Policy Research Working Paper 11052 How Do Migrants Fare in Africa’s Urban Labor Markets? Luc Christiaensen Michael Keenan Agriculture and Food Global Department January 2025 Policy Research Working Paper 11052 Abstract Africa’s urban population is on the rise, and it is often feared (town or city). Overall, migrants prove to integrate swiftly that migrants are a direct source of urban underdevelop- and well. They integrate better in towns than in cities, and ment under the premise that they are poorly integrated into urban migrants tend to integrate better than rural migrants. urban labor markets. This study examines the validity of this The integration of rural-to-city migrants has been more premise using data from six African countries. It explores challenging. They are able to work more and obtain similar whether destination labor market and welfare outcomes welfare levels as their urban counterparts, but initially they systematically differ between migrants and nonmigrants, face an occupational earnings penalty. Together, the find- and whether the results differ depending on the duration ings do not provide support for policies hindering internal of stay, migrants’ origin (rural or urban), and destination migration to urban areas. This paper is a product of the Agriculture and Food Global Department. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at lchristiaensen@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team How Do Migrants Fare in Africa’s Urban Labor Markets? Luc Christiaensen and Michael Keenan1,2 JEL classifications: J46, J61, O18, R23 Key words: migrant, town, city, Africa, labor market integration 1 Luc Christiaensen is Lead Agriculture Economist at the World Bank (lchristiaensen@worldbank.org); Michael Keenan is Associate Research Fellow at the International Food Policy Research Institute (m.keenan@cgiar.org). 2 Comments by Kibrom Abay, Loren Landau, and Brian Roberts are gratefully acknowledged as is financial support by Cities Alliance, the CGIAR Initiative on National Policies and Strategies, and the World Bank Jobs Multi-Donor Trust Fund. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. 2 1 Introduction Sub-Saharan Africa’s (SSA’s) urban population is expanding rapidly. In 1950, about 14 million people were urban; by 2020 its urban population had grown 38-fold to 542 million, spread over an increasing number of settlements. 3 Most urban centers are relatively small, housing fewer than 100,000 inhabitants, and 11 have more than 5 million (OECD/UN ECA/AfDB, 2022). Forecasts project Africa’s rapid urban expansion to continue until 2050. 4 Through agglomeration economies, urban areas can offer higher incomes and lower poverty levels than rural areas. It makes them a potentially powerful force for economic growth and poverty reduction (Ahimah-Agyakwah et al., 2022; OECD/UN ECA/AfDB, 2022). But fast urban expansion also challenges cities to keep up with infrastructure, social services, and job creation, and often comes with urban disamenities (Lall, et al., 2017). Tighter job markets and urban inequality may ensue (Kollamparambil, 2017), rousing social tensions and even conflict (Østby, 2016). Against this background, migrants, particularly rural-urban migrants, are often viewed as a source of urban underdevelopment. Migrants contribute significantly to urban population growth, on average 40 percent in low- and middle-income countries (Farrell, 2017). 5 They are further considered to be poorer and less educated than urban 3 In 2020, OECD/UN ECA/AfDB (2022) estimated the number of urban agglomerations with 10,000 or more residents at 7,243 at least. 4 The United Nations (2019) is projecting the urban population in SSA to grow by another 800 million people by 2050. Defining urban as continuously built-up and developed areas (with less than 200 meters between buildings) inhabited by at least 10,000 people, as in the Africapolis project of OECD/UN ECA/AfDB, instead of using country-specific, administrative definitions, as the United Nations, likely yields even faster growth. 5 Contrary to popular belief, the larger share of urban growth in developing countries comes from urban natural increase, not internal migration, also in SSA ( Jedwab et al, 2017; Menashe-Oren and Bocquier, 2021). 3 nonmigrants, perceived to settle mainly in informal areas, and often seen to place additional pressure on already stretched urban public services. Migrants can also create labor surpluses and depress urban wages, as originally highlighted by Harris and Todaro (1970). These challenges can be further exacerbated when migrants move to cities for environmental or conflict-related reasons, reinforcing the popular view of migrants as a source of urban underdevelopment. The view of migrants as a source of urban underdevelopment poses a dilemma for policy makers. Migration can help reduce poverty, which is still largely concentrated in rural areas (Beegle and Christiaensen, 2019), though possibly at the expense of urban development. Urban policy makers are particularly concerned, as an influx or steady stream of migrants may challenge them to realize the agglomeration benefits of their cities, especially in SSA where cities are already crowded, disconnected, and costly for citizens and firms (Lall, et al., 2017). Unsurprisingly, public discourse and policy have often favored interventions that directly or indirectly limit or slow down internal migration (Selod and Shilpi, 2021; Todaro 1997; United Nations 2013). There is, however, limited robust evidence on the effect of migrants on urban development. Virtually no studies have explored the indirect or dynamic effect of urban migration (how migrants increase urban unemployment and/or reduce urban wages), which is data demanding and methodologically challenging. 6 The findings from the few available studies, such as Kollamparambil (South Africa, 2017) and Combes et al. (China, 2020), are 6It requires panel data spanning longer time periods or at a minimum, large spatial variation combined with historical data which is often only available for very large countries. Neither are widely available. 4 further mixed and focused on urban inequality (not urban development per se). Both show an increase in urban inequality from rural-urban migration but reflect quite different labor market dynamics. The increase in urban inequality follows rising urban unemployment from rural-urban migration in South Africa (no rising wage inequality). In China, on the other hand, all urban residents see their earnings increase (including unskilled workers), though high-skilled residents gain most. Most of the available evidence on the effect of migration on urban development directly compares social, labor, and welfare outcomes between migrants and urban nonmigrants instead, either nationwide across countries or for specific city cases. It provides prima facie evidence of the extent of migrant integration into the urban fabric. If migrants perform similarly to the urban nonmigrants, it suggests that the city manages to absorb them well, as opposed to adding a burden, at least in a static sense. Looking across 12 African countries, Gollin et al. (2021) find that rural-urban migrants have similar housing characteristics as urban nonmigrants, while employment rates among migrants were at least as high as those among nonmigrants in urban Francophone West Africa in the 1980s and 1990s, (Beauchemin and Bocquier, 2004). Migrants integrate poorly in Techiman, Ghana (Ofori-Boateng, 2017), while they tend to have higher employment levels than urban nonmigrants, work longer hours, but earn lower wages in Jijiga, Ethiopia, with differences in education levels and sex between migrants and nonmigrants explaining the wage gaps (Christiaensen and Lozano-Gracia, 2023). In the cities of Jendouba and Kairouan, Tunisia, migrants are at least as likely to be employed as nonmigrants, with those employed enjoying a 30-36 percent wage premium, and monthly 5 earning differentials rising over time (Amara et al., 2023). In Jinja, Uganda, migrants have higher rates of wage employment and wages similar to or higher (on the city outskirts) than those of nonmigrants (Christiaensen and Lozano-Gracia, 2023). Building on these initial findings about internal migrant labor market integration in urban Africa and drawing on insights from the international migration literature (Abramitzky et al, 2014; Caron and Reeve, 2018), this study examines how well migrants integrate into urban labor markets in Africa. Particularly, it provides a more recent and comprehensive assessment of migrant integration into African urban labor markets using more granular labor market outcomes than employment rates, as well as broader welfare indicators across a wider range of countries. It does so by comparing how migrants fare in terms of this wider range of labor market and welfare indicators compared to nonmigrants. 7 In a subset of the countries (Ethiopia, Tanzania, and Uganda), it also explores how potential differences in labor market and welfare outcomes can be explained by differences in socio- economic characteristics such as educational status, dependency ratios and sector of employment or endogenous location choice. The study also explicitly addresses heterogeneity in integration across migrants. Integration likely differs depending on the duration of the migrant’s stay, but also the migrant’s origin (rural/urban) and destination (town/city). Migrants may integrate poorly initially, for example, but build their human and social capital over time and eventually perform similarly or even better in the labor market than urban nonmigrants (Chen et al., 7Most of the migration literature assesses the outcomes of migrants by comparing them with those left behind at the origin, not those at destination. This addresses another question. 