. This contradicts the 549 prevailing developmental paradigm narrative that schooling displaces work in childhood, an 550 idea which is extended to the proposition that school is the antidote to harmful child labor 551 (Bourdillon, Levison, and Myers 2010). 552 It is possible that low opportunity costs to schooling are more characteristic of contemporary 553 low-income high fertility African populations than historical Europe. Throughout Tanzania, 554 household labor requirements have shrunk in recent years following policies such as 555 villagization (ujamaa), and shifts towards less labor-intensive crops (Varkevisser 1973; Wijsen 556 and Tanner 2002). As we have noted, direct costs (i.e. school fees) have also declined as part 557

• f national efforts to achieve targets of universal education. However, historical analyses of

558 education uptake in Industrial England similarly contradict the view that the promotion of 559 schooling was driven primarily by economic returns to education. It has been argued that the 560 social risks of youth unemployment were a greater concern than child labor, because the 561

28 presence of young, undisciplined people with time on their hands could lead to civil unrest. 562 The promotion of compulsory education was therefore a way of controlling young people’s 563 time, rather than because school would be useful to children (Cunningham 1990; Horrell and 564 Humphries 1995). Thus, schooling may be better considered as a form of cooperative child 565 care, which frees parental time for other productive activities by reducing the burden of child 566 supervision and direct care. 567 If the uptake of education is not that costly, as in this context where both opportunity costs 568 and direct costs of schooling are modest, school enrolment can be high even in the absence 569

• f high returns. This has consequences for fertility decline, because investment in schooling

570 does not necessitate, or even necessarily incentivize, a switch to a ‘quality over quantity’ 571 focused parental investment strategy. Indeed, despite near universal primary school 572 enrolment and growing secondary school attendance in this population, fertility rates remain 573 high, suggesting most families perceive education and high fertility as compatible strategies. 574 Liddell, Barrett and Henzi (2003) also observed high and equal investment in education in 575 South Africa, in combination with high fertility, a situation they describe as ‘bet-hedging’; 576 parents invest in education in the hope that at least one child may benefit, but continue to 577 have many children to provide old-age security, household labor, and because large families 578 remain desirable (Liddell, Barrett, and Henzi 2003). Regarding high levels of fertility in 579 Cameroon, Johnson-Hanks writes: “Parents cannot reliably trade child quality for child 580 quantity, or predict that the foreign models of reproduction that now appear promising will 581 not fall apart tomorrow.” In transitioning contexts, parents therefore face uncertainty and 582 unpredictability, coupled with messages from development programs and policies that may 583 be at odds with their own experience. 584

29

5.2 Gendered caveats

585 While children’s work and the opportunity costs of schooling decrease with market 586 integration, these patterns apply predominantly to boys. In the village, boys and girls do 587 similar levels of, but very different types of work, with boys specializing in farming and 588 herding, and girls in household chores. The opportunity costs of boys’ work appear to be much 589 higher than those of girls, particularly at younger ages, and this is reflected in boys’ lower 590 school enrolment rates in the village. Lower enrollment of boys is an unexpected pattern, 591 given the typically assumed greater economic pay-offs to male wage-labor and the intense 592 international focus on undereducated girls (United Nations 2015a). Yet, other studies have 593 also recently documented a ‘male disadvantage’ in education in pastoralist settings in both 594 Kenya and north-eastern Tanzania (Hedges et al. 2016; Mburu 2016; Siele, Swift, and Kratli 595 2013; Sperling and Galaty 1990). Our results suggest this trend is driven by the relative 596 compatibility of girl’s household chores with school attendance, whereas sending boys to 597 school involves foregoing the income from cattle, or the expense of employing a herdsboy. 598 Market integration, and associated livelihood changes, appears to reduce the demand for 599 children’s farm work, and incentivize investment in education. However, there is little 600 reallocation of boys’ farm work time to other forms of work, whereas girls’ time spent working 601 is relatively unaffected by market integration, particularly at older ages. Despite the fact that 602 girls do more productive work than boys in the town, there are no gender differences in 603

• verall enrolment or in time in education, with girls sacrificing leisure time instead. This

604 contradicts the assumption of a direct trade-off between work and education, and suggests 605 the returns to education may also differ by gender in market-integrated communities. 606

