Planned Population Redistribution and its Impact on Family Formation - - PDF document

Planned Population Redistribution and its Impact on Family Formation in Rwanda

Jessica Marter-Kenyon, MSc and Stuart Sweeney, PhD Department of Geography, University of California at Santa Barbara Submitted to the IPC 2017 on September 30, 2017 DRAFT

Abstract:

For over fifteen years, Rwanda has pursued an aggressive program of villagization (or population redistribution) with the stated intent of moving 90% of its rural population to small, clustered villages (or imidigudu). One of the aims of villagization is to help resolve population pressures, including high rates of rural population growth. In this paper, we empirically investigate the impacts of villagization on family formation (marriage and childbearing). Data for this study come from a household survey (N=2,049) designed and implemented by the authors in 2016-17 in four districts of Rwanda. We use Cox proportional hazard models to evaluate differences in the duration to first union and first birth according to settlement

• type. We find that men and women living in planned villages form unions more quickly than their

counterparts living in isolated settlements (40% and 35% faster, respectively). The effect size of villagization on union formation is moderated by women’s education. We find that women living in planned villages have a duration to first birth 44% shorter than those who have not been villagized. The significant challenges of population growth, environmental change and food insecurity across rural Africa can be confronted through a variety of direct and indirect state interventions. Our study adds rare evidence to an understanding of the effectiveness of resettlement as a demographic strategy in the 21st

• century. This study also adds to the theoretical debate on the impact of social and economic upheaval on

family formation, particularly as far as how planned displacement and resettlement affects this behavior. Keywords: family formation, first union, first birth, population policy, resettlement, villagization, Rwanda

• 1. Introduction:

Fertility transitions are happening in most countries across sub-Saharan Africa. Regardless of concern that the region is lagging, and some national transitions are stalling, fertility rates overall have fallen from 6.8 to 5.1 since 1980 (Bongaarts & Casterline, 2013). Some countries, like Rwanda, have halved their TFRs in the same period and are now nearing the middle phases of the demographic transition. Much

• f this change has been achieved through concerted efforts by national governments aimed at increasing

the age at first union and first birth (aka the onset of family formation). Reduced exposure to pregnancy and marriage have been linked to population policies that target increased educational and economic

• pportunities for girls, increased access to family planning services, and sensitization around ideal family
• size. Studies have also shown that urbanization, industrialization, and socioeconomic development can

catalyze fertility transitions (Bongaarts & Casterline, 2013). Less studied is the effect of population redistribution policy on demographic outcomes including fertility and family formation. This is despite the fact that firstly, sedentarization, villagization, mobility restrictions, and land redistribution policies have been common features of population, environment, and development strategy in many African countries since at least the colonial era; and, secondly, involuntary resettlement programs are well-known to have profound effects (both direct and indirect) for a wide

variety of demographic outcomes including fertility, but also mortality, morbidity, and migration. Population redistribution policies were especially common in Africa from the 1950s-1980s and have reemerged in the last two decades. Rwanda and Ethiopia currently have projects each involving more than ten million farmers. This paper assesses the relationship between villagization and family formation using the case of Rwanda and data from a survey we conducted in 2016-17 with 2,049 rural households living in either traditional or planned communities. Population redistribution is a central element of national development strategy in Rwanda; one of the many desired outcomes of the policy is fertility reduction. Despite successful reductions in fertility over the past decade, population growth and its consequences in the context of climate change and land scarcity are still of great concern, and cited as producing extraordinary circumstances demanding aggressive action in the rural sector (Government of Rwanda, 2009). The government hopes that the villagization program will accelerate the countryside’s demographic transition by ‘modernizing’ its people and land. Although the villagization (or imidugudu) policy has been going on for over fifteen years (tracking the decline of fertility in the country), studies of its effects are few and far between and have dwindled in number since the end of the humanitarian crisis in the early 2000s. None have investigated its demographic outcomes. Our research draws on, and further contributes to, demographic and geographic theory and methods and has important implications for population policies in sub-Saharan Africa and in other developing countries globally. The significant challenges of population growth, environmental change and food insecurity across rural Africa can be confronted through a variety of direct and indirect state

• interventions. There is a long-standing, and ongoing, policy debate about what degree of intervention is

best (May, 2012); our study adds rare evidence to an understanding of the effectiveness of resettlement as a demographic strategy in the 21st century. This study also adds to the theoretical debate on the impact

• f social and economic upheaval on family formation, particularly as far as how planned displacement and

resettlement affects this behavior.

• 2. Background: Population Growth, Policy and Villagization in Rwanda

Fertility reduction is a major pillar of sustainable development policy in Rwanda. At 441 inhabitants/km2, the country has the highest population density in Africa (IFAD, 2017). The government has invested heavily in measures aimed at reducing demographic growth, and continues to do so. Still, the country’s population growth rate of 2.3% is worryingly high: between 2015 and 2025, the total population size is projected to increase from 11.6 million to 14.6 million people (IFAD, 2017). The majority

• f the population is rural and dependent on smallholder agriculture, as is the case across most of the
• continent. Population pressure and land scarcity were linked to the 1994 genocide. Since then, farm sizes

have continued to dwindle: the average landholding is 0.3 hectares (IFAD, 2017). Along with climate change, the threat of future violence adds an additional gravity to the population issue in Rwanda. In a very short space of time, Rwanda has made massive strides towards curbing demographic