6 2018). But migrants do not necessarily face an occupation-based earnings penalty before their incomes and occupations converge (Abramitzky et al., 2014). Second, migrants’ reasons for migrating (e.g., employment or marriage) and their human, financial, and social capital are often correlated with their place of origin (rural/urban). As a result, migrants from rural and other urban areas may integrate differently (de Brauw et al., 2014; Selod and Shilpi, 2021). Finally, higher economic activity and wages in large cities lead to higher expected wages, resulting in larger influxes of migrants (Amare et al., 2021; Selod and Shilpi, 2021) and potentially also larger frictions in integration (Beauchemin and Bocquier, 2004; Busso et al., 2021; Harris and Todaro 1970). To explore this, differences in integration patterns between moves to large cities and towns are explicitly assessed. The findings confirm that migrants contribute substantially to Africa’s urban labor force. They also display substantial heterogeneity in their duration of stay, origin, and destination. Across the six African countries studied, migrants account for about a third of the urban labor force on average. About half of them arrived recently (over the past three years) and about one-third to one-half come from other urban areas. It confirms the importance of moving beyond the traditional singular focus on rural-urban migration when exploring the link between migration and urban development. Migrants in the countries studied further distribute about equally between towns and secondary cities (< 1 million inhabitants) and big cities (>1 million inhabitants) and tend to be younger, more educated, with fewer dependents than the nonmigrants at their destinations, except for rural-city migrants who have more dependents and are significantly less educated. 7 Overall, migrants integrate swiftly and well, with some variation depending on migrant origin and destination. Migrants are at least as likely to work as urban nonmigrants in both large cities and secondary cities and towns across the countries studied. In secondary cities and towns, migrants from both rural and urban areas tend to work a similar number of hours (if not, slightly more hours) and earn similar monthly wage incomes to nonmigrants, which translates into similar household income and consumption levels. In big cities, migrants from urban areas also have similar labor market and welfare outcomes to nonmigrants. The experience of migrants from rural areas to big cities most closely resembles the traditional view on migrants’ lack of integration. Rural-city migrants work longer hours at lower wages, though after 3-4 years, their labor market and welfare outcomes also reach the level of nonmigrants. The paper proceeds as follows. Section 2 describes the information base, lays out the metrics of migrant and urban area used, and presents the socio-economic profile of urban migrants and their share in the African urban labor force. The empirical methodology to examine the extent of urban labor market integration is reviewed in Section 3. Section 4 shows the results of the comparisons between migrants and nonmigrants by duration of stay, migrant origin, and city size. Section 5 concludes. 2 Migrants, Cities and Their Characteristics 2.1 Information Base The study draws on individual and urban-level data. Individual-level data are used to define each individual’s labor market and welfare outcomes, their demographic and socio- 8 economic characteristics as well as their residential status (migrant/nonmigrant), the duration of migrants’ stay (short/long-term) and their origin (rural/urban). The data cover six countries and come from censuses 8 and nationally representative household surveys (Table 1). Employment data are available for all six countries. The survey data for Ethiopia, Tanzania, and Uganda have additional information on hours worked and individual wages (for those employed). Household income and consumption data are further available for Tanzania and Uganda. Each individual-level dataset also contains information on age, sex, education, and sector of employment (agriculture, manufacturing, or service). Data on urban agglomerations come from Africapolis (OECD, 2020). It consistently defines urban agglomerations as continuously built-up areas with a total population of at least 10,000 inhabitants. Official definitions of urban differ across countries and data sources, making it hard to compare results across countries. The agglomeration data used are from 2010 and define 1,645 urban agglomerations across the six study countries. They are used to determine migrant destination by city size. 2.2 Migrant and City Definitions Migrants are defined as individuals who moved from another district or zone (first administration divisions) and are living in urban areas at the time of the survey. Individuals 8The census data are obtained from Integrated Public Use Microdata Series (IPUMS), which publicly provides 10 percent random samples of each census (Minnesota Population Center, 2020). 9 who moved more than 10 years ago are considered urban nonmigrants. 9 Migrants are further categorized by their duration of stay, origin, and destination. Short-term migrants are those who have moved in the past 0-3 years; long-term migrants are those who have moved in the past 3-10 years. In Uganda, distinctions between duration are not made. There is only data on whether individuals lived in the same area in the past five years, rather than ten. In Ethiopia, Tanzania, and Uganda information is also available on migrants’ origin. 10 Rural-urban (urban-urban) migrants are individuals who migrated from rural (urban) districts/zones. The analysis only considers working-age individuals. Appendix Table A1 displays the sample size and corresponding represented populations of each survey/census. Four city-size categories are defined from Africapolis: small towns (10,000 – 20,000 inhabitants), large (or secondary) towns (20,000 – 100,000), small (or secondary) cities (100,000 – 1 million), and large cities (greater than 1 million). 11 A less granular definition considering only other urban areas (small cities, large and small towns) and big cities is used in the main econometric specifications due to sample size and representativeness concerns in the non-census data (Ethiopia, Tanzania, and Uganda). Individuals are mapped to the urban agglomerations using each country’s definition of urban 12 and linking urban residents’ district boundaries to Africapolis’ geo-referenced 9 Place of birth is sometimes also considered in classifying people as migrants. In the tables and figures below, this is the case for Ethiopia and Tanzania. 10 This information is not available in the census data for Ghana, Kenya, and Mali. 11 The big cities in this study are Nairobi, Kisumu, and Mombasa (Kenya), Dar es Salaam (Tanzania), Accra and Kumasi (Ghana), Addis Ababa (Ethiopia), Kampala (Uganda), and Bamako (Mali). 12 In Uganda, where individual residence was provided at the sub-district level, sub-districts (instead of districts) were mapped to the urban agglomerations from Africapolis. 10 data. 13 Particularly, individuals classified as urban based on each country’s definition of urban are assigned to the census/surveys’ administration areas/districts with their geo- referenced boundaries identified from a global database of administrative boundaries (Runfola, et. al., 2020). These districts are linked to the Africapolis urban agglomerations if the urban agglomeration falls within the district’s boundaries. Districts without an Africapolis urban agglomeration are dropped from the dataset, and districts including multiple agglomerations are excluded in Ghana (9 districts), Kenya (9) and Mali (1). 14 The city size distribution thus obtained does not perfectly match the corresponding city size distribution observed in Africapolis, but the focus in this study is also on the within-city comparison of migrant and resident profiles by city size, and not on the city size distribution per se. 15 2.3 Migrants across Cities Migrants contribute substantially to Africa’s urban labor force. They make up about a third of the urban labor force across the six countries studied (Table 2). The small share of migrants in urban Uganda is largely definitional; the data only allows for observing whether 13 Ideally each person in the individual datasets could be linked directly with Africapolis data, but individuals are not geo-referenced in the censuses (Ghana, Kenya, and Mali) and the labor force survey (Ethiopia), and for confidentiality purposes, geo-references are offset in the household survey data (Tanzania, and Uganda). 14 In Kenya, too many districts had agglomerations with mixed city sizes to warrant their automatic exclusion. Therefore, if 70 percent of the total agglomeration area from Africapolis in the specific district consisted of one city size classification, then that district was assigned the dominant city size classification. Nine districts were still dropped because the composition of agglomeration city sizes did not satisfy this criterion. 15 The city size distribution deviates most from the Africapolis city size distribution for Uganda (between 12 and 35 percentage points), while it is almost identical in Ethiopia. In the other countries, the difference across the various size categories mostly ranges between 5 and 15 percentage points. There is a better (albeit still imperfect) match with the city size distribution reported in the World Development Indicators (WDI), which are based on official definitions of urban areas. 11 migrants moved in the past five years, rather than ten. About half of the migrants in urban areas arrived recently (less than 3 years ago) and about 40 percent come from other urban areas, underscoring the importance of looking beyond rural-urban migration when analyzing the effect of migration on urban development. Migrants move about equally to other urban areas and big cities. They make up a larger portion of the urban population in small and big cities (more than 100,000 inhabitants) than in small and large towns (20,000 to 100,000), even making up the majority in the big cities of Tanzania and Kenya. They are slightly more likely to be from rural areas in large towns than in cities and tend to stay for slightly shorter periods of time in towns in all countries, except Tanzania. One reason for the more rural and temporary nature of migrants in towns could be ladder migration (Lucas, 2022). 2.4 Migrants and Nonmigrants Migrants are younger than nonmigrants, irrespective of their origin (rural/urban) or destination (town/city) (Table 3). 16 Migrants moving to towns also live in families with lower dependency ratios than town nonmigrants. This is no longer the case when they move to cities (Table 3, col 2). In fact, rural-big city migrants increase the dependency ratio— consistent with the notion of “migrants burdening the cities’ social services”. 16Dependency ratio defined as the ratio of the number of non-working-age household members over the number of working-age household members (15–64 years old). 12 Migrants to towns (though not those to cities) further substantially augment the skill pool of their destinations. 