30 Previous studies have reported evidence that market integration favors girls’ education, with 607 preferential investment in girls’ schooling due to better opportunities for educated women’s 608 marriage or salaried employment (Bereczkei and Dunbar 1997; Neill 2010). Where bride price 609 is practiced, as it is here, educated girls can have higher bride prices, incentivizing parents to 610 invest in daughters’ schooling (Ashraf et al. 2015). At an individual level, educated women are 611 better protected from poverty and divorce than educated men, and girls, and their families, 612 may see school as a pathway to greater status and opportunities than they would traditionally 613 be allowed (Grant and Behrman 2010; Mcdaniel 2012). Thus, newly emerging benefits may 614 balance out the costs of girls’ education, particularly as school attendance does not require 615 parents to forego girls’ work, resulting in enrolment levels that are similar, or even higher, 616 than boys. This fits with a pattern that is receiving increasing attention, that of a female 617 advantage in education. It is now well documented in Europe and North America that girls do 618 better than boys at all levels of education, and a female advantage is now seen even in areas 619 with large historical gender gaps favoring males (Grant and Behrman 2010; Mcdaniel 2012). 620

5.3 ‘Double shifts’ and idle hands

621 Finally, we emphasize the importance of taking a holistic perspective on children’s time 622

• allocation. A narrow focus on work and school alone conceals the impact of market

623 integration on children’s leisure time. While often overlooked by both theoretical and policy- 624 grounded research on childhood, leisure and social time is an important, and at times 625 dominant component of childhood experience, and this time may have important, yet rarely 626 considered implications for child health, well-being, and achievement (Bock and Johnson 627 2004). 628

31 Our results indicate that girls sacrifice leisure time, and combine education with household 629

• work. This situation, where gender equality in the public sphere (school) has been achieved,

630 at least superficially, but gender differences remain in the private sphere (household), echoes 631 the ‘double shift’ seen in many ‘modern’ economies, in which women combine full-time work 632 with responsibility for unpaid household work and childcare (Hochschild and Machung 1989; 633 Mcdaniel 2012). Though a female advantage in education is increasingly seen, it is important 634 to note that gender pay gaps persist, and the highest status jobs continue to be held by men, 635 and the ‘double shift’ may be an important barrier to women achieving at the same levels as 636

• men. Girls’ household chores may negatively impact their wellbeing, and may interfere with

637 time available for private study or homework, potentially impeding their long-term academic 638 achievement (UNICEF 2016). 639 Conversely, there may also be problems associated with excess leisure. If leisure time results 640 from a lack of alternative time uses, then young people may lack opportunities to gain 641 embodied capital, and may occupy their time in more harmful activities. During fieldwork, 642 discussions with local parents and adolescents frequently mentioned adolescent boys 643 “kuzurura” (wandering the streets) as a problem, associated with risky behaviors such as 644 drinking alcohol, smoking marijuana and engaging in casual sex. For adolescent boys, these 645 groups may be an important way to build social capital and may provide immediate 646

• pportunities for earning cash, but in the long-term may result in poor health and

647 employment outcomes. In this case, underemployment may be associated with worse 648

• utcomes than working (Deb and Rosati 2002).

649

5.4 Limitations

650

32 Self-report data from children may give inaccurate estimates of school attendance and work 651 time, with previous studies indicating that children may overestimate their work hours. For 652 ease of interviewing, and to avoid placing too much burden on young respondents, half-hours 653 were the smallest unit of time collected. This could lead to an under-representation of tasks 654 requiring a small amount of time, such as washing, eating, or small household chores. Children 655 and parents may also value time differently, and activities framed as work by children may be 656 classed as play by adults. Thus, this data provides a ‘child’s eye’ view of children’s 657 contributions to their households. This dataset is also a snapshot of a single day in a child’s 658 life, and so cannot account for seasonal variation, or strategies families may employ to 659 ameliorate the trade-off between work and school, such as working on weekends or during 660 school holidays. However, the approach of using only the previous school day enabled us to 661 compare children’s time allocation based on the specific criteria that all children could have 662 used that time either for education or for work. Additionally, we were not able to interview 663 children absent from the area, and so lack data on children with more ‘extreme’ outcomes, 664 such as those away at boarding school, or working in other locations. These limitations aside, 665 we believe the strength of this study lies in providing a large-sample and holistic case study 666

• f children’s time use in an area broadly representative of contemporary rural Tanzania.