• growth. The total fertility rate (TFR) declined slightly- from 6.2 to 5.8- between 1992 and 2000, in part due

to effects of the genocide. Contraceptive use was very low, and the new government initially held back from initiating any policies to stem population growth: firstly, they did not want to be seen as replicating the previous (genocidal) government’s policies; secondly, family planning, population control, and discussion about demographic growth were all relatively taboo topics for several years following the

genocide (May & Kamarusi, 2009). In 2005, the Rwanda Demographic and Health Survey was released. It showed that the TFR had in fact risen to 6.1. The government stepped into high gear. Today, Rwanda’s fertility and family formation indicators are typically higher than its neighbors in the region (Kenya, Uganda, Tanzania, Burundi, the DRC), and more likely expected from a much higher income country. The TFR (see Figure 1) has dropped- in just ten years- from 6.1 to 4.2 (National Institute of Statistics of Rwanda, 2011). Rwanda is now considered to be at the mid-phase of its fertility transition (Muhoza et al., 2014). The most recent national census shows that the average age of first marriage in Rwanda is the highest in the region and has increased steadily over the last two decades, to 25 years for women and 27 years for men (National Institute of Statistics of Rwanda, 2014). The Demographic and Health Survey (DHS) indicates median age at first birth increased (somewhat modestly, but consistently) from 21.5 to 22.7 from 1992 to 2015, and the age at first marriage for women increased from 20 to 21.9 over the same period (see Figure 2). Studies of Rwanda have linked the rapid transition in fertility, age at first marriage, and age at first birth mainly to the government’s strong commitment after 2005 to stemming growth and its concerted efforts at population policy intervention (Rutayisire, 2014). Support from the international donor community was also critical (Muhoza et al., 2014). Investment in family planning services was a primary focus- it is not uncommon to see the packaging from Depo Prevera shots used as wallpaper in rural

• latrines. Women (rural and urban) using modern contraception rose from 10% in 2005 to 45% in 2010

(National Institute of Statistics of Rwanda, 2011). Sensitization programs focused on land scarcity, population pressure, and the benefits of smaller family sizes. Strategies targeted religious organizations, first reducing their own stigma around birth control (BBC, 2007). A big campaign focused on convincing people that three was the ideal number of a children (Muhoza et al., 2014). By 2010, Rwanda had the lowest desired family size of the region, at 3.3- down from 4.3 just five years prior (Muhoza et al., 2014). Facilitated by the fact that mandatory attendance is required at multiple monthly government meetings (and further assisted by villagization!). The government also links family formation to women’s education, noting that women with a secondary or higher education get married and give birth to their first child three to three and a half years later than women who have never attended school (National Institute of Statistics of Rwanda and ICF International, 2012). Rwanda has also identified population growth as a cross- sectoral issue; its Vision 2020 plan and associated sustainable development strategies aim to help by increasing employment in the non-agricultural sectors, promoting urbanization, increasing the uptake of community-based health insurance (Rutayisire, 2014). One major aspect of the Rwandan government’s plan has been overlooked by most external

• bservers, and completely ignored by demographers. Specifically, Rwanda claims that its national

villagization policy is a key element in controlling population growth. Since 2000, the government has sought to move 90% of the rural population into clustered villages known as imidugudu. Government reports indicate 81% have relocated, amounting to some 7.6 million people (see Figure 3) (Rwanda Housing Authority, 2014; Isaksson, 2013). More conservative estimates would still place this number at around 6 million individuals, or 1.4 million households (Hahirwa, 2013). Villagization is intended to bring modernity- and modern services- to rural Rwanda and, in so doing, to reduce fertility (May & Kamarusi, 2009). Grouping households together is expected to improve the efficiency of land use, infrastructure development, market access, and service delivery. In bringing these ‘benefits of urbanization and modernization’ to rural areas, the government is hoping for subsequent impacts on demographic

• utcomes including fertility and family formation (Government of Rwanda, 2009). No research has linked

the decline to the population redistribution policy, however. This reflects a general absence of pop redistribution and land reform in the literature on family formation and demographic transition in Africa (and elsewhere).

• 3. Theoretical Context (3000)

This section provides a brief overview of the relevant literature. We begin with a discussion of what family formation is, why it matters to governments, and what kinds of policies they engage in to manage it. Next, we provide a discussion of the overlooked role played by population redistribution in catalyzing demographic change, including family formation. Finally, we propose three mechanisms via which the villagization policy in Rwanda may be increasing the age at which both men and women form unions and begin having children. 3.1 Family Formation and Population Policy in Sub-Saharan Africa (800) Family formation occurs when a person forms a union or marriage, or has a child. In the sub- Saharan African context, governments and other development actors are typically interested in increasing both the age at first birth and the age at first union, with the goal of reducing lifetime fertility and overall population growth. While women are technically at-risk of pregnancy from age at menarche to age of menopause, culture norms typically encourage child-bearing within formal or informal unions. This means that the age-at-first-union to menopause or to union dissolution are functionally the at-risk period, which is further constrained by contraceptive use. Any policy that delay the age at first union will be closely tied to age at first birth and will tend to decrease population level fertility by reducing exposure. Another important factor in fertility is ideational norms about family size. If a couple desires a family size of, say, 5 and their age at marriage is relatively late (compared to norms) they can decrease the spacing between children to achieve their desired family size... The sooner a woman has her first child, the longer she is exposed to the hazard of pregnancy. This can increase total fertility rates. For the same reason, increasing the age at first marriage can lead to fertility reductions in societies like Rwanda’s where childbearing (exposure to pregnancy) usually follows union formation (Westoff, 1990). Governments are obviously therefore interested in managing family formation because of its relationship to population growth. But, encouraging the postponement of family formation leads to individual and societal benefits beyond fertility reductions. There is a preponderance of evidence that delaying the age at first birth benefits maternal and child health outcomes, particularly when births to mothers under the age of 18 are avoided (Gibbs et al., 2012). Older parents tend to be more financially capable of providing for their children, which can lead to positive social and economic outcomes at multiple scales. At the national level, “countries with the highest rates of early marriage are also the countries with the highest rates of poverty and highest population growth rates” (Walker, 2012, p. 231). Early marriages (under the age of 18) often lead to decreased labor force participation for women and decreased school attendance for both men and women (Bloom & Reddy, 1986). Managing family formation through policy is not a straightforward task, however, in large part because the decision to form a union, or to bear children, is influenced by a plurality of economic, social, and physical factors (Rutayisire, 2014). Culture and religion play an important role (Jayaraman, 2009). People have to want fewer children, and have access to the methods to meet their preferences (Westoff, 1990). Social networks, and the opinions of neighbors and family members also seem to matter (Yabiku, 2006). Marriage can provide security or economic alleviation for families. Less directly, macro-economic,