17 In towns, the share of migrants with no education is on average 4.5 percentage points lower than this of nonmigrants, while they are also 7.6 percent more likely to have pursued some post-secondary education (Table 3, col 1). Disaggregation by migrant origin suggests the migrant augmentation of the town human capital mostly follows from the higher educational achievements of the urban-town migrants. Rural-town migrants are on average also slightly more educated than town nonmigrants, yet urban-town migrants are much less likely to be uneducated and many more of them have pursued at least some post-secondary education (Table 3, cols 3 and 5). 18 In cities, migrants are on average less educated than nonmigrants, with the difference fully driven by rural-city migrants as urban-city migrants tend to be similarly educated (Table 3, cols 4 and 6). The higher educational achievements of urban-town migrants compared with town residents reveals possible urban sorting of urban citizens – lesser- skilled urban citizens often move to rural or smaller urban areas with smaller skill pools and agglomeration economies, maximizing the return to their skills. In contrast, more educated migrants from towns and rural areas move to cities, where greater agglomeration economies and opportunities for specialization maximize the value of their skills (Young, 2013). Finally, as noted by Potts (2018) and Henderson and Kriticos (2018), the agriculture sector continues to play a nonnegligible role in urban employment, at least outside the big 17 In the remainder of the text, “towns” and “cities” are used as shorthand for “other urban areas (<1 million inhabitants)” and “big cities (> 1 million inhabitants)” respectively. 18 The share of uneducated urban-town migrants in the three SSA countries examined here is 12.5 percentage points lower than the share of uneducated among the town residents; the share with at least some post- secondary education is 18.5 percent points higher. 13 cities and mainly for urban nonmigrants, not migrants. About 25 percent of urban nonmigrants are employed in agriculture in small towns (0 to 20,000 inhabitants) and about 14 percent in large towns and small cities (20,000 to 1 million inhabitants). 19 Town migrants, however, are much more likely to be employed outside agriculture (by 11.2 percentage points compared to town nonmigrants), with the difference more pronounced for urban-town migrants (16.2 percentage points more likely employed outside agriculture compared to 7.9 percentage points for rural-urban migrants). In cities, less than 2 percent of employment is in the agriculture sector, and there are no statistically significant differences between migrants and nonmigrants. 3 Examining Labor Market Integration To examine the extent of urban labor market integration of migrants, this study compares labor outcomes between migrants and nonmigrants along a range of labor market indicators. These include the quantity and quality of employment (such as individual employment status, hours worked, and wages) as well as broader household welfare indicators such as household income and consumption per adult equivalent. This allows a more in-depth understanding of the different channels through which integration may operate. Migrants may, for example, be more likely at work for more hours, but at lower hourly wages, which overall may still result in better income and welfare outcomes. 19 Numbers drawn from the analysis, but not reported in Table 3. 14 To fix ideas, the study first explores in a bivariate way by country whether a migrant (or migrant subgroup) is more likely employed than a nonmigrant. The following equation is estimated: = + 1 ∑ + (1) with denoting the employment status of individual i (1=employed; 0=not employed) and an indicator variable indicating whether individual i is a migrant (1= yes; 0= no) and its type j either by duration (short or long term; n=2), by origin (rural-urban or urban-urban; n=2), or by destination (small town, large town, small city or big city; n=4). is a white noise error term. Equation (1) is estimated using a linear probability model for each country, which corresponds to a country-by-country comparison of unconditional means of employment status by population (sub)groups. Individuals are considered employed if they engage in any activity that generates income (e.g., wage employment or running a business). No distinction is made between formal and informal employment. Subsequently, the following two multivariate equations controlling for a series of individual and destination characteristics are estimated: = + 1 + 2 + 3 + 4 × + 5 × + 6 + 7 + + + (2) whereby can be any of the individual labor market (employment, hours worked, individual monthly wages) or household welfare outcomes (income or consumption) for individual or household i, respectively, in country j. Individual wages, household incomes, 15 and household consumption levels are normalized by their respective national averages, to standardize these monetary measures across countries (since all the countries have different currencies). The monetary measures presented in the paper are thus ratios of unit-level outcomes to national outcomes. Household income and consumption are further adjusted for the number of adult equivalents. captures the number of years an individual has stayed in their destination, ranging from 0 to 10 and coded 0 for nonmigrants. and are respective indicators for whether an individual is a rural-urban or urban-urban migrant. is a big city indicator (=1 if ≥ 1 million inhabitants; 0 otherwise) for big city k in country j. is a vector of individual characteristics (age, age-squared, sex, education, household size, dependency ratios, and sector of employment), are district and country indicator variables respectively, and a white noise error term. The focus in equation (2) is on estimating the effect of migration duration on migrant labor market integration and welfare outcomes and whether integration patterns differs by migrant origin (i.e., whether labor market outcomes, for example, converge more rapidly with those of nonmigrants for urban-urban migrants than for rural-urban migrants if integration were to happen only over time). The coefficients of interest are 1 , and 4 and 5 on and the interaction terms × and × respectively. Equation (3) zooms in on migrant labor market integration by city size at destination and how it differs depending on migrant origin (controlling for migrant duration), with all variables defined as in equation (2). It enables exploring questions such as whether, relative 16 to nonmigrants, rural-urban migrants are more likely to integrate in towns than in cities or whether urban-urban migrants thrive better in towns than in cities. = + 0 + 1 + 2 + 3 × + 4 + 5 × + 6 + + + + (3) The interaction terms in equation (3) capture individual-level (in terms of origin) and urban agglomeration-level (in terms of city size) heterogeneity, which allow for comparisons both within and across city size classifications. Equation (1) is estimated separately for all six countries covered in the study (surveys and censuses) and limited to employment status. Equations (2) and (3) are estimated for the three countries with survey data (Ethiopia, Tanzania, and Uganda), with the household outcomes only examined for Tanzania and Uganda. To benchmark the results, equations (2) and (3) are first estimated excluding the controls ( ) (referred to as ‘unconditional’ comparisons). These unconditional comparisons are important for understanding how well migrants integrate into markets in general. Next, controls for socio-demographics and sector of employment are included to show how well these individual characteristics can explain labor market integration. For Equation (2), margins analysis is conducted to better understand how duration of stay is correlated with integration. Particularly, the estimated coefficients, 4 and 5 are combined with the years of duration ranging from 0 to 10, to show how both rural-urban and urban-urban migrants compare to nonmigrants by duration of stay. Finally, migrants may go to certain cities for unobserved characteristics of that city (for example because they 17 are more buoyant, offering better employment opportunities), so district fixed effects ( ) are further included in equations (2) and (3) to help test the robustness of the findings to individual city characteristics. For both (2) and (3), standard errors are estimated using survey weights and appropriate non-linear regression models. All observations are weighted using population weights provided by each dataset. For employment outcomes, a linear probability model is estimated to obtain the coefficients and standard errors (Wooldridge, 2013). For hours worked, which is a censored outcome variable with a lower bound of zero, the Tobit model is applied. For all other outcome variables (wages, income, and consumption), coefficients and standard errors are estimated using OLS. 4 Integration with Heterogeneity 4.1 At Least as Likely Employed Direct comparison of the employment rates of urban migrants and nonmigrants suggests that, across countries, migrants are more likely to be employed than urban nonmigrants, except for Uganda, where there is no statistically significant difference in employment rates (Table 4). 20 In addition, both short- and long-term migrants have on average higher employment rates than nonmigrants. Yet the difference is on average larger for long-term migrants and in two countries, short-term migrants are also less likely employed than urban nonmigrants (Kenya, Ghana). It suggests, prima facie, improving 20The exception may be linked to Uganda’s definition of migrants as having moved in the past 5 years, rather than the past 10 years as in the other countries. 