667

5.5 Conclusion and Implications

668 Embodied capital models dominate contemporary research into the impact of modernization 669

• n parental investment and reproductive strategies, particularly in evolutionary anthropology

670 and demography (Sear et al. 2016). Yet, available data on patterns of educational investment 671 and children’s work, presented here and elsewhere, provide mixed support for assumptions 672 about the costs and benefits of education, and the consequent motivations for limiting 673

33

• fertility. Indeed, many contemporary low-income populations have both high school

674 enrollment and high fertility, supporting the view that low opportunity costs of schooling are 675 an important explanatory factor behind stalled fertility declines. This conclusion echoes wider 676 concerns that historical processes need not necessarily be reflected in current and future 677 patterns of change (Thornton 2001). 678 The developmental paradigm presents a narrative of economic development, founded 679 partially on classic economic models of the demographic transition, in which education is 680 highly desirable, and provides children with the skills needed to participate in a global 681 economy, and societies with the means to reduce poverty. As a result of this focus on the 682 economic returns to education, interventions have predominantly sought to reduce the costs 683 associated with school, such as removing fees or giving conditional cash transfers. However, 684 getting children into school is not enough to foster meaningful embodied capital that will lead 685 to adult success. Anthropologists can offer a broader, more neutral perspective on what 686 childhood is ‘for’, and what skills could help to bridge the gap between subsistence and 687 market-integrated economies. Better connections between anthropological research, and the 688 development sector could serve to develop interventions to promote locally relevant 689 education and enhance program effectiveness (Gibson & Lawson 2015). For example, policy- 690 makers should be aware that efforts to reach ‘universal education’ by making schooling 691 affordable may ameliorate perceived trade-offs between offspring quantity and quality, 692 reducing incentives for low fertility. 693 Lastly, our analyses make clear that the impact of market integration on childhood cannot be 694 understood without considering gender. Our analyses indicate that relatively low school 695 enrollment for boys in rural areas, especially in herding communities, is driven by opportunity 696

34 costs, while low enrolment among boys in more urban areas is more likely due to poor future 697 employment prospects. Parents in this population, and elsewhere, are increasingly educating 698 daughters, often more than their sons, a pattern which may be driven by relatively low 699

• pportunity costs and emerging employment opportunities for young women. While this

700 positive trend should be celebrated, we caution that for girls, school attendance involves a 701 double shift of household chores and school work, and sacrifice of leisure time, with unknown 702 consequences for their wellbeing. More holistic studies of the costs and benefits of children’s 703 time allocation, that fully explore children’s time beyond the most obviously ‘functional’ 704 behaviors of work and schooling, will provide better understanding of how best to promote 705 positive outcomes across all dimensions of children’s lives. 706

35

Acknowledgements

We thank the National Institute of Medical Research, Mwanza, for supporting this study, our fieldwork team, especially Holo Dick, Pascazia Simon, and Vicky Sawalla who conducted interviews, and Adam Mapuli, Isaac Sengerema, and Christopher Joseph who facilitated household introductions in the villages. We especially thank all our participants from Kisesa and Welamasonga, and the headteachers who allowed us to conduct interviews in their

• schools. This study was funded by the UK Economic and Social Research Council, the Wenner-

Gren Anthropological Foundation, the International Society for Human Ethology, the Parkes Foundation, and the European Human Behaviour and Evolution Association.