social and environmental changes have influenced trends in family formation. Urbanization has been linked to decreasing fertility in SSA, as has rising age at first marriage (Jayarman, 2009). Outbreaks of violence can lead to dramatic fertility declines during the episode, followed by spikes in fertility once peace is established. Alternatively, violence can increase marriage in cases where it is used as a coping mechanism or a strategy to achieve personal security. Economic development and the movement of labor from agriculture to industry have been linked to demographic transitions underpinned by delayed family formation. Despite the magnitude and complexity of the task, governments invest a great deal of time and money in controlling population growth, often successfully (if not always humanely). Some of the determinants of family formation are more readily actionable than others: female education and literacy, and access to contraception are the best-proven levers (Rindfuss & St. John, 1983). Other successful interventions in the effort to decrease exposure to marriage and parenthood include sensitization programs and marriage laws (raising the legal age to 21). With somewhat more difficulty, governments can promote indirect macro-level changes in an effort to increase the age at first union and first birth by increasing the direct and opportunity costs of marriage and parenthood. Some of them may be coercive, some not, just as some of them are very intentional, others not so much. In Tanzania under ujaama, family allowances were limited to the first four children (Hamand, 1982). In Rwanda, the introduction of mandatory health insurance and school fees, and a law prohibiting the division of land under 1 hectare, has drastically increased the costs of children. Governments can try to increase female wage employment, so that there are more reasons for young women to stay unmarried. Investment in road infrastructure can have ‘modernizing’ effects on young people, by increasing their access to economic opportunities involving trade and migration, and their exposure to external ideas. Within this literature, however, the impact of population redistribution policy on family formation has been relatively overlooked; it’s to that we turn next. 3.2 Demographic Impacts of Population Resettlement (700) What is population redistribution and why do states engage in it? Population redistribution is the top-down, state-directed planned resettlement of people, houses, community infrastructure, and land. Policies usually involve tens of thousands or even millions of people. Population redistribution policies have typically targeted rural people, and continue to do so today. The stimulation of rural growth is a key pillar in development plans across the region in part because, despite high rates of urbanization, most people in sub-Saharan African countries continue to live in rural areas. Redistribution policy comes in three forms: extensification, sedentarization, and villagization. Extensification occurs when governments seek to extend a population across a landscape. Mugabe’s Zimbabwe land reforms beginning in the 1980s are a good example of policies that seek to extend access to land for certain groups of people. Sedentarization is relatively more common, occurring when governments seek to immobilize a pastoral or migratory population, for example pastoralists in northern Kenya or pygmies in Central Africa. Finally, villagization, which involves grouping people scattered populations in pre-planned, clustered villages and reorganizing land use in accordance with state goals. In general, villagization policies- like the one explored in this study- seek to ‘modernize’ the rural sector,

• ptimizing land use, catalyzing economic growth from agriculture, promoting off-farm jobs and education,

and improving the efficiency of service delivery. The most extensive and notable (notorious) villagization programs in Africa have occurred in Ethiopia, Tanzania, Mozambique and Rwanda. States engage in planned population redistribution efforts for a number of reasons, which typically overlap. National planners engaging in this kind of social engineering view population distribution as a key factor in the efficiency and success of development intervention, and are usually informed by a faith in science and engineering, aka the goal is ‘modernization’, with all the concomitant benefits. Redistribution is almost always initiated by authoritarian, strong, developmental states, although informed by different political ideologies. Tanzanian villagization in the 1960 and 1970s was part of the state-building vision called ujamaa. Population regroupment was intended to catalayze a socialist, agricultural revolution. In Ethiopia, planned redistribution has been motivated by population pressure as well as to target politically-undesirable groups. Likewise, the British colonial government forced Mau Mau into detention camps- planned villages that no one could leave. In the years immediately following the Rwandan genocide, villagization was used as a method for surveilling outsiders, protecting residents, and rooting out rebels. In the 1990s, one of the most famous theorists of villagization- James C. Scott- called the study of villagization a ‘quaint archaeology’ (Scott, 1998). Little did he know that villagization would crop up again in Ethiopia, Rwanda, and Burundi in the 21st century. What is the theoretical relationship between population redistribution and demographic change? As a form of planned migration, population redistribution is of course an implicitly demographic

• event. Additionally, it entails dramatic changes (both positive and negative) to the locations and patterns
• f everyday activities, as well as to social and economic networks, livelihoods, costs and opportunities.

This has further implications for a variety of population-related processes including rural-urban migration, morbidity and mortality, fertility, and family formation and dissolution. For our purposes, it seems reasonable to expect that villagization could have both direct and indirect impacts on the choices that men and women make regarding union formation and child-bearing. Yet, academic scholarship has largely

• verlooked the demographic dimensions of population redistribution and planned displacement.