18 integration over time. There is no systematic difference in the employment rate gap between migrants and nonmigrants by migrant origin. There is also no systematic pattern in the employment rate difference between migrants and nonmigrants by city size. Bivariate comparison of employment rates thus suggests that migrants are overall at least as likely to be employed as nonmigrants and this holds irrespective of their duration of stay, their origin (rural or urban) or whether their destination is a town or city, though with a hint that integration improves over time. Higher employment rates, however, are only one indicator of labor market integration and overall migrant welfare. Migrants may be highly employed because they must work to earn a living in the absence of a social safety net. They may not be able to afford not to work. Therefore, they may take on more temporary, hazardous, or low-paying jobs. This could lead to fewer hours worked and lower wages, which could translate into lower income and consumption levels. It is thus key to explore how migrants fare compared to nonmigrants along the full range of labor market and welfare indicators (i.e., employment as well as hours worked, wages, income, and consumption). Higher employment rates (or differences in other labor outcomes) may also follow because of migrants are younger and more educated (at least those moving to the towns) or because of other features of the destination not captured by city size. Multivariate analysis, controlling for individual migrant and destination characteristics, helps unpack some of the dynamics driving migrant labor market integration (including discrimination) and how it differs by migration type (recent, long-term; rural, urban; town, city). To better understand the impact of migration duration on integration, a continuous (instead of a categorical) variable of migration is further used. 19 4.2 Rapid Integration Overall While migrants are on average more employed than nonmigrants, regardless of origin (Table 4), this holds especially during the early years of entry (Figure 1, Panels A and B). Rural-urban and urban-urban migrants are 5 and 7 percent more likely to work than nonmigrants in their first year (year 0), respectively, but the gap disappears by their third year (year 2) of stay (Figure 1, Panel 1A). Conditional on migrant and destination characteristics, these differences increase and last a bit longer. This holds especially for rural-urban migrants (Figure 1, Panel 1B). They are also less educated than their urban non- migrant counterparts (averaged across towns and cities). 21 Among those employed, the gap in the number of hours worked (the intensive employment margin) displays a similar pattern, even though less pronounced (Figure 1, Panel 1C). All migrants tend to work slightly longer hours initially, but the gap disappears rapidly. It is also not statistically significant. The initial differences become larger and statistically significant for rural-urban migrants when controlling for migrant and destination characteristics (though not for urban-urban migrants) (Figure 1, Panel 1D). Other factors than differences in the characteristics of migrants and the city characteristics of where migrants choose to go appear at play, possibly related to the larger need or more deliberate intention of rural-urban migrants to work (e.g., migrating specifically to work). 21 Controlling for migrant and destination characteristics, rural-urban and urban-urban migrants are respectively 18 and 10 percent more likely to work in their first year, and the gap does not disappear until the fourth and third year of stay respectively (Figure 1, Panel 1B). With the more educated more likely employed and rural-urban migrants less educated, the effect of being a rural-urban migrant on employment status increases when controlling for educational status (Table A2, cols 1 and 2). 20 Turning to monthly wages, initially longer work hours do not translate into higher monthly wages (Figure 1, Panel 1E). In fact, for rural-urban migrants, the monthly wages are initially (statistically significantly) lower, indicating lower hourly wages. Controlling for migrant and destination characteristics, the effect is not statistically significant (Figure 1, Panel F), suggesting it is likely because of the type of job, not discrimination. Moreover, hourly wages for both rural-urban and urban-urban migrants likely increase over time compared to nonmigrants (Figure 1, Panels D and F). 22 They almost immediately exceed them for urban-urban migrants upon arrival and also do so after a couple of years for rural- urban migrants. Combined with the higher initial employment ratios, this is, overall, indicative of swift labor market integration (rather than discrimination). Finally, considering the more comprehensive welfare indicators, migrants do also not perform worse than nonmigrants in terms of household income and consumption from the beginning (Figure 1, Panel 1G and 1I). From arrival throughout the duration of stay, the gaps are not statistically significantly different. Consistent with the labor market outcomes, urban-urban migrants furthermore tend to better than rural-urban migrants, with the welfare gaps with nonmigrants typically slightly positive for the former and slightly negative for the latter. Conditional regressions suggest lower income for rural-urban migrants the first year, pointing to the lower dependency ratios of migrant households (Figure 1, Panel 1K). 22To see this, note that the individual monthly wage gap (migrants minus nonmigrants) displays an upward gradient (Panel F), while the gap of hours worked (Panel 1D) declines. While neither gradient is statistically significant, the directions are consistent with an increase in migrant hourly wages over time. 21 In sum, during the initial years migrants work more than nonmigrants (both at the extensive and intensive margins), though the differences disappear within a few years. For rural-urban migrants longer working hours do not immediately translate into higher wage earnings, likely because of the initial adoption of lower paying jobs, not discrimination. This does however not translate into statistically lower household incomes or consumption. Urban-urban migrants tend to perform better still with wages, income, and consumption at least as high or higher from the beginning, even after controlling for migrant and destination characteristics. Together the findings suggest that migrants integrate overall quickly and well, also rural-urban migrants, though the latter tend to take a bit more time. 4.3 The Rural-City Migrant Figure 2 presents the unconditional and conditional labor market and welfare gaps between migrants and nonmigrants by destination (town/city) and origin (rural-urban) (controlling for duration of stay). 23 As before, all migrants are more likely employed than nonmigrants, irrespective of their origin (rural/urban), though the effect is most pronounced for rural-city migrants, who also work more hours. 24 Urban-town migrants also work longer, while urban-city migrants tend to work less (not statistically significant) (Figure 2, Panel 2A and 2C). As in Beauchemin and Bocquier (2004) for West Africa in the 1980s, the findings do not support the notion of “migrants joining the ranks of the unemployed”. 23 Detailed regression results are in Tables A5 to A7. 24 Average working hours refer to those employed (Figure 2). 22 For all migrants, the employment gaps further widen when controlling for migrant and destination characteristics (Figure 2, Panel 2B), suggesting that they arise because of other factors inherent to the migrant experience (e.g., migrating specifically to work). For rural migrants, conditional working hours also increase (Figure 2, panel 2D), suggesting that rural migrants are particularly inclined to work, either intentionally or out of necessity. Turning to (unconditional) monthly wage incomes, which reflect both time worked and hourly wages, urban-town migrants earn on average as much as nonmigrant town residents (Figure 2, Panel 2E). Rural-town migrants earn less, despite working similar hours, suggesting lower hourly wages. The wage gaps are larger in big cities. Urban-city migrants earn only half as much as city nonmigrants, but also tend to work fewer hours. The unconditional wage gap is the largest for rural-city migrants, despite working more hours. For most migrants (except the rural-city migrants) the differences basically disappear when controlling for migrant and destination characteristics (Figure 2, panel 2F) or when looking at the broader income measure (household income per adult equivalent) (Figure 2 Panel 2G and 2H). This is most pronounced for both urban and rural migrants moving to towns, but the wage and income gaps are also no longer statistically significant (even though still negative) for urban migrants to cities. The differences between migrants and nonmigrants totally disappear for all migrants (including rural-city migrants) when considering household consumption per adult equivalent (the broader welfare measure) (Figure 2, Panel 3I and 3H). Taken together, the evidence suggests that 1) overall, migrants integrate well, 2) urban migrants do better than rural migrants, and 3) all migrants do better in towns than in 23 cities, where the benchmark is also higher. 25 But rural-city migrants face more challenges and resemble most closely the traditional notion of migration contributing to urban underdevelopment. They work more hours, but earn less, even after controlling for differences in their socio-demographic characteristics compared to city nonmigrants. Overall, it does however not translate into lower consumption, their labor market outcomes tend to converge over time (section 4.2). 4.4 Additional Considerations The findings provide prima facie evidence that, with the possible exception of rural- city migrants, urban migrants integrate swiftly and well into urban labor markets. Yet three issues must further be considered when interpreting the results: 1) sample selectivity, 2) cross-sectional inference, and 3) external validity. First, the study relies only on urban samples, which may introduce selection bias into the results. If unsuccessful migrants disproportionally return to rural areas, then the migrant labor market integration may be overstated because we only observe successful migrants. In SSA, about 33 percent of male rural-urban migrants and 20 percent of female rural-urban migrants return to their rural homes (Cattaneo and Robinson, 2020). However, migrants 25 The city workforce is typically better educated and the labor pool more competitive. Big cities are typically also richer, wealthier, and better serviced with public goods and social services than towns and secondary cities, leading to wage premiums (de Weerdt et al., 2021; Ferré et al., 2012; Gollin et al., 2021; Henderson and Kriticos, 2018). The findings here support these previous studies – monthly wages in big cities are around 20 percent higher than in secondary towns and cities (Table A3). Residents of big cities (migrants and nonmigrants) also work more hours per week than in secondary towns and cities (Table A2). 