Show Summary Details

Page of

PRINTED FROM OXFORD HANDBOOKS ONLINE ( © Oxford University Press, 2018. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

date: 21 October 2019

Community Networks and Migration

Abstract and Keywords

This chapter describes the role that community networks have played, and continue to play, in facilitating migration. Historical case studies are provided from the American South and colonial India. Contemporary case studies describe Mexican migration to the United States and occupational mobility in the recently restructured Indian economy. The statistical problem of identifying network effects is discussed. Potential solutions include instrumental variables and the use of restrictions from the theory to indirectly identify network effects. Migration by select individuals when community networks are active, with its implications for inequality across and within communities, is also discussed. The chapter concludes by suggesting potential areas for future research.

Keywords: community networks, migration, identification of network effects, networks and economic efficiency, inequality, mobility

23.1 Introduction

The process of migration, within and between countries, has historically been characterized and continues to be characterized by the movement of groups of individuals. Depending on the context, these groups are drawn from the same caste, clan, parish, or village in the origin. Preexisting social connections within these groups help migrants in different ways when they move. These connections provide psychological and social support. More importantly, from an economist’s perspective, they provide various forms of economic support at the destination.

Migrants, being newcomers to the destination economy, are especially vulnerable to information and commitment problems that prevent them from participating fully in the market. For example, prospective employers will not hire migrants because they do not know their worth. Banks will not lend to migrants because they are unable to provide security and have yet to establish a personal reputation. Under these conditions, communities can harness preexisting social ties to form migrant networks that overcome these constraints to economic activity. Migrants who have been in the destination labor market for a while and have established a reputation within their firms can provide referrals to capable new arrivals who belong to their network. The community network can also pool its wealth and provide credit to members with promising investment opportunities, with the knowledge that these individuals will repay their loans in the future (to avoid the social sanctions they would face if they reneged on their obligations).

As described in Section 23.2 of this chapter, there is a voluminous literature in social history, sociology, and economics that describes the role played by migrant community networks in supporting their members, both historically and in the contemporary economy. While a rapidly emerging theoretical literature in economics (p. 631) examines how network structure determines the extent to which information and commitment problems can be resolved, in this chapter I will restrict attention to simpler network properties such as size and connectedness. Nevertheless, we will see that it is difficult to provide credible statistical evidence that networks support migration and improve the outcomes of their members at the destination. This is because the size and composition of the migrant network will respond to (unobserved) changes in the destination economy that directly determine the outcomes of its members. Any observed correlation between individual outcomes and network characteristics could in that case be entirely spurious.

Much progress has been made over the past decade in tackling this problem and I will present two approaches in Section 23.3 that have been used in previous research. The first approach exploits exogenous variation in economic conditions at the origin to construct a statistical instrument for network characteristics at the destination. This approach is only valid under very special conditions. The second approach exploits exogenous variation in the population characteristics of the community, together with restrictions from the theory, to identify a role for community networks. This approach can be implemented more generally, and is feasible even when the migrant community network is not observed directly.

While economists traditionally modeled migration as a choice based on differential wages at the origin and the destination, credible statistical evidence that networks matter for migration decisions and outcomes using the methods described above has transformed the literature. It is now common practice for researchers to add a “networks” variable to reconcile the pattern of migration across destinations. The networks variable is typically measured by the stock of migrants from the individual’s origin community at each destination. The pitfalls of such an approach have been noted above, and while networks may well play a crucial role in migration, it is important to incorporate networks appropriately in models of migration and to adequately address the identification problems that are associated with their empirical analysis.

Migrant networks will increase the income of those individuals who move by providing them with credit for business investment and with jobs. This will naturally generate inequality within the population from which the network is drawn as well as across communities. Destination networks will form and grow over time in certain communities, favoring their members, while equally capable migrants in other communities will remain at a persistent disadvantage. Within communities, inequality will increase or decrease depending on who benefits from the migrant network. Because these networks grow over time, inequality within the community will also change over time, not necessarily monotonically, as described in Section 23.4.

The discussion up to this point has focused on migrant networks at the destination. If the community can support a network far away, we would imagine that it could support networks serving different purposes at the origin as well. Once we add origin networks to the mix, the analysis of migration becomes more complex. As described in Section 23.5, while the community may support the movement of groups of its members to a new location, it will discourage the movement of individuals, either because they (p. 632) lose the services such as insurance that are provided by the origin network when they move independently, or because there are explicit restrictions on mobility. This tension could persist even at the destination, once networks span multiple generations, with established members discouraging the next generation from pursuing opportunities elsewhere. A complete characterization of the inter-generational evolution of migrants community networks, across space and occupations, is noticeably lacking in the economics literature. The concluding section will briefly discussion possible directions that this research could take.

23.2 Migrant Networks in Historical Perspective

The development of the United States is associated with the first large-scale movement of workers across national boundaries. During the Age of Mass Migration (1850–1913), the United States received 30 million European immigrants. Abramitzky, Boustan, and Eriksson (2014) calculate that this resulted in 38% of workers in northern cities being foreign-born in 1910. Labor markets in the nineteenth century could be divided into three segments: a stable segment with permanent employment, an unstable segment with periodic short-term unemployment, and a marginal but highly flexible segment characterized by spells of long-term and short-term unemployment (Gordon, Edwards, and Reich 1982). Migrants being newcomers to the U.S. market typically ended up in the unstable and marginal segments, where the uncertain labor demand and the lack of information about their ability and diligence naturally provided an impetus for the formation of ethnic job networks (Conzen 1976; Hoerder 1991).

