کیفیت محله و نتایج بازار کار : شواهدی از ماموریت محله نیمه رندوم مهاجران
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|16254||2014||28 صفحه PDF||سفارش دهید||محاسبه نشده|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Journal of Urban Economics, Volume 79, January 2014, Pages 139–166
Settlement in a socially deprived neighborhood may hamper individual labor market outcomes because of lack of employed or highly skilled contacts. I investigate this hypothesis by exploiting a unique natural experiment that occurred between 1986 and 1998 when refugee immigrants to Denmark were assigned to municipalities quasi-randomly, which successfully addresses the methodological problem of endogenous neighborhood selection. I show that individuals sort into neighborhoods. Taking account of location sorting, living in a socially deprived neighborhood does not affect labor market outcomes of refugee men. Their labor market outcomes are also not affected by the overall employment rate and the overall average skill level in the neighborhood. However, an increase in the average skill level of non-Western immigrant men living in the neighborhood raises their employment probability, while an increase in the employment rate of co-national men living in the neighborhood raises their real annual earnings. This provides quasi-experimental evidence that residence-based job information networks are ethnically stratified.
Widespread use of friends, relatives and acquaintances to search for jobs is a stylized fact.1 Personal contacts may convey information about job vacancies and recommend friends, relatives and acquaintances with similar personal characteristics as themselves to their employer. For employers, job referrals lower the search costs as well as the screening costs of applicants. For employees, job referral may speed up the job-finding process and, as suggested by Dustmann et al. (2011) lead to a short-termed wage premium due to reduced uncertainty about the worker productivity. According to recent social network theories the quality of personal contacts is of key importance for job referral (Montgomery, 1994 and Calvó-Armengol and Jackson, 2004).2 The higher the quality of personal contacts, the more useful the contacts are for job referral. My main research question is whether the quality of the job information network affects labor market outcomes. In the first part of my analysis I provide tentative evidence from the Welfare Research Survey conducted in Denmark in 2006 among representative samples of natives and immigrants that the quality for contacts matters: unemployed respondents whose acquaintances (e.g. friends of friends, co-workers and neighbors) have a high employment rate have a higher job-finding rate, after controlling for other personal characteristics and area characteristics. Job seekers may thus receive information about job vacancies from employed neighbors among others. Therefore, living in a neighborhood with more unemployment may reduce job chances. If so, concentration of unemployed workers in certain neighborhoods increases employment inequality in society (see e.g. Montgomery, 1994). Results from studies using observational data are consistent with neighborhood job referral, i.e. individuals who live in the same or adjacent neighborhoods sometimes refer each other to jobs.3 By contrast, quasi-experimental and experimental studies find little role of neighborhood quality on adult labor outcomes.4 I argue that the Danish Spatial Dispersal Policy on Refugees which operated from 1986 until 1998 is an ideal quasi-experiment for investigation of whether the quality of the neighborhood matters for individual labor market outcomes. At the time of receipt of asylum, placement officers working in the central office of the Danish Refugee Council assigned refugee families to housing in different locations in Denmark, exclusively on the basis of a questionnaire with personal information like household size. The placement officers did not meet face to face with refugees at the time of assignment. I observe all personal characteristics known to the placement officers in the administrative registers used for the analysis and condition on them in the regressions. Conditional on these personal characteristics, characteristics of the neighborhood of assignment are valid instruments for characteristics of the current neighborhood of residence. Moreover, the approximately 15,400 refugee men subjected to the spatial dispersal policy were assigned to as many as 1710 different neighborhoods located in 245 different municipalities.5 In other words, the Danish Spatial Dispersal Policy provides extensive geographic variation in neighborhood characteristics which is an important strength relative to the existing quasi-experimental and experimental studies of neighborhood job referral. Therefore, in the second part of my analysis I use detailed administrative register data for the sample of refugee men who had initially been assigned to a neighborhood by the Danish Refugee Council to provide quasi-experimental evidence on the quality of job information networks on individual labor market outcomes. To proxy for social interactions, I use neighborhood-based networks, i.e. connections between individuals living in the same neighborhood, in line with a number of previous studies.6 In particular, I provide two-stage least squares (2SLS) estimates of the causal effects of living in a socially deprived neighborhood on individual labor market outcomes. I define a neighborhood as socially deprived if the employment rate is at most 60%. As instrument for the indicator for current residence in a socially deprived neighborhood I use an indicator for assignment to a socially deprived neighborhood. Next, I estimate the effects of alternative, continuous measures of neighborhood quality (employment rates, average skill level, mean real annual earnings) and compare ordinary least squares (OLS) and propensity score matching (PSM) estimates to quasi-experimental results from instrumental variables (IV) estimation. Using observational data, the studies by Topa (2001) and Hellerstein et al. (2011) find empirical evidence that residence-based job search networks are ethnically stratified. In that case the quality of the co-ethnic network should matter more for individual labor market outcomes than the overall quality of the neighborhood. In the third part of the analysis, I exploit the Spatial Dispersal Policy on Refugees to provide quasi-experimental evidence on the effects of the quality of the immigrant and co-ethnic networks on labor market outcomes of refugees. As before, I use the detailed administrative register data for the sample of refugee men who had initially been assigned to a neighborhood by the Danish Refugee Council. However, to proxy for social interactions, I follow Borjas (1995) and use neighborhood-based immigrant and co-ethnic networks, i.e. defined as connections between individuals of immigrant origin/the same national origin living in the same neighborhood.7 Specifically, I present 2SLS estimates of the effects of the quality of immigrant/co-ethnic men living in the neighborhood on individual labor market outcomes 2–6 years after immigration. As instruments I use the quality of immigrant/co-ethnic men living in the neighborhood of assignment in the year of assignment and an indicator for no other immigrant/co-ethnic men living in the neighborhood of assignment in the year of assignment. I measure quality in terms of the employment rate and the average skill level. Therefore, my paper combines two strands of literature: empirical network studies which use neighborhood-based networks and empirical network studies which use ethnicity-based networks.8 Previous studies on the impact of residence-based ethnic networks on labor market outcomes (Bertrand et al., 2000, Edin et al., 2003 and Damm, 2009) use metropolitan areas/municipalities as the geographic unit of analysis. The empirical model used in these studies to identify the effect of the size as well as the interaction between the size and the quality of the ethnic group omits the quality of the local ethnic group; the direct effect of the quality of the local ethnic group is assumed to be captured by language/country of origin fixed effect. By instead using neighborhoods, I can disentangle the effect of the quality from the size of the immigrant/co-ethnic network in contrast to previous studies. Finally, to shed light on how long it takes for a new resident to become part of the established networks in the neighborhood, I investigate the speed at which network effects operate in the neighborhood using the detailed administrative registers for refugee men who were initially assigned to neighborhoods by the Danish Refugee Council. The next section provides tentative evidence on the importance of the quality and quantity of contacts on individual employment from the Welfare Research Survey conducted in Denmark in 2006 among random samples of natives and non-Western immigrants. Section 3 describes the construction of a balanced panel of male refugees from administrative register data and presents OLS and PSM estimates of living in a socially deprived neighborhood six years after immigration for the individuals in the balanced panel of male refugees. It then describes the Danish Spatial Dispersal Policy on Refugees and exploits it to provide quasi-experimental evidence on the effects of living in a socially deprived neighborhood and the effects of continuous measures of neighborhood quality on individual labor market outcomes of male refugees. Section 4 presents quasi-experimental evidence on whether residence-based networks of male refugees are ethnically stratified and on the speed at which network effects operate in the neighborhood. Section 5 offers conclusions.
نتیجه گیری انگلیسی
The main purpose of this paper has been to investigate whether the quality of the neighborhood affects the individual’s labor market outcomes. I have shown that both Ordinary Least Squares and Propensity Score Matching estimation give biased estimates of the effect of neighborhood quality. The reason is that individuals sort across neighborhoods in terms of unobserved personal characteristics, violating the assumption of selection on observables. To deal with location sorting, I have exploited a spatial dispersal policy on refugees in Denmark which resulted in quasi-random neighborhood assignment of refugees. My quasi-experimental findings suggest four main conclusions. First, individuals sort across neighborhoods. Second, taking account of sorting, residence in a socially deprived neighborhood (defined as a neighborhood in which the employment rate is not above 60%) does not affect individual labor market outcomes. This finding is in line with previous experimental (Katz et al., 2001, Kling et al., 2007, Sanbonmatsu et al., 2011 and Ludwig et al., 2012) and quasi-experimental findings (Oreopoulos, 2003). Thirdly, a relatively high average skill level of non-Western immigrants living in the neighborhood increases the individual’s employment probability, while a relatively high employment rate among co-nationals living in the neighborhood raises individual real annual earnings. By contrast, the effect of the percentage of co-nationals living in the neighborhood is insignificant. Fourthly, the effects of the quality of non-Western immigrant neighbors are cumulative, whereas the effects of the quality of co-national neighbors are of similar magnitude across years since migration. I interpret these findings as evidence of the following. First, quality – rather than the size – of residence-based networks is important for labor market outcomes, consistent with the theories proposed by Montgomery (1994) and Calvó-Armengol and Jackson (2004). Second, residence-based job information networks are ethnically stratified, in line with evidence from observational studies by Topa (2001) and Hellerstein et al. (2011). Living in a socially deprived neighborhood does not hamper labor market outcomes of non-Western immigrants since – in contrast to my priors – the average educational level of non-Western immigrant men across socially deprived neighborhoods (7.8 years over the 1986–1998) is close to the average across all neighborhoods (8.1 years over the 1986–1998 period). Instead the low employment rates of residents in socially deprived neighborhoods are primarily the result of the large overrepresentation of non-Western immigrants and descendants in socially deprived neighborhoods combined with the generally low employment rate of non-Western immigrants and descendants compared to that of natives. An additional explanation – of second-order importance – is that immigrants who settle in socially deprived neighborhoods are negatively selected from the overall group of immigrants, both in terms of observable and unobservable personal characteristics. Dustmann et al. (2011) find that employment through referral is associated with a short-term wage premium. The economic intuition is that referrals reduce the uncertainty about the workers productivity. However, my finding that the quality of co-national neighbors affects earnings in contrast to the quality of immigrant neighbors suggests that only referral through co-national neighbors increases the individual’s earnings, because only co-nationals know the true worker productivity, for instance from the worker’s skill certificates from the source country. Recall my final finding that the effect of the quality of co-national neighbors is time-constant whereas the effect of the quality of immigrant neighbors increases with time spent in the host country. It suggests that information about job opportunities propagates fast through the co-ethnic network in the neighborhood and more slowly through the overall immigrant network in the neighborhood. My findings are economically significant. Having non-Western immigrant male neighbors whose mean skill level is one year above the average, increases the individual’s employment probability by 1.6 percentage points. Similarly, a ten percentage points increase in the employment rate of co-national men aged 18–60 living in the neighborhood (around the mean) increases the individual’s real annual earnings by 16%. The results have important policy implications for the labor market integration of immigrants. Existing evidence shows that policy makers in countries in which the authorities disperse newly recognized refugees across regions can increase the speed of employment and wage-assimilation of immigrants through refugee settlement in regions with low unemployment (Åslund and Rooth, 2007). This study adds that within high-employment regions, refugees should settle in neighborhoods with relatively high average skill levels among immigrants and relatively high employment rates among co-nationals. However, the intent-to-treat estimates are very low compared to the two-stage least squares estimates (see e.g. Table 7). The reason is that few refugees stayed in the assigned neighborhood for a considerable number of years. Therefore, for such a settlement policy to work, the authorities may have to reduce the relocation rates of refugees, by also considering individual location wishes (given that they coincide with the overall goal of the settlement policy) and/or providing economic incentives to stay for some years. More generally, the findings of this study imply that successful local employment and skills-upgrading policies targeted at a subgroup of immigrants entail positive externalities for their immigrant acquaintances.