Migration and Fertility in the Democratic Republic of Congo

Philip Anglewicz (corresponding author) Assistant Professor, Department of Global Community Health and Behavioral Science, School of Public Health and Tropical Medicine, Tulane University, 1440 Canal St Suite 2200, New Orleans, LA, USA. Email: [email protected] Phone: (504) 988-5428 Fax: (504) 988-5480.

Jamaica Corker Post-Doctoral Researcher, Institute of Demography and Socioeconomics, University of Geneva, Faculty of Sciences of the Company, Uni Mail, 40, bd Pont d'Arve, 1211 Geneva, Switzerland. Email: [email protected]

Patrick Kayembe Professor, Division of Epidemiology and Biostatistics School of Public Health, Kinshasa School of Public Health, University of Kinshasa, BP 11850 Kinshasa 1, Democratic Republic of Congo. Email: [email protected]

Migration and Fertility in the Democratic Republic of Congo

Abstract Kinshasa, one of the world’s megacities, has a rapidly-increasing population due to both natural increase and in-migration. Although migration likely contributes substantially to the population of Kinshasa, little is known about the relationship between migration and fertility outcomes in this setting. We representative data from 2,197 women in Kinshasa and examine whether in-migrants have different fertility levels than lifetime Kinshasa residents, and identify characteristics of migrants that explain differences in fertility. We also use detailed migration histories to create migrant categories to examine whether fertility differs by features of migration history, including the number of lifetime moves, duration in Kinshasa since migration, and migration stream (rural-Kinshasa, urban-Kinshasa). Perhaps surprisingly, we do not find evidence that migrants disproportionately contribute to population in Kinshasa: migrants have fertility that is significantly but only slightly higher than non-migrants. This appears due in part to their lower use of modern contraceptive methods and higher reliance on less- effective traditional methods. This research also demonstrates the value of collecting more detailed characteristics on migration patterns as we find higher fertility among migrants is associated with longer duration in Kinshasa, more lifetime moves, and urban-Kinshasa migration. These findings suggest there are important differences in migrant fertility outcomes among different types of migrants in Kinshasa that should be explored in future research that matches the timing of migration with detailed birth histories.

Introduction

Despite the important contribution of migration to the population of Kinshasa, little is known about the relationship between migration and fertility in this context. The capital of the Democratic Republic of the Congo (DRC), Kinshasa, is one of the world’s “megacities.” With a population of over 11 million, Kinshasa is Africa's third largest city (after Lagos and Cairo) and one of the continent’s most rapidly growing urban areas (United Nations 2015). By 2020 Kinshasa is projected to be Africa’s fastest-growing city in absolute terms, adding four million new inhabitants to its population, at which point it is expected to become the continent’s largest urban agglomeration (UN-Habitat 2010).

Population increase in Kinshasa is due primarily to natural increase (a surplus of births over deaths) and in-migration, rather than geographic expansion or reclassification. Given that migrants are typically relatively young, and women in sub-Saharan Africa (SSA) in their peak reproductive years are most likely to migrate (Brockerhoff and Eu 1993), migration also indirectly affects urban growth because fertility among urban-bound migrants contributes to rates of natural increase in urban areas. Yet little is known about the relationship between migration and natural increase in urban areas, both in general and for the DRC in particular. Most research on fertility in Kinshasa was conducted before the period of prolonged conflict (Shapiro and Tambashe 1998; Shapiro 1996; Shapiro and Tambashe 1994; Bertrand et al. 1990) or after the conflict ended (Shapiro 2004; Shapiro and Tambashe 2003), and did not give specific attention to migrant fertility outcomes, preferences or contraceptive use. To date, research has not explored the complex association between internal migration and fertility, or the implications for Kinshasa’s growth and future policies. With stark differences in fertility between rural and urban areas in the DRC, as throughout most of SSA (Garenne 2008), migration from higher-fertility rural areas to lower- fertility urban areas has the potential to contribute substantially to even higher growth of the population size of urban centers like Kinshasa if migrants as a whole have higher fertility than non-migrants. Alternatively, if migrants have lower fertility than non-migrants in rural areas, migration could contribute positively to slower rates of population growth at the national level.

With the vast majority of urban population growth in the coming decades taking place in cities of the developing world (United Nations 2012), there is a critical need for research on the factors contributing to urban growth in these settings. Here we broadly examine whether migrants are likely to make a substantial contribution in the near future to natural growth in Kinshasa. Specifically, we examine differences in fertility profiles between in-migrants and permanent Kinshasa residents. We then examine individual-level factors that may influence any differential fertility, and seek identify features of the migration process that may contribute to fertility differentials.

Theory: the relationship between fertility and migration in Sub-Saharan Africa (SSA) Research on the relationship between migration, urbanization and fertility remains surprisingly limited (White et al. 2008), especially for migrant particularities of the urban aspects of family planning (FP) programs and fertility differentials (National Research Council 2003). As throughout much of SSA, the dearth of research on migration and fertility in the DRC reflects a general uncertainty of how the process of migration and new urban residence impacts fertility behavior (White et al. 2008) and family planning needs of migrant women (National Research Council 2003). Urban residence is generally expected to reduce fertility because of the higher costs of raising children in cities (Easterlin 1975; White et al. 2008), more favorable attitudes toward smaller families (Cleland and Wilson 1987) and better access to family planning services (Cleland et al. 2006). While urban residence has long been associated with lower fertility in SSA (Shapiro and Tambashe 2002), the relationship between migration to and subsequent residence in new urban areas and fertility outcomes is less clear.

Most of the research on the fertility/migration relationship has focused on international migration, which shares some similarities with patterns of internal migration but also differs in several important aspects. In SSA the vast majority of migration is internal and is likely to eclipse the impact from international in- migration on population or urban growth. In addition, the existing research on migration and health- related outcomes in SSA often uses limited measures of migration. Migration differs by destination and origin (e.g. urban, rural), distance travelled, permanence, duration, and motivation, among other characteristics. Yet even the demographic literature on fertility and migration has often neglected to account for variation in migration characteristics in developing countries and almost universally used simple binary measures of migrant/non-migrant (Chattopadhyay et al. 2006; Omondi and Ayiemba 2005; Brockerhoff and Yang 1994; Oucho et Gould 1993). Recent research has called for greater detail of migration patterns (Deane et al. 2010, Montgomery et al. 2016), a valid concern, but one that is often difficult to respond to due to continued data limitations for internal migration.

