Seeking Similarity: How Immigrants and Natives Manage in the Labor Market ˚ slund, Institute for Evaluation of Labour Market and Olof A Education Policy ðIFAUÞ, Uppsala University, IZA, and CReAM
Lena Hensvik, IFAU and Uppsala University Oskar Nordstro¨m Skans, IFAU, IZA, and Uppsala University We investigate how the interplay between manager and worker origin affects hiring patterns, job separations, and wages. Numerous specifications utilizing a longitudinal matched employer-employee database including 70,000 establishments consistently show that managers are substantially more likely to hire workers of their own origin. Workers who share an origin with their managers earn higher wages and have lower separation rates than dissimilar workers, but this pattern is driven by differences in unobserved worker characteristics. Our findings indicate that the sorting patterns are more likely to be explained by profit-maximizing concerns than by preference-based discrimination. I. Introduction Managers are key players in the labor market. Their practices matter for firm performance, for the overall wage distribution, and for the allocation We are grateful for comments by the editor, Mikael Lindahl, Lena MagnussonTurner, Eva Mo¨rk, Peter Skogman Thoursie, and seminar participants at the World Bank, IFAU, IBF, the 2008 Cost Conference in Paris, the ELE Immigration Conference in Helsinki, the labor lunch workshop at Harvard University, the Nordic Labor Economists meeting in Bergen, the CCP Workshop on Personnel [ Journal of Labor Economics, 2014, vol. 32, no. 3] © 2014 by The University of Chicago. All rights reserved. 0734-306X/2014/3203-0001$10.00
405
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
406
˚ slund et al. A
of skills across firms and industries. But the impact of managerial decisions is also of first-order importance for the individual workers whose potential job offers and wages are determined by the managers they encounter. Taken broadly, the theoretical literature suggests that in the labor market, in-group biases can arise either due to preferences or to profitmaximizing concerns and that these play an important role for explaining sorting and economic outcomes ðe.g., Currarini, Jackson, and Pin 2009; Fang and Moro 2011Þ. In this article, we analyze the role of in-group biases related to managers’ and workers’ immigration backgrounds by investigating the impact of similarity/dissimilarity on hiring patterns, job separations, and wages. Although the poor economic integration of migrant workers in many countries has received a lot of attention, a rarely analyzed fact is that groups with poor labor market performance also tend to be underrepresented among managers. In our population-wide Swedish data set, 7.2% of recruited workers but only 1.5% of managers are foreign-born. The fact that there are five times as many immigrants on the supply side as on the demand side of the matching process suggests that in-group biases not only may explain part of the strong ethnic segregation across workplaces found in many countries but also offer a potential contributing explanation to the immigrant-native performance gap.1 Origin-biased hiring patterns may arise for different reasons ða longer discussion can be found in app. B, available onlineÞ. First, systematic sorting can arise due to preferences.2 In Becker’s ð1957Þ discrimination model, some —but not all—employers are unwilling to hire minority workers at the majority wage simply because they derive disutility from doing so.3 Economics in Arhus, the 2009 Swedish Integration Research Network Conference in Lund, and the 2009 SOLE ðSociety of Labor EconomistsÞ and EEA ðEuropean Economic AreaÞ conferences. We are particularly grateful to Fredrik Heyman, Helena Svaleryd, and Jonas Vlachos for very kindly sharing data on industry-level competition. Financial support from FAS is gratefully acknowledged. The order of the authors is in accordance with the English alphabet and not related to contribu˚ slund, at
[email protected]. tion. Contact the corresponding author, Olof A Information concerning access to the data used in this article is available as supplementary material online. 1 ˚ slund and Skans ð2010Þ for Sweden and Dustmann, Glitz, and Scho¨nberg See A ð2011Þ for Germany on immigrant-native labor market segregation and Hellerstein and Neumark ð2008Þ on racial labor market segregation in the United States. 2 Preferences of both managers and workers may be important for who gets hired. Giuliano, Levine, and Leonard ð2011Þ argue that worker preferences are the key factor for why black managers recruit fewer white applicants in the retail firm they study. 3 Laboratory experiments suggest that people tend to favor/trust others with a similar ethnic background ðe.g., Fershtman and Gneezy 2001; Ahmed 2007Þ. Field experiments point at substantial ethnic discrimination in the hiring procedure against African Americans in the United States ðBertrand and Mullainathan 2004Þ and workers of Middle-Eastern descent in Sweden ðCarlsson and Rooth 2007Þ. For
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
How Immigrants and Natives Manage in the Labor Market
407
General equilibrium models in the search and matching framework ðe.g., Black 1995; Rose´n 2003Þ show that firms whose managers have discriminatory preferences may survive even in economies with free entry of firms. However, since firms with discriminating managers will earn lower profits on average ðsee, e.g., Rose´n 2003Þ, such biases in hiring patterns should still be expected to be lower in markets where low productivity is punished more severely ði.e., when product market competition is more pronouncedÞ. Second, managers may hire workers with a similar background for pure profit-maximizing reasons. Workers may become more productive if they share a language or business culture with their managers due to lower transaction and communication costs ðLazear 1999; den Butter, Masurell, and Mosch 2004Þ. But similarity may also make the process when managers select workers more efficient. Theories of statistical discrimination ðsee Fang and Moro ½2011 for an overviewÞ suggest that managers may experience less noise in productivity signals from workers with a similar background. Managers may therefore prefer to hire workers who are similar to themselves if they are risk averse or if acquiring information is costly. Conversely, it is conceivable that workers receive a noisier signal from managers with a background that differs from their own. A related reason is that job-search networks may provide useful information, allowing the manager to hire better workers or achieve a better match quality when relying on ðpotentially segregatedÞ networks ðsee, e.g., Montgomery 1991; Dustmann et al. 2011Þ.4 Studies relying on cross-sectional data have documented correlations between manager race and the race of hires ðCarrington and Troske 1998; Stoll, Raphael, and Holzer 2004Þ. But it is difficult to distinguish the causal impact of manager characteristics on hiring patterns from spurious relationships generated by nonrandom sorting on other characteristics of firms and workers without access to detailed data on the characteristics of the hiring firms. To facilitate more reliable identification, a number of recent papers have relied on single-firm panel data with detailed accounts of many aspects of the process. Bandiera, Barankay, and Rasul ð2009Þ study the importance of nationality and social connections between the man˚ slund quasi-experimental evidence of discrimination in actual recruitments, see A and Skans ð2012Þ. 4 There is a large and growing empirical literature suggesting that social networks are very important when workers get hired; Ioannides and Loury ð2004Þ provide a survey. Individuals who live in the same residential area are more likely to work together ðBayer, Ross, and Topa 2008Þ, parents help their children to find their first job ðKramarz and Skans 2011Þ, former coworkers share information about new jobs ðCingano and Rosolia 2012Þ, and immigrants with larger exogenous networks are more successful in the labor market ðMunshi 2003Þ. A series of recent papers provide indirect evidence that ethnic labor market networks are important for black and Hispanic workers in the United States ðHellerstein, McInerney, and Neumark 2008a, 2008b, and 2009; Hellerstein and Neumark 2008Þ and for immigrants in Germany ðDustmann et al. 2011Þ.
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
408
˚ slund et al. A
ager and employees for the allocation of jobs within a British fruit-picking farm where workers are allocated to jobs on a day-to-day basis. Two papers by Giuliano, Levine, and Leonard ð2009, 2011Þ document substantial ethnic biases in hiring and firing in a large US retail chain. These recent studies provide compelling evidence of a causal effect of manager race/ethnicity on hiring patterns in the studied firms. The main advantage of these data sets is the detailed longitudinal information on workers and managers combined with high worker turnover. Bandiera et al. ð2009Þ also include precise accounts of worker productivity.5 However, whether the findings from the studied firms can be generalized to other parts of the labor market where jobs may be rationed and turnover is low remains an open issue. Our study complements these earlier studies by analyzing hiring patterns in a very broad set of firms. We use a longitudinal matched employeremployee data set covering 70,000 Swedish establishments across the entire economy during a 9-year period. This allows us to implement various strategies to account for unobserved heterogeneity among workers, managers, and firms, as well as to document how the effect of the manager’s origin varies with respect to establishment and worker characteristics. We also provide evidence on the role played by the managers’ networks of former coworkers. In addition, we investigate how origin similarity affects wages and separations. Our analysis shows that immigrant and native managers differ dramatically in their hiring patterns. Native managers hire on average 6% immigrant workers; the corresponding figure for immigrant managers is 21%. A strong association remains when comparing different establishments within the same five-digit industry and location ðe.g., different pharmacies in the same villageÞ and different establishments within the same firm and location and when studying establishments that change management over time. The estimates are economically and statistically significant throughout these specifications, and they remain so if the composition among incumbent workers is controlled for. We find that similarity matters for both high- and low-skilled workers in establishments of different sizes and in most industries. But the effects 5
A closely related literature studies managers from a gender perspective and shows a positive correlation between female management and female wages ðCarrington and Troske 1995; Hultin and Szulkin 2003Þ. Using a matched employer-employee data set for Portugal, Cardoso and Winter-Ebmer ð2010Þ estimate the effect of within-establishment manager changes and find that female-led firms pay a premium to female workers of almost 5%. Using Swedish data, Hensvik ð2011Þ documents a similar correlation between the share of female managers and the establishment gender wage gap but finds that most of this relationship is attributable to female mangers hiring women with higher ðunobservedÞ skills compared to male managers. She finds no evidence of increased hiring prospects for women in women-led establishments.
