Wage Effects of Search Methods for Immigrants and Natives: the Case of Sweden

[Preliminary work do not quote] January 16, 2006 Wage Effects of Search Methods for Immigrants and Natives: the Case of Sweden by Åsa Olli Segendorf and Dan-Olof Rooth Abstract: Using unique cross-section survey data collected in 1998 and 2003, this study examines whether successful job-search method differ between natives and immigrants, and whether there is a wage difference between the two groups associated with the search method used. It is found that the direct method is the successful search method for about 1/3 of the sample. However, informal methods are even more often the successful one for natives while formal methods are so for immigrants. About forty percent of natives (immigrants) find their job using informal (formal) methods. This holds for both men and women. Next, a wage analysis has been performed, which shows that there is a positive return from using the informal compared to using the formal method for natives. For the immigrant group the results are mixed. Further, when pooling the data for immigrants and natives a large wage discount is found when using the formal search method. A wage discount is found also for immigrant men using the informal search method. Such a wage discount does not apply for immigrants using the direct method. Hence, the study shows no evidence for the argument to increase job search through formal methods, such as public employment agencies. Instead the direct method, applying to companies without jobs add being posted, seems to be a fruitful alternative strategy. To what extent these results are due to selection into job search methods are yet unclear. JEL code: J15, J24, J31, J71 Keywords: Job search, Immigrants, Wage differences

1. Introduction A social network is a useful source of information about jobs, as well as establishing an important link between job-seekers and firms, and a considerable amount of research has been devoted to examining the impact of various search methods on job chances. The empirical evidence suggests that about fifty percent of workers obtain or hear about their jobs through an informal search method, i.e. through family and friends and previous colleagues. 1 However, the empirical evidence as regards the wages rewarded for these jobs is more mixed. For instance, Antoninis (2004) finds a positive wage premium for jobs mediated through informal/social networks while Bentolia et al. (2004) finds the opposite. One reason to expect a somewhat different results in the Swedish context is that every employer must by law report job vacancies centrally which potentially should decrease the problem with asymmetric information on the labor market. 2 Because of this, informal methods could potentially be less important for job search in Sweden. Furthermore, it appears that the productivity of the informal search methods varies between different ethnic groups in society (see Battu et al., 2004, for the UK labor market). The sometimes positive effect of using informal methods does not apply to all groups in society. Immigrants from countries whose culture is geographically and culturally far away from the host country are at a disadvantage when it comes to using informal methods, possibly due to the incidence of high unemployment and thereby low job-mediating efficiency within their own ranks (see for instance Rooth and Ekberg, 2003). The purpose of the present study is twofold. First we will examine whether the methods that were successful in getting the job-seeker a job differed between natives and immigrants. Second, it provides an empirical investigation of how the wage return for using different job search methods varies between natives and immigrants as well as between men and women. 3 To this end we use two unique interview samples that were collected in 1998 and 2003, respectively. Not only do these samples include information on wages and by which method the job was mediated but also includes information on a number of important background variables such as immigrants language skills and the type of occupation respondents are employed in. 1 See Rees (1966), Granovetter (1974), Holzer (1988), Gregg & Wadsworth (1996), Addison & Portugal (2001) and the references in these works. 2 This system does not apply perfectly to the labor market, but the access to information about job vacancies could be expected to be better in Sweden than in the US or the UK. 3 The data collected have focused on immigrant groups that are expected to be marginalized in the labor market, with a particular interest in the situation for immigrants born in non-western countries outside Europe. This will be described in greater detail in section 3. 1

Before proceeding we will state the three measures of job-search methods that are used in the present paper: formal methods, informal methods and direct applications. By getting a job by the formal method means that the job was mediated through the services of an impersonal intermediary, i.e. it was achieved through an employment agency or by answering an advertisement in a journal, a newspaper, on TV or radio, or on the internet. An informal method of job search means exploiting one's social networks or getting a job through an employer making the contact. Thus friends, family, previous co-workers or previous employers are all included in an informal network. Direct application means that the jobseeker makes direct contact with the firm without there being an advertisement about a job. Our results show that informal methods are most important in gaining jobs also in the Swedish context but more so for natives than for immigrants. About forty percent of natives gained a job using an informal method while twenty percent gained a job using a formal method. For immigrants formal and informal methods are equally important in gaining jobs. About one third gained a job using an informal method and another one third gained a job using a formal method. In addition approximately one third, in both ethnic groups, gained a job using the direct method. Further, using the data collected in 1998 indicate that native men receive on average fourteen percent higher wages when getting the job through informal rather than formal or direct methods, while immigrant women receive on average twenty percent higher wages when getting the job through direct method rather than formal or informal methods. However, such a difference in the return to using different search methods is not found for native women or immigrant men. The data collected in 2003 indicate that the return to informal methods compared to formal or direct methods have changed over time for immigrant men. As for the 1998 data, those who started in their present job between 1994 and 1998 and got their job through the informal search method show no sign of having higher wages compared to those who during the same period got their job through the formal or direct method. But those who started their present job between 1999 and 2003 and got their job through the informal method have ten percent higher wages than those who got their job through the formal or direct method. Whether informal methods being more important when gaining the job in the second period is a result of more positive macroeconomic conditions in this period or due to the group being 2