There is a great deal of scholarship on the impact of involuntary resettlement stemming from dam and conservation projects (Cernea, 1994). These have mostly looked at social and economic impacts, and to a lesser extent the impacts on morbidity, mortality, and subsequent migration. Historically, and across nearly all cases, planned resettlement has led to impoverishment, increased landlessness, joblessness, homelessness, marginalization, food insecurity, loss of access to common property and resources, increased morbidity and mortality, mental health decline, social disarticulation and out-migration (Cernea, 1994). Less research has examined the impact of resettlement on fertility or family formation. Where the relationship between population redistribution and fertility is examined, it has been in the context of extensifying land redistribution programs involving previously landless people (see: Gwebu, 2006) or in the context of infrastructure-related displacement (see: Cernea, 1994). In extensification, people often end up increasing family size; the same holds true for sedentarization (cite). As for villagization, studies have explored its effect on variables including social capital, income, livelihood diversification and migration (Cernea, 1994; Isaksson, 2013; McDowell, 2013) but not on fertility or family formation, as far

as we are aware. Still, Tanzanian ujaama and Rwandan imidugudu were, respectively, expected and intended to reduce fertility. 3.3 Villagization and Family Formation in Rwanda: Hypotheses (1500) The existing literature isn’t much help for understanding the theoretical relationship between family formation and villagization in Rwanda. Still, we can advance some hypotheses based on what we know about the impacts of resettlement on the determinants of family formation and an understanding

• f the Rwandan case.

In the Rwandan cultural context, we expect child-bearing (in our case, the age at first birth) to be closely related to union formation (i.e. age at first marriage) and thus overall fertility (Sommers & Uvin, 2011). Overall, we hypothesize villagization to be associated with an increase in both the age at first marriage and the age at first birth. We expect resettlement to induce these effects on family formation (marriage and childbirth) through a process of ‘modernization’ away from traditional patterns of agrarian life; specifically, modernization is expected to occur via changes to: access to services and sensitization programs, livelihoods, and the cost of new housing. These three vectors and their hypothesized effects

• n family formation in the Rwandan context are explored in turn below.

Access to Services and Sensitization Programs: One of the most direct pathways through which the government believes villagization will reduce fertility is by bringing modern services to rural people, especially in the form of family planning and government fertility sensitization programs (REMA, 2009). Increased contraceptive use and reductions in desired family size have already broadly taken place in rural Rwanda, and the World Bank has cited links to health centers in Rwanda as one of the major factors leading to fertility decline (May & Kamarusi, 2009). Villagization is believed to have played an important role in driving changes to fertility in this way (ibid.), although the relationship has not yet been empirically explored. As populations are grouped together, access to services (including health centers), recreation and government sensitization programs are supposed to increase (Government of Rwanda, 2009). In previous studies in other geographic contexts, this ‘modernizing aspect’ of land reform has been shown to reduce fertility, although often after a period

• f five or more years (Oberai, 2011). If villagization in Rwanda has progressed as expected, access to

government services of all kinds will increase with population grouping. We would expect increased access to health services to increase the age at first birth in particular, by assisting women to meet their desired family size and avoid unwanted pregnancies. We expect the age at first marriage will increase as well. Union formation is potentially less affected by increased access to health services, but the broader set of semi-urban benefits (infrastructure, education, recreation, etc.) may serve as a type of ‘blocking mechanism’ to marriage if individual opportunities and life expectations displace some of the benefits of union formation. Urbanization has been linked to decreasing fertility in sub-Saharan Africa, as has rising age at first marriage (cite). Villagization is, in a sense, intended to bring the modernizing benefits of the city to the people. Perhaps, through increased connection to off-farm opportunities, electricity, rural development hubs, roads and markets, villagized people view greater opportunity costs associated with marriage. Livelihood Change:

Villagization inherently transforms the patterns of everyday life for displaced people and, in Rwanda, explicitly aims to reduce farm-based livelihoods in favor of jobs in the non-agricultural wage

• sector. Overall we hypothesize that, in so doing, (and via manifold associated mechanisms) villagization is

leading to an increase in the age at first marriage and first birth. In rural Africa there is a strong connection between livelihood structure and decisions related to family formation and fertility. For example, when agricultural activities are primate in the household, and/or when land sizes are large or increasing, fertility tends to be higher (Oberai, 2011). This is due to a number of factors, but particularly related to the value

• f the on-farm labor of children (ibid.). In Rwanda, however, since villagization seems to be minimizing

the long-term sustainability of smallholder farming as a rural livelihoods strategy (Government of Rwanda, 2009), we expect to see an increase in the age at first birth. For example, land sizes are getting smaller (due to population growth, but also the privatization of communal land, restrictions on minimum land sizes and the sale of private landholdings to fund increasing costs), which reduces the viability of agriculture and the immediate economic value of children to the household and (ideally at the same time) leads to off-farm rural jobs and household economic strategies requiring smaller family size. Or, to the contrary, the dramatic alteration of lives and livelihoods that occurs as a result of population displacements such as villagization can result in negative social and economic shocks that make union formation and child-bearing undesirable for a period of time. In conditions of livelihood decline, family formation is typically put on hold (Rutayisire, 2014). In one study of Rwanda’s program, the farms of refugee returnees resettled in pre-planned villages were found to produce, on average, slightly lower agricultural yields (Kondylis, 2000). Another found no evidence of an effect of imidigudu on increasing diversification of income through non-farm activities: one of the purported benefits of the policy (Isaksson, 2013). Related to this, migration-fertility dynamics and their impact on family formation may come in to play here as well (Oberai, 2011). If more rural youth end up moving to major urban centers (such as the capital, Kigali) because the imidugudu are not successful in delivering positive livelihood change, rural union formation and childbirth is likely to decline. We therefore also expect the structural changes to livelihoods associated with villagization to increase the age at first marriage. This is consistent with the expectations of the Rwandan government (Government of Rwanda, 2009). Cost of New Housing: Villagization in Rwanda also seems likely to delay union formation, age at first birth and subsequent fertility through its significant impacts on the costs of new housing for rural people, especially

