In the absence of well-functioning formal markets, community networks have risen to prominence in many developing countries (Chuang & Schechter, 2014). Membership of community-based networks is exclusive and is usually based on ethnicity or kinship. Interpersonal ties are deep-rooted, often spanning multiple generations. As such, these networks are capable of solving commitment and information problems (Munshi, 2011). These benefits are particularly rewarding for migrants, who are initially constrained from economic activity in their destination economies. Employers cannot easily observe migrants’ ability, and banks cannot recognise their creditworthiness. However, by leveraging their ties with existing members of the network in the destination, through he process of referral, migrants can find employment in their destination labour market. Moreover, the community network can finance entrepreneurs as they know the ability of the entrepreneur and, through social sanctions, can punish defaults.
There is a wide consensus in the literature that benefits accrue to migrants through these two channels: ‘networks primarily solve labour and credit market imperfections at the destination’ (Munshi, 2016, p. 642). A strong theoretical literature supports these two veritable advantages of community networks on migrant outcomes. Montgomery’s (1991) seminal contribution exposits a theoretical framework modelling the interaction between networks and the labour market. It affirms that those who are unemployed and who have high-ability working friends are more likely to find jobs. Since Montgomery’s model, the role of networks has been of perennial interest to labour theorists, who have relaxed his strict homophily assumption, predicting that networks solve the information symmetries plaguing migrants (Galenianos, 2013). Moreover, this theoretical literature has extended the outcome variable from the probability of obtaining a job to a prediction of higher wages for networked individuals (Mortensen & Vishwanath, 1994). In addition, Rudner (1994) details that first-generation entrepreneurs could only effectively compete with the Nakarattar caste in India if financed by community-based networks. This served as an impetus for Munshi’s (2011) theoretical modelling of the important role played by community networks in supporting entrepreneurship.
Overall, the theoretical consensus, excellently reviewed by Jackson (2011) and Beaman (2016), conflate networks’ support for labour and credit markets with migrant outcomes. Further credence is provided by anecdotal evidence from developing countries. For instance, Banerjee and Duflo (2007) identify the presence of community networks that provide between 25% and 90% of total access to loans across thirteen developing countries. However, much research in India has found strong community-based clustering in certain industries, such as dock-workers (Cholia, 1941), railway workshops (Burnett-Hurst, 1925) and electric suppliers (Chandavarkar, 1994), thereby validating the view that networks serve as a pool from which to employ. Patel (1963) finds the most compelling evidence among mill workers in Mumbai: of 500 workers surveyed, 81% reported that members of their caste worked in the textile industry, while 66% said that these ties resulted in their employment.
Despite the logical theoretical predictions and compelling anecdotal evidence of how networks impact labour and credit markets, which in turn affects migrants, there is a lack of credible empirical evidence identifying the causal effects of networks on migrant outcomes. This dearth in the literature is the corollary of econometric issues: ‘it turns out to be extremely difficult to identify network effects in practice’ (Munshi, 2011, p. 1225).
The empirical network literature is plagued by unwieldy regressions that suffer from three sources of endogeneity: (1) omitted variables; (2) measurement error and (3) simultaneity (Boucher & Fortin, 2015). As a result, few research papers uphold the rigorous econometric standards expected. To elucidate the parameter of interest and highlight the articles that provide the most credible evidence of network effects, it is first necessary to lay out the general linear econometric model:
wherein 𝑦 is the outcome for migrant 𝑖 in community 𝑐 at time 𝑡, such as migrant wage or entrepreneur’s profit. C is the size of the community network, and A is a migrant’s ability. The error term is an exogenous labour demand shock. The estimated 𝜓 could be biased, as mentioned earlier, in the presence of endogeneity. All three forms of endogeneity are likely inimical to the successful identification of network effects. Problems with information and networks effects are rife owing to the complexity of data collection of entire network structures: ‘the econometrician has only partial knowledge of the network’ (Boucher & Fortin, 2015, p. 3). Therefore, even the most perspicacious data collector would have measurement errors and incomplete information. Additionally, network models suffer from a special type of simultaneity known as the reflection problem. Manski (1993) details how the simultaneous determination of 𝑦 and 𝐶 render standard econometric approaches inconsistent. The next two sections summarise the current literature frontier that seeks to identify the true network effects for migrant workers’ and migrant entrepreneurs’ outcomes, respectively.