more integrated thus having a larger network is unresolved. These results remain stable when human capital variables and other individual characteristics are controlled for. 4 The paper is organized as follows. In Section 2 a review of the research is presented. In section 3 the data, descriptive statistics and the search method that led to the job are presented. Section 4 analyses wage differences between natives and immigrants in relation to the search method they used. Section 5 includes a discussion and offers some conclusions. 2. A Review of the Research Previous research on job search methods has been geared mainly to U.S. and British samples, and has addressed two questions: first, are informal job-search methods more efficient than formal methods in getting a job and, second, is the quality of the job match better when the employer recruits people through informal rather than formal methods. The studies concerning the efficiency of informal methods have found in the main that such methods generate more job offers and reduce the duration of unemployment significantly more than other methods do (Holzer,1987, 1988; Blau & Robins, 1990; Osberg, 1993, Lindenboom et al, 1994; Addison & Portugal, 2001; Wahba & Zenou, 2003 and Bentolila et al, 2004). As regards the quality of the job match acquired through informal methods, two points are particularly focused: the ability of the worker, and the wage the worker receives. As regards the ability of the worker both the theoretical predictions and the empirical evidence are, again, fairly clear. 5 The use of social contacts in hiring practices provides more able workers than other channels, since they provide more precise and, sometimes, better information about the worker. 6 Other papers examine the effect of the methods used on earnings/wages. However, these results are less clear than the quality of the job match. Granovetter (1974), Corcoran et al 4 It should be clear to the reader already at this point that we probably do not estimate the causal effect of using a specific search method since the choice of method might be endogenous. For instance social skill or ability (which are unobserved to us) might stimulate the creation of networks and to have, through the network variable, a separate effect on wages. The common solution to this endogeneity problem is to find an instrument, but such are rarely available. The 2003 data had in fact one potential instrumental variable. The respondents in this data are refugee immigrants and were randomly distributed across Swedish municipalities upon arrival (see Åslund and Rooth, 2005, for a study using this exogenous variation). Hence, the number of people of the respondents own nationality in their first location, which is exogenous due to the placement scheme, could then be used as an instrument for the type of method that got the job. Unfortunately, we did not find a correlation between the size of their own nationality and the method that got the job. Instead, we have relied on including specific control variables in order to reduce this potential endogeneity problem. 5 See the theoretical studies by Rees (1966), Saloner (1985), Snellman (1998) and Olli Segendorf (2005). 6 See Datcher (1983), Snellman (1998), Kugler (2002) and Calvo-Armengol & Jackson (2004). 3

(1980), Simon & Warner (1992), Kugler (2002), Arrow & Borzekowski (2004) and Antoninis (2004) all find a positive wage premium for jobs mediated by informal methods while Bentolila et al (2004), Antoninis (2004), Pistaferri (1999) and Pellizzari (2003) observe wage discounts. Very little research has studied if the benefit of using various search methods differs between (marginalized) immigrants and natives. Most of the empirical evidence about the effects of informal methods for ethnic groups concerns Whites, Blacks and Hispanics in the US, and the information acquired is not directly applicable to studies of immigrants, since although originally immigrants, the members of all three groups may well have lived in the US for many generations. However, a few studies have empirically studied job search by immigrants. Frijters et al (2003) examined native and immigrant male job-seekers in the UK. They found that the job search of immigrants in the UK is less successful than that of UK-born whites. The relative failure of immigrants to exit unemployment in their study cannot generally be explained by differences in the choice of their main job search method or in observable characteristics. On the contrary Battu et al (2004) find big differences in benefits from the methods used across immigrant/ethnic groups in the UK. They find that those who are least assimilated also benefit least from using informal networks as a job-search strategy. In a Swedish context the study by Behtoui (2004) utilizes an important comparison of results. His data covers the period 1992 to 1998 and he interviewed, in January each year, immigrants and natives who had found jobs within the last year. The data included information on the search method that obtained the job, the wage received as well as information on a number of human capital variables. He finds that the wage return to using informal methods is four percent for native men and two percent for native women compared to the formal method. For immigrant men and women, born in non-western countries, the corresponding numbers are four and one percent, respectively. Finally, Olli Segendorf (2005) finds that immigrants use informal methods when searching for jobs to a greater extent than natives. 4