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
How Immigrants and Natives Manage in the Labor Market
409
tend to be larger in the private sector and when product market competition is more pronounced. Although managers with a greater prior exposure to immigrant coworkers appear to recruit more immigrants, we find that the differences in past exposure between managers of different origin is too small to explain the differences in recruitment patterns between native and immigrant managers. In addition, we show that similarity, on average, also matters in cases where managers recruit among their former coworkers. We also find that workers who share an origin with their managers earn higher wages and have lower exit rates than their coworkers. However, when we take individual heterogeneity into account, we find no direct effect of manager-worker similarity on either wages or worker turnover. These results suggest that managers who recruit workers with whom they share background may be able to attract workers with better ðportableÞ unobserved qualities. The results stand in some contrast to predictions from preference-based discrimination theories, which suggest that managers who discriminate should use a lower productivity threshold for workers of the preferred type. Taken at face value, these results, and the finding that the biases tend to be more pronounced in the for-profit sector, indicate that profit-maximizing concerns, potentially related to the difficulties in selecting productive employees, are a more likely candidate for explaining the sorting patterns than is preference-based discrimination. The remainder of this article is structured as follows. Section II describes the institutional background. Section III presents the data. Section IV provides some descriptive patterns and sample statistics. Sections V and VI present the empirical results along with the respective methodological approaches. Section V presents the results on the impact of manager origin on the origin of hires, including how previous interactions in the labor market affect the probability of hiring people with similar/dissimilar origin. Section VI analyzes the impact of origin similarity on wages and job turnover. Section VII concludes. II. Immigrants in the Swedish Labor Market Since 1960, the number of first-generation immigrants living in Sweden has grown from 300,000 to 1,400,000 in 2011. Today, the foreign-born constitute about 15% of Sweden’s 9 million residents and define most of the country’s diversity in terms of origin or ethnicity. More than 40% of the immigrant population originates from non-Western countries outside Europe, and a substantial fraction of the European migrants have arrived from outside the European Union ðEUÞ. As in many other Western countries, the labor market position of the immigrant population has deteriorated during the past 30 years. In the 1950s and 1960s, labor migration from the Nordic countries ðesp. FinlandÞ and continental Europe dominated the inflow. Immigration then gradually shifted toward refugees and family reunification migrants, many times
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
˚ slund et al. A
410
from developing and geographically distant countries ðe.g., Chile in the 1970s, Iran from the 1980s, Somalia and former Yugoslavia in the early 1990s, and Iraq in the 1990s and 2000sÞ. Even though natives on average perform better in the labor market than almost all groups of migrants, the great divider seems to be between those of Western and non-Western origin. In 2002 ðin the midst of our observation period; see belowÞ, the employment rate among natives was 76.8%. The corresponding figure for European Union (EU)/European Economic Area (EEA) migrants was 69.3%, compared to 53.5% among those born outside Europe. Wage differences are less dramatic, but they follow the same pattern: the average monthly ðfull-timeÞ wage among natives was SEK 22,250 in 2002; for immigrants from non-European countries, it was SEK 19,050, while EU migrants had an average wage almost identical to the one received by natives.6 For a further discussion of these differences and their possible causes, see Eriksson ð2010Þ. III. Data Our primary source of data is a Swedish linked employer-employee database ðRAMSÞ covering the period 1985–2005. By combining these data with additional administrative data sources, we are able to track managers, employees, and establishments over time and link each of these subjects to detailed information on individual demographic characteristics ðgender, age, region of birth, education, and place of residenceÞ as well as to basic information about each establishment ðlocation, industry, and sectorÞ. Our main working data set includes all newly recruited workers in establishments with fewer than 50 employees during the period 1997–2005, together with information on the immigration status of each worker and manager. The rationale for restricting the analysis to small- and medium-sized establishments is that it is more likely that the managers we identify are directly involved in the hiring and firing decisions in such establishments.7 A. Origin of Workers and Managers In the main analysis, we group the individuals into two categories: ðiÞ workers of Western origin, that is, natives and immigrants from Western countries; ðiiÞ immigrant workers of non-Western origin. For convenience, we label the groups “Natives” and “Immigrants.” This division corresponds 6 Figures for employment and unemployment come from the Swedish labor force surveys. Wages are calculated from the LINDA database ðsee Edin and Fredriksson 2000Þ, which contains a 3% representative sample of the Swedish population. 7 Since previous research shows that segregation is most prevalent among small˚ slund and Skans 2010Þ, our results are to medium-sized establishments ðsee, e.g., A not necessarily representative for larger establishments. However, given that the data show that the median worker is employed in an establishment with 52 employees ð2001Þ, we cover a substantial part of the workforce.
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
How Immigrants and Natives Manage in the Labor Market
411
well to the main divider in terms of differences in labor market outcomes and also to the public perception of “being foreign” ðsee, e.g., Mella and Palm 2009Þ. In a separate heterogeneity analysis, we also divide the data into four different groups: Native, Western, Eastern Europe, and NonEurope. ðTable A1 in app. A shows the full list of countries as well as the fraction of hires from each country or set of countriesÞ. B. Wages Our wage data come from a register ðStrukturlo¨nestatistikenÞ containing occupations and wages for all employees in the public sector and for a large sample ðdrawn at the firm levelÞ of employees in the private sector. The data cover 1997–2005 and include annual information of fulltime-equivalent monthly wages ðincluding additions and supplementsÞ, measured in September or November each year. The data have an underrepresentation of establishments belonging to smaller private firms due to an oversampling of large firms. But, although the sampling probabilities are small for small firms, many large firms have small establishments, and our final data set thus covers approximately 30% of all smalland medium-sized private establishments.8 C. Managers To identify managers, we use occupational data from the abovementioned register, structured according to the Swedish Standard for Classification of Occupations ðSSYKÞ, which is based on international standards ðISCO-88Þ. The first digit in the occupational code divides the data into 10 major occupational levels based on the skill requirements and with a specific number for managerial positions. Using additional digits, we can also distinguish between top and middle managers. Our manager definition is based on the following hierarchical criteria: ð1Þ top manager, ð2Þ middle manager, and ð3Þ highest wage. In case there are multiple observations fulfilling the same criterion, we use lower-ranked criteria to identify the manager ðe.g., the middle manager with the highest wageÞ.9 This strategy is likely to introduce some measurement error in the manager code. However, it is reassuring that 83% ð78%Þ of the managers identified by their occupational 8 Detailed accounts and an analysis of heterogeneous effects are available in the ˚ slund, Hensvik, and Skans 2009Þ. working paper version of this article ðA 9 To increase sample size, the manager classification ðnot the wage regressionsÞ uses also information from population-wide data on estimated monthly wages ðsee, e.g., Skans et al. ½2009 for proceduresÞ. If an establishment is sampled at two separate points in time with the same manager, the same person is assumed to be manager also in the years in between ðprovided he or she is at the establishmentÞ. If the sample data identify a manager in one year, the same person is assumed to be manager in all continuously preceding and following years in which he or she has the highest estimated wage ðand where the establishment is not sampledÞ.
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
˚ slund et al. A
412
status as top ðmiddleÞ managers are also the highest wage earners in their establishment. We also report estimates separately by type of manager classification, and the results do not differ in general. The main analysis excludes firms run by self-employed owners, but in a robustness analysis we also include these firms and treat the owners as managers ðownership is available for all establishments run by the self-employedÞ. D. Data on Recruited Workers We define workers who received remuneration from the establishment in a given year but not during any of the preceding 5 years as a newly hired worker. We disregard individuals earning below the 10th percentile of the overall annual earnings distribution in order to avoid classifying very loosely connected ði.e., working a few hours within the yearÞ workers as new hires. We are primarily interested in recruitments within continuing establishments and therefore require that the establishment existed in the preceding year. For the same reason we also classify an establishment as new ðand remove itÞ if more than two-thirds of the workforce changed ðin either directionÞ from one year to the next.10 E. Data on Former Colleagues The long panel of individuals and establishments allows us to identify each manager’s set of coworkers in previous jobs dating back to 1985. We use this to ðiÞ measure manager exposure to immigrant coworkers in the past by calculating the fraction of immigrants among all other workers at every ðpastÞ establishment the manager was employed by since 1985; ðiiÞ create a data set including each manager’s set of previous colleagues in the cases where the manager hired a previous colleague. For the second aim, we restricted the data to individuals who worked together in an establishment with fewer than 100 employees. The reason is that we want it to be likely that the two agents interacted at the old establishment so that they were able to eliminate uncertainty about each other’s productivity. We do not, however, put any restrictions on the manager’s occupational level at the past establishment. Thus, he or she did not have to be a manager at the previous job as long as the manager and the hire received compensation from the same establishment during the same year at some point from 1985 and up to the new recruitment.11 We also require that the 10 When checking that new hires did not receive earnings from the same establishment in the past, we use the original establishment identification number in order to make sure that the hires were really externally recruited. 11 Because every establishment where a future manager is observed with a future hire will be blown up by its size times the number of future managers, a size restriction is necessary also for computational reasons.
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
How Immigrants and Natives Manage in the Labor Market
413
manager worked at the ðnewÞ establishment at least 1 year prior to the recruitment. IV. Descriptive Patterns Figure 1 presents the share of immigrant hires and managers, overall and by industry. Two important facts emerge. First, immigrants are underrepresented among managers in relation to their share of newly hired workers ð“entrants”Þ, which implies that the agents of the labor market matching process are unevenly distributed. Whereas 7.2% of the entrants are immigrants, the corresponding number for managers is 1.5%.12 Second, there is systematic sorting across industries both among entrants and managers; the representation of immigrants is, for example, much higher in hotels and restaurants, in mining, and in manufacturing than in other industries. Industries with fewer immigrant managers also recruit a lower share of immigrant workers. Table 1 shows that the sorting at the industry level also carries down to the establishment level. Establishments with immigrant managers ðcol. 3Þ employ a substantially larger share of immigrants than establishments with native managers ðcol. 2Þ. This holds also for the entrants, and the magnitudes are striking: the share of immigrants hired under immigrant management is 21%, compared to 6% under native management. Thus, immigrant managers hire immigrant workers with a more than three times higher probability than native managers. The table also shows that immigrants manage smaller establishments, hire fewer individuals, and operate in more immigrant-dense municipalities than native managers. About half of the immigrant workers who are hired by immigrant managers are from countries other than the country the manager is from ðnot shown in the tableÞ. Turning to the manager characteristics, we see that a substantial share of both the native and the immigrant managers are identified by their wage.13 Even if the vast majority of managers identified as top or middle managers are also the highest wage earners, this may be a concern for some of our estimates since the wage-based classification may be influenced by measurement errors. Our empirical section therefore also presents the key results separately by type of manager. 12 Official statistics from the 2007 labor force surveys confirm this picture for the overall population of employees: 6% of all native employees are managers, whereas the figure is less than 2% for immigrants from non-EU/EEA countries. 13 A much larger fraction of the immigrant managers are self-employed owners, which is in line with the fact that self-employment is comparatively common among immigrants. Owner-led establishments are excluded from the baseline sample. Table A3 presents robustness checks including this group of managers.
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
FIG. 1.—Share of immigrant managers and hired immigrants in different industries Table 1 Sample Statistics Manager’s Origin
Immigrant share among new entrants Immigrant share in establishment Immigrant share in five-digit industry by municipality Immigrant share among manager’s former coworkers* New entrants/year Standard deviation Establishment size Manager type: Top manager Middle manager Highest wage Manager origin: Native ðtreated as “natives”Þ Western countries ðtreated as “natives”Þ Eastern Europe ðtreated as “immigrants”Þ Other ðtreated as “immigrants”Þ Number of observations Number of establishments
All ð1Þ
Native ð2Þ
Immigrant ð3Þ
.06 .04 .05 .03 5.17 3.74 25.6
.06 .04 .05 .03 5.17 3.74 25.6
.21 .17 .11 .11 5.42 3.34 25.3
.15 .24 .58
.15 .27 .58
.17 .15 .68
.95 .04 .01 .01 757,278 71,284
.96 .04
745,660 69,577
.57 .43 11,618 1,012
NOTE.—The sample consists of all establishments that hired at least one individual during the period 1997–2005. * The share of immigrants among former coworkers is calculated from all previous establishments ðfrom 1985Þ of each manager.
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
How Immigrants and Natives Manage in the Labor Market
415
V. Hiring Patterns A. Empirical Framework Our aim is to identify the causal impact of manager origin on the probability that new hires are immigrants. To this end we use data on positions being filled and estimate linear probability models of the following type: Prðjob j is filled by an immigrantÞ 5 gMj 1 Xj b;
ð1Þ
where Mj is a dummy variable taking the value one if the manager responsible for job j is an immigrant and zero otherwise, and Xj is a vector of control variables. In order to remove potential confounding factors, we use several identification strategies, all implemented as variations of controls in the X-vector of equation ð1Þ. These are described below together with the results. B. Baseline Results Table 2 presents results from estimated linear probability models based on equation ð1Þ. The dependent variable is the probability that a job is filled by a worker of immigrant origin, and the covariate of interest is a dummy for whether the manager is an immigrant. All specifications include year dummies to account for national trends in recruitment patterns and establishment size dummies in intervals of 10 employees. Other controls vary between columns. Columns 1–3 of table 2 add controls step-wise, accounting for regional sorting at the municipal level ðcol. 2Þ and industry sorting within municipalities ðcol. 3Þ. The results confirm a very strong association between the manager’s and the recruited worker’s immigrant statuses. Note that the specification presented in column 3 is fairly stringent: we compare similar firms in the same locations by including fixed effects for each combina-
Table 2 Cross-Sectional Estimates of the Effects of Manager Origin Probability That Job Is Filled by an Immigrant Immigrant manager
ð1Þ
ð2Þ
ð3Þ
.143 ð.007Þ
.111 ð.006Þ
.083 ð.007Þ
Establishment immigrant share R2 Fixed effects
.008 Y
ð4Þ
ð5Þ
.066 .058 ð.005Þ ð.005Þ .609 .515 ð.009Þ ð.009Þ .035 .185 .055 .068 YM YMI Y YM Y
ð6Þ .051 ð.006Þ .396 ð.011Þ .195 MI
NOTE.—Each column represents a separate regression. Fixed effects: Y 5 year, M 5 municipality, I 5 five-digit NACE Industry. All regressions control for establishment size dummies of 10 employee intervals. The establishment immigrant share excludes the manager and the new hires. Sample is establishments with 2–50 employees. Observations 5 757,278. Standard errors robust for clustering at the establishment level are shown in parentheses. Mean dependent variable is .064.
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
416
˚ slund et al. A
tion of year, municipality, and five-digit ðNACEÞ industry. The five-digit industry codes are very detailed, and identification therefore comes from, for example, comparisons of hiring patterns between different pharmacies ðcode 52310Þ or taxi businesses ð60220Þ located in the same area ðthe average municipality has 30,000 residentsÞ. The estimate of 0.083 implies that an immigrant manager is more than twice as likely to hire an immigrant compared to a native manager. Thus, manager’s origin is highly correlated with the origin of entrants even between establishments that both produce similar goods or services and recruit their workers from the same local labor market. In order to account for remaining unobservable confounders, columns 4–6 of table 2 include the share of immigrants among the establishments’ incumbent employees ðexcluding the manager and entrantsÞ as a covariate. This measure of workforce composition serves as a proxy for omitted establishment-specific factors, for example, customer preferences, and, not surprisingly, it strongly predicts hiring patterns. The coefficient of interest is also reduced, but the estimate is still large ð0.066Þ and highly significant.14 Hence, also when we compare two firms in the same industry, year, and geographic area and also take into account the demographic composition of the incumbent workers, the probability that the newly hired is an immigrant is almost 80% higher if the manager is also of immigrant origin.15 The presented results all exclude establishments in firms run by selfemployed workers. The estimates are much larger when these are included, suggesting that self-employed workers to a very large extent hire workers of the same origin ðtable A3Þ. However, since it is likely that many of these firms essentially are family-run businesses where the hiring process may be very different from the rest of the economy, we focus on the sample without owners throughout the rest of this article. When interpreting these results, it is important to note that they do not imply perfect, or even deterministically increasing, segregation over time. The reason is that the job durations of both managers and workers are finite, and the sorting is less than perfect. Thus, even if firms with an immigrant manager ðand a high share of immigrant workersÞ tend to hire more immigrants, they will not necessarily end up having a homogenous 14 We have also verified that the origin of the manager is significantly more important than the origin of other workers by reestimating the model with the manager included in the share of incumbent immigrant workers. 15 Interestingly, our estimated effect from the share of immigrant coworkers is not far off from what Dustmann et al. ð2011Þ found for Germany in a similar specification. Replacing municipalities with neighborhood indicators ð“SAMS”Þ, which on average contain a population of 1,000 inhabitants, does not alter the conclusions. The estimate corresponding to col. 5 of table 2 is 0.052 ð0.005Þ.
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
How Immigrants and Natives Manage in the Labor Market
417
workforce since some workers will leave and some of the workers who replace those who leave will be natives. C. Robustness Checks Even though the cross-sectional specifications presented above are quite rich, unobserved characteristics that are correlated both with the origin of the manager and the hire may still be a concern. To test the validity of our results, we perform a series of robustness tests addressing endogenous establishment selection of managers as well as trends and shocks in hiring patterns. In addition, we analyze potential concerns regarding the strategy we use in order to identify managers. First, to handle unobserved factors at the establishment level, we estimate specifications relying on establishments that changed from native to immigrant management ðor vice versaÞ during our sample period. Second, we discuss specifications analyzing the role of trends within these establishments. Third, in order to remove ðpotentially year-specificÞ unobserved heterogeneity at the firm level, we use data from firms with multiple establishments in the same location. The different samples are similar in terms of establishment size but differ in industry composition ðtable A2Þ. Establishments that change manager origin are, for example, located in more immigrant-dense areas. For comparison, we therefore also show baseline model estimates on the firm and establishment fixed effects samples. We have also established that the results are robust to using a conditional logit model to handle the various sets of fixed effects instead of relying on the linear probability model. We do, however, choose to focus the discussion on the estimates from the linear probability models since the interpretation of the magnitudes is more straightforward.16 1. Establishment Fixed Effects The identifying sample for the establishment fixed effects model only includes establishments where a native manager is replaced by an immigrant manager or vice versa. This reduces the sample size substantially. To reduce the risk of mismeasuring manager changes ðparticularly in the “highest wage” categoryÞ, we restrict the sample to establishments that change manager origin only once during the observation period, and we also require that each manager is observed more than 1 year. This restriction removes about half
16 The point estimates ðSEÞ from the conditional logit corresponding to cols. 1, 3, and 5 of table 3 are 0.344 ð0.039Þ, 0.353 ð0.115Þ, and 0.412 ð0.122Þ. Note, however, that the magnitudes are not directly comparable to the estimates reported in table 3.
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
˚ slund et al. A
418
of the changes in manager origin.17 Due to the thinness of the establishment fixed effects sample, we base the comparison with table 2 on the model without the very data-demanding regional five-digit industry interactions ði.e., the estimate corresponds to col. 5 in table 2Þ. The point estimate for the establishment fixed effect sample is somewhat smaller ð0.039Þ than for the full sample ð0.051Þ, but the estimate is still substantial and the difference is not statistically significant. We then include establishment fixed effects to remove all time-constant establishment characteristics. However, this exposes the model to potentially confounding regional changes over time. We therefore include the fraction of immigrant employees in other establishments in the same year, municipality, and five-digit industry in order to account for regional shocks at the industry level.18 The estimates reported in column 3 of table 2 show that establishments that change manager origin alter their recruiting patterns by 4.1 percentage points in favor of workers with the origin of the new manager. This is not far from the baseline model estimates for the same sample ðcol. 2Þ, which supports the plausibility of the baseline results presented above. We have also estimated both the baseline model and the establishment fixed effects model by management type. The results are presented in the lower panel of table 3. Reassuringly, we find large and significant effects for all types of managers. 2. Establishment-Level Trends A potential remaining concern is establishment-specific trends in hiring patterns that may generate a spurious relationship between the origin of the manager and the origin of newly hired workers. Note, however, that such trends must indeed be specific to the very organization, since we control for changes at the regional-times-industry level. A potential concern could be that establishments with an increasing frequency of immigrant ðnativeÞ entrants more often will change into immigrant ðnativeÞ management, which would introduce an upward bias to our estimates. To address this concern, we have performed a number of robustness tests. First, we conducted a placebo exercise in the fixed effects model, using only pre-change data and assuming that the manager change took
17 Because we have data on a yearly basis, we are unable to observe whether management changed before or after other jobs were filled that year. We therefore follow Cardoso and Winter-Ebmer ð2010Þ and exclude the year of the management change. 18 The model does not include the share of immigrants among incumbents since the establishment fixed effect presumably captures the unobserved heterogeneity and to avoid exposing the model to concerns regarding lagged dependent variables in fixed effects models estimated on short panels.