The empirical literature finds that community networks play a significant role in increasing migrants’ probability of obtaining jobs and securing higher wages. Although there is consensus on the existence of benefits for networked migrants, the literature is discordant about how to identify the network effects. The most naïve studies simply use controls to identify effects. For instance, Giles, Park and Cai (2006) find that involuntarily unemployed urban Chinese who belong to a network are more likely to be reemployed when controlling socio-economic covariates such as age, demographic characteristics and education. However, under such an approach, the orthogonality condition will likely not hold, 𝐸(𝐶, 𝜀) ≠ 0, since shocks to labour demand, 𝜀, will also affect the size of the network, 𝐶. Furthermore, regardless of the number of controls, unobservable factors such as migrants’ self-determination will, through the omitted variable endogeneity issue, bias the parameter, 𝜓. Similar studies using controls to identify network effects suffer from the same econometric issues and thus provide non-credible evidence on migrant outcomes.
Baeman and Magruder (2012) employ a more sophisticated approach involving a job referral field experiment in Kolkata, India. This ingenious innovation is an exception to the egregious attempts at sophistication in approaches seen in the extant literature. Varying the incentives for individuals to refer high-ability workers, they find that given the right incentives, networks provide effective screening for firms as high-ability workers are referred. This paper provides clear evidence of the existence of a mechanism by which networks can effect outcomes in the labour market, namely, through an information channel. Despite the neat innovation, the paper fails to convince the reader that, in reality, networks harness this information and improve the outcomes of migrant workers: there are issues regarding the external validity of the laboratory experiment to reality.
The best evidence for surmounting the identification problem is seen in papers using an instrumental variables approach. These papers are rare in the burgeoning literature but provide the clearest causal estimates of network effects. There are two clear stand out papers in the literature: Munshi’s (2003) empirical analysis of Mexican migrants moving to the United States and Luke’s and Munshi’s (2006) empirical exploration of within country migration to urban areas in Kenya. Both papers conclude that networks have a significant role in improving migrants’ income and employment levels.
This review analyses Munshi’s (2003) paper as it is the quintessential application of this methodology for it achieves success without lengthily falsification tests and heuristic arguments, owing to its clearer instrument identification. Its originality stems from its use of rainfall shocks in Mexican origin communities to instrument for the size of the Mexican community network in the United States. The identification problem section of this review explained why estimates could be spurious if network size and outcomes are both determined by a missing variable. It is plausible that some demand shock in the United States would indeed jointly determine both 𝑦 and 𝐶 in equation 1. Therefore, the use of an instrument is justified in this context. For this approach to be employed the instrument denoted 𝑍 must satisfy two conditions: instrumental relevance, 𝐸(𝐶, 𝑍) ≠ 0, and the exclusion restriction, 𝐸(𝑍, 𝜀) = 0.
The paper reports the reduced form results showing the significant correlation between rainfall shocks in Mexico and size of the community network. Moreover, it is reasonable to assume that rainfall shocks far from the U.S. border have no effect on labour demand shocks in the U.S. The paper uses fixed effects to control for selection into the network and restrictions are placed on the data in lieu with theory to control for individual experience effects. The results show that a random increase in size of the U.S. Mexican community increases the probability that any networked individual will attain a non-agricultural job. This is evidence not only of increased employment but of increased wages as non-agricultural jobs pay substantially more.
Similar to the previous section there is evidence that community networks play a significant role in expanding entrepreneurs access to credit. Again, earlier studies fail to overcome the identification challenge. The use of controls to identify the networks effects produces ponderous and biased regressions. For instance, Falchamps (1999) investigates the extent to which community networks support access to trade and bank credit in Nairobi and Harare. His analysis confirms that deeper network ties are associated with greater access to both forms of credit. This is a confusing result as the network variable in his regressions is measured by the number of social visits with a supplier. One would not necessarily expect therefore this coefficient to be positive and significant when regressed upon access to bank credit. The significance of this parameter suggests that an omitted relevant variable is biasing the results. This is likely the case as Falchamps (1999) does not have a variable measuring the entrepreneurs’ abilities. Therefore, the omission of entrepreneurial ability could result in a spurious network effect.
Later studies, have surmounted this particular issue of unobserved heterogeneity by using panel data fixed effects. Fisman’s (2003) article is arguably the most successful attempt in the literature employing this methodology. Using data from World Bank surveys of firms in African economies he shows that European entrepreneurs have greater access to trade credit than African entrepreneurs owing to strong European network effects. However, this article is problematic for it uses a much weaker definition of networks than the community-based networks described in this literature review. Fisman’s (2003) networks are not based on ethnicity or kinship of some form instead refer to business networks. Evidently, this confounds the identification of community network effects. It is likely that his empirical observation is the result not of network effects but because of differences in business norms. Moreover, such a specification will not capture any time-varying unobserved factors which could potentially bias the parameter of interest.