3. Data and descriptive Statistics The empirical analysis utilizes two Swedish cross-section samples of which the first, the Nordic Network Project (NNP) was collected in 1998 and include 606 interviews used in this study, including both natives and immigrants of both sexes. The second data set, Immigrant Labour market outcomes, Language skills and Social networks (ILLS), include 267 interviews with four immigrant groups and was collected in 2003. The two data sets are identical in the sense that they include questions on the search method that obtained the job as well as wage information. In both surveys employed individuals were asked: Which method was successful in getting you your present job?. Six possible answers were allowed: 1) through an ad, 2) through the official employment agency, 3) being contacted by the firm, 4) through friends and relatives, 5) by contacting the firm, and 6) other ways. Alternative 1 and 2 represent formal methods, alternatives 3 and 4 informal methods and alternative 5 is a direct method. Respondents could only choose one answer. The above information will be illustrated in a descriptive manner in section 3.3 and used as the explanatory variables at focus in the empirical analysis in section 4. The outcome variable in the empirical part of the paper will be the natural logarithm of the monthly wage for the NNP data and hourly wage for the ILLS data. We have chosen the type of income/wage measure most often used by the respondents in each study. Respondents reported their hourly, weekly, monthly or annual wages for the jobs they obtained, and these are translated into monthly/hourly wages. 7 3.1 The data sets In this section the two data sets will be explained at length. The Nordic Network Project NNP consists of individuals who were registered with the National Labor Market Board (AMS) and who had left unemployment for work during April and May 1998. Those aged between 25 and 60 were selected and were then interviewed in June 1998. The age restriction was adopted in order to reduce the number of people who left job search to study or due to 7 The NNP data lack information on hours worked and therefore the hourly and weekly wage was multiplied by the days and/or weeks that a month contains while the yearly wage was divided with twelve in order to correspond to a monthly wage. A fraction of 20 percent of the sample reported working only part time. For this group the calculation of the monthly wage might be somewhat biased. However, the results did not change when this group was excluded from the analysis. Hence, the empirical analysis therefore includes a control for working part time. The ILLS data contain full information on hours worked which makes it possible to calculate an hourly wage. 5

early retirement. The initial sample included 2,000 persons. However, the final sample used in this study was reduced to 606 persons. The main reason for the reduction in the sample is that some people did not answer: the non-response rate was 33.4 percent. Second, there was excess coverage: individuals who only changed their category of search but did not actually exit unemployment were included. Excess coverage amounted to 33.4 percent which gave a final sample consisting of 888 individuals. Third, not all the observations included full information about the variables of interest. Some respondents reported zero wages and stated that they were unemployed despite the fact that they had reported a successful job-search method. This suggests that the job they obtained lasted only a very short time, which reduced the sample even further. However, an analysis of the drop-outs does not show any selection as regards sex, age, civil status or education compared to the sample being analysed. 8 Only immigrants that were born outside Europe (excluding those who have immigrated from North-America or Oceania) were included, since this is the most marginalized group on the Swedish labour market. However, it would have been desirable to divide this group into subgroups, since it is well documented that there are differences in labor market outcomes across various ethnic groups (see Rooth and Ekberg, 2003). This has not been possible, due to the small sample size. Immigrant Labor market outcomes, Language skills and Social networks In the spring of 2003, 559 interviews were carried out within the ILLS project. The primary purpose was to measure language skills and social networks among four immigrant groups. The interviewees are men who obtained a permanent residence permit in Sweden during the 1987-89 period for refugee reasons, who had immigrated from Ethiopia/Eritrea, Chile, Iran or Romania and who were between 20 and 40 years of age at the time of immigration. Thus, at the time of the interview the group had spent at least fourteen years in Sweden. The dropout rate of ILLS is relatively large, its participant frequency being only fifty-one percent. 9 The lowest participant frequency was that of the Ethiopians, whose participation rate was only thirty-seven percent. By comparing data from different databases it has, however, been possible to analyze to what extent the dropout tends to be selective (see Rooth, 2004). The analysis shows that among those who have chosen to participate in the study the figures for educational background and earlier position in the labour market are somewhat more 8 For a detailed examination of the sample drop-outs, see Olli Segendorf (2005). 9 The total number of interviewees was 559 individuals out of 1,100 selected. The sample is stratified along country of origin randomly selecting 275 individuals from each country. All calculations have taken this sampling strategy into account using sampling weights. 6

