Assessing the role of social networks on migrant labor market outcomes: evidence from a representative immigrant survey

Working Paper Series Assessing the Role of Social Networks on Migrant Labor Market Outcomes: Evidence from a Representative Immigrant Survey __________________________________________ Cátia Batista Nova School of Business and Economics - Universidade Nova de Lisboa CReAM, IZA and NOVAFRICA Ana Isabel Costa Nova School of Business and Economics - Universidade Nova de Lisboa NOVAFRICA ISSN 2183-0843 Working Paper No 1601 April 2016 NOVAFRICA Working Paper Series NOVAFRICA | Nova School of Business and Economics|Faculdade de Economia | Universidade Nova de Lisboa Campus de Campolide | 1099-032 Lisboa – Portugal | T: (+351)213 801 673 | F: (+351) 213 801 105 | www.novafrica.org Any opinions expressed here are those of the author(s) and not those of NOVAFRICA. Research published in this series may include views on policy, but the center itself takes no institutional policy positions. NOVAFRICA is a knowledge center created by the Nova School of Business and Economics of the Nova University of Lisbon. Its mission is to produce distinctive expertise on business and economic development in Africa. A particular focus is on Portuguese-speaking Africa, i.e., Angola, Cape Verde, Guinea-Bissau, Mozambique, and Sao Tome and Principe. The Center aims to produce knowledge and disseminate it through research projects, publications, policy advice, seminars, conferences and other events. NOVAFRICA Working Papers often represent preliminary work and are circulated to encourage discussion. Citation of such a paper should account for its provisional character. A revised version may be available directly from the author. Assessing the Role of Social Networks on Migrant Labor Market Outcomes: Evidence from a Representative Immigrant Survey  Cátia Batista† and Ana Isabel Costa‡ April 2016 Abstract What role do social networks play in determining migrant labor market outcomes? We examine this question using data from a random sample of 1500 immigrants living in Ireland. We propose a theoretical model formally predicting that immigrants with more contacts have additional access to job offers, and are therefore better able to become employed and choose higher paid jobs. Our empirical analysis confirms these findings, while focusing more generally on the relationship between migrants’ social networks and a variety of labor market outcomes (namely wages, employment, occupational choice and job security), contrary to the literature. We find evidence that having one more contact in the network is associated with an increase of 11pp in the probability of being employed and with an increase of about 100 euros in the average salary. However, our data is not suggestive of a network size effect on occupational choice and job security. Our findings are robust to sample selection and other endogeneity concerns. JEL classification: D8, F22, J3 Keywords: Social Networks; International Migration; Wage Determination; Labor Market Integration  The authors are grateful for helpful suggestions from Ana Balcão Reis, Ana Garcia, Bien Balocating, Francesco Cestari, Julia Seither, Matilde Grácio, Pedro Brinca, José Tavares, and Teresa Molina-Millán. This paper makes use of a representative immigrant survey generously funded by the EU NORFACE Programme on International Migration, under the lead of Principal Investigators Catia Batista and Gaia Narciso. † Nova School of Business and Economics, CReAM, IZA and NOVAFRICA. Email: catia.batis[email protected]t ‡ Nova School of Business and Economics, and NOVAFRICA. Email: costa.anai[email protected]m 1 1. Introduction The study of social networks emerged in the economics literature as a way to explain information mismatches and other market frictions. Early empirical studies show that about half of employed individuals rely on family and friends to find jobs or to have access to job information (Gregg and Wadsworth, 1996; Addison and Portugal, 2001). The main theoretical hypothesis linking social networks to labor market outcomes proposes the existence of an informal channel, usually formed by relatives, friends and acquaintances, that provides individuals information not available through formal sources. Such channel mitigates job search frictions in two distinct ways: by making available information to job seekers about employment opportunities, while also providing employers references about the workers. In the recent decades, a growing literature has investigated the role of social networks on migrant outcomes. In one of its most prominent findings, migrants tend to cluster in groups and share information among them (see Edin et al, 2003 and Munshi, 2003), which illustrates the particular importance of social networks for migration. Being newcomers to the labor market, migrants are in need of information about job openings and the characteristics of the labor market, particularly upon arrival to the host country. On the other hand, networks may also be useful from the perspective of the employer who often lacks information about the newly arrived migrants. Despite the expectable importance of social networks on migrant