24 may also return to their homes as part of a welfare maximization strategy. They can engage in circular migration whereby they earn and save by earning wage premiums in urban areas and return to their rural homes once their targets are reached (Constant, 2020). Whether return migration is disproportionately driven by unsuccessful migration or welfare gains determines whether unobserved return migration means that labor integration is overstated or understated. Given data limitations, this is hard to establish empirically. Evidence on return migration is scarce, also in developed countries. A case study in Tanzania shows that men are more likely to return because of poor labor market outcomes, while women are more likely to return because of failed marriages (Hirvonen and Lilleør, 2015). In India, older and wealthier male migrants are more likely to return than other migrants, suggesting that urban migration is part of a welfare maximization strategy (Dhar and Bhagat, 2021). Yet more direct evidence is needed to understand whether these mechanisms are likely to bias integration results. For international migrants in France (1975-1999), and uniquely and explicitly examining the effect of selective remigration on cross-sectional measures of immigrant labor market integration, Caron and Reeve (2018) do not find a significant effect. Selection bias can also play a role in distorting the observed outcomes of urban nonmigrants. If geographic sorting patterns as suggested by Young (2013) play out, then the lesser-able urban residents move to rural areas and the more able residents stay in urban areas. This would lead to an underestimate of integration into urban areas. Controlling for education helps mitigate this issue. 25 Second, integration differences by duration of stay are also prone to bias from cross- sectional analysis. For example, more recent migrants may be lower skilled than past migrants. Convergence in migrant labor market integration over time observed in cross- sectional data, may then be attributed inaccurately to duration of stay, while it is in fact due to the (unobserved) decline in skill levels across subsequent migrant cohorts, as has been documented, for example, for international migrants from Europe to the US during the second half of the 19th century (Abramitzky et al., 2014). Controlling for educational attainment (and age) in the results reported above helps control for such bias. Nonetheless, it cannot be excluded. As enrollment rates in Africa increased over the past couple of decades, schooling quality also decreased. To the extent that this affects the educational achievements of migrants in the samples studied here, 26 it may bias the results downward, with the relative swift migrant integration observed in the cross-sectional labor market data representing a lower bound instead. Finally, the study focuses on six African countries – two from West Africa and four from East Africa. However, for more granular labor market and welfare outcomes, only East African countries are included. The results therefore mostly reflect the experience of migrants in East African countries, which have the lowest urbanization rates and the fastest growing urban populations in Africa. In East Africa, rural-urban migration contributes twice as much to urban growth as natural population increase (Baeumler et al., 2020). 26Many migrants in the sample had already completed their education before the primary enrollment expansion took off in the 2000s. Furthermore, the commonly held perception of an access-quality trade-off in Africa has less empirical support than previously believed to be the case (Taylor and Spaull, 2015; Valente 2019), with substantial variation across countries (Filmer 2023). 26 Since urban populations are rapidly growing in these countries, the pressures of urban growth on labor markets are likely higher than in countries with slower growing urban populations. Testing the predictions of the Harris-Todaro model of higher unemployment with rural-urban migration, Busso et al. (2021) show that they are most applicable among workers with only primary education, conditions which are more likely to hold in big cities, surrounded by large rural hinterlands disproportionately populated by young adults more prone to migrate. Such conditions are more frequent in less urbanized countries than in high urbanized settings. Against this background, the migrant integration findings obtained here should arguably be even better in settings where urbanization and urban growth are more reliant on natural population growth and less driven by rural-urban migration. The findings are consistent with the experience of West African migrants in the 1980s and 1990s, when urbanization and urban growth in West Africa were similar to the urbanization and urban growth in East Africa in the 2000s and 2010s (Beauchemin and Bocquier, 2004). 5 Conclusion Eighty-five percent of African countries have measures in place to slow internal migration to urban areas (United Nations, 2013). These measures have been adopted with the premise, put forth by Harris-Todaro models, that internal migrants (particularly, rural- urban migrants) lead to underdevelopment in urban centers. Little work has been done to assess the underlying assumptions of these theories, or even more rudimentarily, the extent to which internal migrants manage to integrate in Africa’s urban centers, the focus of this 27 study. Building on the international migration literature (Abramitzky et al, 2014; Caron and Reeve, 2018) as well as initial findings about internal migrant labor market integration in urban West Africa (Beauchemin and Bocquier, 2004), this study examines how well migrants integrate into urban labor markets across six African countries, expanding the scope to include urban-urban migrants and differentiating between integration in towns and cities and the duration of stay. The empirical findings indicate that migrants to Africa’s urban centers integrate, overall, swiftly and well into Africa’s urban labor markets. Migrants coming from other urban areas tend to do better than rural migrants, and migrant integration in towns is smoother than integration in cities. Rural-city migration most closely resembles the Harris-Todaro notion of migrants exacerbating the urban labor market challenges. Even so, while experiencing an occupational earnings penalty, rural-city migrants are still more likely to be employed and work longer hours than local nonmigrants and their consumption levels are on par with those of their local counterparts. Rural-city migrants also represent only part of the urban migrant population. There are no immediate detectable signs of migrant discrimination or migrant selectivity bias. As such, the findings do not provide support for policies hindering internal migration to urban areas. However, an important research agenda remains. More work is needed, covering a broader set of geographies for wider external validity as well as panel data analysis to better protect against the potential bias in integration estimates from cross- sectional studies. The increasing availability of standardized longitudinal datasets for African countries (e.g., DHS and LSMS) renders such analysis increasingly feasible. The 28 current study abstracted from those datasets to examine the dynamic, indirect effects of internal migration on urban labor markets. Combes and Gobillon (2015) provide solid methodological and empirical guidance for doing so. 29 References Abramitzky, R., Boustan, L.P., and Eriksson, K., (2014). A Nation of Immigrants: Assimilation and Economic Outcomes. Journal of Political Economy 122-3: 467-506. Africapolis 2021 (https://africapolis.org/data, consulted April 2021). Ahimah-Agyakwah, S., Nketiah-Amponsah, E., and Agyire-Tettey, F., (2022). Urbanization and Poverty in Sub-Saharan Africa: Evidence from Dynamic Panel Data Analysis of Selected Urbanizing Countries. Cogent Economics and Finance, 10: 2109282. Amara, M., Lozano Gracia, N., and Christiaensen, L. (2023). Are Migrants Second-Class Citizens in the Urban Labor Markets? Evidence from Two Secondary Cities in Tunisia. World Bank. Paper Presented at the Third Annual Conference of the Regional Research Network of Central Banks of the Middle East and North Africa, September 14-15, 2023 Amare, M., Abay, K. A., Arndt, C., and Shiferaw, B. (2021). Youth Migration Decisions in Sub- Saharan Africa : Satellite-Based Empirical Evidence from Nigeria. Population and Development Review, 47(1), 151–179. Baeumler, A., D’Aoust, O., Gapihan, A., Goga, S., Lakovits, C., Restrepo Cavadid, P., Singh, G., and Terraza, H. (2020). Demographic Trends and Urbanization. In S. Wahba, A. Wellenstein, & M. B. Das (Eds.), Demographic Trends and Urbanization (pp. 24–44). Beauchemin, C., and Bocquier, P. (2004). Migration and urbanisation in francophone west Africa: An overview of the recent empirical evidence. Urban Studies, 41(11), 2245–2272. Beegle, K., and Christiaensen, L. (2019). Eds. Accelerating poverty reduction in Africa. World Bank: Washington D.C. Busso, M., Chauvin, J. P., and Herrera L., N. (2021). Rural-urban migration at high 30 urbanization levels. Regional Science and Urban Economics, January. Caron, L., and Reeve, P. (2018). Whose Integration Do We Measure? Immigrants' Remigration and Labour Market Integration in France. Population 73(3): 481-518. Cattaneo, A., and Robinson, S. (2020). Multiple moves and return migration within developing countries: A comparative analysis. Population, Space and Place, 26(7), 1–21. Chen, B., Liu, D., and Lu, M. (2018). City size, migration and urban inequality in China. China Economic Review, 51(August 2016), 42–58. Christiaensen, L., and Lozano-gracia, N. (2023). Migrants, Markets, and Mayors: Rising above the Employment Challenge in Africa’s Secondary Cities (L. Christiaensen & N. Lozano- gracia (eds.); 1st ed.). AFD and World Bank. Combes, P. P., Démurger, S., Li, S., and Wang, J. (2020). Unequal migration and urbanisation gains in China. Journal of Development Economics, 142, January 102328. Combes, P. P., and Gobillon, L. (2015). The Empirics of Agglomeration Economies. In Handbook of Regional and Urban Economics (1st ed., Vol. 5). Elsevier B.V. Constant, A. F. (2020). Time-Space Dynamics of Return and Circular Migration. In Handbook of Labor, Human Resources and Population Economics. de Brauw, A., Mueller, V., and Lee, H. L. (2014). The role of rural-urban migration in the structural transformation of Sub-Saharan Africa. World Development, 63, 33–42. de Weerdt, J., Christiaensen, L., and Kanbur, R. (2021). When Distance Drives Destination, Towns Can Stimulate Development. In IZA Discussion Papers (No. 14157; IZA Discussion Papers). Dhar, B., and Bhagat, R. B. (2021). Return migration in India: internal and international 31 dimensions. Migration and Development, 10(1), 107–121. Ethiopia, C. S. A. of. (2014). Labour Force Survey Dataset Version 1.0. Central Statistical Authority of Ethiopia. Farrell. K. (2017). The Rapid Urban Growth Triad: A New Conceptual Framework for Examining the Urban Transition in Developing Countries.” Sustainability 9(8): 1407. Ferré, C., Ferreira, F. H. ., and Lanjouw, P. (2012). Is There a Metropolitan Bias? The relationship between poverty and city size in a selection of developing countries. The World Bank Economic Review, 26(3), 351–382. Filmer, D., (2023). Long-Lived Consequences of Rapid Scale-Up. The Case of Free Primary Education in Six Sub-Saharan African Countries. Internationl Journal of Educational Development. 