positive. However, the differences are fairly small. The analysis also shows a great similarity in observable outcome variables (like the employment/unemployment rate) for the participating group compared to several other immigrant groups with a similar background. It is therefore likely that the results of this study are representative of the immigrant groups investigated. The ILLS data depart from the NNP data in that some individuals have been employed for several years when being interviewed in 2003 while the individuals in the NNP data just recently got their job. Hence, these individuals differ as regards job tenure and therefore the data is divided between those who got their present job between 1994 and 1998 (which is expected to correspond to the NNP period) and those who got their present job between 1999 and 2004. This division makes it possible to study whether the wage return to using specific search methods have changed over time and makes a comparison of the two data sets more transparent. 10 The final sample used in this study was reduced to 267 persons. First, the main reason for the reduction in the sample is that some people did not have a job in 2003 which reduced the sample to 388 individuals. Out of these another 121 individuals where employed prior to 1994 which gives us the final sample. 3.2 Common and specific covariates In addition to self-reported information on wages and search methods both data sets contain information on individual characteristics such as age, gender (ILLS only includes men), country/region of birth 11 (country/region fixed effects), years since migration, region of residence (county fixed effects) and the highest education level completed (divided into three broad categories; added as fixed effects). These variables are referred to as the common Xs, or just CX, in the empirical section since they are common to both data sets. In addition the NNP data includes information on being married (being 1 if yes), having children under seven years of age (being 1 if yes), previous unemployment experience (being 1 if yes), previous job experience (being 1 if yes), unemployment benefits (being 1 if receiving such), time in last category of search (in months), and working only part time (being 1 if yes). 10 In the empirical section we discuss whether the observed difference over time is due to a changing labor market (that the labor market changes or that the migrants become more integrated) or are due to a different composition of the sample in the two periods. 11 Separating the data into being born in Africa, Asia, South America and in the Soviet union for the NNP data and separating the data into being born in Ethiopia, Chile, Iran or Romania for the ILLS data. 7

The ILLS data includes information on job tenure (yearly fixed effects), language skills (measured as good or not good by the interviewer) and the type of occupation for the present job (divided into high-, medium- and low-skilled occupations; added as fixed effects). These variables which are specific to each data set are referred to as the specific Xs, or just SX, in the empirical section. These additional regressors are included to explain the difference in the wages obtained as between natives and immigrants and to help to break any existing correlation between the error term and the search method variable. 12 A thorough discussion of how these variables might affect wages is found in Olli Segendorf (2005). Table A1 presents summary statistics on these covariates for all individuals in the sample, and separately for natives and immigrants. There are significant differences between the two groups in almost all variables. In the NNP data set immigrants are more often married and more often have at least one child under seven. More immigrants have elementary school, and fewer have secondary school, as their highest level of education. On the other hand, a larger proportion of the immigrants have had a university education, namely a quarter of them compared to a fifth for the natives. A considerable proportion of the natives have been unemployed before: as many as threequarters of the sample, while the proportion for the immigrants is smaller, around 60 percent. As to previous job experience, the same pattern emerges, whereby almost all the natives have had a job before while a fifth of the immigrant group had not. A larger proportion of natives live in areas that are not dominated by a big city, and had received unemployment benefit while they were unemployed. Therefore it is important to standardize the regressions by taking these variables into account. The ILLS data will not be commented on as regards differences towards natives in the NNP dat set. Instead we focus on differences within this data set. Those who got their job after 1998 have a wider distribution as regards education, they are to a greater extent located in the Stockholm region, and consists to a larger extent of immigrants from Chile and Iran compared to those who got their job before 1999. No matter what year they got their job only a tiny fraction, less than seven percent, have a low skilled occupation while thirty percent have a high skilled occupation. Also, about seventy percent speak the Swedish language well. 12 Or at least make it possible to gain knowledge on how sensitive the results are to different empirical specifications. 8

3.3 How do they find their present job? As mentioned in the introduction, formal methods imply that jobs are mediated through the services of an impersonal intermediary and informal methods imply that the job-seekers make use of their social networks or get their jobs through the employer. This is the measure of informal methods used predominantly in the literature. Direct application means that the jobseeker contacts the firm directly. The category direct application also covers respondents who answered that they used some method other than those listed. 13 There are several hypotheses as to why differences could be expected between natives and immigrants in the use of search methods. Formal methods could favour immigrants since they involve more objective criteria, and fewer subjective ones, whereby applicants are judged compared to informal methods. This is an argument for immigrants to expect higher wages when getting the job through formal methods. Formal methods could on the other hand be less productive for an immigrant if the employer employs discriminating behavior and thus select only natives when hiring. Informal methods are likely to be less favourable to immigrants relative to natives in one respect; the networks of contacts available to them may be less useful in the context of job search. Since their networks have a higher unemployment rate and their networks are thus more likely to be composed of other unemployed people. One the other hand the informal method might be fruitful in job search since the individual can get a personal recommendation. The results presented in Table 1 show that the major differences in successful job-search method in the NNP data lies in-between immigrants and natives and that men and women within their ethnical group show a similar hiring pattern. The single method that generated most jobs for all groups (in both data sets) was applying directly to the employer, which differs only slightly between the groups. About one third of the respondents in the NNP data, i.e. both native and immigrant men and women, found their job with the direct method. For immigrant men in the ILLS data this number is slightly smaller, about twenty-five percent. There might be several explanations to why the direct method is so efficient. For instance, some vacancies might only be registered internally (within the firm) awaiting the right job-seeker. 14 This gives the job-seeker who contacts the firm directly an advantage 13 Only four percent in the NNP data and none in the ILLS data. 14 There are anecdotal evidence showing that the hiring pattern of firms have changed since the economic crisis that started in 1992 in Sweden and that firms do not centrally register vacancies due to the large amount of applicants following an ad. This makes the direct application-method more important even though the empirical evidence of this is scarce apart from Olli Segendorf (2005) and the present study. 9