The evidence to date on the association of urban migration and fertility in SSA has been mixed. Most cross-sectional studies have pointed to a positive association of migration to urban areas and fertility decline (Omandi and Ayiemba 2005; Brockerhoff 1998; Brockerhoff 1995), both for migrants themselves (Brockerhoff and Yang 1994) and for subsequent generations born in the urban place of destination (White et al. 2005), and two recent longitudinal studies of migration and fertility in Ghana likewise found evidence of lower fertility among rural-to-urban migrants (Chattopadhyay et al. 2006; White et al. 2008). At least two older studies of SSA, however, found no association or migration and fertility decline or

4 even an association of migration and residence in urban areas with increased fertility (Cleveland 1991; Lee 1992).

Four main theoretical approaches explain the relationship of internal migration and fertility: 1) selection 2) adaptation, 3) socialization, and 4) disruption. According to the selection hypothesis, migrants as a self-selected group for whom lower fertility preferences may be part of the motivation to move to a new area (Kulu 2005); migrants differ in key aspects from non-migrants even before migrating, notably by sharing characteristics highly correlated with the decision to migration to urban areas (higher education levels and aspirations, social mobility desires, higher socio-economic status) that are known to influence fertility directly and indirectly (White et al. 2005) . Thus women with profiles or desires similar to those the place of destination – including for education, fertility preferences, willingness to use modern method – may have been more likely have had lower fertility outcomes regardless of whether or not they migrated (Rokicki et al. 2014).

The socialization hypothesis, on the other hand, predicts that the fertility of migrants will primarily reflect fertility preferences dominant in their place of origin, with the assumption that norms about fertility are learned at a young age, and any changes in fertility behavior among migrants will only occur over the longer-term, for example among second generation migrants (White et al. 2005). Following the socialization theory, if migrant areas of origin are typically rural, rural-to-urban migrants will therefore have higher fertility than non-migrants at the place of destination (Kulu 2005).

The adaptation hypothesis, on the other hand, is based on the idea of a faster re-socialization and adaptation to fertility behaviors at the destination city. Like the socialization theory, adaptation also hypothesizes that migrant fertility will eventually come to resemble the dominant patterns of the destination location but predicts a faster convergence (Kulu 2005) and that different fertility levels will be seen among the migrants themselves. The adaptation hypothesis also implies a time element, suggesting that migrant fertility preferences may erode with longer durations in new areas of residence with different fertility norms (Kulu 2003, Williams et al. 2013). In SSA, the adaptation theory generally assumes improved knowledge of sources of family planning in urban areas (Brockerhoff 1995), and that fertility rates would be lower in urban areas following migration because of the increased acceptance of and access to contraception and abortion in urban areas (Shapiro and Tambashe 1994).

Finally, the disruption hypothesis postulates that migrants’ fertility behavior will change in the period immediately prior to or following a residential change as a result of the disruption in economic and social

5 support and family unification often involved in the process of migration itself (Kulu 2005). Disruption effects are related to the process of migration, rather than residence in a new area, and is measured only in the few years immediately surrounding a move.

Evidence has been found in support of all of these theories, and often several concurrently, implying that outcomes are context-dependent and not mutually exclusive (Kulu 2005, Brockerhoff and Yang 1994). For example, migrant selectivity and adaptation were suggested as the reasons why urban in-migrants were found to have fertility behavior similar to that in destination cities in coastal Ghana (White et al. 2008), while disruption, adaptation and socializing were cited as the primary mechanisms driving lower fertility among first- and second-generation migrants in the interior city of Kumasi (White et al. 2005). Brockerhoff and Yang (1994) argued there was little evidence of selectivity but did find support for the disruption and adaptation in explaining lower urban migrant fertility in six SSA countries. In a separate study, Brockerhoff (1995) found again that for most rural-urban migrants in 13 SSA countries, fertility declined immediately after migration and remained low, supporting the disruption and adaptation hypotheses. The complementary and contradictory nature of these theories, even within the same region, suggests that different mechanisms may be dominant in varying contexts and that no single hypothesis can explain fertility differences among urban migrants in all settings.

Despite previous research on these theories, much remains unknown about the migration/fertility association in most of SSA, due largely to limited data on migration. Although different migration streams (urban-rural, rural-urban, urban-urban, rural-rural, etc.) and the duration of time at destination have been previously examined on occasion (e.g., Corker 2015; Chattopadhyay et al. 2006 Kulu 2005; Brockerhoff and Yang 1994), they have not been consistently studied enough to draw conclusions. Identifying the influence of adaptation may partly depend on the duration at residence, with women residing for longer periods more likely to adopt the fertility profile of the destination, an aspect overlooked in research that considers migrant designation by place of birth only and thus may consider women who moved as very young children "migrants". Similarly, disruption may be measured in part by the number of lifetime moves, with more moves associated with more frequent periods of disruption and lower fertility.

Setting The DRC is Africa’s third most-populous country, with an estimated population of just over 77 million people according to UN estimates for 2015 (United Nations 2015). The DRC is also one of the region's fastest growing countries and is projected by the UN to be one of the 10 most populous countries in the

6 world by 2050. The DRC is consistently among the very poorest countries in the world, and ranked 186 out of 187 countries in the 2014 Human Development Index (UNDP 2014). The country only recently emerged from nearly a decade of prolonged conflict, from roughly from 1996-2004, that devastated the country's economy and infrastructure and is generally considered to be the deadliest conflict since World War II (Coghlan et al. 2008), although uncertainty remains around the true casualty count (Spagat et al. 2009). Given its country profile, it is not surprising that the DRC has among the highest fertility rates in the world: most recent Demographic and Health Survey (DHS) from 2013-14 estimated a country-level TFR of 6.6, a slight increase since the 6.3 TFR from the 2007 DHS.