102871. Gollin, D., Kirchberger, M., and Lagakos, D. (2021). Do urban wage premia reflect lower amenities? Evidence from Africa. Journal of Urban Economics, 121(July 2020), 103301. Harris, J. R., and Todaro, M. P. (1970). Migration , Unemployment and Developmnent: A Two- Sector Analysis. The American Economic Review, 60(1), 126–142. Henderson, J.V. , and Kriticos, S. (2018). The Development of the African System of Cities. Annual Review of Economics, 10, 287–314. Hirvonen, K., and Lilleør, H. B. (2015). Going back home: Internal return migration in rural Tanzania. World Development, 70, 186–202. Jedwab, R., Christiaensen, L., and Gindelsky, M. (2017). Demography, Urbanization and De- velopment: Rural Push, Urban Pull and… Urban Push? Journal of Urban Economics, 98, 6-16. 32 Kollamparambil, U. (2017). Impact of Internal In-Migration on Income Inequality in Receiving Areas: A District Level Study of South Africa. Journal of Development Studies, 53(12), 2145–2163. Lall. S., Henderson, J.V., and Venables, A.J.. 2017. Africa's Cities: Opening Doors to the World. World Bank, Washington D.C. Lucas, Robert. 2022. Crossing the Divide: Rural to Urban Migration in Developing Countries. Oxford: Oxford University Press. Menashe-Oren, A., and Bocquier, P. (2021). Urbanization Is No Longer Driven by Migration in Low- and Middle-Income Countries (1985–2015). Population and Development Review, 0, 1–2. Minnesota Population Center. (2020). Integrated Public Use Microdata Series, International: Version 7.3 [dataset]. IPUMS. National Bureau of Statistics. (2015). Tanzania National Panel Survey 2008-2015, Uniform Panel Dataset (NPS-UPD) 2008-2015. Ref. TZA_2008-2014_NPS-UPD_v01_M. OECD. (2020). Africa’s Urbanisation Dynamics 2020: Africapolis, Mapping a New Urban Geography, West African Studies. https://www.oecd- ilibrary.org/content/publication/b6bccb81-en OECD/UN ECA/AfDB (2022), Africa’s Urbanisation Dynamics 2022: The Economic Power of Africa’s Cities, West African Studies, OECD Publishing, Paris, https://doi.org/10.1787/3834ed5b-en. Ofori-Boateng, Kenneth (2017). Analysis of Severity of Poverty and Social Cohesion among the Urban Poor Migrants in Ghana. Journal of Poverty 21 (3): 265–87. 33 Østby, G. (2016). Rural–urban migration, inequality and urban social disorder: Evidence from African and Asian cities. Conflict Management and Peace Science, 33(5), 491–515. Potts, D. (2018). Urban data and definitions in sub-Saharan Africa: Mismatches between the pace of urbanisation and employment and livelihood change. Urban Studies, 55(5), 965– 986. Runfola, D. et al., (2020). geoBoundaries: A global database of political administrative boundaries. PLoS ONE, 15(4). Selod, H., and Shilpi, F. (2021). Regional Science and Urban Economics Rural-urban migration in developing countries : Lessons from the literature. Regional Science and Urban Economics, 91, 103713. Taylor, S., and Spaull, N. (2015). Measuring Access to Learning Over a Period of Increased Access to Schooling: The Case of Southern and Eastern Africa since 2000. International Journal of Educational Development. 41 (March): 47-59. Todaro, Michael P. 1997. Urbanization, Unemployment and Migration in Africa: Theory and Policy. New York: Population Council.Uganda Bureau of Statistics (UBOS). (2018). Uganda National Household Survey 2016/2017. UBOS. United Nations. (2013). World Population Policies. New York: UN Department of Economic and Social Affairs. https://esa.un.org/PopPolicy/wpp_datasets.aspx. United Nations. (2019). World Urbanization Prospects: The 2018 Revision. United Nations: New York.Vernon Henderson, J., and Kriticos, S. (2018). The Development of the African System of Cities. Annual Review of Economics, 10, 287–314. Valente, C., (2019). Primary Education Expansion and Quality of Schooling. Economics of 34 Education Review, 101913. Wang, W., and Fan, C. (2012). Migrant workers’ integration in Urban China: Experiences in employment, social adaptation, and self-identity. Eurasian Geography and Economics, 53(6), 731–749. Wooldridge, J. M. (2013). Introductory Econometrics: A Modern Approach. (5th ed.). South- Western Cengage Learning. Young, A. (2013). Inequality, The Urban-Rural Gap, and Migration. The Quarterly Journal of Economics, 128(4), 1727–1786. 35 Tables Table 1: Individual-Level Data Sources Country Data Source (year) Data Available Ethiopia Labour Force Survey (2013) Employment status, additional individual labor market outcomes, migration duration, migration origin Ghana Census (2010) Employment status, migration duration Kenya Census (2009) Employment status, migration duration Mali Census (2009) Employment status, migration duration Tanzania National Panel Survey (2010) 27 Employment status, additional individual labor market outcomes, household outcomes, migration duration, migration origin Uganda National Household Survey (2016) Employment status, additional individual labor market outcomes, household outcomes, migration origin 27 Wave 2 of the ISA-LSMS Tanzania National Panel Survey is chosen because surveys in Wave 2 were carried out in 2010, which corresponds perfectly with the Africapolis data used to define urban agglomerations (see below) and corresponds closely with the census years for Ghana, Kenya, and Mali – 2010, 2009, and 2009, respectively. 36 Table 2: Population Shares by Migrant Class Working age population Small Towns: Large Towns: Small Cities: Big cities: Total (15-64 year-old) 0-20k 20k-100k 100k-1000k (>1 million) Migrant share of urban population Ethiopia1 (2013) 46 45 43 25 40 Tanzania 1 (2010) 20 18 36 53 32 Uganda2 (2016) 09 12 17 16 13 Ghana (2010) 26 23 25 40 31 Kenya (2009) 33 29 37 60 47 Mali (2009) 28 23 26 42 35 Country average 27 25 31 39 33 Short term (0-3 years) Ethiopia1 (2013) 39 39 37 35 38 Tanzania (2010) 1 44 45 48 58 52 Ghana (2010) 49 48 49 46 47 Kenya (2009) 54 56 54 51 53 Mali (2009) 51 51 46 49 50 Country average 53 54 50 50 51 Share of rural-urban migrants Ethiopia (2013) 69 57 47 54 58 Tanzania (2010)1) 72 86 72 77 77 Uganda (2016) 2) 38 54 50 55 47 Country average 60 66 56 62 61 Notes: Unless specified otherwise, a person is a migrant if he moved to an area less than 10 years ago. 1) migrants are people who moved into a zone (Ethiopia) or district (Tanzania) that is not their birth district less than 10 years ago; 2) migrants are people who moved to a district less than 5 years ago. Data sources: Ethiopia (Labor force survey); Tanzania and Uganda (Living standard measurement surveys); Ghana, Kenya, and Mali (Censuses). 37 Table 3: Socio-economic Characteristics by Migrant Origin and City Size Share difference (%points) All migrants R-U migrants U-U migrants between urban migrant and (6 countries) (3 countries) (3 countries) nonmigrant Town City Town City Town City(6) (1) (2) (3) (4) (5) Age (years) -4.14*** -4.51*** -4.49*** -4.44*** -3.58*** -4.66*** Dependency ratio -13.15*** 2.08 -11.26*** 4.28** -16.10*** -2.48 No education -4.46*** 7.99*** 0.46 10.54*** -12.45*** -2.72 Post-secondary education 7.58*** -8.46*** 0.88* -12.67*** 18.47*** 0.23 Employed in nonagriculture 11.16*** 0.15 7.90*** 0.27 16.16*** -0.07 Notes: Country coverage for all migrants: Ethiopia (2013), Tanzania (2010), Uganda (2016), Ghana (2010), Kenya (2009), Mali (2009); Country coverage for rural-urban (R-U) and urban-urban(U-U) migrants: Ethiopia (2013), Tanzania (2010), Uganda (2016). Towns and cities refer to other urban (< 1 million inhabitants) and big cities (>1million inhabitants) respectively. Differences assessed using student T-Tests with survey design taken into account. *, **, *** signify statistically significant difference at the 10%, 5%, and 1% level, respectively. 38 . Table 4: Employment gaps by migrant duration, origin, and destination Probability migrant is more All urban Short-term Long-term Rural- Urban- Small town Large town Small city Large city employed (15–64-year-old) migrants [0-3 yrs] (3-10 yrs] urban urban (0-20k) (20k-100k) (100k-1 mil) (>1 million) Ethiopia 2013 0.09*** 0.09*** 0.08*** 0.10*** 0.08*** 0.08*** 0.06*** 0.06*** 0.09*** Tanzania 2010 0.05** 0.05** 0.05* 0.04 0.04 0.12** -0.01 0.22*** 0.09*** Uganda 2016 0.02 - - 0.02 0.01 -0.01 0.02 -0.01 0.02 Ghana 2010 0.01*** -0.03*** 0.06*** - - -0.01*** -0.00 -0.09*** 0.02*** Kenya 2009 0.04*** -0.01*** 0.07*** - - 0.04*** 0.08*** 0.06*** 0.02*** Mali 2009 0.02*** 0.04*** 0.02*** - - 0.03*** 0.04*** -0.03** 0.02*** Country average 0.04 0.03 0.06 0.05 0.04 0.04 0.03 0.04 0.04 Notes: Reported numbers are the coefficients of a linear probability model, regressing being employed on a constant and being a migrant (or migrant subgroups). The coefficients reflect how much a migrant (or migrant subgroup) is more likely to be employed on average than an urban native. For Uganda, migrants are people who have resided in an area less than 5 years. Information on the origin of migrants was not available for Ghana, Kenya, Mali. The slight difference in employment rates between those for all urban migrants in Tanzania and their rural-urban and urban-urban subgroups (both lower) arises from the slight difference in the underlying samples. Not all urban migrants in Tanzania could be classified by their origin. *, **, *** signify statistically significant difference at the 10%, 5%, and 1% level, respectively. 