36 Table 1: Activities mentioned during time allocation interview Code Activity (% of children who mentioned activity) Education Going to school, including travel time (71.7); studying (18.6); tuition (3.1) Chores Washing dishes (40.1); fetching water (38.2); cooking (27.7); sweeping (20.6); washing clothes (6.9); going to the market (6.7); lighting the fire (3.7); cleaning (3.3); collecting firewood (3.1); carrying baby (2.3); washing baby (1.7); going to the shop (2.1); running errands (1.9); milling flour (1.4); mopping (0.9); going to the mill (0.9); processing cassava (0.9); watching children (0.9); processing corn (0.6); food preparation (0.5); milling rice (0.2); folding clothes (0.2); tidying (0.2) Farm work Farming (15.0); herding (6.4); milking cows (1.7); picking vegetables (0.9); picking grass (0.4); animal care (0.4); watering crops (0.3); harvesting rice (0.2); weeding crops (0.2) Leisure Sleeping (100.0); eating (99.7); washing (82.2); resting (65.9); playing (27.2); walking (4.1); watching TV (3.1); drinking uji (gruel) (2.3); drinking tea (1.5); praying (0.9); visiting (0.8); going to church/mosque/funeral (0.7); taking medicine (0.5); watching football (0.5); going to hospital (0.4); having hair braided (0.2) Market work Petty trading (selling peanuts/sugarcane/cassava/uji/ tomatoes/soap/tea/doughnuts etc.) (1.1); working at shop (0.9); hauling sand (0.5); chopping wood (0.2); dancing (0.2); making things to sell (baskets/rope/bricks/doughnuts/ice lollies/CDs) (0.5); running market errands (0.3); working at hotel (0.2); being a DJ (0.1); mending shoes (0.1)

37 Table 2: Sample size and description of child education outcomes and household characteristics. Village Town Total Sample: Number of households 234 222 456 Number of children aged 7–19 768 619 1,387 Number of children interviewed 740 538 1,278 Mean household size (SD) 8.0 (2.9) 7.1 (3.2) 7.6 (3.1) Mean number of children aged 7–19 per household (SD) 3.3 (1.7) 2.7 (1.8) 3.0 (1.7) Education outcomes: Ever enrolled (%) 702 (91.4) 608 (98.2) 1,310 (94.5) Currently enrolled (%) 574 (74.7) 528 (85.3) 1,102 (79.5) Progressed (%; 14–19-year-olds) 80 (30.9) 196 (72.3) 276 (52.1) Attended on previous day (%; currently enrolled and followed up

• nly)

490 (87.2) 417 (88.7) 907 (87.9) Mean years of education (SD; previously enrolled) 5.9 (2.5) 8.1 (2.8) 6.6 (2.8) % households: Household head has some education 71.7 85.9 78.7 Owning land 95.3 72.5 84.2 Growing crops 96.2 47.3 72.4 Owning cattle 43.6 7.2 25.9 With a formal business or shop 6.0 36.5 20.8 Household member has a salaried job 1.7 12.6 7.0 With electricity 2.1 50.5 25.7 With water source on own land 3.4 36.0 19.3 Classed as ‘severely food insecure’ 50.4 48.4 49.5

38 Table 3: Results from logistic regression models of educational outcomes; ever enrolled in school, currently enrolled in school (whole sample), and progressed to secondary school (for 14–19 year olds only). Ever enrolled Currently enrolled Progressed (14–19-year-olds) Town 12.22** 7.00** 5.86** [3.67,40.72] [3.83,12.82] [3.40,10.10] Female 1.98* 1.82* 1.27 [1.06,3.72] [1.14,2.88] [.74,2.18] Residence#gender interaction .27+ .41* 1.22 [.06,1.20] [.20,.87] [.56,2.64] Household food security 1.05+ 1.04* 1.06** [.99,1.10] [1.01,1.08] [1.02,1.09] Age 1.74** .58** 1.25** [1.50,2.02] [.54,.63] [1.12,1.41] Constant .01** 2099.90** .00** [.00,.07] [578.02,7628.80] [.00,.03] Random intercept for household .96 .83 [.48,1.90] [.52,1.34] N 1,367 1,367 523 + p<0.10, * p<0.05, ** p<0.01 Exponentiated coefficients presented; 95% confidence intervals in brackets