Accounts by contemporary observers and an extensive social history literature indicate that friends and kin from the origin community in Europe played an important role in securing jobs for migrants in the U.S. labor market in the nineteenth century and the first quarter of the twentieth century. As an immigrant put it, “The only way you got a job [was] through somebody at work who got you in” (Bodnar, Simon, and Weber 1982: 56). Early historical studies used census data, which provide occupations and country of birth, to identify ethnic clusters in particular locations and occupations (Hutchinson 1956; Gordon, Edwards, and Reich 1982). More recently, social historians have linked parish registers and county data in specific European sending communities to census and church records in the United States to construct the entire chain of migration from those communities as it unfolded over time. This research has documented the formation of new settlements by pioneering migrants, the subsequent channeling of migrants from the origin community in Europe to these settlements, as well as the movement of groups from the original settlement to new satellite colonies elsewhere in the United States (Gjerde 1985; Kamphoefner 1987; Bodnar 1985).

Migration from Europe ceased in 1913, but it was soon replaced by the movement of African-Americans from the rural South to northern cities. The first major movement (p. 633) of blacks out of the South commenced in 1916. Over 400,000 blacks moved to the North between 1916 and 1918, exceeding the total number who moved in the preceding 40 years. During the first phase of the Great Migration, running from 1916 to 1930, over one million blacks (one-tenth the black population of the United States) moved to northern cities (Marks 1983). This movement was driven by both pull and push factors. The increased demand for labor in the wartime economy coupled with the closing of European immigration gave blacks new labor market opportunities (Mandle 1978; Gottlieb 1987). Around the same time, the boll weevil beetle infestation reduced the demand for labor in southern cotton-growing counties (Marks 1989). Adverse economic conditions in the South, together with segregation and racial violence, encouraged many blacks to leave (Tolnay and Beck 1990). Their movement was facilitated by the penetration of the railroad into the deep South (Wright 1986). A confluence of favorable and unfavorable circumstances thus set the stage for one of the largest internal migrations in history.

Although external sources of information such as newspapers and recruiting agents played an important role in jump-starting the migration process, and agencies such as the Urban League provided migrants with housing and job assistance at the destination, networks linking southern communities to specific northern cities, and to neighborhoods within those cities, soon emerged (Gottlieb 1987; Marks 1991; Carrington, Detragiache, and Vishwanath 1996). “[These] networks stimulated, facilitated, and helped shape the migration process at all stages from the dissemination of information through the black South to the settlement of black southerners in northern cities” (Grossman 1991: 67).

The large-scale movement of labor in the United States, supported by migrant networks, was being replicated in other parts of the world as economies industrialized and cities grew in the nineteenth century. For example, Mumbai’s industrial economy in the late nineteenth century and through the first half of the twentieth century was characterized by wide fluctuations in the demand for labor (Chandavarkar 1994). Frequent job turnover will naturally give rise to labor market networks, particularly when the quality of a freshly hired worker is difficult to assess and performance-contingent wage contracts cannot be implemented. The presence of such recruitment networks has indeed been documented by numerous historians studying Mumbai’s economy prior to independence in 1947 (Chandavarkar 1994; Morris 1965; Burnett-Hurst 1925). These networks appear to have been organized around the jobber, a foreman who was in charge of a work gang in the mill, factory, dockyard, or construction site, and more importantly also in charge of labor recruitment.

Given the information and enforcement problems that are associated with the recruitment of short-term labor, it is not surprising that the “jobber had to lean on social connections outside his workplace such as his kinship and neighborhood connections” (Chandavarkar 1994: 107). Here the endogamous caste served as a natural social unit from which to recruit labor. The primary marriage rule in Hindu society, which recent genetic evidence indicates is over 2,000 years old (Moorjani et al. 2013), is that individuals must marry within their caste or jati. Muslims follow the same pattern of (p. 634) endogamous marriage within their biradaris, while converts to Christianity continue to marry within their original jatis. Marriage ties within these kinship groups, formed over many generations, strengthen information flows and improve enforcement. The resulting formation of networks drawn from these kinship communities led to a fragmentation of Mumbai’s labor market along social lines.

The presence of caste clusters in the textile mills, for example, has been well documented. Gokhale’s (1957) survey of textile workers in a single mill in the 1950s showed that Maratha and Kunbi (both middle caste) men were evenly distributed throughout the mill, while Harijan (low caste) men were employed for the most part in the spinning section. Consistent with the presence of local (jobber-specific) networks, the caste clusters that were observed in particular mills often differed from the general pattern for the industry as a whole. The same sort of caste-based clustering has been documented among Mumbai’s dock workers (Cholia 1941), construction workers, and in the railway workshops (Burnett-Hurst 1925), the leather and dyeing industries, and in the Bombay Municipal Corporation and the Bombay Electric Supply and Transportation Company (Chandavarkar 1994).

Although most historical accounts of caste-based networking in Indian cities are situated prior to independence in 1947, a few studies conducted over the subsequent decades in India indicate that these patterns persisted over many generations. For example, Patel (1963) surveyed 500 mill workers in Mumbai in 1961–62 and found that 81% had relatives or members of their caste in the textile industry. 50% of the workers got jobs in the mills through the influence of their relatives, and 16% through their friends, many of whom would have belonged to the same caste. Forty years later, Munshi and Rosenzweig (2006) surveyed the parents of schoolchildren residing in the same area of the city. Sixty-eight percent of the fathers employed in working-class occupations reported that they received help from a relative or member of their caste in finding their first job, while 44% of fathers in white-collar occupations reported such help.

Labor market networks continue to be active in cities throughout the world, most often among migrant populations. For example, Rees (1966) reports that informal sources accounted for 80% of all hires in blue-collar occupations and 50% of all hires in white-collar occupations in an early study set in Chicago. We would expect social ties to play an even stronger role for migrants in the United States Indeed, over 70% of the undocumented Mexicans, and a slightly higher proportion of the Central Americans, that Chavez (1992) interviewed in 1986 found work through referrals from friends and relatives. Similar patterns have been found in contemporary studies of Salvadoran immigrants (Menjivar 2000), Guatemalan immigrants (Hagan 1994), and Chinese immigrants (Nee and Nee 1972; Zhou 1992). Individual respondents in the Mexican Migration Project (MMP), discussed in greater detail below, were asked how they obtained employment on their last visit to the United States; relatives (35%) and friends or paisanos from the origin village in Mexico (35%) account for the bulk of job referrals.