• youth. Under the villagization policy, all new rural housing construction must take place in an umudugudu,
• r planned village; in addition all dwellings must be built according to a standardized set of size, roofing

and material requirements which, ultimately, are more expensive. According to a recent study examining the aspirations of Rwandan youth, this has made “the quest toward completing a house virtually impossible for nearly all poor male youth in rural Rwanda” (Sommers & Uvin, 2011 p. 4). The major demographic impact comes from the fact that there is a strong cultural imperative for Rwandan males to

• wn a house before they are considered adult men and, thus, eligible to marry (ibid.). In addition, marriage

is typically a cultural prerequisite to child-bearing in Rwanda (Rutayisire, 2014). Hence, the difficulty that young men face in building their houses inhibits their ability to reach ‘adulthood’ and is likely to impede both union (marriage) formation and age at first birth. However, since all construction outside of villages is forbidden, it seems that all young people would be affected by this consequence of the policy, not just the ones already living in a planned village. For our analysis, therefore, we may be most likely to see increases in the age of first marriage and first birth (and decreases in subsequent fertility) associated with

regions exhibiting a high degree of villagization. We assume, in this case, that the extent of villagization in a particular region is related to the commitment of the local government to enforcing the policy and encouraging relocation and that regulations regarding housing construction are unequally applied across the country. It is important to note that these three vectors are only those that have been advanced by the Rwandan government (access to services and livelihood change) and the sole academic study that investigates the relationship between villagization and family formation in Rwanda (cost of housing), namely Sommers & Uvin (2011) which uses qualitative interviews with Rwandan youth to advance the theory that villagization leads to delayed union formation due to the prohibitively high costs of buying land and building a home in a village. Many other mechanisms are likely at work. Villagization reduces access to common resources, and forces farmers to tend land they may not be familiar with or have to travel farther to get to. This can produce changes to the value of children, and marriage, and to the resources available to achieve both. Villagization entails dramatic spatial transformation of social networks as well as livelihoods and patterns of daily life. There may initially be less social cohesion in villages, and mistrust as Rwandans are not particularly used to having next-door neighbors. In interviews, we were told that- in areas that are highly villagized- single people remaining in homes outside the village are less likely to encounter marriage opportunities. Much more work is needed to elucidate the connections between villagization and family formation in Rwanda.

• 4. Data and Methods:

The aim of this study is to understand whether people living in imidugudu villages prolong family formation (first union and first birth) relative to their counterparts living in ‘traditional’ villages. Our primary source of data is a household survey we conducted from 2016-17 with 2,049 rural Rwandan

• households. The survey contains two modules: the first gathers data on household demographics,

migration history, land use and assets from the male head of household (or the female head, in the case

• f women-headed households); the second module- modified from the Demographic and Health Survey

(DHS)- gathers information about child-bearing and union history from the female head of household, assuming she exists. The survey defined “union” as either marriage or cohabitation, and it was therefore up to the respondent herself to determine whether a particular relationship counted- in her mind- as a “union”. Cohabitation is not legally recognized in Rwanda, so rights to property ownership, custody over children, and inheritance are upheld only under civil marriages (Jayaraman, 2009). Despite the potential for insecurity in these relationships, and the government’s effort to promote civil marriage, many couples still practice cohabitation. The costs of the ceremonies associated with legal marriage are often prohibitively

• high. In other cases, couples may not view the need to involve formal institutions in marriage rites and

are instead satisfied with the recognition provided by the traditional approach to union formation in Rwanda, which involves the payment of a bride price and an agreement between the two families (National Institute of Statistics of Rwanda, 2014). Although there are certainly functional differences between formal and informal partnership in Rwanda, we are comfortable collapsing them both into the “union” measure because they carry the same import in terms of increased exposure to pregnancy and as a marker of the transition to adulthood (rural women live with their families until a partnership is formed). Informal or common-law unions are still recognized by the community and stem from a long tradition in Rwanda of public recognition of marriage through individual and familial consent. In 2010,

30% of unions in Rwanda were informal (National Institute of Statistics of Rwanda, 2014). We use marriage, partnership, and union interchangeably in the rest of the paper. The sample population is partitioned into those living in imidugudu developments and those living in non-villagized, ‘traditional’ settlements. Rwanda is divided administratively into four provinces (plus the city of Kigali), 30 districts, 416 sectors, 2148 cells and 14837 villages. We selected four districts, one in each of the four provinces, and further clustered our sample at the cell level. These districts were chosen because, together, they represent ten of the twelve Rwandan agroecological zones and therefore provide us a representative cross-section of the impacts of villagization on family formation and fertility under different livelihood settings. From the Government of Rwanda, we secured a listing of imidugudu villages with the exact geographic locations of their centerpoints. We randomly selected 14 cells in each district (~20% of the total cells) and, within each cell, selected five planned village sites and five ‘non- village’ locations. Cells and village sites were both chosen to select dispersed, non-adjacent locations. Non-village locations were selected using an inhibition spatial point process such that a non-village point is located, at minimum, beyond a threshold distance from villages and other non-village points already