39

References

Abebe, Tatek. 2007. “Changing Livelihoods, Changing Childhoods: Patterns of Children’s Work in Rural Southern Ethiopia.” Children’s Geographies 5 (1–2): 77–93. Ansell, Nicola. 2005. Children, Youth and Development. Routledge. Ashraf, Nava, Natalie Bau, Nathan Nunn, and Alessandra Voena. 2015. “Bride Price and Female Education.” Assaad, Ragui, Deborah Levison, and Nadia Zibani. 2010. “The Effect of Domestic Work on Girls’ Schooling: Evidence from Egypt.” Feminist Economics 16 (1): 79–128. Bacolod, Marigee P., and Priya Ranjan. 2008. “Why Children Work, Attend School, or Stay Idle: Theory and Evidence.” Economic Development and Cultural Change 56 (4): 791–828. Becker, Gary S. 1960. “An Economic Analysis of Fertility.” In Demographic and Economic Change in Developed Countries, edited by NBER (National Bureau of Economic Research), 209–40. Columbia University Press. Bedasso, B. E. 2008. “Investing in Education as a Means and as an End: Exploring Microfoundations of the MDGs.” United Nations Economic Commission for Africa. Beegle, Kathleen, Rajeev H Dehejia, Roberta Gatti, and Sofya Krutikova. 2008. “The Consequences of Child Labor: Evidence from Longitudinal Data in Rural Tanzania.” 4677. Bereczkei, Tamas, and Robin Dunbar. 1997. “Female-Biased Reproductive Strategies in a Hungarian Gypsy Population.” Proceedings of the Royal Society London B 264: 17–22. ———. 2002. “Helping-at-the-Nest and Sex-Biased Parental Investment in a Hungarian Gypsy

40 Population.” Current Anthropology 43 (5): 804–9. Biggeri, M., L. Guarcello, S. Lyon, and F. C. Rosati. 2003. “The Puzzle of ‘idle’ Children: Neither in School nor Performing Economic Activity. Evidence from Six Countries.” Bird, Douglas W., and Rebecca Bliege Bird. 2002. “Children on the Reef: Slow Learning or Strategic Foraging.” Human Nature 13 (2): 269–97. Bledsoe, Caroline, John Casterline, Jennifer Johnson-Kuhn, and John Haaga. 1999. Critical Perspectives on Education and Fertility in the Developing World. Washington, DC. Bliege Bird, Rebecca, and Douglas W. Bird. 2002. “Constraints of Knowing or Constraints of Growing?: Fishing and Collecting by the Children of Mer.” Human Nature 13 (2): 239–67. Bock, John. 1999. “Evolutionary Approaches to Population: Implications for Research and Policy.” Population and Environment 21 (2): 193–222. ———. 2002a. “Evolutionary Demography and Intrahousehold Time Allocation: School Attendance and Child Labor among the Okavango Delta Peoples of Botswana.” American Journal of Human Biology : The Official Journal of the Human Biology Council 14 (2): 206– 21. ———. 2002b. “Learning, Life History and Productivity: Children’s Lives in the Okavango Delta, Botswana.” Human Nature 13 (2): 161–97. Bock, John, and Sara E. Johnson. 2004. “Subsistence Ecology and Play among the Okavango Delta Peoples of Botswana.” Human Nature 15 (1): 63–81. Borgerhoff Mulder, Monique. 1998. “Brothers and Sisters.” Human Nature 9 (2): 119–61.

41 Bourdillon, Michael, Deborah Levison, and William Myers. 2010. Rights and Wrongs of Children’s Work. Rutgers University Press. Bourdillon, Michael, Deborah Levison, Ben White, and William E Myers. 2015. “A Place for Work in Children’s Lives?” Brunette, Waylon, Mitchell Sundt, Nicola Dell, Rohit Chaudhri, Nathan Breit, and Gaetano

• Borriello. 2013. “Open Data Kit 2.0 : Expanding and Refining Information Services for

Developing Regions.” In ACM HotMobile. Jekyll Island, Georgia, USA. Buis, Maarten L. 2017. “FMLOGIT: Stata Module Fitting a Fractional Multinomial Logit Model by Quasi Maximum Likelihood.” Stata Software Component. Cain, Mead T. 1977. “The Economic Activities of Children in a Village in Bangladesh.” Population and Development Review 3 (3): 201–27. Clarke, P. 2008. “When Can Group-Level Clustering Be Ignored? Multi-Level versus Single- Level Models with Sparse Data.” Journal of Epidemiology & Community Health 62: 752– 58. Coates, Jennifer, Anne Swindale, and Paula Bilinsky. 2007. “Household Food Insecurity Access Scale (HFIAS) for Measurement of Household Food Access: Indicator Guide (v. 3).” Cronk, Lee. 1989. “Low Socioeconomic Status and Female-Biased Parental Investment: The Mukogodo Example.” American Anthropologist 91 (2): 414–429. Cunningham, Hugh. 1990. “The Employment and Unemployment of Children in England 1680- 1851.” Past & Present 126: 115–50. Deb, P., and F. C. Rosati. 2002. “Determinants of Child Labor and School Attendance: The Role

42

• f Household Unobservables.”