(p. 635) 23.3 Identifying Migrant Networks

While direct evidence on the support provided by migrant networks is sometimes available, it will often be the case that a causal link between observed network characteristics and individual outcomes will need to be established. The statistical problem that arises when attempting to make this connection (which is also discussed by Boucher and Fortin in this handbook) is parsimoniously summarized by the following equation,

y ict =β X ct + ω i + ϵ ct ,
(23.1) where yict is an outcome for migrant i, belonging to community c, in period t, such as employment or wages, that is potentially determined by the network and Xct measures the size of the community network in that period. Individuals benefit from larger networks because more referrals from fellow members are available. At the same time, there will be more competition for available referrals in a larger network. If the first effect dominates, then the hypothesis is that β>0 . ωi measures the individual’s ability, which directly determines his outcome in the destination labor market, and ϵct is an exogenous labor demand shock. Both ωi and ϵct are unobserved by the econometrician. Notice that the demand shock has a community, c, subscript. This reflects the idea that individual migrants from a given origin location could be endowed with specific skills that channel them into particular segments of the labor market even when networks are absent.

The estimated β coefficient in equation (23.1) will be biased. This is because the size of the network, Xct, will respond to demand shocks at the destination, ϵct, violating the orthogonality condition. Selective migration will also bias the estimated network effect. If there is positive selection on ability (i.e., higher ability individuals are more likely to migrate), then an increase in Xct implies that the marginal migrant will have lower ability: E(ωi) is decreasing in Xct. If there is negative selection, E(ωi) will be increasing in Xct. Least squares estimation of the preceding equation will thus be associated with both omitted variable and selection bias.

The difficulty in interpreting the estimated β coefficient as a network effect, discussed above, plagues much of the recent cross-country literature on migration; for example, Beine et al. (2011), Bertoli and Fernandez-Huertas Moraga (2012), Docquier et al. (2014). This literature includes the stock of migrants from the origin country in each destination country to measure the location-specific network size. The same approach has been used to examine networks within the United States (Patel and Vella 2013). The limitation of this strategy is that the stock could instead reflect the unobserved match between skills acquired at the origin and skills needed at the destination. Where this match is better, the flow of migrants over time and, hence, their stock, will be larger. With panel data, it is possible to include fixed effects for each origin-destination pair. However, changes in the stock of migrants at a given destination will still reflect (p. 636) unobserved demand shocks, as discussed above, complicating the network inference problem.

One solution to this problem is to find a statistical instrument that determines Xct but is uncorrelated with ϵct. A major advantage of working with migration data is that the size of the network will be determined by pull factors from the destination as well as push factors from the origin. Origin characteristics or shocks that generate exogenous variation in the size of the migrant network, but are uncorrelated with demand shocks at the destination, will thus be valid instruments. Note, however, that we would still need to include individual fixed effects when estimating equation (23.1) because E(ωi) will respond to changes in Xct, whether or not they are exogenously determined.

Munshi (2003) shows how network effects can be consistently estimated in the context of immigrant Mexican labor in the United States. Migration from Mexico tends to be recurrent, with individuals working in the United States for spells of three to four years and then returning. Panel data from the MMP can be used to study the labor market outcomes in the United States of a sample of individuals drawn from different Mexican origin communities (villages) over multiple migration spells. The idea is to assess whether the same individual does better in spells where he has access to a larger network in the United States. The MMP collected information from a large number of communities (see Massey et al. 1987 for a description of these data). Each community was surveyed once only and retrospective information over many years was collected from approximately 200 individuals. This information included the location of the individual in each year (United States or Mexico) and his labor market outcome (employment, job-type). Munshi measures the size of the community network in the United States in a given year by the fraction of sampled individuals in the community who were located in the United States in that year. To test for network effects, the sample is restricted to person-years in the United States. In the most basic specification, corresponding to equation (23.1), we would regress each individual’s labor market outcome on the contemporaneous size of his community network, including fixed effects in the regression.

Once fixed effects are included, we are effectively assessing the effect of changes in network size on changes in labor market outcomes (i.e., is the individual more likely to be employed [and holding a better job] in years in which his network in the United States is relatively large). However, we know from the discussion above that even if a positive correlation is obtained, this correlation could be entirely spurious if individual labor market outcomes and the size of the community network are jointly determined by (unobserved) economic conditions in the United States. To estimate the causal effect of networks on individual outcomes, we need to find a statistical instrument for network size. A valid instrument in this context will generate changes in network size but will be uncorrelated with direct determinants of individual labor market outcomes in the United States. Munshi’s innovation is to use rainfall in Mexican origin communities, or more correctly rainfall shocks once fixed effects are included, as instruments for network size in the United States.

(p. 637) In practice, network effects will depend on their size and their vintage, since migrants who have been in the United States longer are more established and better positioned to provide referrals. Instead of simply including the size of the network as the key regressor, a more sophisticated specification would thus include the fraction of sampled individuals who recently arrived in the United States and the corresponding fraction for established migrants, separately as regressors.

Table 23.1, Column 1 reports the estimated network effects, based on the instrumental variable regression described above. The number of established migrants in the destination network has a strong effect on the individual’s labor market outcome, which is measured by employment. In contrast, the number of recent migrants does not significantly affect his employment. Column 2 reports the corresponding reduced-form estimates. The established migrants would have moved in response to negative rainfall shocks many years ago and, as expected, we see that distant-past rainfall has a negative and significant effect on employment. In contrast, recent-past rainfall, which determines the number of recent migrants, does not significantly affect employment.