• sited. Once the locations were chosen we imported them to Google Earth and, using satellite imagery,

ensured that each of the selected points was fairly close to an inhabited structure. Location manipulation was not necessary in the case of village locations (all of the government centerpoints were indeed located in settlements). In the case of non-village locations, some manual manipulation was necessary. For example, if a selected location turned out to be in a forest or field, then we moved it to the nearest house. Once the points were finalized, we downloaded their coordinate locations into GPS devices. Surveyors were then able to navigate to their preselected locations. This introduced a greater degree of randomness in household selection than is normally possible in rural settings and in the absence of population rosters. At each location, four surveys were conducted for a total of 40 households in each cell. The total sample size, number of districts and cells within districts sampled, and number of households per village or non- village point were not based on power calculations. Instead those decisions were based on trying to both maximize the sample size while maintaining variation in areas sampled but also with an eye towards efficiency in data collection by the survey team. Our primary goal is to measure the impact of villagization on family formation. Our expected result is that the indicator variable will be significant and that exposure to villagization will increase the duration to both first birth and first marriage among both men and women. Again, we theorize that this effect stems from the combined influence of increased access to family planning and sensitization programs, the costs to individuals and families associated with villagization, and the restructuring of life and space. We use Cox proportional hazard models with outcome variables being the duration to first marriage, measured in years since age 10, and duration to first birth measured in months from the start of the

• marriage. Our approach allows us to partition the sample and, with information on timing of migration to

the imidugudu, we can also identify the approximate period of exposure. We then use the settlement- type indicator as a time-dependent covariate in our hazard models. We include other covariates that are known to influence the timing of first birth and first marriage: specifically, education level and socio- economic status (ubudehe category).

• 5. Results and Discussion:

Summary statistics:

In the end, our sample size for duration to first union was 353 men and 443 women. Our sample size for duration to first birth was 377 women. We excluded data from the district of Gatsibo from both analyses, because of the character of villagization there, which renders it distinct from the other three districts surveyed. Gatsibo was populated recently, mainly by Tutsi refugees returning from Uganda and Tanzania following the genocide. Villages were constructed for them as part of an emergency housing program, not because of the national imidigudu policy under study here. We also excluded any cases with missing data. Because we only asked female respondents for their union histories, and then interpolated the men’s duration to union, the male data does not include single, widowed, divorced, or separated men,

• r married men whose wives were unavailable to respond to the survey. This also required us to restrict

the sample for the analysis to men who were not more than five years older than their spouse: we assume that in cases where men are relatively close in age to their partner this means it was his first marriage as well. The mean age of the women (men) in our sample ranged from 52.2 (39.2) years in formal settlements to 56.4 (42.1) years in isolated settlements. The population living in villages seems to skew slightly younger; this makes sense because older people are reportedly less willing to resettle, and younger people are likely to be more attracted by village life. The reason the average age of the female populations we sampled is so much higher than the average male age is likely an artifact of the survey

• design. Since we collected union histories exclusively from women, we could only include men who were

currently married, whereas women could be included even if they were widowed or divorced. Men- especially those who would now be 40-70 years old- died disproportionately during the civil war and

• genocide. Their widows, along with divorcees- both groups usually represent older women- are still

included in the sample, which increases the mean age. The mean age at first marriage for the women (men) surveyed is 20.9 (22.2) in imidugudu and 20.7 (21.9) in isolated settlements. The median education for women is two years in both settlement types, compared to five years for men living in villages and four years for men in isolated houses. Age at first union Results The analysis strongly supports our hypothesis that the duration to first marriage/cohabitation would take longer in ‘formal’ villages relative to ‘isolated’ settlements. We are still exploring various model specifications, but so far the main effect is holding steady. Marriage/ partnership is delayed for both men and women living in villages. In every model we ran, where you lived matters. Figure 4 displays the marriage survival curves for men and women according to settlement type. Table 1 shows the three models we tested. First, accounting only for the effect of settlement type, we found that men in isolated villages marry 40% faster than men in ‘formal’ villages. Women in isolated villages marry 35% faster than women living in ‘formal’ villages. When using time-dependent residuals to account for inter-generational effects, we found that- for both men and women- union formation occurred 27% faster. These inter-generational effects include increased access to education, the sensitization program and investment in family planning since 2005. Second, we incorporate education (both the individual’s years of education, and their partner’s years of education) and socio-economic (ubudehe) category as co-variates. The ubudehe category is not significant, nor is men’s education. However, we find that women’s education moderates

the effect size. Specifically, when accounting for their partner’s level of education, men in isolated villages marry 22% more quickly than men in imidugudu. Women in isolated villages marry 27% more quickly than their villagized counterparts when accounting for their own level of education. The effect of women’s education on duration to first union is similar for both men and women: each year of additional schooling reduces the risk by 4%. Thirdly, we fit a model with a different measure of education- in this case we measure whether the individual’s and their partner’s education is less than 5 years, which is approximately the time at which children move into secondary education. We also include ubudehe category as a

• covariate. Again, socioeconomic categories were not significant, and nor was men’s education, but

women’s education was. Men living in isolated villages get married 21% faster (and women in isolated villages 25% faster) when accounting for the effect of undereducated women. The effect of a woman having <5 years of education is 17% for men and 20% for women, almost as large as the effect of

• villagization. We are continuing to explore specifications. Initial results show that, when controlling for

intergenerational effect, the same overall effect of villagization on union formation is observed, but men’s education becomes significant. Discussion Women marry earlier than men, and nearly everyone is married by age 35, consistent with other