before those who only uses the employment agency and social networks as a job-seeking device. Table 1. The search method that lead to the job 1998 and 2003. Separate for men and women, and also for immigrants and natives. Percent. # # $ % & ' & $ ( ) ) * ' & & # # # ) # $ + $ $# Note: The ILLS data is divided between those who got their present job between 1994 and 1998 and those who got their present job between 1999 and 2004. # Further, about forty percent of immigrants in the ILLS data and about thirty percent in the NNP data (both men and women) got their job through a formal method while about one third got their job using the informal method. For natives (both men and women) twenty percent got their job through a formal method while as many as forty-seven percent of native men and forty percent of native women got their job using the informal method. 15 The difference between natives and immigrants in the method gaining the job is due to the fact that natives are more often contacted by the employer in gaining the job (informal search method) while immigrants more often have gained the job using the employment agency (formal search method). This fact indicate that immigrants have to rely more on formal methods due to a lower contact rate by previous employers compared to natives. In the introduction we discussed that informal methods might be less important in the Swedish labor market compared to, for instance, the US one due to employers being obliged to report vacancies centrally. While US studies (see section 2) report that about fifty percent of successful job search being mediated through the informal methods the number is 15 Olli Segendorf (2005) provides a crude measure of the efficiency of each method by comparing the share using a specific job search method and the share being successful using that method. Contacting the employer direct is found to be the most efficient method for both natives and immigrants with rates of 0.55 and 0.45 respectively. Answering advertisements is the least efficient method for both immigrants and natives with a jobsuccess/job-search rate of 0.11 and 0.27 respectively. These statistics are merely descriptive and no attempt is made to control for differences in individual characteristics. 10

approximately the same for natives in Sweden. This indicate that informal networks are as important in gaining jobs in the more regulated Swedish labor market as in the relatively unregulated US labor market. Before preceding to the empirical analysis it is important to remember that in the case of the two groups compared in this study, natives have higher employment rates and lower unemployment rates than immigrants. And since this study is concerned only with those who were successful in the job-search process, the group of immigrants born in non-western countries is likely to be more (positively) selective than the native group. To what extent this selective character of the data affect the results is not possible to analyse. The reader should be aware of this when interpreting the results. 16 16 In a similar manner the reader should be aware of that individuals born in countries outside Europe are overrepresented in the NNP data, due to the sampling scheme. Hence, the share of immigrants in this data set does not comply with the share of immigrants born in these regions in the overall population. 11

4. Analysis of Wage Effects of Search Methods As could be seen from the descriptive statistics in section 3, Table 1, the successful job search method for immigrants and natives, and men and women, showed major differences between the ethnical groups but similar patterns between men and women in the groups. Table A1 reweals that immigrants receive lower wages than natives no matter which search method they use. In the present section the wage returns from using various methods will be examined. The main emphasis will be on the return of search methods for natives respectively for immigrants and the examinations will be performed for each group separately. Still, it is of general interest to compare the two groups in the same frame, whereby this will also be done. First, in Model (1) shown in Table 2, we examine the wage difference for natives respectively immigrants using informal (INF) and direct application (DIR) methods without any common or specific controls to get a crude correlation between using the various methods and the wage. Second, in Model (2), those control variables that are common for both samples (CX) will be controlled for apart from search methods. This implies that it is Model (2) that is best suited in order to compare the NNP and the ILLS samples with each other. In Model (3) also those variables that are specific to each sample (SX) are accounted for. The empirical specification is thus: lnwage = β 0 + β1inf + β2dir + β3cx + β4sx where the wage measure corresponds to self-reported monthly/hourly wages. 4.1 The returns to informal compared to formal methods We estimate the returns to informal compared to formal methods separately for natives and immigrants. For native men the wage return from getting the job through an informal method compared to a formal method is 13.5 percent, see model (3), Table 2. This indicates that native men uses social contacts as a job search device and that these contacts yield a good return. The wage return for native women from using informal methods as compared to formal ones is smaller than what is found for men, with a return of 7.9 percent. Thus native women also successfully use social contacts when applying for jobs. However, women do not experience the same payoff from social contacts as men do. One explanation for this might be that more men are in positions in the workplace where they can influence the decision to hire 12