The implicit assumption in the preceding argument is that if there is a drought in the origin village, the demand for labor will decline, with an accompanying increase in migration from Mexico to the United States. Providing support for this assumption, we see in Column 3 that recent-past rainfall has a positive and significant effect on employment in the Mexican origin village, whereas the effect of distant-past rainfall is smaller in magnitude. Local rainfall in Mexican communities far from the border has no impact on the U.S. labor market. However, it has a strong effect on the number of migrants, and these migrants, in turn, improve outcomes for their network members years later when they are established. The estimated network effect is large in magnitude; if the networks were shut down but migration flows remained unchanged, unemployment would increase from 4% to 10%. Complementing this finding, the prevalence of preferred (more remunerative) non-agricultural jobs would decline from 51% to 32%.

One alternative interpretation of the results in Table 23.1 is that they reflect an individual experience effect; the individuals who moved in response to the negative rainfall shock years ago are now doing better themselves. However, when Munshi restricts the sample to individuals who arrived recently in the United States, he finds that the estimated network effects are even larger. This is exactly what the theory would predict, since newcomers to the foreign labor market benefit the most from referrals.

The preceding example provides a general framework for identifying network effects. Panel data (and fixed effects) allow the econometrician to control for selection into the network. Rainfall shocks in the origin location generate exogenous variation in the size and the vintage of the network in the destination labor market. The theory is used to place additional restrictions on the data that rule out the alternative explanation based on an individual experience effect. As predicted, recent arrivals benefit more from the network, while established migrants contribute disproportionately to the network. Munshi’s application is exceptionally well suited to testing for network effects because both panel data and a clean source of variation in network size (by vintage) is available. (p. 638) It is, however, possible to identify network effects even when this is not the case, as long as there is exogenous variation in the population characteristics of the community, by deriving and testing additional predictions from the theory. The example that follows shows how this can be done.

The setting for this example is the American South in the decades after Emancipation in 1865. Chay and Munshi’s (2014) objective is to assess whether and where African Americans were able to overcome centuries of social dislocation and form new networks once they were free. Their analysis starts with the observation that cropping patterns varied substantially across southern counties, during and after slavery. Where labor-intensive crops such as tobacco, cotton, rice, and sugarcane were grown, black slaves would have lived and worked in large plantations. Their numbers on these plantations would have been large enough to support cooperative arrangements even during slavery. In contrast, where crops such as wheat and corn were grown, blacks were dispersed more widely, living and working in small family farms. Restricted social interaction across these farms would have prevented cooperative arrangements from forming. Blacks could interact without restriction after Emancipation, but the strength and frequency of these interactions would have been limited by spatial proximity, which was determined, once again, by cropping patterns. Social connectedness would thus have been greater in southern counties where labor intensive crops were grown, both during and after slavery. Greater connectedness would have supported higher levels of cooperation, resulting in larger networks drawn from the population. These larger networks would, in turn, have allowed blacks to work more effectively as a group to achieve common objectives.

Southern blacks had two significant opportunities to work together in the decades after Emancipation. First, blacks were able to vote and to elect their own leaders during and just after Reconstruction, 1870–1890. Second, blacks were able to leave the South and find jobs in northern cities during the Great Migration, 1916–1930. Based on the theory, more connected populations would have supported the formation of larger networks of black activists during Reconstruction and larger networks of black workers moving together to northern cities during the Great Migration. This, in turn, would have given rise to greater overall political participation and migration.

Population connectedness  network size  political participation ( migration )

While a positive relationship between population connectedness and particular outcomes during Reconstruction and the Great Migration is consistent with the presence of underlying (unobserved) black networks, other explanations are available. For example, racial conflict could have been greater in counties where labor-intensive plantation crops were grown, encouraging individual black voters to turn out during Reconstruction and to move independently to northern cities during the Great Migration. Alternatively, adverse economic conditions in these counties could have encouraged greater migration without requiring a role for black cooperation. Chay and Munshi’s strategy to identify the presence of underlying networks takes advantage of the additional prediction of their theory, which is that networks will only form (p. 639) above a threshold level of population connectedness. There should thus be no association between the outcomes of interest—political participation and migration—and population connectedness up to a threshold and a positive association thereafter.

Figure 23.1 reports the relationship between population connectedness and (separately) black political participation and migration. Population connectedness is measured by the fraction of cultivated land in the county that was allocated to labor-intensive plantation crops in 1890, midway between Reconstruction and the Great Migration, adjusting for differences in labor intensity across those crops. Plantation size in 1860 (before slavery) is smoothly increasing in this measure of population connectedness. That measure is by construction equal to the population density of black farm workers in 1890. It thus represents both the social capital that was carried forward from the period of slavery as well as the strength and frequency of social interactions in the period after slavery. Black political participation is measured by the number of Republican votes in the 1872 presidential election, since blacks would have voted almost exclusively for the Republican Party (the party of the Union) at that time (Morrison 1987). Some whites would also have voted for the Republican Party at this time. This would confound the test of the theory if the number of white votes varied systematically with our measure of black population connectedness. Robustness tests reported below will rule out this potential source of confounding variation. The black migration measure is derived from inter-censal changes in the black population between 1910 and 1930 (recall that the Great Migration commenced in 1916), adjusting for natural changes due to births and deaths. It appears from Figure 23.1 that the specific nonlinearity implied by the theory, characterized by a slope discontinuity at a threshold, is obtained for both political participation and migration.

Chay and Munshi construct a statistical estimator that allows them to formally test whether the data-generating process underlying a particular outcome is consistent with the theory. Based on this test they verify that both relationships reported in Figure 23.1 are consistent with the theory. In addition, they show formally that the specific nonlinearity implied by their theory of network formation is also obtained for the following outcomes: (i) the election of black leaders during Reconstruction, which complements the pattern of voting and which would not be obtained if the results were driven by white Republican votes, (ii) church congregation size in black denominations, which is the most direct available measure of network size, and (iii) the clustering of black migrants in northern destination cities. In contrast, this nonlinearity is not obtained for (i) Republican votes after Reconstruction when blacks were effectively disfranchised, (ii) black migration prior to 1916, (iii) white migration, and (iv) church congregation size in non-black denominations.