Esquivel, Valeria, Debbie Budlender, Nancy Folbre, and Indira Hirway. 2008. “Explorations: Time-Use Surveys in the South.” Feminist Economics 14 (3): 107–52.

• FAO. 2014. “Gender Inequalities in Rural Employment in Tanzania Mainland: An Overview.”

Rome. GEM Report. 2015. “Tanzania Abolishes Secondary School Fees. But Does Anything Come for Free?” World Education Blog. Gibbons, Elizabeth D., Friedrich Huebler, and Edilberto Loaiza. 2005. “Child Labour, Education, and the Principle of Non-Discrimination.” November. New York. Gibson, Mhairi A., and David W. Lawson. 2011. “‘Modernization’ Increases Parental Investment and Sibling Resource Competition: Evidence from a Rural Development Initiative in Ethiopia.” Evolution and Human Behavior 32 (2). Elsevier Inc.: 97–105. ———. 2015. “Applying Evolutionary Anthropology.” Evolutionary Anthropology 24 (1): 3–14. Gibson, Mhairi A., and Rebecca Sear. 2010. “Does Wealth Increase Parental Investment Biases in Child Education?” Current Anthropology 51 (5): 693–701. Goodman, A., I. Koupil, and D. W. Lawson. 2012. “Low Fertility Increases Descendant Socioeconomic Position but Reduces Long-Term Fitness in a Modern Post-Industrial Society.” Proceedings of the Royal Society B: Biological Sciences 279 (August): 4342–51. Grant, Monica J., and Jere R. Behrman. 2010. “Gender Gaps in Educational Attainment in Less Developed Countries.” Population and Development Review2 36 (1): 71–89.

43 Gurven, Michael, and Hillard Kaplan. 2006. “Determinants of Time Allocation across the Lifespan: A Theoretical Model and an Application to the Machiguenga and Piro of Peru.” Human Nature 17 (1): 1–49. Gurven, Michael, Hillard Kaplan, and Maguin Gutierrez. 2006. “How Long Does It Take to Become a Proficient Hunter? Implications for the Evolution of Extended Development and Long Life Span.” Journal of Human Evolution 51 (5): 454–70. Hedges, S., M. Borgerhoff Mulder, S. James, and D.W. Lawson. 2016. “Sending Children to School: Rural Livelihoods and Parental Investment in Education in Northern Tanzania.” Evolution and Human Behavior 37 (2). Hivos/Twaweza. 2014. “Are Our Children Learning? Literacy and Numeracy Across East Africa.” Nairobi, Kenya. Hochschild, Arlie Russell, and Anne Machung. 1989. The Second Shift: Working Families and the Revolution at Home. London: Piatkus. Horrell, Sara, and Jane Humphries. 1995. “‘The Exploitation of Little Children’: Child Labor and the Family Economy in the Industrial Revolution.” Explorations in Economic History. Ilahi, Nadeem. 2000. “The Intra-Household Allocation of Time and Tasks: What Have We Learnt from the Empirical Literature?” 13. Gender and Development.

• ILO. 2004. “Child Labour Statistics: Manual on Methodologies for Data Collection through

Surveys.” Geneva. ———. 2013. “Marking Progress against Child Labour.” Janzen, Sarah A. 2015. “Child Labor Measurement: Who Should We Ask?”