The sharp fertility differential between Kinshasa and the rest of the DRC, including other urban areas, makes it an ideal context in which to examine the relationship between migration and urban residence with fertility outcomes. Since the 1960s, fertility has remained stubbornly high throughout the DRC, although Kinshasa is a notable exception (Romaniuk 2011): the capital has a TFR of 4.2, considerably lower than other urban (5.4) and rural (7.3) areas of the DRC (Ministère du Plan et Suivi de la Mise en oeuvre de la Révolution de la Modernité (MPSMRM), Ministère de la Santé Publique (MSP) and ICF International, 2014). However, Kinshasa saw an increase in TFR between the 2007 and 2013-14 DHS, from 3.7 to 4.2 (MPSMRM, MSP and ICF International, 2014); rural areas also had an increase in TFR over the same period, from 7.0 to 7.3, but urban areas as a whole held steady at 5.4. Modern method use is extremely low throughout the DRC, at just 7.8% for the country as a whole, with condom use accounting for half of the modern contraceptive prevalence rate (MCPR) (MPSMRM, MSP and ICF International, 2014). The MCPR in Kinshasa in 2013-14 was 19% for Kinshasa, making it more than double the national average but still among the lowest of all capital cities in SSA. Traditional method use, on the other hand, is comparatively high in DRC and generally exceeds the MCPR, including in Kinshasa, where 25.7% of women resort to traditional methods compared to 12.6% for the DRC as a whole. Like many countries in SSA, there are significant differences in fertility by population density.

Data and Methods Data Our data comes from Performance, Monitoring, and Accountability 2020 project (PMA2020). PMA2020 currently operates in ten countries worldwide and was established in part to measure uptake of contraceptive use in many of the world’s most populous countries (http://www.pma2020.org/). To achieve this aim, PMA2020 collects representative data on an annual basis for a range of fertility and family planning-related measures.

We use the first round of data collection in Kinshasa, DRC under PMA2020, collected from November 2013 to January 2014. This survey is representative of Kinshasa: it uses a two-stage cluster sampling approach, in which the study first randomly selected census enumeration areas within Kinshasa, then conducted a listing of households in these EAs, and randomly selected 33 households within each EA and interviewed all women of reproductive ages within the household. For selected women, PMA2020 administered a survey that included basic demographic information and extensive information on fertility history and preferences, and contraceptive use. This sampling approach is the same as that used for the DHS, thereby making it comparable to DHS data for Kinshasa.

An important feature of the round 1 of the PMA2020 study in Kinshasa is the inclusion of an extensive residence/migration history, which contains information on all locations in which a woman previously lived for six months or more, and selected characteristics of these former residences. The survey captures the year and month in which the respondent moved to and from a particular area, along with reasons for the migration, urban or rural designation, number of friends and family known in the location before moving, and the household composition for the household in each residence. These data offer advantages over migration data collected by DHS, which collects only data on date of last move and type of place of origin location type of childhood place of residence (although the latter is not collected consistently across all surveys). We use this information to designate women were born outside of Kinshasa as migrants. Individuals who were born in and then moved within Kinshasa are not considered migrants.

Analytic methods We conduct our analysis in several steps. We begin by tabulating background characteristics for our study population, including age, marital status, level of education, number of lifetime births, and household wealth. Household wealth is measured using a constructed wealth index based on ownership of 25 household durable assets, house and roof material, livestock ownership and water source; PMA used similar measures as the most recent DHS in DRC. Creation of the wealth index is achieved by using principal component analysis (Filmer & Pritchett, 2001), which is then converted into quintiles (a linear measure of wealth yielded similar results). We also focus on several measures of fertility desires and family planning, including contraceptive use (modern and traditional), unmet need, desire to not have another child, and whether the last birth was unintended. We test for statistically significant differences in these characteristics between migrants and non-migrants.

Next, we utilize our migration histories to examine several characteristics of migration, most of which are typically not included in studies of migration and fertility. We tabulate the number of lifetime moves

8 among our respondents, the duration of time spent living in Kinshasa, migration stream (rural-urban, urban-urban), and the age at migration.

We then use multivariate Poisson regressions with the number of cumulative births at the time of the survey as the dependent variable. The main independent variable of interest is the binary category of migration, compared to non-migrants. To explore whether any fertility differences among migrants are related to differences in age, education or other background characteristics, we also control for characteristics that significantly differ by migration status and likely affect fertility, specifically: age, level of education, marital status, and household wealth. We first run our regression for the full sample and then run the same regression models separately for migrants and non-migrants to examine if the factors associated with fertility differ between these groups.

We next measure differences in family planning and fertility preferences by migration status by running three logistic regressions where the dependent variables are overall contraceptive use, traditional contraceptive use, and modern contraceptive use. Then we examine fertility-related attitudes, specifically unmet need for family planning, desire to have another child, and whether the most recent birth was unintended. We run separate logistic regressions to examine whether these measures differ by migration status.

Finally, we examine particular characteristics of the migrants, specifically the number of lifetime moves, the duration of time living in Kinshasa, and if the most recent previous residence was urban or rural, and investigate whether these characteristics are associated with lifetime births. Number of lifetime moves is measured as categorical, with those moving from one, two, and three or more times compared to the reference group of non-migrants. Duration of time in Kinshasa is categorical, according to short-term, medium-term and longer-term residence in Kinshasa: 0-5 years, 6-20 years, and 21 or more years resided in Kinshasa (alternative categorizations yielded results that were not substantively different). The same regression controls from previous models are included here.

Results Table 1 shows background characteristics for migrants and non-migrants. Of the approximately 2,200 women in the sample, 340 (15.5%) are classified as migrants. Several significant socio-economic differences between migrants and non-migrants are evident in Table 1. Most notably, migrants are older and more likely to be married than non-migrants. Somewhat surprisingly, migrants and non-migrants have nearly identical levels of education: the only significant difference for no education, 8.5% of

9 migrants compared to 5.4% of non-migrants, with all other categories similar in proportion and not statistically significantly different. No significant differences were found for wealth status between migrants and non-migrants.