labor market outcomes, there is still a lack of robust empirical evidence on this research question. This paper attempts to fill this gap by using a random sample of 1500 immigrants from 110 different nations residing in Ireland in 2010. We use detailed survey data for migrants and their contacts, which enables us to measure the social network for each migrant, and to find in this way empirical evidence linking networks to the labor market outcomes of immigrants. The core hypothesis of this study is that, immigrants with larger networks are more exposed to information and 2 therefore are more likely to be employed and to hold higher paid jobs. We also analyze other labor market outcomes such as occupational choice and job security (meaning the job stability offered by the type of job contract). Our paper contributes to the previous literature by studying the role of migrant networks at various dimensions of labor market integration. We examine three main questions: (i) is the probability of being employed higher for individuals with more contacts; (ii) what is the causal effect of the network size on subsequent wages: do immigrants with larger networks have higher wages; (iii) is there a selection process with the choice of working among immigrants. Identification of the causal effects of network size on migrant outcomes is complicated by the possibility of several endogeneity problems. First, the suspected causal relationship between the migrant network size and the respective labor market outcome can be simultaneously determined by unobservable characteristics. Second, endogeneity can also occur through a selection process that leads some immigrants to choose to participate in the labor market while others do not – which results in a sample selection problem in that it is not possible to observe the wages of those who do not work. In order to tackle these possible estimation biases, and to ensure the robustness of our results we estimate several distinct models: Linear Probability Model (LPM), Two-Stage Least Squares (2SLS), Heckit model and IV-Heckit model. We find evidence that immigrants with a larger network size are more likely to be employed and earn higher salaries. Moreover, our empirical analysis confirms that there is, in fact, a self-selection process influencing the decision of work. However, our data indicates that network size does not seem to influence occupational choice or the job security of immigrants. Our results are in line with the previous findings in the literature related to wages 3 and employment. The impact of the network size on the occupation and job security are new findings to the literature. The structure of the remainder of the paper is organized as follows. Section 2 reviews the main results found in the literature. Section 3 shows a theoretical framework to model the impact of the network size on the employment and wages of immigrants. Section 4 describes the methodology used and the identification issues. Section 5 presents the data used and the descriptive statistics. Finally, Section 6 includes the empirical results and section 7 concludes. 2. Literature Review This section presents a systematic analysis of the previous findings on the topic and relates the past literature with our empirical strategy. We summarize the impact of social networks both on migration and on labor market outcomes. As initially proposed by Sjaastad (1962) and Harris and Todaro (1970), migration is often an uncertain and risky investment. However, migration networks can lower the costs and uncertainty of follower migrants, as verified by McKenzie and Rapoport (2010). Umblijs (2012) presents theoretical and empirical evidence that if potential migrants have access to a network at the destination, more risk-averse migrants will migrate than if they had no access to these networks 1 . Calvo-Armengol and Zenou (2005) were one of the first authors to put forward a theoretical model of the role of networks on labor market outcomes particularly wages and employment. The model accounts for a job-matching process, in which workers find jobs through their social network, and shows that in the steady-state, labor market outcomes are positively correlated across time and across agents within a network. Nonetheless, it assumes that networks are exogenous, a strong assumption that has not been validated by empirical research (a few examples include Munshi, 2003 and An, 2015). 1 For further evidence of the importance of social networks for migration, see: Beine et al. (2011), Batista and Umblijs (2014), Batista and Umblijs (2016), Batista et. al (2016). 