44 Johnson-Hanks, Jennifer. 2004. “Uncertainty and the Second Space: Modern Birth Timing and the Dilemma of Education.” European Journal of Population 20: 351–73. Kaplan, Hillard. 1996. “Theory of Fertility and Parental Investment.” Yearbook of Physical Anthropology 39: 91–135. Kaplan, Hillard, and John Bock. 2001. “Fertility Theory: The Embodied-Capital Theory of Life History Evolution.” In International Encyclopedia of the Social and Behavioural Sciences, edited by J. M. Hoem, 1:5561–68. Elsevier Science. Kaplan, Hillard, John Bock, and Paul L. Hooper. 2015. “Fertility Theory: Embodied-Capital Theory of Life History Evolution.” In International Encyclopedia of the Social & Behavioral Sciences, Second Edi, 28–34. Elsevier. Kaplan, Hillard, Jane B. Lancaster, Sara E. Johnson, and John A. Bock. 1995. “Does Observed Fertility Maximise Fitness among New Mexican Men? A Test of an Optimality Model and a New Theory of Parental Investment in the Embodied Capital of Offspring.” Human Nature 6 (4): 325–60. Kishamawe, Coleman, Raphael Isingo, Baltazar Mtenga, Basia Zaba, Jim Todd, Benjamin Clark, John Changalucha, and Mark Urassa. 2015. “Health & Demographic Surveillance System Profile: The Magu Health and Demographic Surveillance System (Magu HDSS).” International Journal of Epidemiology, no. September: 1851–61. Kramer, Karen L. 2002. “Variation in Juvenile Dependence: Helping Behavior among Maya Children.” Human Nature 13 (2): 299–325. Kramer, Karen L. 2005. “Children’s Help and the Pace of Reproduction: Cooperative Breeding

45 in Humans.” Evolutionary Anthropology: Issues, News, and Reviews 14 (6): 224–37. Lancy, David F. 2012. “The Chore Curriculum.” In African Children at Work: Working and Learning in Growing Up for Life, edited by Gerd Spittler and Michael F. C. Bourdillon, 23–

• 56. Lit Verlag.

Lancy, David F. 2010. “Learning ‘From Nobody’: The Limited Role of Teaching in Folk Models

• f Children’s Development.” Childhood in the Past 3: 79–106.

Lawson, David W., and Monique Borgerhoff Mulder. 2016. “The Offspring Quantity-Quality Trade-off and Human Fertility Variation.” Philosophical Transactions of the Royal Society

• f London, B 371 (20150145).

Levison, Deborah, Karine S Moe, and Felicia Marie Knaul. 2001. “Youth Education and Work in Mexico.” World Development 29 (1): 167–88. Liddell, Christine, Louise Barrett, and Peter Henzi. 2003. “Parental Investment in Schooling: Evidence from a Subsistence Farming Community in South Africa.” International Journal

• f Psychology 38 (1): 54–63.

Malipula, Mrisho. 2015. “Depoliticised Ethnicity in Tanzania: A Structural and Historical Narrative.” Afrika Focus 27 (2): 49–70. Mattison, Siobhán M., and Dawn B. Neill. 2013. “The Effects of Residential Ecology on Patterns

• f Child Work and Mother’s Reproductive Success among Indo-Fijians.” Evolution and

Human Behavior 34 (3). Elsevier B.V.: 207–15. Mattison, Siobhán M., and Rebecca Sear. 2016. “Modernizing Evolutionary Anthropology.” Human Nature 27 (4): 335–50.

46 Mburu, Samuel. 2016. “Effects of Livestock Herd Migration on Child Schooling in Marsabit District, Kenya.” Compare: A Journal of Comparative and International Education. Mcdaniel, Anne. 2012. “Women’s Advantage in Higher Education: Towards Understanding A Global Phenomenon.” Sociology Compass 6 (7): 581–95.

• MoHCDGEC. 2016. “Tanzania Demographic and Health Survey and Malaria Indicator Survey,

2015-16.” Dar es Salaam, Tanzania and Rockville, Maryland, USA. Mukudi, Edith. 2002. “Gender and Education in Africa.” Comparative Education Review 46 (2): 234–41. Murdock, George P., and Caterina Provost. 1973. “Factors in the Division of Labor by Sex: A Cross-Cultural Analysis.” Ethnology 12 (2): 203–25. Mwinsheikhe, Halima Mohammed. 2009. “Spare No Means: Battling with the English/Kiswahili Dilemma in Tanzanian Secondary School Classrooms.” In Languages and Education in Africa: A Comparative and Transdisciplinary Analysis, edited by Birgit Brock-Utne and Ingse Skattum, 223–36. Symposium Books Ltd. Neill, Dawn B. 2010. “Expanding Opportunity Structures: Parental Investments in Education, Migration, and Extrinsic Risk Reduction among Indo-Fijians.” Human Nature 21 (2): 165– 85. ———. 2011. “Urbanization and Daughter-Biased Parental Investment in Fiji.” Human Nature 22 (1–2): 139–55. Nieuwenhuys, Olga. 1993. “To Read and Not to Eat: South Indian Children between Secondary School and Work.” Childhood 1: 100–109.