Characteristics related to fertility preferences show greater differences between the two groups. Although levels of contraceptive use are similar (32.3% for migrants and 35.0% for non-migrants), traditional method use is higher and modern method use is lower among migrants. While nearly a quarter of migrant women (24.1%) have an unmet need for contraception, only 16.5% of non-migrants do. Notably, 39.2% of migrants say they want no more children while only 26.7% of non-migrants expressed the same desire.

According to migration characteristics (Table 2), most migrants moved more than once: these migrants averaged 2.0 moves in their lifetime. As with migration patterns elsewhere, the majority of women moved at around young adulthood: the average age at migration was 18.1 years old, and nearly 70% of migrants moved between ages 10 and 30. Most had been residing in Kinshasa for over a decade: the average duration spent in Kinshasa is 12.6 years. Migrant women were more likely to originate from an urban (8.4% of all women, 54.1% of migrants) than rural area (7.1% total, 45.9% of migrants). Of the 340 migrants, 7 moved from another country (five from the Republic of the Congo and two from Rwanda).

Although migrant fertility is significantly higher than that for non-migrants in Kinshasa, it is not substantially different: 4.76 compared to 4.00. The difference in cumulative fertility between migrants and non-migrants is thus relatively small (less than one child per woman) and perhaps less than would be expected given the much higher fertility from the migrants' places of origin (for urban and rural areas taken as a whole). The fertility differences, however, are nonetheless still significant, leading us to ask: are the higher fertility among migrants could be due to the differences in age, education or other background characteristics? Results shown in Table 3 suggest that even after controlling for differences in background characteristics, migrants have slightly but significantly higher cumulative births when individual-level characteristics are held constant.

Somewhat surprisingly, given the differences in fertility outcomes, characteristics associated with fertility included in our models are fairly similar for migrants and non-migrants. Poisson regression results stratified by migration status show the same non-linear relationship between the two groups for age and fertility, the same negative relationship between educational attainment and fertility, and lower fertility for unmarried women.

There appear to be two important differences in fertility determinants between migrants and non- migrants, however: while wealthier non-migrant women have fewer children, there is no significant relationship between wealth and fertility among migrants. We also included an interaction between migration and wealth in the pooled sample (statistically significant at p<0.02), which supports these results for differences between migrants and non-migrants in the relationship between wealth and fertility (results not shown). Another difference between migrants and non-migrants is the relationship between separated/divorced/ widowed and fertility, which is statistically significant for non-migrants but not for migrants. An interaction between separated/divorced/widowed was not statistically significant in the pooled sample, so we conclude that this difference between the stratified models for migrants and non- migrants is due to the small sample size of this category for migrants (n=17), which may not be sufficient to detect a statistically significant relationship. This may also represent relatively high rates of remarriage for women who are widowed or divorced in SSA, as they are less likely to spend much a large proportion of their reproductive lives unmarried (De Walque & Kline 2012).

Results for fertility preferences and family planning-related characteristics may yield some insight into the differences in fertility between migrants and non-migrants. First, we find no significant differences in overall contraceptive use between migrants and non-migrants (Table 4). When this is disaggregated into contraceptive method type, however, we find that migrant women are significantly less likely to use modern contraception and significantly more likely to be using traditional methods compared to Kinshasa natives. Second, the difference in fertility does not appear to be related to greater preferences for limiting, as we see no evidence of a significant difference between migrants and non-migrants with regard to the desire to not have another child (Table 5). Similarly, there are no significant differences between migrants and non-migrants in unmet need or whether a woman's last birth was unintended.

Finally, our results indicate fertility is associated with several characteristics of migration. As shown in Table 6, we find that longer duration in Kinshasa is associated with higher fertility (even after controlling for age), albeit in only slightly: migrants who had lived in Kinshasa for 21 or more years had 0.12 more children than permanent Kinshasa residents, while there are no significant differences compared to natives for women who migrated within the past 20 years. Similarly, we see a positive relationship between number of lifetime moves and cumulative fertility (Table 7), suggesting an association between frequent moves and higher fertility. Finally, while there are no significant differences in fertility between non-migrants and rural-Kinshasa migrants, we find significantly higher fertility for urban-Kinshasa

11 migrants compared to non-migrants, shown in Table 8. In results not shown, there were no significant differences in fertility by age at migration.

Discussion We find that migrants into Kinshasa have fertility that is remarkably similar to those of non-migrants. The difference in cumulative fertility of migrants is statistically significant but only slightly higher than women who have spent their whole lives in the city. These differences are apparent before and after controlling for individual characteristics, suggesting that even though migrants systematically differ from the general population with regard to average age and marital status, differences in fertility are not explained by differences in background characteristics.

Based on analysis of family planning-related outcomes, it appears that the slightly higher fertility of migrants is partly due to their greater use of less effective contraception: migrants are more likely to use traditional methods instead of modern contraceptive methods. This pattern is particularly noteworthy as it is not a result of continued contraceptive practices from place of origin, as women in rural and small urban areas use traditional methods at a much lower rate than do women in Kinshasa, and implies that migrants are taking up traditional method use after migrating. Our data are not able to determine specifically whether high rate of traditional method use among migrants is a preference or results from barriers to obtaining modern methods; although there were no significant differences in knowledge of a location to obtain family planning methods (results not shown), migrants may face more barriers to accessing family planning services (and are less likely to use modern methods than non-migrants) (Irani et al. 2013; Shapiro and Tambashe 1994), suggesting that many migrants may turn to traditional methods in the absence of available modern methods. This finding is in line with previous research on contraceptive use and migration in Kinshasa, using data from 1990, which also found that migration status was significantly related to a decreased likelihood of using modern methods but not for using any method of contraception (Shapiro and Tambashe 1994), suggesting that a higher likelihood of traditional than modern method use by migrants in Kinshasa is long-held trend.