new personnel. Assuming that women networks contain mostly other women, these could be expected to be less productive than male networks. - Table 2 here Even though that for native women the return to informal methods, compared to formal methods, is similar to the ones for native men (but smaller) they are insignificant. The reason for this might be the small sample size which makes the standard errors large rather than that the return is small in size, i.e. with a larger sample size the standard errors could possibly be smaller and the return then being significantly different from zero. Hence, due to the small sample size we discuss sizeable estimates even if they are not significantly different from zero in a statistic sense. For immigrant men in the NNP-sample the return from getting the job through informal compared to formal methods is negative, -12.9 percent in Model (3) while immigrant women have a small positive return. Both these estimates are insignificant. This indicate that the social contacts immigrants possesses reward a low payoff in wages. Perhaps this is a result of a difference between immigrants and natives in how the job was mediated. As found in Table 1 a smaller share of immigrants got their job by being contacted by the employer and a larger share by getting the job through the employment agency compared to natives. Both these factors possibly act in lowering the productivity advantage between the informal and formal search methods for immigrants as compared to natives. For the ILLS-samples the return getting the job through informal methods compared to formal varies between minus 2.3 percent for those immigrants who got a job between 1994-98 to a positive 9.7 percent for those who got a job between 1999-2003. Thus, the return to informal search methods for immigrants who got their present job between 1999 and 2004 is similar to the return for natives in the NNP-sample. 17 One possible explanation for this disparity in results between immigrant groups in the two data sets is that they are heterogeneous as regards access to productive networks. A second possible explanation is that those immigrants in the ILLS-sample that got a job in the mid 90s and those in the NNP data experienced a greater downturn in the economy when searching for 17 The results, both for the informal and the direct method, are generally stable for other individual characteristics and human capital variables included in the model. Since the NNP-data does not include information about occupation the results for the different methods might be indicators for different search channels for different occupations, where the occupations themselves pay different wages. However, the ILLSdata do include type of occupation. To the extent that the results obtained from that sample are generalizable they indicate that the results are not affected by adding type of occupation to the list of regressors. 13

jobs than what the later cohort did. This implies that in a recession more people of ones network are unemployed or in firms that lay-off employees. Hence, such an explanation could also explain the difference between immigrants and natives in the NNP data if native network has a greater attachment to the labour market and therefore is less hurt by a downturn of the economy than immigrant network. A third possible explanation is that when searching for jobs immigrants with longer time of residence have become more integrated and has a larger network. 18 4.2 The returns to direct application methods compared to formal In this section we estimate the returns to using a direct application as compared to using formal methods, separately for natives and immigrants (see Table 2). The results for the direct application method show a small positive estimate, but insignificant, for native men and women and for immigrants in the ILLS data. For immigrant men in the NNP-sample there is instead a small negative effect, but insignificant. Interestingly a large positive estimate, and significant, is found for immigrant women in the NNP-sample using the direct method. It thus seems as if women born outside Europe have some special skill or special feature which is valuable to employers. This skill is known to the women who, without previously known contacts or the firm placing an ad, are hired. Further research into this area would be of interest. 4.3 Comparing wage differentials between natives and immigrants The NNP data set also provide a picture of wage differences between natives and immigrants within search methods, see Table 3. 19 In the following specification natives and immigrants are merged into one data set but separated between men and women. In the empirical specification, the influence of being an immigrant (IMM), the methods used (INF and DIR), interaction terms between immigrants and the method used (INF*IMM and DIR*IMM), those control variables that are to be found in both samples (CX) and finally those control variables that are specific to the NNP-sample (SX). The empirical specification is thus: lnwage = β 0 + β1imm + β2inf + β3dir + β4inf * IMM + β5dir * IMM + β6cx + β7sx 18 However, this latter explanation is not supported by the NNP data since the immigration year variable interacted with search methods is not significantly different from zero. 19 The specification in Table 3 is somewhat more restrictive than the specification in Table 2 since we have not allowed the parameter estimates to vary between immigrants and natives. Further, we do not control for region of birth since we only include one immigrant dummy variable. 14

The native-immigrant wage difference for men is approximately sixteen percent when obtaining the job through the informal method, about ten percent when obtaining the job through the formal method, and close to zero when obtaining the job by the direct method. For women, the pattern is somewhat different. The native-immigrant wage difference for women is approximately eighteen percent when obtaining the job through the formal method, close to zero when obtaining the job through the informal method, and a negative nineteen percent when obtaining the job by the direct method. - Table 3 here - Both immigrant men and women achieve lower wages getting the job through formal search methods. Hence, wage native-immigrant wage differences seem to arise using public employment agencies. Further, only immigrant men get lower wages getting the job through informal search methods while no such difference is found for women. This result for immigrant women is not in line with our expectations. Especially not since immigrant women to a much lesser extent than native women are being contacted by the employer in getting the job. Perhaps this result is due to the fact that men and women to a large extent are employed in different sectors of the economy (and in different occupations) and that the productivity of informal networks varies between natives and immigrants within occupations and/or between occupations. In this respect it would be desirable to study search method specific nativeimmigrant wage differences within occupations. However this is not possible with the present data. For the direct method we find that immigrant women receive higher wages than native women while we find no difference in the wage between immigrant and native men. In this respect, the direct method seems to be a fruitful way for immigrants to gain jobs in order to minimize wage differences toward natives. About one third have got their jobs through this method so the direct method does not only pay relatively higher wages to immigrants; it is also an important way of getting the job. Whether this is due to selection mechanisms in the sample or the recruiting structure of the firm is an issue for further investigation. 15