47 ———. 1996. “The Paradox of Child Labor and Anthropology.” Annual Review of Anthropology 25 (1): 237–51. Pörtner, Claus C. 2016. “Effects of Parental Absence on Child Labor and School Attendance in the Philippines.” Review of Economics of the Household 14 (1): 103–30.

• PRB. 2016. “Accelerating Family Planning Progress in Tanzania: What Does the 2015-16

Demographic and Health Survey Tell Us?” Pritchett, Lant H. 2013. The Rebirth of Education: Schooling Ain’t Learning. Center for Global Development. Ravallion, Martin, and Quentin T. Wodon. 2000. “Does Child Labour Displace Schooling? Evidence on Behavioural Responses to an Enrollment Subsidy.” The Economic Journal 110 (462). Sear, Rebecca. 2011. “Parenting and Families.” In Evolutionary Psychology: A Critical Introduction, edited by Viren Swami, 215–50. Wiley-Blackwell. Sear, Rebecca, David W. Lawson, Hillard Kaplan, and Mary K. Shenk. 2016. “Understanding Variation in Human Fertility: What Can We Learn from Evolutionary Demography ?” Philosophical Transactions of the Royal Society of London, B 371 (20150144). Sellen, Daniel W. 2003. “Nutritional Consequences of Wealth Differentials in East African Pastoralists: The Case of the Datoga of Northern Tanzania.” Human Ecology 31 (4): 529– 70. Sieff, Daniela F. 1997. “Herding Strategies of the Datoga Pastoralists of Tanzania: Is Household Labor a Limiting Factor.” Human Ecology 25 (4): 519–44.

48 Siele, D., J. Swift, and S. Kratli. 2013. “Reaching Pastoralists with Formal Education: A Distance- Learning Strategy for Kenya.” In Pastoralism and Development in Africa: Dynamic Change at the Margins, edited by A. Catley, J. Lind, and I. Scoones, 206–14. Routledge. Sperling, L., and J. G. Galaty. 1990. “Cattle, Culture, and Economy: Dynamics in East African Pastoralism.” In The World of Pastoralism: Herding Systems in Comparative Perspective, edited by J. G. Galaty and D. L. Johnson, 69–98. New York: Guildford Press. Thornton, Arland. 2001. “Developmental Paradigm, Reading History Sideways, and Family Change.” Demography 38 (4): 449–65. Turke, Paul W. 1988. “Helpers at the Nest: Childcare Networks on Ifaluk.” In Human Reproductive Behaviour: A Darwinian Perspective, edited by Laura Betzig, Monique Borgerhoff Mulder, and Paul W. Turke, 173–88. Cambridge University Press.

• UNESCO. 2011. “Tanzania Education Sector Analysis: Beyond Primary Education, the Quest

for Balanced and Efficient Policy Choices for Human Development and Economic Growth.” Dar Es Salaam. ———. 2015. “Education For All 2000-2015: Achievements and Challenges.” Paris.

• UNICEF. 2016. “Harnessing the Power of Data for Girls: Taking Stock and Looking Ahead to

2030.” New York. United Nations. 2015a. “Sustainable Development Goals Thematic Clusters: Youth, Education and Culture.” ———. 2015b. “United Nations Millennium Development Goals.”

• USDoL. 2013. “Findings on the Worst Forms of Child Labour: Tanzania.” Washington, DC.

49 Varkevisser, Corlien M. 1973. Socialization in a Changing Society: Sukuma Childhood in Rural and Urban Mwanza, Tanzania. The Hague: Centre for the Study of Education in Changing Societies. Wijsen, Frans, and Ralph Tanner. 2002. “I Am Just a Sukuma”: Globalization and Identity Construction in Northwest Tanzania. Zelizer, Viviana A. 1985. Pricing the Priceless Child. Princeton University Press.