4 Later, Wahba and Zenou (2005) extended the model of Calvo-Armengol and Zenou (2005) by differentiating between lowand high-educated individuals and comparing the efficiency of using networks to find a job with other search methods. The underlying assumption made by the authors is that low-educated individuals only use informal job search methods to find a job, while high-educated individuals use both formal and informal methods. Considering population density as a proxy for the size of the network in Egypt, they provide empirical evidence that conditional on being employed, the probability of finding a job through friends and relatives, compared to other search methods, is higher in denser areas than in less denser areas. Given the importance of networks as an informal institution with consequences in the labor market, one would expect its effect to be stronger among migrants (relative to natives) as they are newcomers to the market. Munshi (2003) argues that both migrants and employers at the destination lack full access to information about each other, reason why they are in need of job referrals. Social networks have the role of decreasing the asymmetry of information between both agents. Using a sample of Mexican migrants in the US, the study finds that migrants with larger networks are more likely to be employed and hold higher paying jobs upon arrival. Also, Chen (2009) uses the proportion of labor migrants in the home village as an indicator of the village social network to study the effects of internal migration in China and finds that larger networks are associated with higher wages. Recently, Kerr and Mandorff (2015) provide evidence that immigrants in the US tend to cluster in the same occupations as immigrants from the same nationality. A result verified by Patel and Vella (2007) that find additional evidence of a wage premium for those immigrants who choose to work in the most popular occupations of their networks due to possible market power among these groups. Although the correlation between networks and higher wages may seem intuitive, some other studies have pointed to an opposite direction. For instance, Datcher Loury (2006) and 5 Long et. al (2013) find evidence of a wage penalty among those individuals who use networks to find jobs over other search methods. Their view rests on the assumption that workers who do not obtain better job offers through formal channels use networks to find jobs as a last resort. For that reason, network users will hold lower paid jobs compared to non-users. In our paper, we will further explore the effects of networks in the labor market using a sample of international migrants living in Ireland. By controlling for a variety of individual characteristics, and addressing potential endogeneity problems, we are able to provide novel insight on this issue. Contrary to the existing literature, our main definition of networks is not only based on geographical proximity or place of origin. Our measure of networks includes non-migrants and migrants from the same country, but also from different countries. In spite of the great importance of migrant communities, it is plausible that they also share and are provided with information about the labor market by other residents. This approach is even more pertinent in the context of Ireland as migrants come from a variety of countries (including highly developed and developing economies), mostly work in high-skilled jobs and are, on average, highly educated 2 . 3. Theoretical Model In order to frame our research, we propose an immigrant’s search behavior model, adapted from Patel and Vella (2007), to study the theoretical mechanisms linking networks to wages and employment. Job search is made using formal and informal methods, where the latter refers to the individual network of employed friends and relatives. The model assumes a continuous time framework in which job offers can arrive from either formal or informal channels following a Poisson process and that immigrants, after arriving to a new country, seek to maximize the expected value of discounted future income using discount rate 𝑟. Individuals 2 Central Statistics Office – Ireland (census report for 2011) 6 (2) (3) receive job offers through the formal channel with arrival rate 𝑝, and from the informal channel with arrival rate 𝑝𝑁. As long as the immigrant does not participate in the labor market, he receives compensation 𝑏, which can be interpreted as leisure or a different type of compensation. Wages offered through the formal channel are represented by 𝑤 and have distribution 𝐹(𝑤), while wages offered through the network channel are represented by 𝑤𝑁 and have distribution 𝐹𝑁(𝑤𝑁). The model assumes that, as long as the migrant finds a job with wage 𝑤 or 𝑤𝑁, he will keep it forever. The value of working in a job found through the formal channel and the informal channel is, 𝑊(𝑤)= 𝑤 𝑟 and 𝑊(𝑤𝑁)= 𝑤𝑁 𝑟, respectively. Thus, the flow value of being unemployed, 𝑟𝑈, is: 𝑟𝑈=𝑏+𝑝∫𝑚𝑎𝑥[𝑊(𝑤)−𝑈,0]𝑑𝐹(𝑤)+ 𝑝𝑁∫max [𝑊(𝑤𝑁)−𝑈,0]𝑑𝐹𝑁(𝑤𝑁) (1) The first term on the right-hand side of the equation, 𝑏, stands for the instantaneous payoff of being unemployed, while the second and the third terms represent the expected value from being employed in a job found via the formal and network channels respectively. The decision of entering the labor market is made when the individual finds a job offer with wage realization 𝑤 or 𝑤𝑁 in which the value of being employed, (𝑤) or (𝑤𝑁) is higher than the value of not entering the labor market, U. The larger the network size, the more job offers a given individual will receive through the informal channel. Assuming that the number of people in the network is given by 𝑛 and that the arrival job rate 𝑝 is constant among all individuals in the network, 𝑝𝑁 is expressed by: 𝑝𝑁= ∑𝑝𝑖 𝑖∈𝑛 =𝑛𝑝 The individual will accept the job offer if 𝑤 is at least as large as the reservation wage, 𝑤𝑅. The reservation wage is solved by W(𝑤𝑅) = 𝑈. Noting that 𝑟𝑈=𝑤𝑅 and 𝑊− 𝑈=𝑤−𝑤𝑅 𝑟. Adding this to equation (1), one obtains the following simplified equation for the reservation wage: 13 had to fulfill the following requirements: 1) be 18 years old or older; 2) not Irish or British born 10 ; 3) have arrived in Ireland between 2000 and six months prior to the interview date; and 4) not an Irish or British citizen. The random sampling procedure followed three steps: first, 100 Enumeration Areas (EA) were randomly selected out of the 323 Electoral Districts. Second, 15 households were chosen within each EA. Finally, if the household had more than one member eligible to take part in the survey, the individual respondent was randomly selected based on a next-birthday rule. Due to missing relevant information about eligibility for nine respondents, the final number of immigrants included in the sample is 1,491. 5.2. Defining Network Size In order to estimate the effect of the network size on wages in the context of migration one first needs to define the network size for each immigrant. We follow (Patel and Vella, 2007) and include in the network the individuals that were working at the time the survey was conducted, and therefore were more able to provide inside information about the labor market 11 . We use three different questions in the survey to cover all the contacts of the immigrants in Ireland. The first one concerns the composition of the household members of the respondent 12 : “Please indicate all the persons who belong to this household”. The second question concerns the contacts who the respondent knew living in Ireland before he had moved to the country: “Before coming to Ireland, how many people did you know who were already living in Ireland (at the time you moved)”? 13 ; The third question includes the contacts living in Ireland whom, at the time of the survey, the migrant had most contact with: 10 As reported in Batista and Narciso (2013), “British citizens were excluded due to the close historical ties between Ireland and Great Britain.” 11 We also tested network definitions including individuals that were not working. However, we obtained weaker results, which suggests that labor market information is shared by those who are insiders in the labor market. 12 Household members are defined as those who usually sleep and eat in the same unit. 13 People belonging to the same household are excluded from this question. We decided to include these contacts once they may have been an important channel of information to the migrants both before and after the migration process. 14 “Currently, who are the people (excluding people you live with or people you knew before coming to Ireland) that you have most contact with in Ireland?”. Using this information we are able to observe the size of the representative network for each migrant. 5.3. Descriptive Statistics Table 1 includes the main descriptive statistics about the migrants’ characteristics. The first variable corresponds to the current monthly net wages expressed in euros. As we can observe, only 957 respondents reported a positive wage 14 . Our data indicates that around 63% of the individuals reported to be working as their main occupation. The size of the migrant network is the explanatory variable of main interest. Network size is a discrete variable that assumes values between 1 and 12 in our data. The average number of people in the network is approximately three. The sample is made up of a highly educated class of migrants who had, on average, 15 years of education, which roughly corresponds to a Bachelor’s degree. The average age is approximately 33 years old. The average number of immigrants per country is 10004. Finally, the average monthly household income, excluding the respondent’s, is 1128 euros and around 46.5% of the immigrants have a child. 15 6. Results 6.1. Employment The theoretical model proposed in section 3 predicts that the size of the network is positively correlated with the probability of the immigrant being employed. Table 2 presents the empirical analysis of this prediction. Employed is a binary variable that assumes value 1 if 14 Those individuals for whom we did not observe a positive wage have the following occupations: unpaid housework, student, retired and unemployed or are not allowed to work due to visa issues. 15 The supplementary appendix includes Table A.1. which shows the distribution of foreign-born individuals in the sample by continent and Table A.1., a description of the distribution of the top nationalities. 15 the individual is employed and value 0 if otherwise 16 . We start by using the LPM to study the relationship between the employment status and our main variable of interest, network size 17 . The results indicate that having one more person in the network is associated with an increase of 3 percentage points (pp) in the probability of being employed – an estimate significant at the 1% significance level. Moreover, females are less likely to be employed compared to males, as the gender coefficient is negative and highly significant. Being more educated is also positively correlated with employment, one more year of school increases the probability of being employed by 1.3 pp. Age and age squared are also strong determinants of the probability of being employed, although inversely related. As the immigrant gets older, he is less likely to be employed. Interestingly, the number of years since migration does not seem to influence the employment. The same holds true for marital status. The results are robust once endogeneity concerns are accounted for. The network size effect becomes higher in magnitude suggesting that having one more contact in the network increases the probability of being employed by 11.3 pp and remains positive and significant at the 10% level. Our instrumental variable seems to be valid as it passes the weak identification test: the Cragg-Donald Wald F test presented in the end of the table is 12.249 implying a strong association between the stock of immigrants and the number of contacts in the network size. 18 These empirical results therefore lend empirical support to the first testable implication of our theoretical model. 6.2. Wages 16 We consider as employed individuals who reported to be working as the main occupation. 17 Table A.4. of the supplementary appendix includes the Probit estimations whose sign, significance and magnitude are similar to the LPM estimations. 18 Staiger and Stock (1997) indicate that, for the case of a single endogenous regressor, the instrument is strong if it passes the threshold of 10 in the first-stage F-statistic. 16 The second main prediction of our theoretical model is that individuals’ wages are a positive function of the size of the informal network. To empirically test this hypothesis, we employ three different models: OLS, Heckit and IV-Heckit. The estimated coefficients and respective standard errors are presented in Table 4. Monthly wage appears in the logarithmic form given the skewness present in the variable (1.9318, a considerable right-hand skewness). We begin our analysis by introducing the OLS estimation results (Column 1). Network size is statistically significant at the 5% level with a coefficient of 0.0260. That is, having one more person in the network size is associated with a 2.60 percent increase in wages. Individual’s gender does not seem to be correlated with the salary earned by the migrant. The coefficient is negative (suggesting that, on average, females tend to receive less than males), although it is not significant. Moreover, the number of years of schooling is positively correlated with the wages earned. One more year in school is associated with a 2.14 pp increase in wages, a result significant at the 10% level. The number of years in Ireland since migration, is also an important positive determinant of wages, indicative of the presence of a process of acquisition of human capital in the host country. Age and age squared are not correlated with wages in the OLS regression. Finally, married immigrants do not earn more on average compared with other marital status. In addition to the main individual characteristics explaining wages, namely gender, years of schooling, years in Ireland, marital status, age and continent fixed effects, we include further controls that may also potentially affecting wages, to increase the comparability of the different estimation models we use. Although the OLS estimator allows one to have a first insight on the relation between network size and wages, it may yield biased estimates as sample selection and other endogeneity concerns might be active in our sample. To test and correct for the hypothesis of sample selection, we employ a Heckman selection model, included in the second column of 17 Table 4. The inverse Mills ratio representing the latent selection factor is negative and statistically significant, indicating that having a child and a higher household income is negatively correlated with the probability of entering the labor market. We can conclude that there is a selection bias on unobserved characteristics that turns the sample of those migrants for whom we observe wages different from the remaining ones. Considering that unobserved characteristics of the immigrants reflect their unobserved ability, the Heckit estimate of network effects reveals that column (1) overestimated the network effects of less able people who gain more from the informal channels. A brief reflection on why less able people take more advantage from their networks leads us to consider that lower earning-ability migrants may be relying more on their social network as they find it more difficult to acquire jobs through formal methods. Comparing the OLS with the Heckit estimates, we can see that when we take into account selection in the decision of working on unobserved ability, the network effect is still significant at the 10% level and our main coefficient of interest is now 0.0196, lower than the OLS estimate. After correcting for sample selection, age and age squared become significant in opposite directions. While age is positively related with the migrants’ wages, age squared influences wages in a negative way. This implies a diminishing marginal effect of age, i.e., as the migrant gets older, the effect of age on wages lessens. This result is robust to the IVHeckit model. All the other coefficient estimates remain similar to the OLS model. Column 3 presents the results of the Heckit model with the network size instrumented, our IV-Heckit model. After controlling for endogeneity, we obtain our best estimate of the causal effect of networks on wages, which is still positive and significant at the 10% level. Having one more person in the network is associated with an increase of 6.46 percent in wages. This result suggests that both the OLS and Heckit model underestimate the effects of 18 network size. Our instrumental variable is significant and valid with a first-stage F value of 20.845. IV-Heckit results are more robust when compared to the other two models, as it simultaneously deals with the two main sources of potential estimation biases in our empirical analysis. Nevertheless, the robustness of our main coefficient of interest in the three models provides clear evidence that the size of migrant social networks has a positive impact on labor market outcomes. Table A.3. of the supplementary appendix presents the first-stage regressions of the Heckit and IV-Heckit models. 6.3. Occupational Choice and Job Security In addition to employment and wages, we also examine the impact of network size on other labor market outcomes, namely occupational choice and job security. As mentioned in Patel and Vella (2007) and Kerr and Mandorff (2015), immigrants tend to choose the same occupations as immigrants from the same background, and therefore migrants belonging to the same network may enjoy a large market power in some given sectors of the economy. In our empirical analysis we test if having more contacts can influence individuals to choose a certain type of occupation. Tables 4 to 6 include the estimations of the immigrants’ main occupations. Our data rules out the hypothesis that having more contacts can influence migrants to choose a certain occupation over the other ones, as the significance of the results is not robust to endogeneity or to the inclusion of continent fixed effects in the regression. However, the results suggest some evidence of a possible effect related with low-skilled jobs (Table 4) for which the coefficients are significant if we not account for continent fixed effects and remains positive in four equations. In what concerns job security, i.e., the employment stability offered by the type of contract, we examine the probability of having a certain type of job contract when the 19 network size is larger. Table 7 presents the results for the different types of job contracts: permanent, temporary, no contract or even self-employment. We find weak evidence that network size is correlated with permanent and temporary contracts or of being self-employed. Evidence is stronger for individuals who do not have any employment contract. Interestingly, networks seem to lead immigrants to other types of job situation rather than jobs with no contract as the coefficient is significant and negative. This is, having one more contact decreases the probability of having a job with no contract by 17 pp. For this reason, networks seem to play an important role in decreasing precarious work. 7. Concluding Remarks This paper examined the impact of larger network sizes on immigrant’s wages, probability of being employed, occupational choice and job security. We find evidence that having one more contact in the host country’s labor market is associated with an increase of 11pp in the probability of being employed and with an increase of about 100 euros in the average salary. However, the evidence related to the impact of networks on job security and occupational choice is rather weak or non-existent. Expanding the previous findings in the literature that have been mainly focused on developing countries, this study concludes that networks are also important to migrants in developed countries. Information flows can play an important role for immigrants in a new labor market, a central conclusion for policy makers who wish to create better migration experiences. Our results may reflect demand and supply influences in the labor market. On the demand side, employers wanting to reduce their screening costs will employ immigrants based on referrals or, if they are satisfied with the performance of a certain type of workers, they are more likely to hire people from the same background. From the supply perspective, upon arriving in the host country, most immigrants are still seeking for job opportunities (as our data suggests, only 13.48% of them had already a job offer when they moved to Ireland). If 20 individuals are in contact with more people who are employed, it is likely that they will have better information about opportunities in the job market. Although we are not able to disentangle such possible mechanisms in our paper, they provide an interesting topic for future research. References Addison John T., and Pedro Portugal. 2002. “Job Search Methods and Outcomes,” Oxford Economic Papers 54, 505-533. An, Weihua. 2015. “Instrumental Variables Estimates of Peer Effects in Social Networks”. Social Science Research Volume 50. Pages 382–394. Batista, Catia, and Gaia Narciso. 2013. “Migrant Remittances and Information Flows: Evidence from a Field Experiment”, IZA Discussion Papers 7839. Batista, Catia, and Janis Umblijs. 2014. “Migration, risk attitudes, and entrepreneurship: Evidence from a representative immigrant survey”. IZA Journal of Migration, 3: 1-25. Batista, Catia, and Janis Umblijs. 2016. “Do Migrants Send Remittances as a Way of Insurance? Evidence from a Representative Immigrant Survey”, Oxford Economic Papers, 68 (1): 108-130. Batista, Catia, and Pedro Vicente. 2011. “Do Migrants Improve Governance at Home? Evidence from a Voting Experiment”. World Bank Economic Review, 25(1): 77-104. Batista, Catia, Julia Seither, and Pedro C. Vicente, 2016. “Migration, Institutions and Social Networks in Mozambique”, NOVAFRICA Working Paper. Beine, Michel, Frédéric Docquier and Çağlar Özden (2011) “Dissecting Network Externalities in International Migration”, CESifo Working Paper: Labour Markets, No. 3333. Calvo-Armengol, Antoni, and Yves Zenou. 2005. “Job Matching, Social Network and Word-of-Mouth Communication”, Journal of Urban Economics 57, 500 – 522. Cameron, A. Colin, and Pravin K. Trivedi. 2005. Microeconometrics: Methods and Applications. Cambridge: Cambridge University Press. Chen, Yunsong. 2014. “Do Networks Pay Off Among Internal Migrants in China?” Chinese Social Review, Volume 45, Issue 1, pages 28-54. Cortes, Patricia. 2008. "The Effects of Low Skilled Immigration on US prices: evidence from CPI data", Journal of Political Economy, vol. 116, n:3. Datcher Loury. 2006. “Some Contacts Are More Equal than Others: Informal Networks, Job Tenure, and Wages”, Journal of Labour, 24, 299-318. 21 Edin, P., P.Fredriksson, and P.Aslund. 2003. "Ethnic Enclaves and the Economic Success of Immigrants—Evidence From a Natural Experiment," Quarterly Journal of Economics, Vol. 118, No. 1, pp. 329—357. Gregg, Paul, and Jonathan Wadsworth. 1996. “How Effective are State Employment Agencies? Jobcentre Use and Job Matching in Britain”, Oxford Bulletin of Economics and Statistics, 58, 443-457. Harris, John, and Michael Todaro. 1970. “Migration, Unemployment and Development: A two-sector analysis”, American Economic Review 60: 126–42. Kerr, William, and Martin Mandorff. 2015. “Social Networks, Ethnicity, and Entrepreneurship”, National Bureau of Economic Research, Working Paper n. 21597. Long, Wenjin, Simon Appleton, and Lina Song. 2013. “Job Contact Networks and Wages of Rural-Urban Migrants in China”, IZA Discussion Papers 7577, Institute for the Study of Labor. McKenzie, David, and Hillel Rapoport. 2010. "Self-Selection Patterns in Mexico-U.S. Migration: The Role of Migration Networks," The Review of Economics and Statistics, MIT Press, vol. 92(4):811-821. Munshi, Kaivan. 2003. “Networks in the Modern Economy: Mexican Migrants in the U.S. Labor Market,” The Quarterly Journal of Economics, MIT Press, vol. 118(2), pages 5495999. Patel, Krishna, and Francis Vella. 2007. “Immigrant Networks and Their Implications for Occupational Choice and Wages” IZA Discussion Papers 3217, Institute for the Study of Labor (IZA). Sjaastad, Larry. 1962. “The Costs and Returns of Human Migration”, Journal of Political Economy 70:80. Staiger, Douglas, and James H. Stock. 1997. “Instrumental Variables Regression with Weak Instruments,” Econometrica, 65, 557–86. Umblijs, Janis. 2012. “The Effect of Networks and Risk Attitudes on the Dynamics of Migration”, Oxford IMI Working Papers, 54/2012. Wahba, Jackline and Yves Zenou. 2005. “Density, Social Networks and Job Search Methods: Theory and Application to Egypt”, Journal of Development Economics, Volume 78, Issue 2, Pages 443-473. Heckman, James. 1979. “Sample Selection Bias as a Specification Error”, Econometrica, 47: 153-161. 22 Table 1. Selected Descriptive Statistics Table 2: Employed as the binary dependent variable Wooldridge, J. M. 2006. Introductory Econometrics. A Modern Approach. (Third Edition). Thomson South-Western. 24 Umblijs, Janis. 2012. “The Effect of Networks and Risk Attitudes on the Dynamics of Migration”, Oxford IMI Working Papers, 54/2012. Appendix: LPM (1) IV (2) Network Size 0.0288*** (0.00615) 0.113* (0.0665) Female -0.0866*** (0.0257) -0.0828*** (0.0277) Years of Schooling 0.0133** (0.00475) 0.0113* (0.00556) Years in Ireland -0.000491 (0.00486) -0.00630 (0.00669) Age 0.0417*** (0.00942) 0.0433*** (0.0101) Age^2 -0.000447*** (0.000126) -0.000458*** (0.000135) Married -0.0128 (0.0284) -0.0526 (0.0425) Constant -0.361** (0.172) -0.647** (0.262) Continent Fixed Effects YES YES Variables Obs. Mean Std. Dev. Max. Min. Main Dependent Variables Monthly net wage (euros) 957 1665 987.9 10500 100 Employed (dummy) 1481 0.627 0.484 1 0 Main Independent Variables Network Size 1280 3.077 2.165 12 1 Female (dummy) 1491 0.541 0.498 1 0 Years of schooling 1483 14.69 2.798 17 0 Years in Ireland 1489 5.348 2.863 11 0 Married (dummy) 1491 0.424 0.494 1 0 Age 1491 32.59 8.025 72 18 Instrumental Variable Stock of immigrants in Ireland (year 2000) 1485 10004 12919 44633 2 Exclusion Restrictions Monthly Household Income (euros) 1077 1128 1746 17500 0 Having children (dummy) 1491 0.465 0.499 1 0