• studies. Of all four groups, women in isolated villages get married the fastest. Per our theoretical

framework, this may be due to the fact that women in in formal villages associate greater opportunity costs with marriage, have more educational opportunities and increased access to the opportunities provided by village life. It is also possible that there is a kind of neighborhood effect: perhaps the men living in isolated areas are more readily able to build houses, or it’s easier to illegally live in extended family arrangements away from direct government surveillance. The negative effect of villagization on union formation is greatest for women between the ages of ~16 and 25 or 26 (see Figure 4). This is especially significant because it may be helping to avoid early marriage (the most damaging for social

• utcomes) and allowing young women more time to pursue educational and economic opportunities. The

effect likely begins to taper off around age 26 because of the social pressures to get married. Women’s education has the expected effect in terms of prolonging union formation, which is consistent with the literature on age at first marriage in Rwanda and the rest of sub-Saharan Africa. Men’s education was not significant in our models, although future specifications may change this result. Ubudehe category- which is a locally-specific measure of socioeconomic status- was also insignificant across all models. It may be that the distinctions between ubudehe categories are simply too small to matter when it comes to major, expensive life events like marriage and resettlement. Men living in formal villages get married the slowest of all four groups. We can’t say for certain, but this may support the hypothesis from Sommers & Uvin (2011) that men, who must provide a home if they intend to form a union, are especially penalized by the increased costs of housing associated with

• villagization. All new houses are supposed to be built in formal villages, so it is somewhat surprising to

find young couples living in isolated settlements. Perhaps men in isolated settlements purchased or inherited a home outside the village, or built on their family’s land against government regulation. Duration from union to first birth Results

Figure 5 shows the survival curve for first birth (the odd drop-offs are due to digit preference, i.e. women answering with a year, rather than months, when asked when she gave birth to her first child; this will be adjusted in the final paper). Since we asked women for the years in which their marriage and birth events occurred (not the exact dates), we lack some precision. Figure 5 suggests that, in both types

• f settlements, most women have their first birth during the second year of their marriage. Almost all

women (90%) living in formal villages have had a child within three years of getting married. In informal villages, only ~80% have had children by year three. Our hypothesis was that the duration from union to first birth would take longer in formal villages relative to isolated settlements. In fact, we found the opposite. Women living in imidugudu settlements gave birth to their first children 44% more quickly after cohabitation/ marriage compared to their counterparts living outside of villages. Neither education nor ubudehe category was significant. Discussion We assumed that women living in formal villages would be more likely to delay childbirth following union formation due to greater access to family planning services, government sensitization programs, schooling and other modes of education, and the off-farm job opportunities provided by proximity to rural development hubs and infrastructure. We were therefore surprised to find that partnered women living in imidugudu have a shorter duration from union to their first birth (44% faster). We are in the process of conducting qualitative fieldwork aimed at elucidating the mechanisms underpinning this relationship. One hypothesis is that women living in formal villages are speeding up their postmarital pregnancies because they get married later than women in isolated homes, and that this dominates improved access to family planning. In other words, perhaps villagized couples make a temporal adjustment in order to hit their target family size (which, in Rwanda, is 3.3 children). Perhaps having a child provides some form of security in villagized settlements- either because the man may be migrating away, there are more opportunities for infidelity in the marriage, or life is less secure in some

• regard. Or, in contrast, maybe it’s the people in isolated settlements who need more time to prepare for

the costs and responsibilities of raising children. In a cross-national study of developing countries, Mensch et al. (2003) found that age at first marriage had increased overall, but that the gap between marriage and first birth had narrowed; they linked this to the need for women to prove their fecundity. There are limitations to both analyses. The need to interpolate men’s union history from their wives’ limited the size of our sample and its representativeness to the general population. Our models provide little understanding of why we observe these differences; we cannot attribute differences in family formation directly to villagization. Additional covariates would of course be useful. For example, access to and use of modern contraception and other forms of family planning; whether the family home was constructed, purchased, or inherited; and livelihood or income diversity. It would also be helpful to understand where individuals were living before the union or birth event. A woman who spent her childhood in a village, then moved after marriage to an isolated settlement, might have different expectations and opportunities regarding marriage and pregnancy compared with a woman who was never directly exposed to villagization. Qualitative fieldwork is necessary to understand the mechanisms underlying the relationships revealed through our analyses.

• 6. Conclusion:

 Summary of findings

 Importance of the research

• Rural population redistribution policies involving villagization- despite their multiple

failures in the recent past- have persisted in certain areas of the developing world. Today, the lives and livelihoods of tens of millions of people are impacted by villagization and

• ther rural resettlement schemes (McDowell, 2013). Worldwide, 83% of countries are

dissatisfied with their population distributions; in Africa, 77% of countries with available data have policies explicitly targeted at reducing or managing internal migration (May, 2012).  Future research

• A. References

BBC (2007).”Rwanda moves to limit family size”. http://news.bbc.co.uk/2/hi/africa/6358381.stm (Accessed 9/28/2017) Bloom & Reddy (1986) Cernea, M. (1994). African population resettlement in a global context. In: Involuntary resettlement in

• Africa. Selected papers from a conference on environment and settlement issues in Africa, edited by

Cynthia C. Cook. Washington, D.C., World Bank, 1994. 11-32. (World Bank Technical Paper No. 227; Africa Technical Department Series Gibbs, C. M., Wendt, A., Peters, S., & Hogue, C. J. (2012). The Impact of Early Age at First Childbirth on Maternal and Infant Health. Paediatric and Perinatal Epidemiology, 26(0 1), 259–284. http://doi.org/10.1111/j.1365-3016.2012.01290.x Government of Rwanda (2009). Updated National Human Settlements Policy. Kigali, Rwanda. Government of Rwanda (2012). Rwanda Population and Housing Census. Kigali, Rwanda. Government of Rwanda (2013). “Reducing Vulnerability to Climate Change in North West Rwanda Through Community Based Adaptation”. MINIRENA Project Proposal to the Adaptation Fund. February 2013, Kigali, Rwanda. Gwebu, T.D. (2006). Intra-Rural Fertility Determinants in Zimbabwe: A Path Analysis. African Population Studies 21(1): 71-91. Hahirwa, Gumira (2014). Resistance to Reforms: settlement and agricultural reforms in post-genocide