39 Figures Figure 1: Labor market outcome and welfare gaps by duration of stay and migrant origin. Employment gap Panel 1A Panel 1B Gap in hours worked Panel 1C Panel 1D 40 Wage gap Panel 1E Panel 1F Income/adult equivalent gap Panel 1G Panel 1H Consumption/adult equivalent gap Panel 1I Panel 1K Notes: Country coverage: Ethiopia (except income and consumption per adult equivalent), Tanzania, Uganda. Coefficients and standard errors estimated using OLS (except for hours worked (Tobit)), as specified in Equation 2, controlling for country fixed effects, and considering survey design. All regressions are run at the individual level. Employment regressions include the entire working age population. Hours worked refers to the total hours worked during the last week in Ethiopia and Uganda, but only hours worked in wage employment in Tanzania. Hours worked regressions only include the employed population. Wages refer to individual monthly wages. Wage regressions only include the wage-earning 41 population. For income and consumption (which are only observed at the household level), all household members are assigned the same household income/consumption. Households with a migrant are classified as migrant households of the corresponding migrant type. If migrants are from varying origin (rural/urban) or destination (town/small city vs. big city) within the same household, they are assigned rural. A value of 0 corresponds to equality between the population of interest (either R-U migrants or U-U migrants) and nonmigrants within the city size of interest (either town/secondary city or large city). 95% confidence intervals are shown. Full regression output can be found in Tables A2-A4. 42 Figure 2: Labor market outcomes and welfare gaps by migrant destination and origin Employment gap Panel 2A Panel 2B Gap in hours worked Panel 2C Panel 2D 43 Wage Gap Panel 2E Panel 2F Income/adult equivalent gap Panel 2G Panel 2H Consumption/adult equivalent gap Panel 2I Panel 2J Notes: Country coverage: Ethiopia (except income and consumption per adult equivalent), Tanzania, Uganda. Coefficients and standard errors estimated using OLS (except for hours worked (Tobit)), as specified in Equation 3, controlling for country fixed effects, and considering survey 44 design. All regressions are run at the individual level. Employment regressions include the entire working age population. Hours worked refers to the total hours worked in the last week in Ethiopia and Uganda, but only hours worked in wage employment in Tanzania. Hours worked regressions only include the employed population. Wages refer to individual monthly wages. Wage regressions only include the wage-earning population. For income and consumption (which are only observed at the household level), all household members are assigned the same household income/consumption. Households with a migrant are classified as migrant households of the corresponding migrant type. If migrants are from varying origin (rural/urban) or destination (town/small city vs. big city) within the same household, they are assigned rural. A value of 0 corresponds to equality between the population of interest (either R-U migrants or U-U migrants) and nonmigrants within the city size of interest (either Secondary Towns and Cities or Large Cities). 95 confidence intervals are shown. Full regression output can be found in Tables A5-A7. 45 Appendix Table A1: Samples Size by Country and Migrant Duration and Origin Country Short Term Long Term Rural-Urban Urban-Urban Total Urban Urban Migrants Migrants (R-U) Migrants (U-U) Migrants Migrants Nonmigrants Ethiopia (2013) 10,662 13,533 12,727 10,937 24,195 54,797 (1,378,877) (1,704,031) (1,775,770) (1,268,008) (3,082,909) (6,233,829) Tanzania (2010) 719 554 818 264 1,273 2,407 (1,139,393) (1,089,760) (1,504,804) (410,728) (2,229,153) (4,964,702) Uganda (2016) - - 1,660 1,320 2,980 33,172 - - (992,542) (785,520) (1,778,062) (17,569,741) Ghana (2010) 118,391 126,281 - - 244,672 518,356 (1,184,219) (1,262,811) - - (2,447,031) (5,183,249) Kenya (2009) 146,863 133,737 - - 280,600 453,602 (1,468,404) (1,337,716) - - (2,806,120) (4,535,900) Mali (2009) 27,360 31,912 - - 59,272 123,157 (273,644) (319,068) - - (592,712) (1,231,578) Notes: Census and survey samples sizes are shown in first line for each country, with the corresponding represented populations, calculated via survey weights, displayed below in parentheses. Differences in total of RU and UU migrants and ST and LT migrants within countries is due to missing data on migrant origin. Total migration is calculated using the sum of short- and long-term migrants, except in case of Uganda, where the sum of rural-urban and urban-urban migrants are used. Population weights are derived from each survey source and are calculated such that the data is nationally representative at the urban level for working age residents. 46 Table A2: Differences in employment and hours worked by duration of stay and migrant origin Employed Hours Worked (1) (2) (3) (4) (5) (6) (7) Duration of Stay (Years) 0.0163* 0.0212** 0.0221*** 0.573 0.625 0.258 0.419 (0.00865) (0.00825) (0.00823) (0.556) (0.548) (0.542) (0.543) Rural-urban migrant (1=yes) 0.0529*** 0.177*** 0.183*** 1.737 4.234*** 3.807*** 3.393*** (0.0153) (0.0150) (0.0144) (1.155) (1.224) (1.230) (1.079) Urban-urban migrant (1=yes) 0.0701*** 0.0962*** 0.104*** 2.164** 1.493 0.739 0.588 (0.0145) (0.0145) (0.0140) (1.032) (1.084) (1.063) (0.932) R-U Migrant X Duration of Stay (Years) -0.0110 -0.0341*** -0.0352*** -0.865 -1.189** -0.801 -0.946* (0.00906) (0.00865) (0.00860) (0.592) (0.588) (0.582) (0.573) U-U Migrant X Duration of Stay (Years) -0.0175* -0.0296*** -0.0318*** -0.591 -0.711 -0.329 -0.573 (0.00911) (0.00869) (0.00863) (0.592) (0.585) (0.578) (0.572) Big city (1=yes) -0.0941*** -0.126*** -0.132*** 9.197*** 7.514*** 6.276*** 7.533*** (0.00775) (0.00678) (0.00657) (0.610) (0.625) (0.647) (0.462) Sex (male=1) - 0.133*** 0.133*** - 4.030*** 4.420*** 4.609*** (0.00665) (0.00626) (0.550) (0.540) (0.457) Age - 0.0666*** 0.0671*** - 0.942*** 0.796*** 0.720*** (0.00152) (0.00144) (0.140) (0.140) (0.112) Age^2 - -0.000770*** -0.000778*** - -0.0120*** -0.0100*** -0.00934*** (0.0000207) (0.0000197) (0.00187) (0.00185) (0.00150) Some primary - 0.0732*** 0.0694*** - -0.237 -0.0948 -0.285 (0.0117) (0.0108) (0.920) (0.898) (0.795) Primary or any secondary - 0.109*** 0.109*** - 2.850** 1.864 1.285 (0.0112) (0.0107) (1.167) (1.160) (1.003) Secondary completed - 0.118*** 0.127*** - 7.029*** 5.407*** 4.763*** (0.0134) (0.0131) (1.209) (1.208) (1.113) Any post-secondary - 0.211*** 0.215*** - 5.288*** 3.264*** 3.108*** (0.0105) (0.0102) (0.848) (0.880) (0.800) Manufacturing (1 = yes) - - - - - 7.635*** 6.600*** (0.877) (0.731) Service (1 = yes) - - - - - 9.546*** 8.304*** (0.728) (0.623) Constant 0.634*** -0.731*** -0.758*** 32.74*** 12.14*** 7.676*** 11.02*** (0.00425) (0.0273) (0.0260) (0.306) (2.673) (2.603) (2.181) Country Fixed Effects Yes Yes Yes Yes Yes Yes Yes District Fixed Effects No No Yes No No No Yes R2 0.0129 0.227 0.257 N 91047 89590 89590 45788 44955 44922 44922 Notes: Country coverage: Ethiopia, Tanzania, Uganda; Hours worked refers to the total hours worked in the last week in Ethiopia and Uganda, but only hours worked in wage employment in Tanzania. Hours worked regressions in Columns 4-7 only include the employed population. Regressions control for country fixed effects; errors are corrected for survey design and regressions estimated with Linear Probability Model (Columns 1-3) and Tobit (4-7) using survey weights and following the specification in Equation 2. Coefficients are reported with standard errors reported below in parentheses. *, **, *** signify statistically significant difference at the 10%, 5%, and 1% level, respectively. 47 Table A3: Differences in individual wages by duration of stay and migrant origin Wage Index (1) (2) (3) (4) Duration of Stay (Years) 0.0342 0.0639 0.0244 -0.0394 (0.0843) (0.0815) (0.0783) (0.0857) Rural-urban migrant (1=yes) -0.580*** -0.110* -0.107* -0.102 (0.0650) (0.0647) (0.0642) (0.0665) Urban-urban migrant (1=yes) -0.0375 0.0453 0.0464 0.0454 (0.0883) (0.0883) (0.0880) (0.0861) R-U Migrant X Duration of Stay (Years) 0.0194 -0.0599 -0.0207 0.0483 (0.0851) (0.0823) (0.0791) (0.0862) U-U Migrant X Duration of Stay (Years) -0.00964 -0.0631 -0.0222 0.0416 (0.0858) (0.0829) (0.0799) (0.0875) Big city (1=yes) 0.189*** 0.206*** 0.193*** 0.117*** (0.0555) (0.0534) (0.0524) (0.0358) Sex (male=1) - 0.546*** 0.544*** 0.543*** (0.0443) (0.0440) (0.0444) Age - 0.0685*** 0.0684*** 0.0673*** (0.0125) (0.0124) (0.0125) Age^2 - -0.000701*** -0.000705*** -0.000684*** (0.000177) (0.000174) (0.000174) Some primary - -0.110 -0.126* -0.0730 (0.0752) (0.0736) (0.0769) Primary or any secondary - 0.0452 0.0169 0.0706 (0.0798) (0.0800) (0.0825) Secondary completed - 0.339*** 0.304*** 0.336*** (0.0908) (0.0904) (0.0967) Any post-secondary - 0.980*** 0.914*** 0.952*** (0.0914) (0.0823) (0.0826) Manufacturing (1 = yes) - - 0.329** 0.176 (0.128) (0.118) Service (1 = yes) - - 0.262** 0.116 (0.118) (0.106) Constant 1.061*** -1.084*** -1.302*** -1.133*** (0.0288) (0.209) (0.226) (0.230) Country Fixed Effects Yes Yes Yes Yes District Fixed Effects No No No Yes R2 0.0790 0.190 0.191 0.254 N 26761 26524 26501 26501 Notes: Country coverage: Ethiopia, Tanzania, Uganda. Only wage-earning population is included. Individual wages are indices, whereby the value of each observation is normalized by its respective country average to make them comparable across countries. When multiplied by 100, coefficients reported in columns 1-4 can be interpreted as the percent increase/decrease compared to the country average. Coefficients are reported with standard errors reported below in parentheses. Estimated by OLS controlling for survey design using survey weights following the specification in Equation 2. *, **, *** signify statistically significant difference at the 10%, 5%, and 1% level, respectively. 