Falchamps (1999) and Fisman (2003), despite their lack of econometric robustness, confirm that networks do have an affect on credit markets. Although, this literature review is concerned with the effect that networks have on migrants, and conceivably there are differences in network support for migrants to those living at the origin. Woodruff and Zenteno (2007) use an instrumental variables approach to assess the causality of community network support on migrant entrepreneurs. Using the Mexican National Survey of Microenterprises it is found that migrant entrepreneurs’ capital stock is positively associated with strong migrant networks from their origin. The results are amplified when historical rail lines instrument for current migration rates from entrepreneurs’ place of origin. This is based theoretically on the argument that those historically those living closer to rail lines were more likely to migrate and hence today have stronger migrant networks relating back to their origin communities. The validity of this result is somewhat challenging owing to the fact it requires persistent effects of historical migration and there is little evidence on mechanisms supporting such long-run effects. Nevertheless, this paper leads the empirical literature on networks and migrant entrepreneurs’ outcomes finding higher profits of migrant entrepreneurs with strong community networks.
Overall, it is found that the consensus amongst theoretical predictions and anecdotal evidence of community networks on migrant outcomes extends to the empirical literature. This review has emphasised the struggle to identify network effects econometrically. Although, it should give readers much optimism for the future as robust empirical approaches to identifying network causality have been developed in the past decade. Network economics is becoming an increasingly important subfield and there is a nascent literature forming around networks and the process of development. Therefore, the coming decade of research in this area will likely consolidate empirical identification of network effects and begin to answer questions on how these networks affect the aggregate growth models.
Banerjee, A., & Duflo, E. (2007). The economic lives of the poor. Journal of Economic Perspectives, 21, 141–168.
Beaman, L. (2016). Social Networks and the Labour Market. In Y. Bramoullé, A. Galeotti, & B. Rogers, The Oxford Handbook of the Economics of Networks (pp. 649-674). New York: Oxford University Press.
Beaman, L., & Magruder, J. (2012). Who gets the job referral? Evidence from a social networks experiment. American Economic Review, 102, 3574–3593.
Boucher, V., & Fortin, B. (2015). Some Challenges in the Empirics of the Effects of Networks. Bonn: Institute for the Study of Labor.
Burnett-Hurst, A. (1925). Labour and Housing in Bombay: A Study in the Economic Conditions of the Wage-Earning Classes of Bombay. P.S. King and Son.
Chandavarkar, R. (1994). The Origins of Industrial Capitalism in India: Business Strategies and the Working Classes in Bombay, 1900–1940. Cambridge: Cambridge University Press.
Cholia, R. (1941). Dock Labourers in Bombay.
Chuang, Y., & Schechter, L. (2014). Social Networks In Developing Countries. Annual Review of Resource Economics, 7(1), 1-23.
Galenianos, M. (2013). Learning about match quality and the use of referrals. Review of Economic Dynamics, 16, 668-690.
Giles, J., Park, A., & Cai, F. (2006). Reemployment of Dislocated Workers in Urban China: The Roles of Information and Incentives. . Michigan State University, Department of Economics.
Jackson, M. (2011). An Overview of Social Networks and Economic Applications. In J. Benhabib, A. Bisin, & M. Jackson, Handbook of Social Economics (Vol. 1, pp. 511-585). Elsevier.
Luke, N., & Munshi, K. (2006). New Roles for Marriage in Urban Africa: Kinship Networks and the Labor Market in Kenya. The Review of Economics and Statistics, 88(2), 264–282.
Manski, C. (1993). Identification of endogenous social effects: The reflection problem. The Review of Economic Studies , 60(3), 531–542.
Montgomery, J. (1991). Social networks and labor market outcomes: Toward an economic analysis. American Economic Review, 81, 1408-1418.
Mortensen, D., & Vishwanath, T. (1994). Personal contacts and earnings: It is who you know! Labour Economics, 1, 187-201.
Munshi, K. (2003). Networks in the Modern Economy: Mexican Migrants in the U.S. Labor Market. Quarterly Journal of Economics, 118(2), 549–597.
Munshi, K. (2011). Labor and Credit Networks in Developing Economies. In J. Benhabib, & A. J. Bisin, Handbook of Social Economics (Vol. 1, pp. 1123-1254). Elsevier.
Munshi, K. (2011). Strength in Numbers: Networks as a solution to Occupational Traps. Review of Economic Studies, 78, 1069-1101.
Munshi, K. (2016). Community Networks and Migration. In Y. Bramoullé, A. Galeotti, & B. Rogers,
The Oxford Handbook of the Economics of Networks (pp. 630-649). New York: Oxford University Press.
Patel, K. (1963). Rural Labor in Industrial Bombay. Popular Prakashan.
Rudner, D. (1994). Caste and Capitalism in Colonial India: The Nattukottai Chettiars. Berkeley: University of Calfornia Press.