The higher fertility among migrants may also be due to the absence of a relationship between wealth and fertility among migrants: while wealthier non-migrant women have fewer children in Kinshasa, there is no such relationship among migrants. This could represent the fact that migrant women in Kinshasa who are relatively wealthier have become so more recently and thus current wealth may not be associated with prior fertility outcomes. The higher wealth levels are overwhelmingly found in Kinshasa (at a national level, among those in the richest wealth quintile in the most recent DHS, 75% of reside in Kinshasa and

12 only 2% in rural areas), suggesting that migrants may become wealthier (in terms of assets) following migration.

We find that our more-detailed measures of migration contribute to our knowledge on the relationship between migrant category and fertility in Kinshasa. Surprisingly, cumulative fertility is higher among migrants who have lived longest in Kinshasa, even after controlling for age. We also find that migration does not appear to result in lower fertility due to disruption, as those with more lifetime moves is associated with higher fertility in our sample; this could potentially reflect instead the disruptive effect of migration on the protective effect of breastfeeding or access to family planning. Finally, we find that the higher fertility among migrants appears to be driven primarily by urban-Kinshasa migrants, who have significantly higher fertility than non-migrants, while no difference is found for rural migrants into Kinshasa. This may reflect different patterns of adaptation between these two groups or stronger selection of wealthier migrants or those seeking higher education from rural areas. It could also result from different age patterns fertility and different patterns of return migration of these two groups, if for example a greater proportion of women migrate at younger ages from rural areas (for schooling or work) and have low fertility while in Kinshasa but are also more likely to return to their villages during peak childbearing years while migrants from urban areas are more likely to stay in Kinshasa permanently. It is important to note that the order of events between migration and fertility cannot be accurately measured with this dataset (such data are very rare in any setting, particularly SSA), but our nuanced measures, which are also seldom available, nonetheless yield insight into migration processes in SSA.

Our findings on lower urban migrant fertility compared to those in the places of origin are consistent with much of the recent literature on SSA that have found evidence of lower fertility outcomes among rural-to- urban migrants compared to rural non-migrants (Omandi and Ayiemba 2005; Chattopadhyay, White et al. 2006; Brokerhoff 1995) and are closely aligned with those from one of the most recent studies from West Africa on internal migration to a capital city, which also found remarkably similar fertility outcomes between migrants and non-migrants in Accra, Ghana (Rokicki et al. 2014). Prior research suggests an increased use of modern contraception after urban migration impacts lower fertility (Brokerhoff 1995); while this may be the case for Kinshasa migrants as well, it appears that the higher fertility may be partly explained by greater reliance on traditional methods. Earlier research in Kinshasa also found that migrant women were less likely to report having had an abortion, suggesting higher abortion rates among Kinshasa natives may also play a role in their lower fertility outcomes (Shapiro and Tambashe 1994).

Overall, our results do not provide definitive support for any any single hypothesis of fertility and migration; they do support adaptation and selection. Although we cannot measure it directly, we find that the most compelling hypothesis in this context to be a selection effect among migrants to Kinshasa. We suspect this selection effect based largely on the similar education profiles of migrants and non-migrants, which was particularly surprising given the differences in average years of education among women in Kinshasa (10.4), compared to other urban areas of the DRC (8.0) and rural DRC (4.5) (MPSMRM, MSP and ICF International, 2014). As few other cities in Kinshasa have universities and many rural areas have only primary schools, in order to attain the similar education profile of non-migrants, most migrants either had relatively high educational attainment at their places of origin prior to moving (unlikely, if not impossible, in most rural areas) or migrated to Kinshasa in order to increase their educational attainment – both scenarios which imply selection in terms of education. The similar fertility between migrant and non-migrants also hints at adaptation following migration (although we cannot distinguish clearly between fertility prior to and following migration), particularly as migrant rates of contraceptive use, while they lag behind those of non-migrants, are much higher than for women in their places or origin, suggesting they are adapting their contraceptive practices following relocation in Kinshasa.

Conversely, the substantially lower fertility of migrants in Kinshasa compared to their places of origin lends little support to the socialization hypothesis, which predicts that migrants will maintain fertility closer to that found in their place of origin and that any changes in fertility behavior will only happen over the long term. The fact that migrant TFR for the two years previous to the study (4.76) is much closer to that of women native to Kinshasa (4.18) than to either women in other urban areas (6.03) or in rural areas of the DRC (7.25) (comparisons made using DHS data using two-year ASFRs) does not suggest that women who settle in Kinshasa perpetuate the fertility norms from villages or urban area of origin. We cannot comment directly on the disruption hypothesis, as the period surrounding migration is not captured by our data, but the lack of a relationship between number of moves and fertility suggests that disruption is unlikely to be a factor in cumulative fertility outcomes among migrants.

We also tested other possible explanations for higher fertility among migrants for which no significant associations were found. There did not appear to be differences in knowledge of family planning: migrants were no different from non-migrants in knowledge of a location to obtain family planning methods, although it is possible they may face other barriers to access. Similarly, as mentioned above, there was no statistically significant relationship between age at migration and fertility.

It is important to be clear about the goals of our research: we seek to identify differences in cumulative fertility between migrants and non-migrants as a result of residence in Kinshasa and explore possible explanations for these differences. We do not aim to measure the effect of migration on fertility: without complete birth histories for respondents, we cannot identify the order of events between births and migration and thus cannot determine whether the higher fertility found for migrants is a result of higher fertility prior to migration or higher levels of fertility by migrants after re-locating to Kinshasa; longitudinal panel data with measures before and after migration would be necessary to address this potential bias. We area also not able to evaluate differences in employment in the formal and non-formal sectors, which research in Kinshasa prior to the period of conflict found to be highly correlated with contraceptive use and access to abortion (Shapiro & Tambashe 1994). Thus, it is also possible that differences in employment sector characteristics between migrants and non-migrants may also play a role in explaining the fertility differentials. Last, although one of the main advantages of the PMA2020 Kinshasa dataset is its comparability to the DHS data for Kinshasa, because the 2014 DRC DHS did not include variables on migration (as these were dropped from the DHS Round 6 core questionnaire in most countries), we are unable to compare our migrant and non-migrant profiles or fertility outcomes directly with those from the DHS. Although our fertility estimates for Kinshasa as whole match closely those from the DHS, thus indicating concurrence, it would be ideal to have more direct comparisons by migration status and we hope that the migration variables will be included in future DHS core questionnaires.