48 Table A4: Differences in household income and consumption per adult equivalent by migrant duration of stay and origin. Income per adult equivalent Consumption per adult equivalent (1) (2) (3) (4) (5) (6) (7) (8) Duration of Stay (Years) 0.0757 0.0727 0.0194 -0.0148 0.0767 0.0542 0.0442 0.00871 (0.0689) (0.0637) (0.0613) (0.0639) (0.0627) (0.0535) (0.0548) (0.0430) Rural-urban migrant (1=yes) -0.256 -0.219 -0.455 -0.561** 0.0244 0.0312 -0.0196 -0.0463 (0.281) (0.274) (0.290) (0.265) (0.0899) (0.0898) (0.103) (0.116) Urban-urban migrant (1=yes) 0.197 -0.0434 0.0883 0.0207 0.388 0.260 0.240 0.177 (0.328) (0.343) (0.359) (0.332) (0.243) (0.217) (0.247) (0.241) R-U Migrant X Duration of Stay (Yrs) 0.0242 0.0265 0.128 0.170** -0.0786 -0.0519 -0.0140 0.0219 (0.0895) (0.0850) (0.0873) (0.0864) (0.0648) (0.0560) (0.0599) (0.0508) U-U Migrant X Duration of Stay (Yrs) -0.00016 0.0291 0.0211 0.0642 -0.106 -0.0670 -0.0527 -0.0219 (0.106) (0.103) (0.0968) (0.0957) (0.0724) (0.0629) (0.0674) (0.0579) Big city (1=yes) 0.891*** 0.685*** 0.539*** 0.186 0.585*** 0.486*** 0.457*** -0.146 (0.155) (0.150) (0.164) (0.863) (0.0843) (0.0779) (0.0927) (0.228) Some primary - -0.371 -0.268 -0.158 - -0.0925 0.0142 -0.0736 (0.408) (0.351) (0.365) (0.0838) (0.100) (0.120) Primary or any secondary - 0.168 0.138 0.0883 - 0.0463 0.0550 -0.115 (0.403) (0.361) (0.374) (0.0750) (0.0924) (0.117) Secondary completed - 0.410 0.290 0.232 - 0.244*** 0.224** 0.0442 (0.413) (0.363) (0.377) (0.0888) (0.103) (0.128) Any post-secondary - 2.178*** 2.070*** 1.993*** - 1.515*** 1.506*** 1.313*** (0.438) (0.407) (0.403) (0.175) (0.182) (0.184) Manufacturing (1 = yes) - - 0.0161 -0.0853 - - 0.178** 0.0833 (0.263) (0.218) (0.0830) (0.0902) Service (1 = yes) - - 0.259 0.208 - - 0.380*** 0.298*** (0.221) (0.166) (0.0575) (0.0561) HH Size -0.121*** -0.144*** -0.125*** -0.115*** -0.0419*** -0.0601*** -0.0600*** -0.0573*** (0.0221) (0.0222) (0.0230) (0.0237) (0.0138) (0.0132) (0.0159) (0.0179) Dependency Ratio -0.00155* -0.000023 -0.00166* -0.00197** -0.00250*** -0.00149*** -0.0015*** -0.00137*** (0.000931) (0.000915) (0.000850) (0.000777) (0.000325) (0.000278) (0.000346) (0.000326) Constant 3.269*** 3.094*** 2.789*** 2.916*** 1.570*** 1.532*** 1.200*** 1.193*** (0.183) (0.389) (0.390) (0.698) (0.0863) (0.0835) (0.110) (0.163) Country Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes District Fixed Effects No No No Yes No No No Yes R2 0.0998 0.162 0.177 0.272 0.0890 0.174 0.176 0.211 N 4375 4375 3866 3866 4630 4630 4130 4130 Notes: Country coverage: Tanzania, Uganda; Dependent variables normalized by their respective country average for comparability across countries. When multiplied by 100, coefficients reported in columns 1-8 can be interpreted as the percent increase/decrease as compared to the country average. Regression run at individual level for working age population, with all household members assigned the same household income/consumption. Households with a migrant are classified as migrant households of the corresponding migrant type. If migrants are from varying origin (rural/urban) or destination (town/small city vs. big city) within the same household, they are assigned rural. Estimated by OLS controlling for survey design using survey weights following the specification in Equation 2. *, **, *** signify statistically significant difference at the 10%, 5%, and 1% level, respectively. 49 Table A5: Differences in employment and hours worked by migrant origin and city size Employed Hours Worked (1) (2) (3) (4) (5) (6) (7) Duration of Stay (Years) 0.00383** -0.00941*** -0.0100*** -0.128 -0.302** -0.303** -0.324*** (0.00189) (0.00182) (0.00173) (0.138) (0.143) (0.142) (0.123) Rural-urban migrant (1=yes) 0.0478*** 0.147*** 0.154*** 0.591 2.100* 1.741 1.691* (0.0146) (0.0135) (0.0127) (1.125) (1.179) (1.183) (1.005) Urban-urban migrant (1=yes) 0.0496*** 0.0795*** 0.0850*** 3.873*** 2.958*** 2.097** 1.547* (0.0130) (0.0132) (0.0128) (0.894) (0.956) (0.953) (0.808) Rural-urban migrant x big city 0.0344 0.0512** 0.0506** 1.736 3.900** 4.094** 3.322** (0.0220) (0.0205) (0.0203) (1.714) (1.754) (1.758) (1.638) Urban-urban migrant x big city -0.00433 0.0666*** 0.0642*** -6.072*** -3.232* -2.400 -1.564 (0.0236) (0.0215) (0.0214) (1.670) (1.730) (1.702) (1.610) Big city (1=yes) -0.1000*** -0.141*** -0.147*** 9.532*** 7.173*** 5.797*** 7.187*** (0.00865) (0.00744) (0.00742) (0.600) (0.624) (0.644) (0.535) Sex (male=1) - 0.133*** 0.133*** - 3.993*** 4.392*** 4.574*** (0.00666) (0.00626) (0.550) (0.540) (0.458) Age - 0.0666*** 0.0671*** - 0.934*** 0.791*** 0.714*** (0.00153) (0.00144) (0.140) (0.139) (0.112) Age^2 - -0.000771*** -0.000779*** - -0.0120*** -0.0100*** -0.00928*** (0.0000207) (0.0000197) (0.00187) (0.00185) (0.00150) Some primary - 0.0726*** 0.0690*** - -0.304 -0.158 -0.313 (0.0117) (0.0108) (0.920) (0.897) (0.796) Primary or any secondary - 0.109*** 0.110*** - 2.801** 1.820 1.267 (0.0112) (0.0107) (1.170) (1.163) (1.005) Secondary completed - 0.120*** 0.129*** - 7.074*** 5.465*** 4.827*** (0.0134) (0.0131) (1.217) (1.216) (1.117) Any post-secondary - 0.214*** 0.219*** - 5.278*** 3.275*** 3.137*** (0.0106) (0.0102) (0.855) (0.887) (0.806) Manufacturing (1 = yes) - - - - - 7.643*** 6.578*** (0.875) (0.731) Service (1 = yes) - - - - - 9.555*** 8.286*** (0.728) (0.624) Constant 0.637*** -0.724*** -0.752*** 32.66*** 12.54*** 8.018*** 11.30*** (0.00460) (0.0273) (0.0260) (0.320) (2.663) (2.593) (2.174) Country Fixed Effects Yes Yes Yes Yes Yes Yes Yes District Fixed Effects No No Yes No No No Yes R2 0.0129 0.227 0.257 N 91047 89590 89590 45788 44955 44922 44922 Notes: Country coverage: Ethiopia, Tanzania, Uganda; Hours worked refers to the total hours worked in the last week in Ethiopia and Uganda, but only hours worked in wage employment in Tanzania. Hours worked regressions in Columns 4-7 only include the employed population. Regressions control for country fixed effects; errors are corrected for survey design and regressions estimated with Linear Probability Model (Columns 1-3) and Tobit (4-7), controlling for survey design using survey weights following the specification in Equation 3. Coefficients are reported with standard errors reported below in parentheses. *, **, *** signify statistically significant difference at the 10%, 5%, and 1% level, respectively. 50 Table A6: Differences in individual wages by migrant origin and city size Wage Index (1) (2) (3) (4) Duration of Stay (Years) 0.0411*** 0.00446 0.00246 0.00300 (0.00862) (0.00844) (0.00837) (0.00867) Rural-urban migrant (1=yes) -0.416*** -0.0511 -0.0496 -0.0286 (0.0701) (0.0684) (0.0693) (0.0658) Urban-urban migrant (1=yes) 0.0348 0.0836 0.0965 0.0698 (0.0687) (0.0677) (0.0668) (0.0650) Rural-urban migrant x big city -0.333*** -0.182* -0.184* -0.167 (0.113) (0.106) (0.108) (0.110) Urban-urban migrant x big city -0.466*** -0.178 -0.190 -0.104 (0.150) (0.148) (0.150) (0.137) Big city (1=yes) 0.313*** 0.267*** 0.261*** 0.163*** (0.0734) (0.0694) (0.0696) (0.0412) Sex (male=1) - 0.543*** 0.545*** 0.545*** (0.0443) (0.0446) (0.0451) Age - 0.0679*** 0.0665*** 0.0677*** (0.0125) (0.0125) (0.0126) Age^2 - -0.000694*** -0.000684*** -0.000687*** (0.000176) (0.000176) (0.000174) Some primary - -0.104 -0.125 -0.0735 (0.0755) (0.0761) (0.0797) Primary or any secondary - 0.0467 0.00641 0.0591 (0.0803) (0.0831) (0.0859) Secondary completed - 0.336*** 0.287*** 0.317*** (0.0907) (0.0928) (0.0998) Any post-secondary - 0.976*** 0.899*** 0.939*** (0.0920) (0.0841) (0.0850) Manufacturing (1 = yes) - - 0.333*** 0.181 (0.129) (0.114) Service (1 = yes) - - 0.268** 0.121 (0.118) (0.104) Constant 1.002*** -1.099*** -1.290*** -1.151** (0.0341) (0.208) (0.230) (0.236) Country Fixed Effects Yes Yes Yes Yes District Fixed Effects No No No Yes R2 0.0814 0.190 0.189 0.253 N 26761 26524 26457 26457 Notes: Country coverage: Ethiopia, Tanzania, Uganda. Only wage-earning population is included. Individual wages are indices, whereby the value of each observation is normalized by its respective country average to make them comparable across countries. When multiplied by 100, coefficients reported in columns 1-4 can be interpreted as the percent increase/decrease compared to the country average. Coefficients are reported with standard errors reported below in parentheses. Estimated by OLS controlling for survey design using survey weights following the specification in Equation 3. *, **, *** signify statistically significant difference at the 10%, 5%, and 1% level, respectively. 51 Table A7: Differences in household income and consumption per adult equivalent by migrant origin and city size Income per adult equivalent Consumption per adult equivalent (1) (2) (3) (4) (5) (6) (7) (8) Duration of Stay (Years) 0.0849** 0.0941** 0.0946** 0.0990*** -0.00100 0.00425 0.0221 0.0144 (0.0400) (0.0397) (0.0409) (0.0355) (0.0162) (0.0148) (0.0176) (0.0171) Rural-urban migrant (1=yes) 0.122 0.0390 0.0798 0.0484 0.0518 0.0245 0.0281 -0.0390 (0.290) (0.287) (0.301) (0.277) (0.0913) (0.0862) (0.0934) (0.0932) Urban-urban migrant (1=yes) 0.336 0.169 -0.0696 -0.0230 0.335* 0.242 0.239 0.139 (0.327) (0.333) (0.366) (0.339) (0.197) (0.178) (0.214) (0.217) Rural-urban migrant x big city -0.925*** -0.701** -0.860** -0.967*** -0.151 -0.0393 -0.0965 0.112 (0.358) (0.349) (0.381) (0.363) (0.195) (0.187) (0.219) (0.259) Urban-urban migrant x big city -0.624 -0.559 -0.277 -0.486 -0.286 -0.214 -0.340 -0.129 (0.449) (0.444) (0.468) (0.441) (0.226) (0.215) (0.246) (0.256) Big city (1=yes) 1.261*** 0.984*** 0.824*** 0.858 0.663*** 0.525*** 0.520*** -0.222 (0.195) (0.186) (0.202) (0.874) (0.120) (0.111) (0.127) (0.292) HH Size -0.119*** -0.142*** -0.121*** -0.114*** -0.0423*** -0.0607*** -0.0598*** -0.0574*** (0.0219) (0.0222) (0.0228) (0.0232) (0.0136) (0.0130) (0.0155) (0.0175) Dependency Ratio -0.00145 0.0000294 -0.00152* -0.00182** -0.00253*** -0.00151*** -0.00154*** -0.00136*** (0.000928) (0.000913) (0.000851) (0.000781) (0.000338) (0.000284) (0.000344) (0.000330) Some primary - -0.351 -0.260 -0.154 - -0.0854 0.0344 -0.0644 (0.412) (0.366) (0.371) (0.0847) (0.102) (0.120) Primary or any secondary - 0.157 0.112 0.0674 - 0.0500 0.0673 -0.106 (0.407) (0.375) (0.379) (0.0758) (0.0923) (0.115) Secondary completed - 0.395 0.257 0.210 - 0.249*** 0.237** 0.0543 (0.417) (0.377) (0.382) (0.0890) (0.103) (0.125) Any post-secondary - 2.150*** 2.035*** 1.955*** - 1.523*** 1.517*** 1.323*** (0.441) (0.417) (0.407) (0.178) (0.186) (0.185) Manufacturing (1 = yes) - - -0.0205 -0.0903 - - 0.173** 0.0876 (0.265) (0.219) (0.0828) (0.0902) Service (1 = yes) - - 0.232 0.203 - - 0.381*** 0.296*** (0.221) (0.167) (0.0579) (0.0561) Constant 3.205*** 3.051*** 2.753*** 2.919*** 1.563*** 1.526*** 1.176*** 1.184*** (0.184) (0.394) (0.404) (0.699) (0.0843) (0.0839) (0.113) (0.163) Country Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes District Fixed Effects No No No Yes No No No Yes R2 0.0998 0.162 0.177 0.272 0.0890 0.174 0.176 0.211 N 4375 4375 3866 3866 4630 4630 4130 4130 Notes: Country coverage: Tanzania, Uganda; Dependent variables normalized by their respective country average for comparability across countries. When multiplied by 100, coefficients reported in columns 1-8 can be interpreted as the percent increase/decrease as compared to the country average. Regression run at individual level for working age population, with all household members assigned the same household income/consumption. Households with a migrant are classified as migrant households of the corresponding migrant type. If migrants are from varying origin (rural/urban) or destination (town/small city vs. big city) within the same household, they are assigned rural. Estimated by OLS controlling for survey design using survey weights following the specification in Equation 1. *, **, *** signify statistically significant difference at the 10%, 5%, and 1% level, respectively.