• Rwanda. Doctoral Thesis. University of Gothenburg, Faculty of Social Sciences, School of Global Studies,

Peace and Development Research Hamand (1982) Snags in reaching the villages. People 9(2): 31-3. IFAD, 2017

Isaksson, A.S. (2013). Manipulating the Rural Landscape: Villagisation and Income Generation in

• Rwanda. Journal of African Economies 22(3): 394–436.

Jayaraman, 2009 Kondylis, F. (2006). Rwanda: the costs of conflict for subsistence households. CentrePiece, Autumn 2006,

• pp. 26-27.

May, J. and Kamarusi, A. (2009). Demographic Growth and Development Prospects in Rwanda: Implications for the World Bank. Kigali, June 4-16, 2009 May, J. (2012). World Population Policies: Their Origin, Evolution, and Impact. Springer: Washington D.C. McDowell, C. (2013). Climate-Change Adaptation and Mitigation: Implications for Land Acquisition and Population Relocation. Development Policy Review 31(6): 677-695. Mensch, B. S. (2005). The transition to marriage. In C. B. Lloyd (Ed.), Growing up global: The changing transitions to adulthood in developing countries (pp. 416–505). Washington, DC: National Academies Press Muhoza et al. (2014). Variations in Musahara, H. (2006). Improving Tenure Security for the Rural Poor: Rwanda Country Case Study. Working Paper #7, Legal Empowerment of the Poor. National Institute of Statistics of Rwanda (2014). RPHC4 Marital Status and Nuptuality National Institute of Statistics of Rwanda (2011), Rwanda Demographic and Health Survey 2010, Preliminary Report, Calverton, MD, USA: Measure DHS and ICF Macro. National Institute of Statistics of Rwanda and ICF International. 2012. 2010 Rwanda Demographic and Health Survey: Key Findings. Calverton, Maryland, USA: NISR and ICF International. Oberai, A.S. (2011). Assessing the Demographic Impact of Development Projects. Routledge: New York. Ronald R. Rindfuss and Craig St. John. Journal of Marriage and Family. Vol. 45, No. 3 (Aug., 1983), pp. 553-565 Rwanda Environment Management Authority (REMA) (2009). Rwanda State of Environment and Outlook

• Report. Government of Rwanda. Kigali, Rwanda.

Rwanda Housing Authority (2014) (http://www.rha.gov.rw/index.php?id=469&tx_ttnews%5Btt_news%5D=129) Rutayisire, P. (2014). Changes in Fertility Decline in Rwanda: A Decomposition Analysis. International Journal of Population Research. Volume 2014, pp. 1-10.

Scott, James C. (1998). Seeing Like A State: How Certain Schemes to Improve the Human Condition Have

• Failed. New Haven: Yale University Press.

Sommers, M. & Uvin, P. (2011). Youth in Rwanda and Burundi: Competing Visions. United States Institute of Peace Special Report. Uwimbara, P. and Lamaru, R. (2011). Compelling Factors of Urbanization and Rural-Urban Migration in

• Rwanda. Rwanda Journal, 22(B): 9-26.

Van Leeuwen, M. (2001). Rwanda's Imidugudu programme and earlier experiences with villagisation and resettlement in East Africa. The Journal of Modern African Studies 39(4): 623-644. Westoff, C.F. (1990). Age at marriage, age at first birth, and fertility in Africa. Yabiku, S. T. (2006). Neighbors and neighborhoods: Effects on marriage timing. Popul Res Policy Rev, 25(4), 305–327.

• B. Figures:

Figure 1: Total Fertility Rate in Rwanda (1992-2015) with 2015 TFR references for neighboring countries

Total Fertility Rate

Year T F R 6.2 5.8 6.1 5.5 4.6 4.2

Kenya (3.9) Uganda (5.7) Tanzania (5.2)

1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 1 2 3 4 5 6 7

Figure 2: Age at First Birth and First Marriage in Rwanda (1992-2015) Figure 3: Percent of the Rural Population in Rwanda Living in Imidigudu (1999-2015)

Family Formation

Year A g e 20 20.7 20.7 21.4 21.9 21.5 22 22 22.3 22.4 22.7 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 1 5 1 7 1 9 2 1 2 3 2 5 age at first birth age at first marriage

Imidugudu Policy: Percent of Rural Population Villagized

Year P e rce n t 20 20 20 22 39 54 72 81 1999 2001 2003 2005 2007 2009 2011 2013 1 0 3 0 5 0 7 0 9 0

Figure 4: Survival Curve, Age at First Union: Formal Vs. Isolated Villages

First Union

Age P ro p o rtio n n o t in u n io n , S (x) 10 15 20 25 30 35 0 .0 0 .2 0 .4 0 .6 0 .8 1 .0 Women, formal villages Women, isolated villages Men, formal villages Men, isolated villages

First Birth

Age P ro p o rtio n a t p a rity 0 , S (x ) 6 12 18 24 30 36 0 .0 0 .2 0 .4 0 .6 0 .8 1 .0 formal villages isolated villages

Figure 5: Survival Curve, Age at First Birth: Formal Vs. Isolated Villages (Women)