5. Conclusion In this paper we have investigated the successful job-search methods of natives and immigrants, and have looked at the wage premiums of using these search methods for the two groups. To explore these issues, sample statistics show the successful job search methods and OLS is used to examine differences in wages between the groups. Wage differentials by search method are interesting because they provide a reason for job-seekers preferring a particular search method. The empirical analysis utilizes two unique cross-section survey data collected during 1998 and 2003 in which individuals born outside Europe are overrepresented compared to the share they represent in the population. The results from the first part of the analysis show that a larger fraction of immigrants found jobs through formal and direct methods than with informal methods compared to natives. One third of all jobs were found by direct application methods, and every fifth by friends and relatives. As to the wage gain, the various methods were examined separately for natives and immigrants as well as all individuals together in the same regressions. There is an overall wage discount for male immigrants that persist after individual characteristics, humancapital variables and the successful job-search method are controlled for which serves to emphasize the poor payoff from their network. Native men get a higher wage when they use informal methods than when they use formal ones. Native women also get a higher wage from using informal methods but not to the same extent as men do. Whether it is differences in the expected payoff that evoke the propensity to search for jobs in certain manners for the two groups, or if it is the opposite, the search patterns that evoke the different payoffs is not possible to tell in this study. In order to establish the direction of the causality more data and information would be needed. Neither is it established if it is unobserved productivity that drives the results. That is, do native male worker earn more due to that they are more able than other worker, and thus social contacts are more prone to pass on information to these contacts? This paper does not provide evidence of the cause of the immigrants wage discount. Regardless of the cause, however, be it employer discrimination, inadequate language skill for immigrants or some other unobservable characteristics that explains the results, there is clearly an immigrant-penalty in Sweden. More detailed investigations of the causes of this are needed. 16

References: Åslund, O. and Rooth, D-O. (2005) Do When and Where Matter? Initial Labor Market Conditions and Immigrant Earnings, forthcoming in Economic Journal. Addison, J.T. and P. Portugal (2001) Job Search Methods and Outcomes, Oxford Economic Paper, 54, 505-533. Antononis, M. (2004) The Wage Effect from the Use of Personal Contacts as Hiring Channels, Centre for the Study of African Economies, No. 203. Arrow, J.A. and R. Borzekowski (2004) Limited Network Connections and the Distribution of Wages, Finance and Economics Discussion Series paper 2004-41. Battu, H., Seaman, P. and Y. Zenou (2004) Job Contact Networks and the Ethnic Minorities, Working Paper No. 628, The Research Institute of Industrial Economics, Stockholm. Behtoui, A. (2004) Informal Recruitment Methods and Disadvantages of Immigrants in the Swedish Labour Market, Department of Thematic Studies (Ethnic Studies), Linköping University. Bentolila, S., Michelacci, C. and J. Suárez (2004) Social Contacts and Occupational Choice, CEPR Discussion Paper No. 4308. Blau, D.M. and P.K. Robins (1990) Job Search Outcomes for the Employed and Unemployed, Journal of Political Economy, 98, 637-655. Calvó-Armengol, A. and M.O. Jackson (2004) The Effect of Social Networks on Employment and Inequality, American Economic Review, 94, 426-454. Corcoran, M., Datcher, L. and G.J. Duncan (1980) Most Workers Find Jobs Through Word of Mouth, Monthly Labor Review, 103, 33-35. Datcher, L. (1983) The Impact of Informal Networks on Quit Behavior, Review of Economics and Statistics, 55, 491-495. Frijters, P., Shields, M.A. and S. Wheatly-Price (2003) Immigrant Job Search in the UK: Evidence from Panel Data, IZA Discussion Paper No. 902. Granovetter, M. (1974) Getting a Job, Chicago: The University of Chicago Press. Gregg, P. and J. Wadensworth (1996) How Effective are State Employment Agencies? Jobcentre Use and Job Matching in Britain, Oxford Bulletin of Economic and Statistics, 58, 443-457. Holzer, H.J. (1987) Informal Job Search and Black Youth Unemployment, American Economic Review, 77, 446-452. Holzer, H.J. (1988) Search Method Use by the Unemployed Youth, Journal of Labor Economics, 6, 1-20. 17