No single alternative can explain the specific nonlinear relationship between population connectedness and outcomes associated with underlying networks, obtained for blacks alone at particular points in time. The nonlinear relationship that is obtained for black church congregation size and the clustering of black migrants in northern destinations, in particular, provide direct support for the hypothesis that blacks were able to work together to achieve common objectives in counties where population (p. 640) connectedness exceeded a threshold. If black migration decisions were based on factors that did not include a coordination externality, then the probability of moving to the same destination would not track migration levels so closely. Once again, the (implied) magnitude of the estimated network effect is large; for example, over half of the migrants to the North came from the third of Southern blacks who lived in the most connected counties, while less than 10% came from the third in the least-connected counties.

Empirical analyses of migrant networks have utilized both strategies described above. A series of papers on Mexican migration to the United States use historical migration patterns, which in turn were determined by initial railroad placement, to predict current migration (Woodruff and Zenteno 2007; McKenzie and Rapoport 2007, 2012). Beaman (2012) uses variation in the placement of refugees across locations in the United States to estimate network effects. It is possible that railroads were not placed randomly and that historical migration patterns are associated with unobserved community characteristics that directly determine labor market outcomes in the United States today. Similarly, it is possible that refugees were assigned to cities where they had a comparative advantage, resulting in spatial clustering by national origin and, hence, a potentially spurious network effect. Although these instruments plausibly satisfy the exclusion restriction, additional support is required to credibly identify network effects. Chay and Munshi (2014) face the same problem in their analysis, since pre-determined cropping patterns could be associated with individual or county characteristics that directly determine migration decisions. Their strategy to identify a role for networks is to exploit additional predictions from the theory; specifically, the non linear relationship between black population connectedness and migration when networks are active. In general, what is needed are additional theoretical predictions on variation in the network effect over cohorts of migrants, as in Munshi (2003), or across subpopulations, as in Chay and Munshi (2014). The studies listed above do derive and test such restrictions, providing credible evidence that networks are active in different contexts.

23.4 Migrant Networks and Inequality

The discussion up to this point has focused on the role of migrant networks in improving the economic outcomes of their members. If these networks are effective, they will naturally have consequences for inequality; across communities with networks of varying strength, as well as within communities depending on who migrates.

Migrant selection is the subject of a large and active literature in international migration. The starting point for this literature is Borjas (1987), who extends the Roy (1951) model to derive predictions for selection by education (or ability) that depends on wage inequality in the sending and receiving country. If wages are increasing more (less) steeply in education in the sending country, then there will be negative (positive) selection on education. As summarized in Abramitzky, Boustan, and Eriksson (2012), (p. 641) empirical tests from across the world find mixed support for this prediction. Moreover, the nature of migrant selection from Mexico to the United States, a topic of great policy interest that has received much research attention, remains unresolved (Chiquar and Hanson 2005; Cuecuecha 2005; Orrenius and Zavodny 2005; Mishra 2007; Ibarran and Lubotsky 2007; Fernandez-Huertas Moraga 2011, 2013).

One way to resolve this apparent ambiguity is to allow migration costs to vary with education. For example, Chiquar and Hanson document positive selection on education for migrants from Mexico to the United States despite the fact that wages are increasing more steeply with education in Mexico. Their response is to augment the Borjas model by allowing moving costs to be decreasing in education. This is evidently unsatisfactory; once moving costs are allowed to vary with education, any result can be explained. In addition, the augmented Borjas model cannot explain the dynamics of selection from Mexican origin communities. As documented in McKenzie and Rapoport (2007), while there is positive selection initially in their sample of communities, this is replaced by negative selection over time. As I show below, all the results in the literature can be easily reconciled by adding networks to the Roy model.

For ease of exposition, suppose that there are two education levels: low (L) and high (H). Less educated workers are channeled into low-skill occupations, while more educated workers are channeled into high-skill occupations. Without networks, the wages at the origin (O) and the destination (D) for the two types of workers are denoted by W e O , W e D ,e  {L,H} , respectively. A worker will choose to migrate if W e D c W e O , where the distribution of the moving cost, c, is independent of education and is characterized by the function F(c). Because we have assumed that the distribution of moving costs is the same for both types of workers, there will be positive selection if W H D W H O > W L D W L O ; i.e. if W H D W L D > W H O W L O , and negative selection if the sign is reversed, as in the Borjas model.

A natural way to introduce migrant networks into this simple model is to allow them to increase wages at the destination. As noted earlier, labor market networks tend to be concentrated in blue-collar occupations. This is because educational credentials are a good indicator of competence in white-collar occupations, but not necessarily so in blue-collar occupations. Moreover, production tends to take place in teams in these occupations, making it difficult for the firm to attribute effort or competence to individuals on the job. Networks of socially connected workers can overcome both the information and the enforcement problems that arise with team production. This is incorporated in the theoretical framework by allowing the low-skill wage at the destination, W L D , to be increasing in the size of the migrant network, which, in turn, is increasing over time.

Suppose that W H D W H O > W L D W L O to begin with, before migrant networks have had a chance to form. This implies that there will be positive migrant selection out of all communities. However, if a migrant network does subsequently form in a given community, the right-hand side of the preceding inequality will increase over time, possibly resulting in a switch in the sign of the inequality. As documented by McKenzie and Rapoport, there will be positive migrant selection and an increase in (p. 642) inequality within the community in the early stages, followed by negative selection and a decline in inequality once the destination network exceeds a threshold size. Note, however, that this dynamic pattern will not be obtained in all communities. Recall from Chay and Munshi (2014) that migrant networks will only form when population connectedness exceeds a threshold. Migrant networks will thus increase inequality across communities, and this inequality will increase over time as migrant networks strengthen in some communities but not others. Moreover, we could obtain positive or negative selection on average in a sample of communities at a given point in time, depending on their population characteristics and the stage of development of their migrant networks. Our Roy model with migrant networks is thus easily able to generate the dynamic and the cross-sectional patterns of migrant selection that have been documented in the literature.