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
How Immigrants and Natives Manage in the Labor Market
419
Table 3 Cross-Sample Comparison of Baseline Estimates Probability That Job Is Filled by an Immigrant Baseline Model ð1Þ Sample
Baseline Model ð2Þ
Establishment Fixed Effects Model ð3Þ
Baseline Model ð4Þ
Full Establishments Establishments Private firms sample that change that change with multiple manager manager establishments A. Immigrant manager .051 .037 .041 .051 ð.006Þ ð.014Þ ð.013Þ ð.013Þ B. Effect by manager type: Top manager .063 .038 .069 .057 ð.016Þ ð.028Þ ð.030Þ ð.026Þ Middle manager .034 .023 .051 .025 ð.016Þ ð.030Þ ð.030Þ ð.024Þ Highest wage .051 .042 .031 .063 ð.007Þ ð.020Þ ð.016Þ ð.020Þ Observations 757,278 7,468 7,468 155,095 Establishment immigrant share Yes Yes Yes YMI immigrant share Yes Fixed effects YMI Y M Y, E YMI
Firm Fixed Effects Model ð5Þ Private firms with multiple establishments .044 ð.014Þ
.043 ð.027Þ .025 ð.030Þ .054 ð.021Þ 155,095 Yes
YMIF
NOTE.—Each column represents a separate regression. Fixed effects: Y 5 year, M 5 municipality, I 5 five-digit NACE Industry, E 5 establishment, F 5 firm. All regressions control for establishment size dummies of 10 employee intervals and year fixed effects. The share of immigrants is the share when the manager is excluded. Sample is establishments with 2–50 employees. The coefficients in the lower part of the table are obtained from interactions with dummies indicating the type of manager. Standard errors robust for clustering at the establishment level are shown in parentheses.
place 1 year prior to the actual change. The estimate is very small ð0.004Þ and insignificant. Second, we estimated various models with linear trends centered on the year of the manager change for the interval ½26, 6. The results are displayed in table A4. The models, which allow for different slopes depending on the direction of the change ðimmigrant to native or vice versaÞ, produce results that are well in line with the baseline estimates. We also reestimated the model allowing the slopes to differ before-after the change ðby direction, thus four trendsÞ following the setup of regression discontinuity models. Finally, we also tried estimating models with establishment fixed effects and a set of establishment-specific linear trends. The estimates are robust to these variations with one exception; when including both establishment fixed effects and specific trends, the estimates become sensitive to the source of manager classification. When managers classified from having the highest wage are included in the sample, estimates are close to zero and insignifi-
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
˚ slund et al. A
420
cant. But focusing on the more reliable occupational codes ðtop and middle managersÞ, we get results that are almost identical to the baseline estimates for the same sample. Our interpretation of this pattern is that the very tight specification with establishment-level trends and fixed effects increases the role of measurement errors in the “highest wage” category. Altogether, however, we interpret the results as indicating that establishment-level trends should not be a crucial concern regarding the validity of the main estimates. 3. Firm Fixed Effects As an additional robustness test we have analyzed differences in hiring patterns between establishments that are run by the same firm. Here we are able to allow for arbitrary changes in firm-level recruitment patterns over time without imposing the linear ði.e., trendÞ functional form. The model includes fixed effects for the combination of year, firm, municipality, and industry of the establishment. Thus, we compare establishments with the same firm- and year-specific culture, involved in a similar production process, and located at the same local labor market. Given the year interaction, this specification fully accounts for unobserved time effects or trends at the firm level ðe.g., changes in a firm’s human resource policiesÞ. The specification also controls parametrically for differences in the composition of the workforce ðcf. table 2Þ to capture unobserved differences between establishments within the same firm. We exclude the public sector since the firm is a private sector concept. As shown by table A2 in appendix A, the used sample ðdefined as establishments belonging to private sector firms with multiple establishments operating within the same municipality and industryÞ is often found in consumer services, for example, retailers and banks. The estimate of ð0.044Þ presented in column 5 of table 3 is very similar and not statistically different from that of the baseline model for the corresponding sample. Again, we also see similar point estimates for different types of managers, although precision is an issue for individual estimates.19 D. Heterogeneous Effects This section presents results from models that rely on more detailed definitions of manager and worker origin and models that investigate whether the impact of worker-manager similarity varies by establishment and worker characteristics. We retrieve results from the regional-industry fixed effects model ðcf. col. 6 of table 2Þ as well as from the establishment fixed effects 19 As an additional robustness analysis we have reestimated the baseline model using only very large establishments ð1001Þ, and the effect there is very close to zero, which suggests that managers only matter when they actually are involved in the hiring decisions.
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
How Immigrants and Natives Manage in the Labor Market
421
model ðcf. col. 3 of table 3Þ. The estimates are presented in appendix tables A5 and A6. First ðsee table A5Þ, we split the foreign-born into three groups: Western Immigrants, Eastern-European Immigrants, and Non-European Immigrants. We estimate four linear probability models, one for the probability of hiring a worker of each of the four origin groups. In all cases we let the reference category be native managers. The first finding in the table is that originating in the same broad region increases the probability of recruitment; same-origin matches are always more prevalent relative to matches with other managers. In addition, with one exception, point estimates in columns 2–4 stay positive, suggesting that foreign-born in general tend to match also with foreign-born of descent other than their own. Many of the estimates are, however, small and insignificant. The most striking impression is that natives are much more likely to be hired by native managers than by managers of other origins and that the probability of a nonWestern worker being hired is much higher under non-Western compared to native management. Thus, it appears that differences between “true” natives and non-Western immigrants are an important driving force behind the results reported in the baseline two-group setting, which could reflect that the degree of dissimilarity is greatest between these groups. In table A6, we allow the effect to vary with respect to worker, establishment, and manager characteristics. Overall we find large, positive, and significant ðwhen using the baseline modelÞ effects from manager origin on the origin of entrants in establishments of all sizes and in most industries. The effects are similar for male and female workers and for workers with and without college education ðalthough somewhat stronger for low-educatedÞ. The absence of notable differences in the gender dimension remains when we instead consider manager gender.20 However, the results of the establishment fixed effects model suggest that similarity is more influential when low-educated managers recruit. One possible interpretation is that highly educated managers have a lower cost of digesting information to reduce uncertainty about dissimilar workers. We find indications of similarity biases in most parts of the economy; point estimates are positive in 11 out of 12 industries, whereof 7 are significant at the 5% level ðdetailed results are available on requestÞ. But there is also notable heterogeneity in the magnitudes of the impact: manager’s origin matters significantly more in the private than in the public sector despite the fact that recruitments in the Swedish public sector are as decentralized and informal as in the private sector. There are no direct institutional barriers preventing public sector managers from hiring workers 20
This result differs from some interesting experimental results on discrimination. Measuring discriminatory behavior in an experimental trust game, Fershtman and Gneezy ð2001Þ find that only men, and not women, engage in ethnic discrimination.
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
˚ slund et al. A
422
who are similar to themselves, and previous studies have also found evidence of ethnic discrimination in recruitments to public sector jobs in Swe˚ slund and Skans 2012Þ. den ðsee A Since Becker ð1957Þ, it is well known that product market competition is a factor that can drive out discriminatory behavior based on preferences, and for this reason it is interesting to see how our estimated effects vary with the market environment of the recruiting firms. To this end, we use a measure of market competition based on Boone ð2008Þ, which uses the within-industry elasticity of profits with respect to marginal costs. Thus, it captures how severely the market punishes inefficient firm behavior, which makes it suitable for the purpose of our study. The values of the indices are taken from Heyman, Svaleryd, and Vlachos ð2012Þ, which studies associations between product market competition, gender workforce composition, and firm takeovers in Sweden. We refer to that paper for details about the construction of the index.21 There is substantial variation in the competitiveness across industries: it is highest in manufacturing, retail, and the hotel and restaurant business and lowest in electricity, gas and water supply, and the financial sector.22 We interact manager origin with an indicator for whether the establishment operates in a high- or low-competition industry, separated by the median level of competition. The point estimates are larger for entrants in industries with more pronounced product market competition, a result that also holds in the establishment fixed effects model. Taken at face value, this indicates that the main effects may be driven by economic optimization rather than by preferences for similarity. E. Managers’ Previous Coworkers To investigate possible sources of the hiring bias documented above, this section analyzes the relationship between past establishment-level interactions and current hiring patterns. To this end, we use the data on former coworkers described in Section III. First, we examine whether having worked with immigrants in the past affects the probability of hiring a migrant worker. Second, we study hiring biases when managers hire from a pool of former coworkers. 21
Since we do not have access to the measure for all years in our observation period, we use the average competition across years between 1996 and 2002. Heyman et al. ð2012Þ report the rank correlations between competition and its 1-year lag and 12-year lag to be 0.89 and 0.72, respectively. Hence, the cross-industry pattern of product market competition is fairly stable over time. 22 Finer industry codes reveal that competition is highest in apparel and leather products ð18/19Þ, manufacturing of nonmetallic mineral products, and metals and pulp and paper products ð26/21/27Þ. It is lowest in electricity, gas, steam and water supply, collection and distribution of water and water transport ð40/41/61Þ, and financial intermediation ð65Þ, as well as renting of machinery and equipment without operator and of personal and household goods.
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
How Immigrants and Natives Manage in the Labor Market
423
As seen in table 1, immigrant managers on average have a much higher fraction of immigrants among their previous coworkers than native managers. Interactions in the establishment arguably bring valuable information that could make it easier to evaluate candidates in future recruitments and also increase one’s contact with a given group of workers. If this is true in the origin dimension, we would expect to see that managers are more likely to hire immigrants when they have interacted more with immigrants at previous jobs. To investigate this hypothesis, we construct a variable capturing the immigrant share of each manager’s former coworkers ðfrom 1985; see Sec. IIIÞ, which we include as a control in the industry and establishment fixed effects model. The results displayed in table A6 show that the past share of immigrant coworkers is positively associated with the probability of hiring an immigrant. Although the specification does not rule out the possibility that the result depends on unobserved heterogeneity on the manager side ðe.g., preferences for working with immigrantsÞ, the result is consistent with self-propagating segregation through on-the-job network formation. Importantly, though, most of the impact of manager’s origin remains after accounting for the composition of past coworkers. This means that differences in ðthis aspect ofÞ immigrant and native managers’ work histories are too small to explain ðin a statistical senseÞ the overall importance of manager origin. A related issue is how managers act when actually hiring former coworkers.23 Assuming that work-related information is revealed when working together, information asymmetries should be less of a concern in this case. In the empirical setup described below, we model the hiring decision as a choice from a given pool of workers encountered at a past establishment in a given year. This means that we pre-specify potential pools ðnetworksÞ from which the managers can recruit and analyze the selection within this pool. By including a fixed effect for each network, we focus identification to selection within these pre-specified networks. Our model estimates the probability of an individual worker being hired as a function of manager and worker origin. Formally, we estimate the following model: Prðworker i; from the managers network; is recruited to job jÞ 5 rEi 1 fEi Mj 1 mj ;
ð2Þ
where E is an indicator for whether worker i is an immigrant, M is an indicator for whether the manager who recruited for job j was an immigrant, and mj is a fixed effect for the manager-specific pool/network. The fixed effect is defined by the interaction between each job and each set of Among all new hires in our sample during the period 1997–2005, 4% ðaround 39,000Þ had previously worked together ðat an establishment with fewer than 100 employeesÞ with the recruiting manager. 23
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
424
˚ slund et al. A
previous colleagues from a past establishment of the hiring manager ðfor details on data creation, see Sec. IIIÞ.24 The fixed effects thus handle all unobserved differences that are shared between workers in the same network.25 The model does not include an explicit dummy for manager origin since it is captured by the network fixed effects. The coefficient of interest f measures whether similarity with the ðfutureÞ manager increases the probability to be hired relative to other workers within the same network.26 A positive f estimate would therefore indicate that managers are more likely to recruit former coworkers who are of similar origin even when information asymmetries are substantially reduced. Results presented in table 4 suggest that workers are more likely to be hired by managers they worked with in the past if they share the same origin. The effect is strongest when immigrant workers and managers are from the same source country group ðas indicated by the coefficient on “same immigrant country group”Þ, but there are also significant cross-group effects.27 It is reassuring to find that the results hold if we divide the networks by skill level by limiting the comparison to former colleagues of similar skill as the workerðsÞ actually recruited ðcol. 3Þ. An interesting result is the zero coefficients on the “worker immigrant” dummy ðcorresponding to r in eq. ½2Þ. This estimate suggests that immigrant workers, on average, are recruited as often as native workers when native managers hire former coworkers. This could be driven by a higher average propensity among immigrant workers to use job-based networks 24 If two managers are linked to and recruit from the same pool/network, there will thus be two fixed effects. Since we require variation within each fixed effect, the effective sample will be restricted to cases where someone, but not everyone, from the network was hired and where the pool/network contains both immigrants and natives. The data set includes all workers at past establishments meeting these criteria. 25 A similar approach is used in Cingano and Rosolia ð2012Þ, which studies how the employment rate in networks of former coworkers affect reemployment probabilities for newly displaced workers. The study accounts for selective sorting into establishments by including closing-firm fixed effects. 26 The strategy is similar to Bayer et al. ð2008Þ, which analyzes whether individuals are more likely to work together if they live in the same block than individuals who live in the same census tract but not in the same block. The analogue is that we treat the previous establishment as the census tract and estimate whether a worker who belongs to the same ethnicity ðblockÞ is more likely to follow the manager than a worker belonging to the same previous establishment ðtractÞ but not to the same ethnicity. Kramarz and Skans ð2011Þ uses a similar strategy to study whether parents are more likely to hire their own children than other individuals who belong to the same cohort and graduate from the same school, class, and field of study. 27 The variable “same immigrant country group” is defined from the 27 country groups reported in table A1.
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
How Immigrants and Natives Manage in the Labor Market
425
Table 4 Fixed Effects Estimates of the Effect of Origin among Former Colleagues All ð1Þ Worker and manager immigrants Same immigrant country group Worker immigrant Manager immigrant Fixed effects Observations R2
All ð2Þ
Same Skill Group ð3Þ
.048 ð.007Þ
.020 .021 ð.007Þ ð.010Þ .098 .111 ð.015Þ ð.020Þ 2.002 2.002 2.001 ð.003Þ ð.003Þ ð.004Þ Captured by establishment fixed effects E E ES 646,036 646,036 406,784 .209 .209 .134
NOTE.—Dependent variable is the probability that worker i, from the manager’s network, is recruited to job j. Each column represents a separate regression. Fixed effects: E 5 establishment, S 5 skill level. The sample includes all workers at establishments where at least one individual followed the manager to a new establishment, and the outcome variable is taking the value of one if the worker is the “follower” and zero otherwise. The last column includes establishment- skilllevel fixed effects, where “skill level” is a dummy taking a value of one if the individual has at least some college education and zero otherwise. A worker was hired by an immigrant ðnativeÞ former coworker ðnow managerÞ in 751 ð37,527Þ cases.
or it could be the result of unbiased recruitments when native managers hire their former coworkers. The latter interpretation would imply that the impact of similarity ðfÞ only appears because immigrant managers are more likely to recruit immigrant workers than native workers when they hire within their job-based network.28 VI. Wages and Separations In the introduction, we provided an overview of theories predicting manager-employee sorting based on similarity ðwe discuss these in more detail in the online app. BÞ. Because it is impossible to distinguish between these theories solely based on the hiring pattern analyzed above, we also examine how manager-employee similarity affects wages and worker turnover in order to shed further light on the underlying processes. The wage differences between immigrants and natives in the Swedish labor market are difficult to explain with observable differences in qual28 Assuming that our main effect in table 2 is symmetric, such that both immigrant and native managers are biased when the applicant pool does not consist of former coworkers, this result is consistent with information assymmetries reducing origin bias among native managers. As noted above, however, it is also possible that the effect is driven by employee behavior. If immigrant employees always have a higher propensity to follow former coworkers but native employees dislike working for immigrant managers, this could generate the zero effect under native management but positive effect under immigrant management.
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
˚ slund et al. A
426
ifications or occupational differences.29 Available evidence suggests that firm-specific factors determine a substantial fraction of average wage increases in Sweden ðsee, e.g., Skans, Edin, and Holmlund 2009; Carlsson, Messina, and Skans 2011Þ. We examine whether manager-employee similarity predicts the immigrant-native wage gap by estimating regressions of the following form: yijt 5 dMjt 1 vEi Mjt 1 jEi 1 Xjt b 1 εijt ;
ð3Þ
where yijt is either the log monthly wage of worker i in job j in year t or a dummy taking the value one if the worker leaves the firm in t 1 1 and does not return within 3 years, Ei is an indicator for whether worker i is an immigrant, Mjt is an indicator for whether the manager in job j at time t is an immigrant, Xjt is a vector of control variables, and εijt is the error term. The coefficient of main interest is v, which captures the impact of manageremployee similarity on wages ðor separationsÞ. Table 5 reports the results. Panel A reports the average impact of employee-manager similarity on wages and panel B shows the results on separations. The regressions are based on a sample of all employees in establishments with a workforce below the size restriction in our main analysis ð50 employeesÞ where managers are identified. Because separations are defined as workers who are not observed within the next 3 years, we focus the analysis on the period 1997–2002. The first column shows estimates where we compare wages for workers employed at the same time in the same local labor market ðwithin municipality, year, and five-digit industryÞ by managers of different origin. We account for worker characteristics ðgender, age, education, and five tenure-group dummiesÞ, five establishment-size categories, and the establishment immigrant share. We also examine whether there is a wage premium of similarity if immigrant managers and workers come from the same more narrowly defined country group ðsee table A1 in app. A for a list of countries within each groupÞ. The estimate on manager-worker similarity in the wage regression suggests that there is no association between similarity and wages. However, the estimate turns positive when we account for establishment fixed effects in column 2, and it is particularly strong if immigrant managers and workers are from the same group of countries.30 The estimate suggests that workers with the same origin as the man29 In the beginning of our observation period, the wage gap between natives and immigrants from outside Europe amounted to 22% ð8%Þ for men ðwomenÞ when adjusting for age, schooling, and experience and 11% ð6%Þ when also accounting for occupational differences. The difference between natives in immigrants from Europe was 10% ð2%Þ for men ðwomenÞ and 6% ð2%Þ within occupations ðLe Grand and Szulkin 2002Þ. 30 We have estimated models including same country group for Western Europeans as well. The wage effect of belonging to the same Western country group
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
How Immigrants and Natives Manage in the Labor Market
427
Table 5 The Effect of Similarity on Wages and Separations
A. Wages: Immigrant manager Immigrant worker Both immigrants Both immigrants, same immigrant country group B. Separations: Immigrant manager Immigrant worker Both immigrants Both immigrants, same immigrant country group Fixed effects
OLS ð1Þ
Establishment Fixed Effects ð2Þ
Worker Fixed Effects ð3Þ
Worker Fixed Effects 1 Occupation ð4Þ
2.009 ð.002Þ 2.067 ð.001Þ .008 ð.006Þ
2.004 ð.002Þ 2.065 ð.001Þ .015 ð.006Þ
2.001 ð.001Þ Absorbed by fixed effect 2.003 ð.003Þ
2.001 ð.001Þ Absorbed by fixed effect 2.003 ð.003Þ
.005 ð.013Þ
.030 ð.012Þ
.005 ð.003Þ
.005 ð.003Þ
.007 ð.003Þ .016 ð.001Þ 2.015 ð.005Þ
.003 ð.004Þ .016 ð.001Þ 2.009 ð.006Þ
2.001 ð.004Þ Absorbed by fixed effect .007 ð.011Þ
2.002 ð.004Þ Absorbed by fixed effect .007 ð.011Þ
2.012 ð.009Þ YMI
2.010 ð.010Þ Y, E
2.006 ð.021Þ Y, W
2.007 ð.021Þ Y, W
NOTE.—Each column represents a separate regression. Fixed effects: Y 5 year, M 5 municipality, I 5 five-digit NACE Industry, E 5 establishment, W 5 worker. Other controls include gender, age, age2, education level, five tenure-group dummies, five establishment size categories, and the establishment immigrant share. The dependent variable in panel A is the log monthly wage adjusted to full-time. The dependent variable in panel B is an indicator for whether the worker separates before year t 1 1. Standard errors robust for clustering at the establishment level are shown in parentheses. The number of observations is 2,954,007 ð1997–2002Þ. The separation rate is 0.065. Occupation is defined at the twodigit level ð31 occupation groupsÞ.
ager receive 1.5% higher wages than coworkers not sharing immigrant background with the manager and an additional wage premium of 3.0% if sharing the same country group origin. Column 3 includes worker fixed effects instead of the establishment fixed effects. This reduces the estimates to zero, suggesting that the previously estimated wage premium reflects worker sorting rather than differential wage compensation based on worker-manager similarity. The point estimate for the combined effect of belonging to the same immigrant group as the manager ði.e., the sum of the two estimates of interestÞ falls from 4.5% to 0.2% when individual fixed effects are accounted for, which suggests that sorting is a quantitatively important factor. In column 4, we estiis small and insignificant. For expositional reasons we therefore decided to leave these estimates out.
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
428
˚ slund et al. A
mate models with worker and occupational fixed effects, and the results are stable, suggesting that the zero effect is not due to counteracting occupational sorting. We have also estimated the model using both worker and establishment fixed effects, and the estimates are again very similar to those with only worker fixed effects.31 We have also examined whether the association varies over the employment spell by including an interaction between manager-employee similarity and employee tenure in job j. We see some tendency toward lower entry wages followed by a somewhat more rapid wage growth for workers sharing origin with the manager, although the differences are small in magnitude and not statistically significant. The results indicate that managers who recruit workers with a similar background attract workers with a higher ðpersistentÞ wage potential as captured by the fixed effect. We have also explicitly analyzed the “effect” of similarity on the workers pre-employment wage for the sample of job-to-job movers. Consistent with the results presented in table 5, we find that workers who are recruited by managers with a similar origin had higher wages in their previous jobs.32 The lower part of the table shows estimates of identical models where the outcome instead is the probability of separation. As for the wage estimates, column 1 shows a significant negative relationship between managerworker similarity and the probability of job separation. But accounting for establishment fixed effects does not have the same impact as for wages. In other words, establishments where migrants manage migrants are lowwage establishments, but they are not necessarily establishments with higher turnover.33 But, as with the wage analysis, accounting for worker fixed effects reduces the coefficient to zero. This finding suggests that managers who recruit workers with a similar background recruit workers with a lower tendency for mobility.34 In line with the wage regressions, these results are insensitive to the inclusion of occupational dummies and workers from the same The estimates from the two-way fixed effects model are 20.0002 ðimmigrant managerÞ, 20.0011 ðboth immigrantsÞ, and 0.0051 ðboth immigrants, same immigrant country groupÞ. 32 The association between the “both immigrant” dummy on the entrant’s wage in the previous job is 0.048 ½0.015 in a regression that accounts for entrant characteristics ðage, age2, education and genderÞ, establishment size, establishment fixed effects, entrant origin, and manager origin. The result is nearly the same if we condition on the occupation of the entrant. 33 The estimates not shown in the table reveal that immigrants have a significantly higher baseline turnover rate ð0.038 ½0.001Þ in the first column of table 5. Turnover is also higher among women and is decreasing with age and establishment tenure. 34 Unfortunately, we cannot distinguish between voluntary quits and firings. How˚ slund et al. 2009Þ we show that the impact is unlikely to ever, in the working paper ðA 31
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
How Immigrants and Natives Manage in the Labor Market
429
country group as the manager do not have lower turnover rates than other workers. In terms of interpreting the estimates, we find it interesting that none of the results suggest that the productivity cutoff is lower for workers who share a background with the manager. On the contrary, our results suggest that workers with a similar background on average have better unobserved characteristics, at least as measured by the individual fixed effects. The results thus appear to lean toward productivity concerns rather than preference discrimination as a driving force behind the results. Although it is difficult to pinpoint the exact source of potential productivity gains from hiring similar workers, the effects tend to suggest that similarity provides benefits in terms of selecting inherently more productive workers ðhigher wage potential and lower separation ratesÞ than in terms of finding workers with a higher idiosyncratic match quality since the effects disappear when we control for individual fixed effects. VII. Conclusion Managers play an important role in the labor market. Their actions affect the careers of individual workers and influence labor market segregation and inequality. We investigate the existence of in-group bias in this setting, specifically whether origin similarity matters for hirings, separations, and wages. Our analysis provides strong evidence that the manager’s origin matters for recruitment patterns. Jobs are disproportionately often filled by workers who share a background with the recruiting manager. This pattern holds in a large set of specifications, utilizing variation in several dimensions to control for observed and unobserved characteristics and trends, suggesting that we actually capture a causal effect of manager origin. These results are consistent with racial and ethnic hiring biases documented in singlefirm studies by Bandiera et al. ð2009Þ and Giuliano et al. ð2009Þ. We also present evidence suggesting that although past exposure to immigrant workers appears to be associated with future recruitment patterns, the differences in past exposure between immigrants and native managers are too small to be an important explanation for the observed recruitment patterns. In addition, we show that managers, on average, are more likely to recruit similar workers also when they recruit among the pool of former coworkers. Finally, we show that similarity matters for wages and separations but does so mainly through sorting. We find large wage and exit differences depending on the degree of similarity between workers and managers, be driven by downsizing firms: restricting the sample to establishments that hired at least one new worker in the following year yields very similar results.
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
430
˚ slund et al. A
but these differences disappear when we account for worker fixed effects. The pattern is most pronounced when managers hire workers belonging to the same country group; these workers earn a 4.5% wage premium relative to their coworkers, a difference that fully can be explained by individual fixed effects. In essence, the results imply that managers are able to attract workers with a higher market wage when they recruit similar workers but that workers do not extract any additional wage gains from similarity.35 Theory points to several potential mechanisms as to why in-group recruitment biases may arise. We attempt to empirically distinguish between two broad sets of mechanisms: preference-based behavior along the lines of Becker ð1957Þ, Black ð1995Þ, and Rose´n ð2003Þ and profit-maximizing concerns. The latter may by the result of direct productivity advantages arising from similarity ðLazear 1999Þ, statistical discrimination due to group-related information advantages ðAigner and Cain 1977; Fang and Moro 2011Þ, or information advantages regarding individual ðMontgomery 1991Þ or match-specific ðDustmann et al. 2011Þ productivity provided by social and professional networks following ethnic delineations. It is hard to pin down the exact relevance of these different theories for why managers are more likely to hire workers who share their origin. Several mechanisms related to the productivity of the match or the preferences of the agents could be at work at the same time. However, our results do leave some suggestive evidence regarding the relative importance of different explanations. Two observations speak against pure preference-based explanations. The first is that we find the bias to be larger in industries where firms face stronger product market competition. The second is that our evidence speaks against the notion that managers use a higher performance threshold when hiring dissimilar workers. On the contrary, wage results suggest that managers hire more workers with higher previous wages and better unobserved characteristics when they recruit workers with whom they share a background. Although separating between different potential sources of profitmaximizing benefits arising from in-group biased recruitments arguably is an even more difficult task than separating between profit-maximizing gains and preference-based hiring procedures, we interpret the evidence as being less in line with theories based on similarity-based idiosyncratic productivity gains than with stories based on the ability to select workers with a high productive capacity overall. Although we find that similar 35
The empirical literature on the wage impact of social job-search networks contains many studies reporting negative wage effects ðe.g., Antoninis 2006; Bentolila, Michelacci, and Suarez 2010Þ alongside studies reporting positive wage effects ðe.g., Loury 2006; Brown et al. 2012Þ from social network recruitments. In that sense the results are not unique.
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
How Immigrants and Natives Manage in the Labor Market
431
workers earn higher wages and have lower separation rates, these differences appear to be driven by time-constant differences between the workers and not by match-specific factors. Similarity is positively correlated with wages through the fixed effects in the wage regressions and negatively correlated with exits through the fixed effects in the exit regressions, suggesting that managers are able to detect or recruit better workers when they share origin with these workers. Our overall impression, therefore, is that the results point in the direction of employee selection models in the statistical discrimination ðFang and Moro 2011Þ or referral hiring ðe.g., Montgomery 1991Þ traditions. However, since our empirical setting partly builds on linking wage growth to job-to-job mobility, the mapping to the settings of these models is not perfect. But drawing on the patterns we document, we do concur with conclusions in the recent review of the personnel economics literature by Oyer and Schaefer ð2011Þ, who emphasize firms’ strategies for employee selection as a particularly fruitful area for future research. Regardless of the structural interpretation, a first-order implication of our results is, however, that the lack of access to key players in the labor market may explain some of the difficulties faced by immigrant workers. Increasing the representation of immigrants in managerial positions could therefore improve other immigrants’ employment prospects. Hence, promoting the careers of already employed immigrants may be an important complement to current integration policies, which nearly exclusively focus on getting the nonemployed into work. Appendix A Table A1 Countries and Regions Region Natives: Native Western
Countries Included
Share of Hires
0-Sweden Born outside Sweden: 1-Finland 2-Denmark 3-Norway 1 Iceland 4-Great Britain 1 Ireland 5-Germany 6-Mediterranean Europe ðGreece 1 Italy 1 Spain 1 Portugal 1 the Vatican 1 Monaco 1 Malta 1 San MarinoÞ
89.7
7-Other Europe ðAndorra 1 Belgium 1 France 1 Liechtenstein 1 Luxemburg 1 The Netherlands 1 Switzerland 1 AustriaÞ 8-United States 1 Canada
.2
1.9 .3 .5 .2 .3 .2
.1
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Table A1 (Continued ) Region Immigrants: Eastern Europe
Non-Western, non-Europe
Countries Included 9-Bosnia-Herzegovina 10-Former Yugoslavia ðYugoslavia 1 Croatia 1 Macedonia 1 SloveniaÞ 11-Poland 12-The Baltic states ðEstonia 1 Latvia 1 LithuaniaÞ 13-Eastern Europe 1 ðRumania 1 Former USSR 1 Bulgaria 1 AlbaniaÞ 14-Eastern Europe 2 ðHungary 1 The Former CzechoslovakiaÞ 15-Mexico and Central America 16-Chile 17-Other South America ðArgentina 1 Bolivia 1 Peru1Colombia1Uruguay1Ecuador1Guyana1 Paraguay 1 Surinam 1 VenezuelaÞ 18-African Horn ðEthiopia 1 Somalia 1 Sudan 1 DjiboutiÞ 19-North Africa 1 Middle East ðLebanon 1 Syria 1 Morocco 1 Tunisia 1 Egypt 1 Algeria 1 Israel 1 Palestine 1 Jordan 1 South Yemen 1 Yemen 1 United Arab Emirates 1 Kuwait 1 Bahrain 1 Qatar 1 Saudi Arabia 1 CyprusÞ 20-Other African ðall African countries not included elsewhereÞ 21-Iran 22-Iraq 23-Turkey 24-East Asia ð Japan 1 China 1 Korea 1 Hong Kong 1 TaiwanÞ 25-SoutheastAsiaðVietnam1Thailand1Philippines1 Malaysia 1 Laos 1 Burma 1 Indonesia 1 SingaporeÞ 26-Other Asia ðSri Lanka 1 Bangladesh 1 India 1 Afghanistan 1 Pakistan 1 Brunei 1 Bhutan 1 Kampuchea 1 the Maldives 1 Mongolia 1 Nepal 1 Oman 1 SikkimÞ 27-Oceania ðAustralia 1 New Zealand, etc.Þ
Share of Hires .7 .8 .5 .1 .4 .2 .1 .4 .3
.3 .6
.2 .8 .5 .3 .1 .3
.2
.0
432
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Table A2 Sample Statistics for Robustness Specifications
Sample Establishment size Immigrant share in five-digit industry by municipality New entrants/year Standard deviation Manager type: Top manager Middle manager Highest wage Industry: Agriculture, hunting, and forestry Fishing Mining and quarrying Manufacturing Electricity, gas, and water supply Construction Wholesale and retail sale Hotels and restaurants Transport, storage, and communication Financial intermediation Real estate, renting, and business activities Public administration and defense Education Health and social work Other community, social, and personal service activities Observations
Baseline Sample, All Establishments ð1Þ 25.6
Establishment Fixed Effects Firm Fixed Effects Sample, Sample, Private Establishments Firms with Multiple That Change Establishments Manager ð2Þ ð3Þ 24.4
24.2
.05 5.2 3.77
.05 4.7 3.59
.09 6.1 4.61
.15 .26 .58
.21 .38 .41
.18 .21 .62
.9 .0 .0 2.9
1.0 .0 .0 4.0
.4 .0 .0 1.4
.7 3.1 14.1 1.6
.3 4.6 40.7 3.0
.1 .8 18.9 3.1
4.6 3.8
4.9 16.9
2.5 2.5
6.4
12.0
4.7
5.3 18.4 31.3
3.0 5.9
3.9 12.1 44.2
6.9 757, 278
3.7 155,095
5.3 7,468
NOTE.—Column 1 reports sample characteristics for the overall sample of hires, whereas cols. 2 and 3 show the characteristics for the samples used in the robustness specifications reported in table 3 in the main text. The level of observation is the individual, and hence the table shows the fraction of new hires in each industry. The three samples correspond to 95,910 ðcol. 1Þ, 17,706 ðcol. 2Þ, and 372 ðcol. 3Þ establishments, respectively.
433
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Table A3 Results Including Owners Immigrant manager
ð1Þ
ð2Þ
ð3Þ
.371 ð.007Þ
.334 ð.006Þ
.247 ð.007Þ
Establishment immigrant share R2 Fixed effects
.060 Y
ð4Þ
ð5Þ
.136 .137 ð.004Þ ð.005Þ .607 .553 ð.006Þ ð.006Þ .088 .253 .130 .142 Y YM Y YM YMI
ð6Þ .124 ð.005Þ .465 ð.008Þ .274 MI
NOTE.—The dependent variable is the probability that the job is filled by an immigrant. Observations 5 833,383. Each column represents a separate regression. Based on data from 99,376 establishments. The share of immigrants excludes the manager. Fixed effects: Y 5 year, M 5 municipality, I 5 five-digit NACE Industry. All regressions control for establishment size dummies of 10 employee intervals. Standard errors robust for clustering at the establishment level are shown in parentheses. The mean dependent variable is .07.
434
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
435
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c). Y 2 trends
Y
Y 4 trends
Yes
Establishment trends .063 ð.021Þ 7,468 .080
Yes Y, E
Establishment FE .062 ð.019Þ 3,576 .224
Yes Y, E Establishment specific ðlinearÞ
Establishment FE 1 establishment trends .063 ð.032Þ 3,576 .293
Establishments That Change Manager, No Highest Wage Managers ð5Þ
NOTE.—The dependent variable is the probability that the job is filled by an immigrant. Each column represents a separate regression. Fixed effects: Y 5 year and E 5 establishment. Standard errors robust for clustering at the establishment level are shown in parentheses. All regressions control for establishment size dummies of 10-employee intervals and year fixed effects ðFEÞ. In these models the immigrant establishment share is a lagged dependent variable and is therefore not included. Instead we control for the fraction of immigrant employees in other establishments in the same municipality, industry ðfive-digitÞ, and year.
Yes
Establishment trends .065 ð.020Þ 7,468 .080
Yes
.048 ð.010Þ 7,468 .079
Immigrant manager
Observations R2 Establishment immigrant share Neighborhood immigrant share Fixed effects Trends
Baseline
Specification
Sample
Establishments That Change Manager, No Establishments That Establishments That Establishments That Highest Wage Change Manager Change Manager Change Manager Managers ð1Þ ð2Þ ð3Þ ð4Þ
Table A4 Establishment Fixed Effects with Trends
Table A5 Heterogeneity: Origin Groups Hired Worker’s Origin
Manager’s Origin
Native ð1Þ
Western Countries ð2Þ
Eastern Europe Non-Western, ð3Þ Non-Europe
A. Baseline model: Native Western countries Eastern Europe Non-Western, non-Europe Fixed effects R2 B. Establishment fixed effects model: Native Western countries Eastern Europe Non-Western, non-Europe Fixed effects R2
2.021 ð.004Þ 2.043 ð.008Þ 2.069 ð.009Þ YMI Y .195
2.045 ð.030Þ 2.042 ð.014Þ 2.018 ð.019Þ Y, E .227
.014 ð.002Þ .004 ð.004Þ .006 ð.004Þ MI .157
.002 ð.002Þ .023 ð.004Þ .006 ð.005Þ YMI .158
.006 ð.002Þ .016 ð.006Þ .056 ð.008Þ YMI .171
.016 ð.018Þ .020 ð.008Þ .004 ð.009Þ Y, E .162
.012 ð.027Þ .016 ð.011Þ .033 ð.017Þ Y, E .218
.017 ð.021Þ .006 ð.007Þ 2.018 ð.009Þ Y, E .116
NOTE.—Observations for panel A 5 757,278; observations for panel B 5 7,468. Each column represents a separate regression. Y 5 year, M 5 municipality, I 5 five-digit NACE Industry. Standard errors robust for clustering at the establishment level are shown in parentheses. The dependent variable indicates whether the new hire belongs to each of the groups in cols. 1– 4. All regressions control for establishment size dummies of 10-employee intervals. The specification in the upper panel is the same industry-fixedeffects specification as col. 6 in table 2. The specification in the lower panel corresponds to the one of col. 3 in table 3.
Table A6 Heterogeneity Baseline Sample, All Establishments Main effect Worker characteristics: Education level: College No college Gender: Male Female
Establishment Fixed Effects Sample, Establishments That Change Manager
Estimate
SE
Estimate
SE
.051
.006
.041
.013
.043 .059
.007 .008
.030 .052
.015 .018
.077 .042
.010 .006
.024 .048
.020 .015
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
How Immigrants and Natives Manage in the Labor Market
437
Table A6 (Continued )
Baseline Sample, All Establishments Establishment characteristics: Establishment size: 2–9 10–19 20–29 30–39 40– 49 Competition in two-digit industry:* High Low Sector: Private Public Manager characteristics: Education level: College No college Gender: Male Female Manager past interactions:y Immigrant manager ðbaselineÞ Immigrant share among manager’s former coworkers ðcontrolÞ Establishment size dummies Establishment immigrant share Neighborhood immigrant share Fixed effects
Establishment Fixed Effects Sample, Establishments That Change Manager
Estimate
SE
Estimate
SE
.079 .043 .058 .034 .051
.015 .010 .012 .012 .013
.074 .017 .064 .010 .086
.037 .021 .024 .028 .044
.069 .037
.009 .007
.048 .035
.018 .020
.067 .045
.011 .007
.091 .024
.026 .015
.051 .047
.007 .011
.029 .090
.015 .029
.047 .053
.008 .008
.040 .042
.018 .020
.037
.006
.040
.016
.023
.029
.227 Yes Yes YMI
.106 Yes Yes Y, E
NOTE.—Standard errors robust for clustering at the establishment level are shown in parentheses. Y 5 year, M 5 municipality, I 5 five-digit NACE Industry. The table reports results from seven integrated regressions. Thus the reported coefficients show the differential effect obtained from combining the main effect of manager origin and its interaction with the variable of interest. The standard errors are calculated using the nlcom command in STATA. All regressions control for establishment size dummies of 10employee intervals and year fixed effects. * Groups separated at the median level of the average industry competition between 1996 and 2002. y We include the share of each manager’s former coworkers as a control; hence, we do not interact this share with manager origin. The share is calculated using the working history from 1985. It excludes coworkers in the current establishment.
References Ahmed, Ali M. 2007. Group identity, social distance and intergroup bias. Journal of Economic Psychology 28, no. 3:324–37. Aigner, Dennis J., and Glen C. Cain. 1977. Statistical theories of discrimination in the labor market. Industrial Labor Relations Review 30, no. 2:175–87.
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
438
˚ slund et al. A
Antoninis, Manos. 2006. The wage effects from the use of personal contacts as hiring channels. Journal of Economic Behavior and Organization 59, no. 1:133– 46. ˚ slund, Olof, Lena Hensvik, and Oskar N. Skans. 2009. Seeking simiA larity: How immigrants and natives manage at the labor market. IFAU Working Paper no. 2009:24, IFAU, Uppsala, Sweden. ˚ slund, Olof, and Oskar N. Skans. 2010. Will I see you at work? Ethnic A workplace segregation in Sweden, 1985–2002. Industrial and Labor Relations Review 63, no. 3:471–93. ———. 2012. Do anonymous job application procedures level the playing field? Industrial and Labor Relations Review 65, no. 1:82–107. Bandiera, Oriana, Iwan Barankay, and Imran Rasul. 2009. Social connections and incentives in the workplace: Evidence from personnel data. Econometrica 77, no. 4:1047–94. Bayer, Patrick, Stephen Ross, and Giorgio Topa. 2008. Place of work and place of residence: Informal hiring networks and labor market outcomes. Journal of Political Economy 116, no. 6:1150–96. Beaman, Lori, and Jeremy Magruder. 2012. Who gets the job referral? Evidence from a social networks experiment. American Economic Review 102, no. 7:3574–93. Becker, Gary S. 1957. The economics of discrimination. Chicago: Chicago University Press. Bentolila, Samuel, Claudio Michelacci, and Javier Suarez. 2010. Social contacts and occupational choice. Economica 77, no. 305:20–45. Bertrand, Marianne, and Sendhil Mullainathan. 2004. Are Emily and Greg more employable than Lakisha and Jamal? A field experiment on labor market discrimination. American Economic Review 94, no. 4:991–1013. Black, Dan A. 1995. Discrimination in an equilibrium search model. Journal of Labor Economics 13, no. 2:309–34. Boone, Jan. 2008. A new way to measure competition. Economic Journal 118, no. 531:1245–61. Brown, Meta, Elizabeth Setren, and Giorgio Topa. 2012. Do informal referrals lead to better matches? Evidence from a firm’s employee referral system. Staff Reports 543, Federal Reserve Bank of New York. Calvo-Armengol, Antoni, and Matthew O. Jackson. 2004. The effects of social networks on employment and inequality. American Economic Review 94, no. 3:426–54. Cardoso, Ana R., and Rudolf Winter-Ebmer. 2010. Female-led firms and gender wage policies. Industrial and Labor Relations Review 64, no. 1: 143–63. Carlsson Mikael, Julian Messina, and Oskar N. Skans. 2011. Wage adjustment and productivity shocks. IZA Working Paper 5719, Institute for the Study of Labor ðIZAÞ, Bonn.
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
How Immigrants and Natives Manage in the Labor Market
439
Carlsson, Magnus, and Dan-Olof Rooth. 2007. Evidence of discrimination in the Swedish labor market using experimental data. Labour Economics 14, no. 4:16–29. Carrington, William J., and Kenneth R. Troske. 1995. Gender segregation in small firms. Journal of Human Resources 30, no. 3:503–33. ———. 1998. Interfirm segregation and the black/white wage gap. Journal of Labor Economics 16, no. 2:231–60. Cingano, Federico, and Alfonso Rosolia. 2012. People I know: Job search and social networks. Journal of Labor Economics 30, no. 2:291–332. Currarini, Sergio, Matthew O. Jackson, and Paolo Pin. 2009. An economic model of friendship: Homophily, minorities, and segregation. Econometrica 77, no. 4:1003–45. den Butter, Frank A. G., Enno Masurell, and Robert H. J. Mosch. 2004. The economics of co–ethnic employment. Discussion Paper 2004-027/3, Tinbergen Institute, Rotterdam. Dustmann, Christian, Albert Glitz, and Uta Scho¨nberg. 2011. Referralbased job search networks. IZA Working Paper 5777, Institute for the Study of Labor ðIZAÞ, Bonn. Edin, Per-Anders, and Peter Fredriksson. 2000. LINDA–Longitudinal INdividual DAta for Sweden. Working Paper 2000:19, Department of Economics, Uppsala University. Eriksson, Stefan. 2010. Utrikes fo¨dda pa˚ den svenska arbetsmarknaden ½Foreign born in the Swedish labor market. Bilaga 4 till La˚ngtidsutredningen 2011, in Va¨gen till arbete–Arbetsmarknadspolitik, utbildning och arbetsmarknadsintegration ½The road to work: Labor market policies, training, and labor market integration. SOU 2010:88, Fritzes, Stockholm. Fang, Hanming, and Andrea Moro. 2011. Theories of statistical discrimination and affirmative action: A survey. In Handbook of social economics, vol. 1A, ed. Jess Benhabib, Matthew O. Jackson, and Alberto Bisin, 133–200. Amsterdam: Elsevier. Fershtman, Chaim, and Uri Gneezy. 2001. Discrimination in a segmented society: An experimental approach. Quarterly Journal of Economics 116, no. 1:351–77. Giuliano, Laura, David I. Levine, and Jonathan Leonard. 2009. Manager race and the race of new hires. Journal of Labor Economics 27, no. 4:589– 631. ———. 2011. Racial bias in the manager-employee relationship: An analysis of quits, dismissals, and promotions at a large retail firm. Journal of Human Resources 46, no. 1:26–52. Granovetter, Mark S. 1973. The strength of weak ties. American Journal of Sociology 78, no. 6:1360–80. Hellerstein, Judith K., Melissa McInerney, and David Neumark. 2008a. Neighbors and co-workers: The impact of residential labor market net-
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
440
˚ slund et al. A
works. NBER Working Paper no. 14201, National Bureau of Economic Research, Cambridge, MA. ———. 2008b. Spatial mismatch or racial mismatch? Journal of Urban Economics. 64, no. 2:467–79. ———. 2009. Spatial mismatch, immigrant networks, and Hispanic employment in the United States. NBER Working Paper no. 15398, National Bureau of Economic Research, Cambridge, MA. Hellerstein, Judith K., and David Neumark. 2008. Workplace segregation in the United States: Race, ethnicity, and skill. Review of Economics and Statistics 90, no. 3:459–77. Hensvik, Lena. 2011. Manager impartiality: Worker-firm matching and the gender wage gap. IFAU Working Paper 2011:22, IFAU, Uppsala, Sweden. Heyman, Fredrik, Helena Svaleryd, and Jonas Vlachos. 2012. Competition, takeovers, and gender discrimination. Journal of Industrial and Labor Relations Review 66, no. 2:409–32. Hultin, Mia, and Ryszard Szulkin. 2003. Mechanisms of inequality, unequal access to organizational power, and the gender wage gap. European Sociological Review 19, no. 2:143–59. Ioannides, Yannis M., and Linda D. Loury. 2004. Job information, neighborhood effects, and inequality. Journal of Economic Literature 42, no. 4:1056–93. Kramarz, Francis, and Oskar N. Skans. 2011. With a little help from my . . . parents? Family networks and youth labor market entry. IFAU Working Paper 2011:18, IFAU, Uppsala, Sweden. Lazear, Edward P. 1999. Culture and language. Journal of Political Economy 107, no. 6:95–126. Le Grand, Carl, and Ryszard Szulkin. 2002. Permanent disadvantage or gradual integration: Explaining the immigrant-native earnings gap in Sweden. Labour 16, no. 1:37–64. Loury, Linda D. 2006. Some contacts are more equal than others: Information networks, job tenure, and wages. Journal of Labor Economics 24, no. 2:299–318. Mella, Orlando, and Irving Palm. 2009. Ma˚ngfaldsbarometern 2009. Attityderna till etnisk ma˚ngfald i Sverige ½Diversity Index: Attitudes toward ethnic diversity in Sweden, 2009. Department of Sociology, Uppsala University, Uppsala, Sweden. Montgomery, James D. 1991. Social networks and labor-market outcomes: Toward an economic analysis. American Economic Review 81, no. 5: 1407–18. Munshi, Kaivan. 2003. Networks in the modern economy: Mexican migrants in the U.S. labor market. Quarterly Journal of Economics 118, no. 2:549–99.
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
How Immigrants and Natives Manage in the Labor Market
441
Oyer, Paul, and Scott Schaefer. 2011. Personnel economics: Hiring and incentives. In Handbook of labor economics, ed. Orley Ashenfelter and David Card, 1769–1823. Amsterdam: Elsevier. ˚ sa. 2003. Search, bargaining, and employer discrimination. JourRose´n, A nal of Labor Economics 21, no. 4:807–29. Skans, Oskar N., Per-Anders Edin, and Bertil Holmlund. 2009. Wage dispersion between and within plants: Sweden, 1985–2000. In The structure of wages: An international comparison, ed. Edward P. Lazear, and Kathryn L. Shaw. Chicago: National Bureau of Economic Research and University of Chicago Press. Stoll, Michael A., Steven Raphael, and Harry J. Holzer. 2004. Black job applicants and the hiring officer’s race. Industrial and Labor Relations Review 57, no. 2:267–87.
This content downloaded from 130.238.166.078 on September 25, 2017 11:02:54 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