While we cannot determine the causal effect of migration on these outcomes without the ability to match the timing of migration with birth histories, this nonetheless signals that there are important differences in migrant fertility in Kinshasa – both between migrants and non-migrants but just as importantly between different categories of migrants – that should be considered in current family planning program approaches and for more detailed study in further research. The complexity of our results reinforces the importance of acknowledging diversity among migrants. Different hypotheses may apply for different migrant groups at different times and in different contexts and the collection of more detailed data on both migration and birth histories may be necessary to better identify dominant theories of fertility and mobility in different contexts.

More generally, our findings suggest that rather than fuel rapid increases in Kinshasa's urban growth, fertility from internal migration to Kinshasa, from urban or rural areas in DRC, will likely have a small impact on the city's natural growth rate given the indications that migrants have fertility that is less than one child per woman more than lifetime Kinshasa residents. Accordingly, continued migration to

Kinshasa may also have a positive contribution, albeit small, to the process of the fertility transition throughout the country, if the city continues to absorb high numbers of migrants who ultimately have significantly and substantially lower fertility than they would likely have had if they remained in their urban or rural places of origin. Nonetheless, the fertility differences found here for migrants, compared to both Kinshasa natives and women in their places of origin, suggests an important influence of migration on fertility among Congolese women who migrate to this megacity. While more complete data on the timing of migration and fertility is needed to tease out the dominant mechanisms at play, our findings strongly suggest that residence in a new urban area may play an important role in lowering fertility rates of the substantial numbers of women who are expected to migrate to Kinshasa from other areas of the DRC in the coming years.

Tables & Figures

Table 1: Background characteristics for migrant and non-migrant women, PMA2020 Project, Kinshasa 2013 Non-migrants Migrants Mean age (SD) 26.8 (0.2) 30.6** (0.5) Level of education No education 5.4% 8.5%* Primary 47.1% 47.1% Middle secondary school 39.7% 37.7% Advanced secondary or higher 7.8% 6.7% Marital status Married/partnered 47.3% 64.1%** Separated/divorced/widowed 3.3% 5.0% Never married 49.4% 30.9%** Wealth index Quintile 1 (lowest) 19.7% 17.9% Quintile 2 20.1% 21.8% Quintile 3 20.3% 20.9% Quintile 4 20.2% 20.0% Quintile 5 (highest) 19.7% 19.4% Fertility-related characteristics Total fertility rate (TFR) 4.00 4.76** Contraceptive use 32.3% 35.0% Traditional contraceptive use 15.9% 22.1%** Modern contraceptive use 16.1% 12.9% Unmet need for contraception 16.5% 24.1%** Do not want another child 26.7% 39.2%** N= 1,857 340 Notes: Difference between migrants and non-migrants significant at *p ≤ 0.10; **p ≤0 .05; ***p ≤ 0.01

Table 2: Migration characteristics, Kinshasa PMA2020 Mean number of moves (SD) 2.0 (0.03) Resided in Kinshasa for 0-9 years 49.7% 10-19 years 26.8% 20-29 years 12.3% 30+ years 11.2% Migration stream

Rural-Kinshasa migrant 45.9% Urban-Kinshasa migrant 54.1% Age at migration 0-9 years old at migration 20.3% 10-19 years old 40.0% 20-29 years old 28.2% 30-39 years old 7.7% 40-49 years old 3.8% N= 340

Table 3: Poisson regression results for differences in lifetime fertility between migrants and non-migrants, PMA2020 Project, Kinshasa 2013 Full Sample Non-Migrants Only Migrants Only Coef SE Coef SE Coef SE Migrant 0.08** 0.04 ------Age 0.27*** 0.02 0.29*** 0.02 0.17*** 0.04 Age2 -0.01*** 0.00 -0.01*** 0.00 -0.00*** 0.00 Level of education

No education (ref.) ------Primary school -0.10* 0.05 -0.10* 0.06 -0.05 0.10 Middle secondary school -0.36*** 0.06 -0.38*** 0.07 -0.24** 0.12 Advanced secondary or higher -0.71*** 0.09 -0.74*** 0.10 -0.61** 0.20 Marital status

Married/partnered (ref.) ------Separated/divorced/widowed -0.21*** 0.06 -0.22** 0.07 -0.18 0.13 Never married -0.98*** 0.06 -0.96*** 0.06 -1.14*** 0.17 Household wealth

Quintile 1 (lowest, ref.) ------Quintile 2 -0.03 0.05 -0.07 0.05 0.11 0.10 Quintile 3 -0.04 0.05 -0.06 0.06 0.05 0.11 Quintile 4 -0.17*** 0.05 -0.22*** 0.06 -0.04 0.11 Quintile 5 (highest) -0.18*** 0.06 -0.23*** 0.06 -0.01 0.14 N= 2197 1857 340 Notes: *p ≤0.10; **p ≤ 0.05; ***p ≤ 0.01

Table 4: Logistic regression results for differences in current contraceptive use for migrants, PMA2020 Project, Kinshasa 2013 Using Any Method Modern Traditional of Contraception Contraception Contraception Odds SE Odds SE Odds SE Migrant 0.98 0.13 0.73* 0.13 1.30* 0.20 Number of births 1.24*** 0.05 1.25*** 0.06 1.12*** 0.05 Age 1.37*** 0.06 1.26*** 0.07 1.32*** 0.07 Age2 0.99*** 0.00 0.99*** 0.00 0.99*** 0.00 Level of education No education (ref.) ------Primary school 1.03 0.23 1.18 0.35 0.93 0.24 Middle secondary school 1.37 0.32 2.02** 0.62 0.89 0.24 Advanced secondary or higher 2.01** 0.57 2.54** 0.92 1.25 0.41 Marital status Married/partnered (ref.) ------Separated/divorced/widowed 0.26*** 0.10 0.34** 0.18 0.31** 0.15 Never married 1.42*** 0.19 1.32* 0.22 1.28 0.20 Household wealth

Quintile 1 (lowest, ref.) ------Quintile 2 1.10 0.18 0.83 0.17 1.36 0.27 Quintile 3 1.41** 0.23 0.89 0.18 1.85*** 0.37 Quintile 4 1.22 0.20 1.05 0.22 1.31 0.28 Quintile 5 (highest) 1.55** 0.26 1.30 0.27 1.49* 0.32 N= 2197 2197 2197 Notes: *p ≤ 0.10; **p ≤ 0.05; ***p ≤ 0.01

Table 5: Logistic regression results for differences in fertility preferences among migrants, PMA2020 Project, Kinshasa 2013 Most recent birth Unmet need Want another child unintended Odds SE Odds SE Odds SE Migrant 1.11 0.18 1.01 0.18 1.07 0.17 Births 1.12*** 0.04 2.21*** 0.13 1.49*** 0.07 Age 0.95*** 0.01 0.64*** 0.04 0.72*** 0.05 Age2 ------1.01*** 0.00 1.00*** 0.00 Level of education No education ------Primary school 1.49 0.37 1.28 0.37 1.42 0.33 Middle secondary school 1.16 0.31 1.12 0.34 1.56* 0.40 Advanced secondary or higher 0.57 0.22 1.46 0.60 0.84 0.31 Marital status Married/partnered

Separated/divorced/widowed 2.99*** 0.77 1.74 0.62 1.22 0.34 Never married ------1.29 0.25 4.09*** 0.78 Household wealth

Quintile 1 (lowest, ref.) ------Quintile 2 1.12 0.21 0.91 0.19 1.43* 0.28 Quintile 3 0.99 0.20 1.09 0.23 0.91 0.18 Quintile 4 0.81 0.17 1.10 0.24 0.94 0.20 Quintile 5 (highest) 1.15 0.26 1.12 0.25 1.09 0.25 N= 1175 2197 1296 Notes: *p ≤ 0.10; **p ≤ 0.05; ***p ≤ 0.01; the quadratic term for age was not statistically significant in the model for unmet need and was therefore left out; unmet need was calculated only for ever-married women, and recent birth unintended was available only for women ever giving birth.

Table 6: Poisson regression results for cumulative fertility by duration of time in Kinshasa, PMA2020 Project, Kinshasa 2013 Coef SE Duration of time in Kinshasa

Non-migrant (born in Kinshasa) (ref.) ------0-5 years 0.11 0.07 6-20 years 0.00 0.06 21+ years 0.12** 0.06 Age 0.27*** 0.02 Age2 -0.01*** 0.00 Level of education No education (ref.) ------Primary -0.10* 0.05 Middle secondary school -0.35*** 0.06 Advanced secondary or higher -0.71*** 0.09 Marital status Married/partnered (ref.) ------Separated/divorced/widowed -0.21*** 0.06 Never married -0.98*** 0.06 Household wealth Quintile 1 (lowest, ref.) ------Quintile 2 -0.03 0.05 Quintile 3 -0.04 0.05 Quintile 4 -0.17*** 0.05 Quintile 5 (highest) -0.18*** 0.06 N= 2197 Notes: *p ≤ 0.10; **p ≤ 0.05; ***p ≤ 0.01

Table 7: Poisson regression results for cumulative fertility by number of moves, PMA2020 Project, Kinshasa 2013 Coef SE Number of moves

Non-migrant (no moves) (ref.) ------One move -0.01 0.05 Two moves 0.07* 0.04 Three+ moves 0.17** 0.09 Age 0.27*** 0.02 Age2 -0.00*** 0.00 Level of education No education (ref.) ------Primary -0.10* 0.05 Middle secondary school -0.35*** 0.06 Advanced secondary or higher -0.71*** 0.09 Marital status Married/partnered (ref.) ------Separated/divorced/widowed -0.21*** 0.06 Never married -0.98*** 0.06 Household wealth

Quintile 1 (lowest, ref.) ------Quintile 2 -0.04 0.05 Quintile 3 -0.04 0.05 Quintile 4 -0.17*** 0.05 Quintile 5 (highest) -0.18*** 0.06 N= 2197 Notes: *p ≤ 0.10; **p ≤ 0.05; ***p ≤ 0.01

Table 8: Poisson regression results for lifetime fertility by migration stream, PMA2020 Project, Kinshasa 2013 Coef SE Migration stream Non-migrant (ref.) ------Rural-Kinshasa migrant 0.05 0.05 Urban-Kinshasa migrant 0.10** 0.05 Age 0.27*** 0.02 Age2 -0.00*** 0.00 Level of education

No education (ref.) ------Primary -0.10* 0.05 Middle secondary school -0.36*** 0.06 Advanced secondary or higher -0.71*** 0.09 Marital status

Married/partnered (ref.) ------Separated/divorced/widowed -0.21*** 0.06 Never married -0.98*** 0.06 Household wealth

Quintile 1 (lowest, ref.) ------Quintile 2 -0.03 0.05 Quintile 3 -0.04 0.05 Quintile 4 -0.17*** 0.05 Quintile 5 (highest) -0.18*** 0.06 N= 2197 Notes: *p ≤ 0.10; **p ≤ 0.05; ***p ≤ 0.01

References Bertrand, J. T., C. Chirhamolekwa, et al. 1990. "Post-partum events and fertility control in Kinshasa, Zaire." Journal of Biosocial Science 22(2): 197-211.

Brockerhoff, M. 1995. "Fertility and Family-Planning in African Cities - the Impact of Female Migration." Journal of Biosocial Science 27(3): 347-358.

Brockerhoff, M., Yang, X. 1994. "Impact of migration on fertility in sub-Saharan Africa." Social Biology 41(1-2): 19-43.

Chattopadhyay, A., M. J. White, C. Debpuur, et al. 2006. Migrant fertility in Ghana: Selection versus adaptation and disruption as causal mechanisms. Population Studies 60(2): 189-203.

Cleland, J., S. Bernstein, A. Ezeh et al. 2006. Family planning: the unfinished agenda. Lancet 368(9549): 1810-1827.

Cleland, J. and C. Wilson 1987. Demand Theories of the Fertility Transition - an Iconoclastic View. Population Studies 41(1): 5-30.

Cleveland, D. A. 1991. Migration in West Africa - a Savanna Village Perspective. Africa 61(2): 222-246.

Corker, J. 2015. Not Just When but Where: Investigating internal migration fertility decline in West Africa. Presented at the Population Association of America Annual Conference, San Diego CA, April- May.

Coghlan, B., P. Ngoy, F. Mulumba, C. Hardy, V. N. Bemo, T. Stewart, J. Lewis, and R. Brennan. 2008. Mortality in the Democratic Republic of Congo: An ongoing crisis. New York: International Rescue Committee.

Deane KD, Parkhurst JO, Johnston D. Linking migration, mobility and HIV. Trop Med Int Health 2010; 15: 1458–63.

de Walque, D., & Kline, R. 2012. The Association Between Remarriage and HIV Infection in 13 Sub‐ Saharan African Countries. Studies in family planning, 43(1), 1-10.

Filmer, D., L. Pritchett. 2001. “Estimating wealth effects without expenditure data or tears: An application to educational enrollments in states of India,” Demography 38(1): 115–132.

Garenne, M. M. 2008. Fertility changes in sub-Saharan Africa (DHS Comparative Reports No. 18). Calverton, MD: Macro International Inc.

Irani, L., I. Speizer, C. Barrington. 2013. Attitudes, beliefs and norms relating to contraceptive use among young migrant and non-migrant adults in urban Dar es Salaam, Tanzania. Global public health, 8(9), 1048-1062.

Kulu, Hill. "Migration and fertility: Competing hypotheses re-examined." European Journal of Population/Revue européenne de Démographie 21.1 (2005): 51-87.

Lee, B. S. 1992. The Influence of Rural-Urban Migration on Migrants Fertility Behavior in Cameroon. International Migration Review 26(4): 1416-1447.

Ministère du Plan et Macro International. 2008. Enquête Démographique et de Santé, République Démocratique du Congo 2007. Calverton, Maryland, U.S.A. : Ministère du Plan et Macro International.

Ministère du Plan et Suivi de la Mise en oeuvre de la Révolution de la Modernité (MPSMRM), Ministère de la Santé Publique (MSP) et ICF International, 2014. Enquête Démographique et de Santé en République Démocratique du Congo 2013-2014. Rockville, Maryland, USA : MPSMRM, MSP et ICF International.

Montgomery, M., D. Balk, Z. Liu, S. Agarwal, E. Jones, and S.Adamo. 2016. Urban Migration of Adolescent Girls: Quantitative Results from Developing Countries. In the International Handbook of Migration and Population Distribution, Michael White, ed., Springer Publishing, pp 573-604.

National Research Council. 2003. Cities Transformed: Demographic Change and Its Implications in the Developing World, The National Acadamies Press.

Omondi, C. O., E. H. Ayiemba. 2005. Fertility differentials in Kenya: The effect of Female Migration. African Population Studies 20(2): 25-42.

Oucho, J.O., W.T.S. Gould. 1993. Internal migration, urbanization and population distribution. In Karen A. Foote, Kenneth H. Hill, and Linda G. Martin (eds.), Demographic Change in Sub-Saharan Africa, pp. 8-67. Washington, DC: National Academy Press.

Rokicki, S., Montana, L., & Fink, G. 2014. Impact of Migration on Fertility and Abortion: Evidence From the Household and Welfare Study of Accra. Demography, 51(6), 2229-2254.

Romaniuk, A. (2011). "Persistence of high fertility in tropical Africa: the case of the Democratic Republic of the Congo." Population and Development Review 37(1): 1-28.

Shapiro, D. Tambashe B.O. 1994. "The impact of women's employment and education on contraceptive use and abortion in Kinshasa, Zaire." Studies in Family Planning 25(2): 96-110.

Shapiro, D., O. Tambashe. 2002. Fertility transition in urban and rural areas of sub-Saharan Africa: Preliminary evidence of a three-stage process. Journal of Africa Policy Studies 8: 103-127.

Spagat, M., A. Mack, T. Cooper, J. Kreutz. 2009. Estimating War Deaths: An arena of contestation. Journal of Conflict Resolution 53(6): 934-950.

United Nations. 2009. World Urbanization Prospects: the 2009 Revisions. New York: Population Division Department of Economics and Social Affair of the United Nations. Available at: http://esa.un.org/unpd/wup/Documents/WUP2009_Highlights_Final.pdf

United Nations, Department of Economic and Social Affairs, Population Division. (2012). World urbanization prospects: The 2011 revision, highlights. New York, NY: United Nations.

United Nations, Department of Economic and Social Affairs, Population Division (2015). World Population Prospects: The 2015 Revision, Volume II: Demographic Profiles (ST/ESA/SER.A/380).

United Nations Human Settlements Program [UN-Habitat]. 2010. The State of African Cities 2010: Governance, Inequality and Urban Land Markets. Nairobi, UN-Habitat.

White M., Tagoe et al. 2002. “Urbanization and the Fertility Transition in Ghana”. Population Research and Policy Review 24(1): 59-83.

White, M.J., Muhidin, J. et al. 2008. Urbanization and fertility: an event-history analysis of coastal Ghana. Demography 45 (4): 803-816.

Williams, J., Ibisomi, L., Sartorius, B., Kahn, K., Collinson, M., Tollman, S., & Garenne, M. 2013. Convergence in fertility of South Africans and Mozambicans in rural South Africa, 1993–2009. Global health action, 6.