McKenzie and Rapoport (2007, 2012) also add networks to the Roy model to generate the observed patterns of selection. However, they choose to embed the network effect in the moving cost, following past work by Carrington, Detragiache, and Vishwanath (1996). One limitation of this approach is that it is at odds with the functions that migrant networks perform. These networks primarily solve labor and credit market imperfections at the destination, once the migrant has arrived. While the community may support the movement of migrants, by providing credit or insurance, these services will be provided by networks situated at the origin. If the empirical analysis exploits exogenous variation in community characteristics, as in Chay and Munshi, this distinction is less relevant; both origin and destination networks will form when population connectedness exceeds a threshold level. However, if the destination network is measured directly, then there will be a disconnect between a theory based on moving costs and the empirical analysis. A second limitation of the moving-cost approach is that its predictions—for example, as in McKenzie and Rapoport (2012)—are sensitive to a number of auxiliary assumptions, such as the distribution of schooling in the population. In contrast, the observed patterns of migrant selection are obtained with minimal and defensible assumptions when the migrant network serves more naturally to improve individual outcomes at the destination.

23.5 The Intergenerational Dynamics of Migration

The discussion up to this point has focused on migrant networks at the destination. However, community networks at the origin will also shape migration. On the one hand, networks at the origin can support migration by financing the move. On the other hand, mutual insurance networks at the origin can restrict the movement of individuals. If an individual moves on his own from the village to the city, he always (p. 643) has an incentive to under-report his income realizations as a way of getting transfers to flow in his direction. Even if this information problem is resolved, he will still have an incentive to renege on his obligation to make transfers to less fortunate community members far away in the village. Munshi and Rosenzweig (2015) show theoretically and empirically that these information and commitment problems can restrict individual migration when formal insurance at the destination is unavailable. These restrictions will be paradoxically amplified when the rural insurance network is well functioning because individual movers have more to lose.

The preceding discussion highlights the important role that origin-based networks can play in supporting or restricting migration. Losing access to the origin network is yet another reason why migrants tend to move in groups, since the presence of a group at the destination will solve the information and enforcement problems discussed above. In some contexts, as with mutual insurance, the origin network will discourage individuals from moving, but may encourage migration by groups of individuals because this diversifies income risk. In other contexts, the origin network will discourage any migration. This will be the case, as in farming communities, when the performance of the origin network depends on the number of members who are based at the origin. When we extend our analysis of destination-based community networks past one generation, an analogous argument applies. A network that has been established at a particular destination will discourage the next generation from moving elsewhere or into new occupations if its performance depends on its size.

Munshi and Rosenzweig (2006) show theoretically that the restrictions on mobility described above can be welfare enhancing when they are established, but can result in a dynamic inefficiency in the long run. To see why this is the case, return to our setup with two types of occupations, skilled and unskilled, and two levels of associated education, high (H) and low (L). Consider a migrant network that has already been established at a particular destination, so moving costs can be ignored. Individual heterogeneity is now in ability, which determines the cost of education. It costs C ¯ e for low-ability individuals to attain high education, whereas the corresponding cost for high-ability individuals is C ¯ e < C ¯ e . We normalize so that the cost of attaining low education is zero.

As above, the unskilled wage, WL, is increasing in the size of the network. We assume that members of the network in a given generation provide referrals to the next generation. Depending on the size of the network in the previous generation, the following conditions are satisfied:

C1.   W H C ¯ e > W L ( 0 ) C2.   W H C ¯ e < W L ( N ).

The first condition says that if a network was not active in the previous generation, then individuals of both types would invest in high education and end up in skilled jobs. The second condition says that if everyone in the community selected into the network in the previous generation (N is the size of the community) then individuals of both types would select low education and end up in the unskilled occupation.

(p. 644) Suppose that the ability distribution is the same across all communities, but communities are exogenously assigned to one equilibrium or the other in the initial period. If conditions C1 and C2 are satisfied, it follows that communities will stay in the initial equilibrium from one generation to the next, with everyone either investing or not investing in education. Now suppose that the high-skill wage starts to increase. When condition C2 just binds, high-ability individuals will be indifferent between investing and not investing in education in those communities where everyone has traditionally selected into the low-skill occupation. If the high-skill wage increases marginally above that level, high-ability individuals will deviate from the traditional equilibrium and invest in education. Overall welfare in the community will decline in this case because the low-ability individuals who remain in the (smaller) network now have a substantially lower wage. This is one reason why blue-collar communities and farming communities, where traditional occupations require a high degree of networking, are often characterized by cultures that discourage occupational and spatial mobility (Elder and Conger 2000; Gans 1962; Kornblum 1974).

While a culture that restricts mobility may have been welfare-enhancing when it was put in place, its persistence can result in a dynamic inefficiency if the skilled wage increases sufficiently. Culture and social norms are very persistent, explaining the common perception that farming and blue-collar communities often stubbornly resist change. This provides a new intergenerational perspective on mobility, since networks in these communities would have historically supported the occupational and spatial mobility of their members. Today, they hold them back. A complete characterization of the relationship between community networks and migration requires attention to networks at the origin and the destination, as well as the dynamic process through which migrant networks form, become established, and then serve as the point of departure for further migration in subsequent generations.

23.6 Conclusion

This chapter describes the role that community networks have played, and continue to play, in facilitating migration. Establishing that these networks improve the outcomes of their members is a challenging statistical problem. Exogenous variation in either network or community characteristics, together with restrictions from the theory, can be used to credibly identify network effects.