Kugler, A.D. (2002) Employee Referrals and Efficency Wages, IZA Discussion Paper No. 633. Lindenboom, M., Van Ours, J. and G. Renes (1994) Matching Employers and Workers: An Empirical Analysis on the Effectiveness of Search, Oxford Economic Papers, 46, 45-67. Olli Segendorf, Å. (2005) Job Search Strategies and Wage Effects for Immigrants, Dissertation 65, Swedish Institute for Social Research, Stockholm University. Osberg, L. (1993) Fishing in Different pools: Job Search Strategies and Job-Finding Success in Canada in the Early 1980s, Journal of Labor Economics, 11, 348-386. Pellizzari, M. (2003) Do Friends and Relatives Really Help in Getting a Good Job?, Mimeo, London School of Economics. Pistaferri, L. (1999) Informal Networks in the Italian Labor Market, Giornale degli Economisti, 58, 355-375. Rees, A. (1966) Information Networks in the Labor Markets, American Economic Review, 56, 559-566. Rooth, D-O. and Ekberg, J. (2003) Unemployment and Earnings for Second Generation Immigrants in Sweden Ethnic Background and Parent Composition, Journal of Population Economics 16 (4), 787-814. Rooth, D-O. (2004) Selektion i IASS-studien, mimeo. Kalmar university. Saloner, G. (1985) Old Boy Networks as Screening Mechanisms, Journal of Labor Economics, 3, 255-267. Simon, C. and J. Warner (1992) Matchmaker, Matchmaker: The Effect of Old Boy Networks on Job Match Quality, Earning, and Tenure, Journal of Labor Economics, 10, 306-330. Snellman, K. (1998) Studies of Appointments and Incentives under Imperfect Information, Dissertation, Åbo Akademi University. Whaba, J. and Y. Zenou (2003) Density, Social Networks and Job Search Methods: Theory and Application to Egypt, Journal of Development Economics. 18

Appendix: Table 2 is found last. Table 3. NNP ),,,, ),,, ),, -). -). -. -). -)). -. + $,,, #,,, ) )$ -$. -$. -$#. -$. -$). -$$. / # $ ) -$. -$$. -$. -$. -$#. -$#., #,, #,, $ # + -. -. -. -. -#. -#., ),, #,,, / -. -#. -. -. -. -. % +% 0 & & & & % +0 & & 1 # + # # # Note: Set CX (common Xs) controls: Age, Age2, Education, Region of residence; set SX (specific Xs) controls: Married, Children under 7, unemployment experience, previous job experience, unemployment benefits, time in last category of search. 19

Table A1. Descriptives. Means. Standard deviations in parentheses. 2 3 4 4 3 3 - #. # $ # -. -. ))- ). )- $. + $-. % & $ # -. ' ' $ -. -. / # -$. % ' $ % ( $ -$. # -$. )-. ) * ( - ). $-). -. # -$). -#. -#. * 5 3 * & 6 )6 $6 6 $6 ) 6 & $ 6 6 # 6 6 # 6 6 7 & $# 6 )6 )6 # 6 $6 6 1 + 3 8 6 $6 6 6 # 6 6 9 $6 6 6 6 ))6 # 6 9 $# 6 6 # 6 6 ))6 6 : )-$. ))-#. -. # -. 1 + 3 ( $) 6 ) 6 ( + # )6 6 5 ( # )6 6 6 6 % # 6 6 * ' 6 6 6 6 1 $6 6 * 6 $6 6 # # 6 ; 5 $6 $6 )6 # 6 7 ' & < ' )$ 6 # $6 ) 6 # 6 2 5 < ' )6 $# 6 # 6 $ $6 7 ' & + 6 $ 6 ) 6 $$6 4 = 8 6 $)6 8 $# )6 $ 6 ; 8 6 6 ' 8 = )6 $6 + $ $# Note: Set CX (common Xs) controls: Age, Age2, Education, Region of residence; set SX (specific Xs) controls: Married, Children under 7, unemployment experience, previous job experience, unemployment benefits, time in last category of search. Wage measure in 1998 and 2003 respectively. 20 The following three regions are considered big city regions in Sweden. 20

Table A2. NNP. Only natives. ), -. -). -. + $,,, #,,, -$. -$. -$#. / # $ -$. -$$. -$., #,, #,, + -. -., ) / -. -#. -. % +% 0 & & % +0 & 1 + # # # Note: Set CX (common Xs) controls: Age, Age2, Education, Region of residence; set SX (specific Xs) controls: Married, Children under 7, unemployment experience, previous job experience, unemployment benefits, time in last category of search. 21

Table 2. + + ),, $, #,, #, ) ),, -$. -$). -$). - $. -. -. -$. -$. -# $. -. -#. - ). / # $ $ $ ) # ) # -$$. -). -). -). -. -. -). -). -$. -#. -# ). -#. % +% 0 & & & & & & & & % +0 & & & & 1 $ # # # $ $ + $ $ $ $# $# $#, + ) ) ) )# $ -#. -$. -$. -. - #. - ). / $ ) ),, #,, )#, -#. -$. -$. -. -. -. % +% 0 & & & & % +0 & & 1 $ $ + Note: Set CX (common Xs to both data sets) controls: Age, Age2, Education, Region of residence; set SX (specific Xs to each data set) controls: NNP: Married, Children under 7, unemployment experience, previous job experience, unemployment benefits, time in last category of search: ILLS:. The baseline model (Model 1-3) for immigrants also control for region of origin and years since migration. Important to note is that the immigrant model estimates in table 2 are not affected by controlling for region of origin or years since migration. This hold for both men and women. All regressions control for parttime work (NNP) or job tenure (ILLS).