Networks are characterized by their size and connectedness (broadly defined) in this chapter. This contrasts with the focus on more detailed measures, often revolving around the centrality of members, in the rapidly growing literature on networks in economics. For example, Nava (2015) surveys an emerging theoretical literature on repeated games and networks that is concerned with the relationship between network structure and equilibrium payoffs in environments where commitment problems are relevant. Similarly, Chaney (2015) surveys the theoretical literature on information (p. 645) diffusion in trade networks in which the structure of these networks determines the extent to which informational frictions can be reduced. One possible direction for future research on migration networks would be to incorporate a more detailed network structure. However, it is important to exercise caution when venturing down this path. There is no powerful empirical fact that lends credence to the idea that these more detailed network structures will add substantially to our understanding of the process of migration. Moreover, credible inference with these detailed network structures is likely to be even more challenging than it is with the simpler structures, based on size and connectedness.

A possibly more fruitful direction for future research would be to more completely characterize the process of network formation and development, as it unfolds over multiple generations within a community, with new networks sometimes breaking away and existing networks dissolving. The theoretical challenges and data requirements for such analyses are substantial. However, in this case, the payoffs are more visible; it is easy to see how this analysis would complement existing models of growth and shed new light on the dynamics of inequality.


Abramitzky, Ran, Leah Platt Boustan, and Katherine Eriksson (2012). “Europe’s tired, poor, huddled masses: Self-selection and economic outcomes in the age of mass migration.” American Economic Review 102(5), 1832–1856.Find this resource:

Abramitzky, Ran, Leah Platt Boustan, and Katherine Eriksson (2014). “A nation of immigrants: Assimilation and economic outcomes in the age of mass migration.” Journal of Political Economy, forthcoming.Find this resource:

Beaman, Lori A. (2008) “Social networks and the dynamics of labor market outcomes: Evidence from refugees resettled in the U.S.” Northwestern University, Department of Economics, typescript.Find this resource:

Beine, Michel, Frederic Docquier, and Caglar Ozden (2011). “Diasporas.” Journal of Development Economics 95, 30–41.Find this resource:

Bertoli, Simone and Jesus Fernandez-Huertas Moraga (2012). “Visa policies, networks and the cliff at the border,” IZA Discussion Papers 7094, Institute for the Study of Labor (IZA).Find this resource:

Bodnar, John (1985). The Transplanted: A History of Immigrants in Urban America. Bloomington: Indiana University Press.Find this resource:

Bodnar, John, Roger Simon, and Michael P. Weber (1982). Lives of Their Own: Blacks, Italians, and Poles in Pittsburgh, 1900–1960. Urbana: University of Illinois Press.Find this resource:

Borjas, George (1987). “Self-selection and the earnings of immigrants.” American Economic Review 77(4), 531–553.Find this resource:

Boucher, Vincent and Bernard Fortin (2015). “Some challenges in the empirics of the effects of networks,” in Jan Bramoulle, Andrea Galleotti, and Brian W. Rogers, eds. Oxford Handbook in the Economics of Networks. Oxford: Oxford University Press.Find this resource:

Burnett-Hurst, A.R. (1925). Labour and Housing in Bombay: A Study in the Economic Conditions of the Wage-Earning Classes of Bombay. London: P.S. King and Son.Find this resource:

Carrington, William J., Enrica Detriagiache, and Tara Vishwanath (1996). “Migration with endogenous moving costs.” American Economic Review 86(4), 909–930.Find this resource:

(p. 646) Chandavarkar, Rajnarayan (1994). The Origins of Industrial Capitalism in India: Business Strategies and the Working Classes in Bombay, 1900–1940. Cambridge: Cambridge University Press.Find this resource:

Chaney, Thomas (2015). “Networks in International Trade.” in Jan Bramoulle, Andrea Galleotti, and Brian W. Rogers, eds. Oxford Handbook in the Economics of Networks. Oxford: Oxford University Press.Find this resource:

Chavez, Leo (1992). Shadowed Lives: Undocumented Immigrants in American Society. Harcourt Brace Jovanovich College Publishers.Find this resource:

Chay, Kenneth and Kaivan Munshi (2014). “Black networks after emancipation: Evidence from Reconstruction and the Great Migration,” University of Cambridge mimeo.Find this resource:

Cholia, R. P. (1941). Dock Labourers in Bombay. Bombay.Find this resource:

Chiquiar, Daniel and Gordon H. Hanson (2005). “International migration, self-selection, and the distribution of wages: Evidence from mexico and the united states.” Journal of Political Economy 113(2), 239–281.Find this resource:

Conzen, Kathleen Neils (1976). Immigrant Milwaukee 1836–1860: Accommodation and Community in a Frontier City. Cambridge, MA: Harvard University Press.Find this resource:

Cuecuecha, Alfredo (2005). “The immigration of educated Mexicans: The role of informal social insurance and migration costs,” ITAM mimeograph.Find this resource:

Docquier, Frederic, Giovanni Peri, and Ilse Ruyssen (2014). “The cross-country determinants of potential and actual migration.” International Migration Review, forthcoming.Find this resource:

Elder Jr., Glen H. and Rand D. Conger (2000). Children of the Land: Adversity and Success in Rural America. Chicago and London: University of Chicago Press.Find this resource:

Fernandez-Huertas Moraga, Jesus (2011). “New evidence on emigrant selection.” Review of Economics and Statistics 93(1), 72–96.Find this resource:

Fernandez-Huertas Moraga, Jesus (2013). “Understanding different migrant selection patterns in rural and urban Mexico.” Journal of Development Economics 103, 182–201.Find this resource:

Gans, Herbert (1962). The Urban Villagers: Group and Class in the Life of Italian-Americans. New York: The Free Press of Glencoe.Find this resource:

Gjerde, Jon (1985). From Peasants to Farmers: The Migration from Balestrand, Norway to the Upper Middle West. Cambridge, England: Cambridge University Press.Find this resource:

Gokhale, R. G. (1957). The Bombay Cotton Mill Worker. Bombay.Find this resource:

Gordon, David M., Richard Edwards, and Michael Reich (1982). Segmented Work, Divided Workers: The Historical Transformation of Labor in the United States. Cambridge England: Cambridge University Press.Find this resource:

Gottlieb, Peter (1991). “Rethinking the Great Migration,” in Joe William Trotter Jr. (Ed.), The Great Migration in Historical Perspective: New Dimensions of Race, Class, and Gender. Bloomington: Indiana University Press.Find this resource:

Grossman, James (1989). Land of Hope: Chicago, Black Southerners, and the Great Migration. Chicago: The University of Chicago Press.Find this resource:

Hagan, Jacqueline Maria. Deciding to be Legal: A Maya Community in Houston. Philadelphia: Temple University Press, 1994.Find this resource:

Hoerder, Dirk (1991). “International labor markets and community building by migrant workers in the Atlantic economies,” in A Century of European Migrants, 1830–1930, eds. Rudolph J. Vecoli and Suzanne M. Sinke, Urbana: University of Illinois Press.Find this resource:

Hutchinson, E. P. (1956). Immigrants and Their Children, 1850–1950. New York: John Wiley and Sons, Inc.Find this resource:

(p. 647) Kamphoefner, Walter D. (1987). The Westfalians: From Germany to Missouri. Princeton, NJ: Princeton University Press.Find this resource:

Ibarran, Pablo and Darren Lubotsky (2007). “Mexican Immigration and self-selection: New evidence from the 2000 Mexican Census.” in George Borjas (Ed.), Mexican Immigration in the United States, Chicago: University of Chicago Press.Find this resource:

Kornblum, William (1974). Blue Collar Community. Chicago and London: University of Chicago Press.Find this resource:

Mandle, Jay R. (1978). The Roots of Black Poverty: The Southern Plantation Economy After the Civil War. Durham, NC: Duke University Press.Find this resource:

Marks, Carole (1983). “Lines of Communication, Recruitment Mechanisms, and the Great Migration of 1916–1918.” Social Problems 31(1), 73–83.Find this resource:

Massy, Douglas, Rafael Alarcon, Jorge Durand, and Humberto Gonzalez (1987). Return to Aztlan: The Social Process of International Migration from Western Mexico. Berkeley: University of California Press.Find this resource:

McKenzie, David and Hillel Rapoport (2007). “Network effects and the dynamics of migration and inequality: Theory and evidence from Mexico.” Journal of Development Economics 84, 1–24.Find this resource:

McKenzie, David and Hillel Rapoport (2012). “Self-Selection patterns in Mexico-U.S. migration: The role of migration networks.” Review of Economics and Statistics 92(4), 811–821.Find this resource:

Menjivar, Cecilia (2000). Fragmented Ties: Salvadoran Immigrant Networks in America. Berkeley: University of California Press.Find this resource:

Mishra, Prachi (2007). “Emigration and wages in source countries: Evidence from Mexico.” Journal of Development Economics 82(1), 180–199.Find this resource:

Morris, Morris David (1965). The Emergence of an Industrial Labor Force in India: A Study of the Bombay Cotton Mills, 1854–1947. Berkeley: University of California Press.Find this resource:

Morrison, Minion K.C. (1987). Black Political Mobilization: Leadership, Power, and Mass Behavior. Albany: State University of New York Press.Find this resource:

Moorjani, Priya, Kumarasamy Thangaraj, Nick Patterson, Mark Lipson, Po-Ru Loh, Periyasamy Govindaraj, Bonnie Berger, David Reich, and Laiji Singh (2013). “Genetic evidence for recent population mixture in India.” American Journal of Human Genetics 93(3), 422–438.Find this resource:

Munshi, Kaivan (2003). “Networks in the modern economy: Mexican migrants in the U.S. labor market.” Quarterly Journal of Economics 118(2), 549–597.Find this resource:

Munshi, Kaivan and Mark Rosenzweig (2006). Traditional institutions meet the modern world: Caste, gender and schooling choice in a globalizing economy.” American Economic Review 96(4), 1225–1252.Find this resource:

Munshi, Kaivan and Mark Rosenzweig (2015). “Networks and misallocation: Insurance, migration, and the rural-urban wage gap.” American Economic Review, forthcoming.Find this resource:

Nava, Francesco (2015). “Repeated games and networks.” In Jan Bramoulle, Andrea Galleotti, and Brian W. Rogers, eds. Oxford Handbook in the Economics of Networks. Oxford: Oxford University Press.Find this resource:

Nee, Victor G. and Brett de Bary Nee (1972). Longtime Californ’: A Documentary Study of an American Chinatown. New York: Pantheon Books.Find this resource:

Orrenius, Pia M. and Madeline Zavodny (2005). “Self-Selection among undocumented immigrants from mexico.” Journal of Development Economics 78(1), 215–240.Find this resource:

Patel, Krishna and Francis Vella (2013). “Immigrant networks and their implications for occupational choice and wages.” Review of Economics and Statistics 95(4), 1249–1277.Find this resource:

(p. 648) Patel, Kunj (1963). Rural Labor in Industrial Bombay. Bombay: Popular Prakashan.Find this resource:

Rees, Albert (1966). “Information networks in labor markets.” American Economic Review (Papers and Proceedings) 56, 559–66.Find this resource:

Roy, A. D. (1951). “Some thoughts on the distribution of earnings.” Oxford Economic Papers 3(2), 135–146.Find this resource:

Tolnay, Stewart E. and E. M. Beck (1990). “Lethal violence and the Great Migration, 1900–1930.” Social Science History 14(3), 347–370.Find this resource:

Woodruff, Christopher and Renee Zenteno (2007). “Migration networks and microenterprises in Mexico.” Journal of Development Economics 82(2), 509–528.Find this resource:

Wright, Gavin (1986). Old South, New South: Revolutions in the Southern Economy Since the Civil War, New York: Basic Books, Inc.Find this resource:

Zhou, Min (1992). Chinatown: The Socioeconomic Potential of an Urban Enclave. Philadelphia: Temple University Press, 1992.Find this resource: