The Assimilation of Balkan and Middle Eastern Refugees in Sweden

(1)

The Assimilation of Balkan and Middle Eastern Refugees in Sweden

- A comparative study of how the country of origin affects the labour market outcome

Authors:

Sofia Fritzson Nelly Sand

Supervisor: Thomas Ericson

Examiner: Mats Hammarstedt Term: VT20

Subject: Economics Level: Bachelor Course code: 2NA11E

(2)

Abstract

This paper investigates the assimilation of Balkan and Middle Eastern refugees in Sweden. The aim of this report is to examine whether an immigrant from a country that has similarities with the host country regarding language, institutions and culture, will have an easier and faster assimilation process compared to an immigrant originating from a more distant country. With the use of regression and linear probability analyses, Balkan immigrants show faster assimilation in employment and earnings than the immigrants originating from the Middle East.

By examining the genders separately, the Balkan males are seen to exhibit a much higher earnings assimilation rate than the Middle Eastern males, while the female rates portray less variation. The initial employment and earnings differentials, relative natives, however, tend to be larger for Balkan immigrants than Middle Eastern, for cohorts arriving before 1990. This trend changes to the opposite scenario after 1990, when Middle Eastern immigrant cohorts face larger disadvantages compared to Balkan groups. Less transferable skills and discrimination against Middle Eastern minority groups could potentially explain the results. Nonetheless, since both immigrant groups experience large disadvantages in earnings and employment at arrival in the country, reaching parity in earnings and employment rate with natives seems distant, regardless of gender and origin.

Keywords: Assimilation, Immigrants, Earnings, Employment, Balkan, Middle East, Sweden

(3)

Acknowledgements

First, we would like to address our gratitude to our supervisor, Thomas Ericson, and examiner, Mats Hammarstedt, for support and guidance throughout the writing process. Further, we send a special thanks to Lina Aldén and Chizheng Miao, who have been very helpful regarding the statistical work and whose guidance have been of much importance for the outcome of the thesis. Finally, we express our gratefulness to our discussants, Eglon Kaja and Jonas Schwerger, for giving their wise comments and advices that have contributed to the improvement of our work.

(4)

1. Introduction

Throughout recent decades the migration has been of great importance for various countries (Lundborg, 2013). This has raised the question of how well the immigrants are able to perform in the host countries’ labour markets. In Sweden, the immigrant composition has shifted from primary domination of European labour-migrants, consisting of immigrants from the Nordic countries, Eastern Europe and Southern Europe, to change towards a more heavily domination of refugee-migrants. These migrants have arrived from all over the world such as Asia, Latin America, Africa, and Europe. However, in recent decades the migration flow has consisted more heavily of Middle Eastern refugee-migrants (Hammarstedt & Shukur, 2007). The transformation from both country composition as well as immigrant type raises the question of how well these migrants are able to assimilate in comparison to European immigrants.

This paper investigates the assimilation in yearly earnings and employment rate of Balkan and Middle Eastern refugees in Sweden. As European countries have similarities regarding human capital, such as language, institutions and cultures, the transferability of skills is argued to be more easily conducted for the European immigrants than for the immigrants that originate from more distant countries (Chiswick & Miller, 2009). This suggests that the Balkan refugees will have an easier and faster assimilation process in Sweden than the Middle Eastern refugees. By doing a comparative study between different refugee-migrants, showing the impact of country of origin, as well as gender, we are able to provide further evidence to policymakers of how some groups may struggle more than others in their assimilation process, and therefore be in more need of assistance in order to positively integrate into the country. The remainder of this paper will proceed in the following way: Section 2 provides a historical overview of how immigration to Sweden has changed during the last decades. Section 3 gives a review of how the assimilation process has been investigated in previous studies. Section 4 describes the theoretical framework of the study. Section 5 provides the panel data and descriptive statistics.

Section 6 gives the methodology, introducing the specification of the models. Section 7 gives a presentation of the results attained. Finally, section 8 gives a discussion and analysis, which is then followed by a short summary and conclusion in section 9.

(6)

2. Historical overview

¹

After the end of World War II and onto the early 1970s, labour-migrants were of great importance for many European countries, Sweden included. As the Swedish industrial sector expanded, the demand for labour increased. This, in turn, attracted low-skilled migrants from Finland and Southern European countries, such as Italy, Greece, and primarily Yugoslavia, as well as some high-skilled migrants from Denmark and Norway, to the country. These migrants proved to easily assimilate into the labour market with high levels of employment and earnings in comparison to natives. There were also additional migrant flows during these times which consisted of East-European refugees who arrived between the 1950s and 1960s. These refugees showed similar positive outcomes regarding good assimilation and high performance in the labour market.

The combination of lower industrial and economic growth, a greater demand for high-skilled labour in the production, as well as stricter migration policy changes in the 1970s, led to the migrant flows slowly shifting from a majority of labour-migrants to a majority of refugees. In the 1970s and 1980s many refugees from Eastern Europe, Africa, Latin America and Asia entered the country. In response to the ongoing wars in the 1990s, such as the conflicts between Iraq and Iran and the Yugoslav wars, the migration flows then started to consist heavily of refugees from the Middle East, mainly Iraq, as well as from former Yugoslavia. During this period, Sweden also experienced a recession resulting in higher levels of unemployment among the workers. This decrease in labour demand had a negative influence on the employment rate of immigrants, as well as the likelihood of attaining employment decreased. In the 2000s, the migrant flows continued to be dominated by Middle Eastern refugees. Throughout the most recent decades, the migration flows have been seen to result in higher levels of unemployment than the previous flows, as well as a more difficult transition into the labour market. This has resulted in an overrepresentation of the immigrant population among low-income jobs. In the year of 2010, one-fifth of the population had a foreign background, in terms of being born in another country or having one or more foreign-born parents. This can be put into contrast to the year of 1940, when only one percent of the population was born outside of Sweden.

1 Historical overview is based upon Hammarstedt & Shukur (2006), Hammarstedt & Shukur (2007), Bevelander

(7)

3. Literature review

To examine the assimilation process of immigrants, most studies investigate the variation in employment and earnings in relation to natives over time.

In a study by Chiswick (1978) the assimilation process of immigrants in the United States is investigated. By analysing cross-section data, he finds that the initial earnings gap between immigrants and natives decline gradually after the time of arrival of the immigrant. The initial gap is argued to be due to the lack of host-country specific knowledge, implying that as time since migration goes by, the immigrants will make larger investments in human capital than natives and thereby increase their productivity and attain a steeper earnings growth. However, a drawback of this study is that it does not follow the same individuals over time and thereby assumes that newly arrived immigrants have a constant level of skills over time.

Borjas (1985) departs from the assimilation model developed by Chiswick and extend it by using two censuses of data and thereby examine cohorts of immigrants and their earnings assimilation processes. This usage of data makes it possible to study the assimilation of the same individuals over time, while at the same time observing the skill quality of different cohorts. The results show that the earnings growth within cohorts is smaller compared to the predictions from the cross-section estimations. As for the skill composition of arriving immigrants, there seem to be a decline in quality for the later cohorts in the study.

Dustmann and Frattini (2011) presents a study comparing employment probability and occupational distribution of European natives and immigrants from different countries.

Immigrants tend to have a disadvantage against natives, in general. The disadvantage seems to be even greater for non-European immigrants, which could be due to institutional and cultural barriers that make it difficult for non-European immigrants to enter the European labour market. The study shows that immigrants in Europe are on average less educated compared to natives, which influence the assimilation and especially the employment rate of immigrants.

The differences in educational level is greater for non-European and North African immigrants than for European immigrants for whom the educational distribution is similar to natives. In the Swedish case, even with equal characteristics of immigrants and natives, the immigrant population has an employment disadvantage of 17 percentage points.

(8)

Adsera and Chiswick (2007) use panel data for the years 1994-2000 to investigate how immigrant earnings depend on gender, country of origin and host country. In most cases immigrants face an earnings disadvantage relative to natives at arrival in the new country, where Sweden shows greater disadvantages than many other European countries. Furthermore, immigrants from non-EU countries tend to be worse off than EU immigrants. For a male immigrant from a country within the European Union, the earnings disadvantage is roughly 37% relative to natives and 49% for EU women. However, non-EU men earn 80% less than natives, while women earn 69% less relative native women. These regression results are based on immigrants with a high school degree and ten years of experience, at time of arrival in Sweden. In general, immigrants who settle in countries with linguistic and cultural similarities earn more compared to those who speak a totally different language. However, the initial earnings gap tends to decrease as time since migration increases introducing a catch-up process.

If all other variables are equal, this process will take 18 years, on average.

Husted et al. (2001) use panel data to study the labour market assimilation, in terms of wages and employment probabilities, of Danish first-generation male immigrants, between 1984- 1995. Refugees tend to have lower initial employment probabilities compared to non-refugees.

The assimilation process of refugees and non-refugees is closely related to the attachment to the labour market. Since labour migrants are often positively selected from the source country, they are expected to have higher earnings capacity compared to refugees. Non-refugees who arrive for economic reasons, may not face as large obstacles in the labour market, language acquisition or in educational attainment as refugees. However, since Denmark has a compressed wage structure and generous benefit systems, both refugees and labour migrants can be negatively selected, which means that mainly low-skilled immigrants will arrive. The results show that immigrants from Africa, Iran, Iraq, Sri Lanka and Palestine (without citizenship) face lower employment probabilities than refugees from European countries, but the employment probability is positively correlated to education level for all groups of measure (native Danes included).

Longva and Raaum (2003) compute a comparison study of the relative earnings growth and assimilation of immigrants from inside versus outside the OECD region in Norway. The earnings assimilation is measured between 1980-1990. The immigrant composition changed remarkably during the first two decades of measure from about 17% non-OECD immigrants in the pre-1965 cohort to almost 50% in the 1980-1989 cohort. The results show evidence of

(9)

converging earnings of different cohorts during 1980-1990. At time of arrival in Norway, the earnings of OECD immigrants are in parity with natives, while non-OECD immigrants earn less relative natives. The results presented could be explained by OECD immigrants’ high transferability of human capital, while non-OECD immigrants face difficulties in the labour market, since a high share of this immigrant group is overrepresented in sectors of low productivity growth. Therefore, the rapid increase in the national unemployment rate in 1990 might hit the earnings of non-OECD immigrants harder than other minority groups.

Hammarstedt and Shukur (2006) use pooled data to investigate earnings assimilation of immigrants in Sweden between 1990-1999. The data consists of immigrants from seven different cohorts between pre 1971 and 1999, originating from different countries. The assimilation rates and the entry disadvantage of immigrants (relative to natives) differ across regions of origin. The descriptive statistics show a decline in the share of immigrants from Southern Europe and non-European countries with positive earnings from labour during the 1990s. This could be due to the increased inflow of refugees from former Yugoslavia and the Middle East, since a great share of those individuals had a hard time finding a job, initially.

The earnings differential at arrival for male individuals migrating from Nordic Countries and Western Europe is smaller than for those migrating from Eastern Europe. Furthermore, non- European immigrants face a greater entry disadvantage compared to European immigrants. In addition, both men and women from Southern Europe and African countries have the fastest assimilation rate of earnings. The recent European cohorts do better than the earlier ones, while non-European immigrants face the opposite scenario. The same applies for female immigrants.

However, none of the groups will catch up with the earnings of native counterparts. Larger differences in culture and less transferable skills for immigrants originating from more distant countries could potentially explain the given results. Discrimination against non-European immigrants in the Swedish labour market and the changed immigrant composition from a high share of return- and labour immigration in previous cohorts to consist more of refugee immigration in recent cohorts, is also relevant.

Hammarstedt and Shukur (2007) take on a quantile regression approach and examine the disparity in earnings between various immigrant groups and natives in Sweden, with a focus on the different levels of the earnings distribution. The results show that immigrants from the higher end of the earnings distribution are performing better than those from the lower end, in comparison to natives. In addition, newly arrived non-European and Southern European

(10)

immigrants, from the lower end of the earnings distribution are seen to be most affected by the economic decline in the country in the 1990s. The problematic labour market assimilation among the non-European immigrants can, in some ways, be explained by their human capital not being easily transferable to the Swedish labour market.

Åslund and Rooth (2007) look at how the assimilation process of refugee-migrants are affected by the labour market conditions they experience when arriving in the country. By using data from the placement policy, conducted around the 1990s by the Swedish government, the authors control for the fact that the choice of settlement is not random. The results of the study show that the assimilation of earnings is highly reliant upon how beneficial the economic settings are, where bad conditions at arrival makes it harder for the immigrants to engage in employment. This means that it is not only the transferability of skills that affects the labour market outcome, but also the initial economic setting.

Lundborg (2013) analyses the assimilation process of refugee-migrants into the Swedish labour market with the use of register data. The results show that the variation in employment between refugees and natives are very large at arrival but declines as time since migration increases.

However, the study finds that employment level never fully reaches parity with the natives regardless of length of residency. In addition, refugees with a more similar culture as the destination country are seen to easier assimilate than those from more distant ones.

An alternative method to analyse the assimilation process is to investigate the impact of language skills. This human capital attainment is argued to be very important for immigrants and is examined by the two following studies.

Dustmann and Fabbri (2003) examine the importance of language proficiency in the assimilation process, regarding employment and earnings, of non-white immigrants in the United Kingdom. With the use of a matching approach and data containing evaluated language proficiencies, the authors can control for errors of measurements and the endogeneity associated with attainment of language skills. The results show that immigrants with a better proficiency has a greater employment probability than those with lower language ability. In addition, the findings show a positive relation between language proficiency and earnings when running the OLS but is however insignificant when errors of measurement are considered. This can be explained by them attaining small samples.

(11)

Berman et al. (2003) analyse the impact of language proficiency on earnings growth for Soviet immigrants in Israel. With the use of a first-difference approach they find that a good language proficiency positively affects the assimilation process of immigrants with higher skills.

However, for low-skilled professions the language proficiency seems to have no effect regarding the assimilation of earnings, meaning that the language proficiency is only of help for the high-skilled immigrants to assimilate their earnings towards the wage levels of natives.

In addition, it provides the intuitive finding that language proficiency might be complementary to occupational skills.

A final important component when considering the assimilation process of immigrants is the disparity between a refugee and a labour-migrant, and how these two groups experience different settings in the destination country.

Brell et al. (2020) gives an overview of the assimilation process of refugees. The authors discuss how refugees and labour-migrants differ regarding selection and thereby may have a different performance in the host country’s labour market. The findings show that refugees have lower rates of employment and earnings in comparison to other immigrants when arriving in the country, but that this gap eventually will fade over decades. The initial gap is explained as due to refugees possessing lower skills and skills that are less transferable to the host country in comparison to other migrant groups. In addition, the results show that some of the refugees attain mental health issues due to traumatic events from the migration or home country, which can be seen as another explanatory factor to the variation in labour market performance between the immigrant groups.

4. Theoretical framework

By viewing previous research, it is evident that a lot of immigrants face an earnings disadvantage when arriving in the destination country. This disadvantage is seen to vary with immigrant type and decline as time since migration increases, which implies that the immigrants face a positive earnings assimilation. The theory regarding this phenomenon originally comes from a study performed by Chiswick in 1978, where the author conducted a model to investigate how long it would take for the immigrant earnings to be in parity with the natives. This model is nowadays known as the assimilation model. Furthermore, immigrants with similar skills, culture and language have exhibited an easier assimilation into the host country’s labour market, compared to those attaining more distant ones. This suggests that

(12)

some groups may struggle more than others in using their home country specific skills and experiences in the new country’s labour market. A final factor from the literature review is that the various immigrant groups tend to face different economic settings when arriving in the host country, which can be related to the theory regarding migration decision. The remainder of this section will be divided into three subcategories, presenting the theoretical aspects of our study.

First, the underlying theory behind refugee migration will be presented (4.1), secondly the human capital model (4.2), and finally, a presentation of the assimilation model is given (4.3).

4.1 Refugee migration

Refugee migration occurs due to various problematic situations in the country of residence.

War, political conflicts and violence, poverty and natural catastrophes are some reasons why people are forced to displace, search safe environments, and settle in foreign countries. In the host country, refugees may face difficulties in the labour and housing market. In addition, cultural and linguistic differences slow down the process of adapting to the new environment.

These challenges can seem to be even greater when people who have witnessed horrifying events, suffer from mental instability. Furthermore, not only each migrating individual is affected by the circumstances, but also the receiving county. Large waves of refugees can put pressure on the political systems and markets in the host country. Economic consequences and increased population density could create xenophobia among permanent residents, who might prejudice against minority groups (Union of International Associations, 2019).

Refugees migrate primarily for safety and the choice of destination country could be connected to where their family or other people with the same ethnic background or nationality have settled. Network theory refers to social connections between migrants, former migrants and non-migrants living in the same areas. Ties through friendship and kinship can increase international movement since it reduces risks and costs of migration. As the size of the network expands, not only those who are best qualified to enter the foreign labour market will migrate, instead, the migration flow will be more characterised by the society of origin and less selective on socioeconomic variables (Massey et al., 1993).

4.2 Transferability of skills

The human capital model is based on the belief that at arrival immigrants possess skills that are not perfectly transferable to the host country’s labour market, which suggests that this lack of certain attributes causes the immigrants to struggle in getting employed and higher wage in the

(13)

destination country. The relevant requirements could be language proficiency, knowledge of the labour market, networks and valid educational degrees in the host country. This implies that the more equal the human capital of the countries are, the easier the migrants will have to assimilate (Borjas, 2020). Larger differences in human capital, on the other hand, hinders the transferability of skills in the way that these immigrants are not fully able to use the skills attained in the country of origin. Therefore, they need to invest more in human capital than natives to be able to attain the same occupations (Chiswick & Miller, 2009). To be able to enter the labour market of the host country investments in human capital need to be made. These can either be small, to transfer the existing skills from the country of origin to the new country.

Either, they can be large, where the immigrant must attain new skills and experiences, since the human capital is too different to be transferable (Chiswick et al., 2006). The most common and arguably important human capital investment is the attainment of skills regarding host- country language. If the language from the country of origin is somewhat similar to the destination country, the proficiency is seen to be easier attained and thereby less investments are needed. However, if the language is more different, the investments will need to be larger.

In addition, this human capital attainment is not only important in the labour market assimilation, but also for the social aspect of the integration. The host-country language proficiency tends to be smaller among the immigrants arriving at an older age, and higher among the young, as well as for the well-educated (Chiswick et al., 2006).

4.3 The assimilation model

Immigrant assimilation has been modelled in previous research by Chiswick (1978) and Carliner (1980) (Borjas, 1994). They study how immigrants’ skills are transferred to the labour market in the host country. Immigrants from the 1970 census is compared to natives in the US and the authors use a cross-section regression model to estimate the assimilation process.

Chiswick studies immigrants that have arrived in the US at different points in time and uses the following regression model:

log wi = Xiθ + δAi + γ0Ii + γ1yi + ԑi (1) In equation (1) wi is defined as worker i’s wage rate; socioeconomic characteristics such as education and region of residence is captured in the vector Xi; Ai is the age of the worker (could be interpreted as years of experience); Ii is a dummy variable equal to one if the individual is an immigrant and zero if the individual is native; yi is years since migration (which is zero for

(14)

native-born workers) and ԑi is an error term. This is a simplified version of the model which could include yi in a higher polynomial degree to indicate that additional years of residence can increase the wage rate in a linear or non-linear way. The coefficient γ0 tells the initial percentage wage differential between natives and immigrants at arrival in the host country, while γ1

explains the speed at which immigrants’ earnings change relative natives. Previous studies reached the conclusion that γ0 is negative and γ1 positive. The results from Chiswick’s analysis indicated that at time of arrival immigrants earn on average 17 percentage points less relative natives. Immigrants have faster wage growth compared to natives and catch up in earnings as years since migration increases. After 30 years of residence in the US, the initial disadvantage for immigrants would, according to Chiswick, turn to an advantage of 11 percentage points and immigrants will outgrow natives. The results are explained by 1) at arrival, immigrants tend to lack host-country specific skills that are important in the American labour market, such as English proficiency and cultural knowledge; and 2) as immigrants invest in human capital and gather country specific skills, their accumulated human capital stock will grow faster relative natives. Chiswick explains the part where immigrants outgrow the earnings of natives by assuming a positive-selection process of migration (Borjas, 1994).

The positive correlation between years since migration and the relative earnings of immigrants could be overestimated. In a study from 1985, Borjas uses panel data instead of cross-section data to account for the probability that different cohorts differ in skill level. This is referred to as cohort effects. Borjas’ main critique to the model developed by Chiswick is that conclusions drawn from a cross-section is less reliable, since immigrants arriving at different points in time can be associated with different characteristics, which can result in different age-earnings profiles relative natives. To remove the difficulties with the first assimilation model by Chiswick, Borjas presents a model including cohorts arriving in the US in 1960, 1980 and 2000, assuming that all immigrants are in their twenties at arrival. The 1960 cohort is the most productive, while the 2000 cohort is the least productive, which implies that the initial relative earnings are the highest for those migrating in 1960 and the lowest for the latest cohort. These suggestions are supported by empirical evidence, which implies that there has been a significant decline in immigrant skill level (Borjas, 2020). When cohort effects are considered, there are less convergence between immigrant and native wages and immigrants’ age-earnings profiles do not tend to catch up with natives to the same extent as Chiswick first found. The cohort effect can arise due to changes in economic and political conditions in the source and host country, respectively. The mix of national origin of immigrant flows can also shape cohort

(15)

effects if the skill composition of individuals from different countries are not equally transferable between source and host country. Furthermore, if the return migration is negatively selected, due to that low-wage immigrants return to their home countries, and not random, it is possible that earlier cohorts have relatively higher earnings than more recently arrived cohorts (Borjas, 1985, 1994). The extended version of the assimilation model for immigrants, presented by Borjas is computed as follows:

log wij = Xjθi + δiAj + αyj + βCj + γiπj + ԑij (2)

The regression model for the native wage level looks as follows:

log wnl = Xlθn + δnAl + γiπj + ԑij (3)

The dependent variable wij is the wage of immigrant j while wnl gives the wage of native l in equation (2) and (3), respectively. X is a vector of socioeconomic characteristics; A is age of the worker; y gives years of residence in the US for the immigrant; C is a dummy variable for calendar year of arrival measuring the cohort effect; π controls for period effects as a dummy variable for year of observation; ԑ is the error term (Borjas, 1994).

5. Data

A comprehensive table regarding variable definitions is found in the Appendices.

In this study panel data is used. This data comes originally from Statistics Sweden (LOUISE/LISA), and contains information regarding a sample of the immigration in Sweden between the years of 1990-1999, including year of birth, date of arrival (which gives information about the years since migration) and country of origin. Furthermore, the data provides information regarding employment status, yearly earnings, gender, age, educational level and civil status of the immigrant population, as well as for a collected sample of comparative natives. In addition, information regarding number of children living at home and area of residence, is also included in the data. The study departs from a primary focus on the working age immigrants of 20-64-year olds, as the labour market assimilation process is of main interest. We define Middle Eastern refugees, as refugee-migrants originating from the countries Iraq, Iran, Turkey, Syria, Saudi-Arabia, Yemen, Qatar, Palestine, Lebanon, Kuwait, Jordan, Israel, Egypt, Cyprus, UAE, Bahrain and the Gaza Strip. The Balkan refugees are defined as refugee-migrants originating from the state formerly known as Yugoslavia, consisting of Bosnia, Montenegro, Serbia, Herzegovina, Croatia, Kosovo, Slovenia and

(16)

Northern Macedonia. Immigrants from the Southern and Eastern European countries Greece, Albania and Bulgaria are also included. Natives are defined as individuals born in Sweden.

Further, the refugee-migrants are divided into five cohorts, such as pre-1980, 1980-1985, 1986- 1990, 1991-1995 and 1996-1999. Each cohort contains a group of immigrants that has arrived in the country during the same years. Three observation points are considered from the census of the immigrant population, in the years of 1990, 1995 and 1999. The first observation (1990) includes the three earliest cohorts. The second observation (1995) includes the pre-1980, the 1980-1985, 1986-1990 as well as the 1991-1995 cohorts. Finally, the 1999 observation includes all five cohorts.

Source: LOUISE/LISA, Statistics Sweden

Table 1 presents the number of immigrants and natives in working age, 20-64-year olds, sorted by region of origin and time of immigration current in 1990, 1995 and 1999. In the first observation, 1990, the Middle Eastern immigrant group consisted of 7,702 individuals residing in Sweden, while the Balkan immigrant group was composed of 6,543 individuals. The reference group contained 31,785 natives. At the second observation, 1995, the number of immigrants from the Middle East had increased to 11,024 individuals, while the number of Balkan immigrants increased to 10,545 individuals. The natives amounted to 32,203 individuals. In the final observation, 1999, the Middle Eastern immigrant group had further grown to a total of 12,321 individuals, while the Balkan immigrant group amounted to 11,353 individuals. The reference group contained 31,882 natives. The variation in immigrant population between the observed years is directly related to individuals entering and exiting

(17)

the working age of the sample, as well as the inclusion of more recent cohorts into the observations. This information portrays a larger representation of Middle Eastern immigrants in Sweden, than of immigrants originating from the Balkan countries. By viewing the different cohorts, it is then visual that the largest influx of the Middle Eastern immigrants came in the years of the 1986-1990 cohort, while a larger flow containing Balkan immigrants arrived between the years of the 1991-1995 cohort. This can be drawn to the refugees seeking safety during the Yugoslav wars and the wars between the Middle Eastern countries, which is earlier mentioned in the section regarding the historical overview (2. Historical overview).

Nonetheless, the pre-1980 cohort for the Balkan countries, which includes the longest time interval of the cohorts, contains the highest number of Balkan immigrants.

Table 2 displays the educational attainment of the immigrant groups and natives (measured as a share of the population) in working age, 20-64-year olds, sorted by region of origin and time of immigration in the years 1990, 1995 and 1999. By studying the three observations it is possible to conclude that there is a larger share of educational acquisition among the immigrants from the Balkan countries, than for the immigrants originating from the Middle East, with the exception of the third educational category (Edu 3), which presents university studies. This trend can also be found by comparing the different cohorts between the two immigrant groups. In addition, the largest share of immigrants from both groups possess a

(18)

degree from upper secondary school, which is corresponding to education level 2 (Edu 2) in the table. Furthermore, through observing the different cohorts for the Balkan countries it is possible to see a trend portraying a decrease in acquisition of low education (Edu 1) between the observed years, 1990, 1995 and 1999. The attainment of upper secondary school (Edu 2) and university studies (Edu 3) are seen to increase. This is similar to the cohorts for the Middle Eastern countries, where the trend portrays a decrease in attainment of low education (Edu 1), and an increase in attainment of high education (Edu 3) over the years of measure. However, the attainment of upper secondary school differs between the two groups, as it in the Middle Eastern cohorts decrease in some cases, and in others increase. By looking at cohort 1991-1995 and 1996-1999 for both groups it is visual that there is a very low attainment in education group 1, contra a very high attainment in education group 3. According to a report about immigrant establishment in Sweden by the Swedish government (2003), this can probably be explained by the collected dataset missing educational information for many individuals. These individuals are expected to have a relatively low educational acquisition and since they are omitted from the values in Table 2, the group Edu 1 get unrealistically low shares for the two latest cohorts. This would cause the dataset to rely on few observations, which in turn affects the reliability of the table. In a case where the educational level of each individual was specified, the shares for Edu 2 and Edu 3 would probably be lower and the share for Edu 1 higher than expressed in Table 2.

(19)

Table 3 presents the employment differential, i.e. the percentage difference in employment rate, between the immigrant groups and the native reference group. The measures are calculated for men and women separately and for each cohort in the observation points 1990, 1995 and 1999.

All presented values are negative, which indicates that the immigrant groups face an employment disadvantage compared to natives for all groups and years of measure. The total average disadvantage for Middle Eastern and Balkan men are lower than for the women.

However, they follow the same fluctuation pattern, that the smallest differentials are measured in 1990 and the highest in 1995. Between these two points, the employment differential increases with 22.1 and 19.7 percentage points respectively for Middle Eastern men and women and 34.4 contra 32.1 percentage points for Balkan men and women.

The Balkan immigrant group face smaller disadvantages, regardless of gender, in all observation years, except for the male 1991-1995 cohort in 1995, where the differential is slightly higher than for the Middle Eastern men. The different cohorts follow a similar pattern as the total values, with the smallest disadvantages in 1990 and highest in 1995, except for the early Balkan cohorts pre-1980 and 1980-1985, where the lowest employment rates are presented in 1999 for both men and women. Comparing the two regions and each cohort, the Balkan immigrants face smaller disadvantages compared to the Middle Eastern in most cases, and in the few cases where the opposite happens, the difference between the groups are relatively small. Since the employment differential increases for most groups of measure between 1990 to 1995, this implies that the employment level for the immigrant groups lowers to a wider extent than for natives. The decreasing employment rates could be a result of the economic recession that occurred during this period (see section 2. Historical overview). A recession in the national economy tend to hit minority groups harder than native counterparts, which the data shows evidence of. The more recent cohorts, i.e. 1991-1995 and 1996-1999, face greater employment disadvantages compared to earlier cohorts, which applies for both Middle Eastern and Balkan immigrants.

The larger influx of refugee immigrants during the later years of measure, combined with the economic crisis, could potentially explain the lowered employment rates. The 1991-1995 cohorts, at observation point 1995, face an 80.5% and 91.3% lower employment probability for Middle Eastern men and women, respectively. Corresponding values for Balkan men and women are 81.0% and 88.3% disadvantage. However, the same Middle Eastern cohort of

(20)

measure in 1999 show a 31.2 (men) and 27.7 (women) percentage points lower differential compared to year 1995. The corresponding Balkan cohort have lowered the negative differential by 52.3 (men) and 49.3 (women) percentage points. This indicates a positive employment assimilation process for immigrants originating from both the Middle East and the Balkan countries during the first years of residence in Sweden. The initial negative differentials for the latest cohort, 1996-1999, are high, consisting of 74.2% and 56.1% for males and 87.9% and 71.3% for females, favouring Balkan. The faster assimilation shown by the Balkan 1991-1995 cohort and the higher initial employment rates for the Balkan 1996-1999 cohort could be due to Middle Eastern immigrants having less transferable skills to the Swedish labour market, or simply ethnic discrimination against Middle Eastern refugees.

Table 4 displays the yearly earnings differentials between the immigrant groups and natives in working age, 20-64-year olds, attaining a positive income from employment (i.e. individuals income from work equal to zero are excluded) in the observation years 1990, 1995 and 1999.

For each time point of observation, the calculations are done for men and women separately.

The average earnings of natives are presented in Swedish Crowns (SEK). In total, the gap for Middle Eastern men fluctuates between a disadvantage of 48.4%, at its lowest in 1990, 60.0%, at its highest in 1995 and 52.2% in 1999. The Middle Eastern women face a smaller disadvantage (compared to males), reaching from 39.2% in 1990, to 50.4% in 1995 and 47.9%

in 1999. Male and female immigrants from the Balkan countries face smaller earnings

(21)

disadvantages compared to their Middle Eastern counterparts, for all years of measure. In 1990, the male Balkan immigrants show a negative earnings gap of 12.9% in 1990 and then it grows larger to 25.5% in 1995 and the final value for 1999 is 19.5%. The female Balkan immigrant group has an earnings disadvantage of 6.3% in 1990 and this number grows over the years to reach 18.9% in 1999. The pattern for the Middle Eastern immigrant group is in line with the employment differentials presented in Table 3, i.e. when the employment rate goes down (up), the earnings differential grows bigger (smaller). When the earnings gap grows larger as employment rates go up, as in the Balkan case for women in 1999, this could possibly be explained by recently arrived immigrants being low-skilled and attaining jobs with lower average wage levels. For both men and women from both regions and the two latest arriving cohorts, the negative earnings differentials tend to be approximately twice as high as for the previous cohorts, pre-1980 and 1980-1990. However, for the 1991-1995 cohort, it lowers from 82.2% in 1995 to 64.6% in 1999 for Middle Eastern men and from 80.2% to 61.5% for Middle Eastern women. The corresponding numbers for Balkan men range from 79.8% in 1995 to 42.5% in 1999 and for women it runs from 80.7% to 48.6%. This pattern evolves as a result of the employment status, which increases remarkably between 1995 to 1999 for this particular cohort. In most measures, the female group show smaller, or close to equal, earnings disadvantages compared to the male group and for the pre-1980 Balkan cohort in 1995, the women are almost in parity with native females. By this time, however, the immigrants have resided in Sweden for over 15 years. As reflected in the summed up total numbers, each cohort originating from the Balkan countries, tend to face smaller earnings disadvantages compared to their Middle Eastern counterparts, regardless of gender or year of migration, except for women in the 1991-1995 cohort. In this case, the Balkan group has a 0.5 percentage points higher disadvantage compared to Middle Eastern females.

6. Methodological framework

The methodological framework of the study is based upon regression analysis. Here the data is examined and tested using various variables in the regressions. This paper will depart from the model applied in the study by Hammarstedt and Shukur (2006), in which the earnings assimilation is tested among the immigrants in Sweden. The model applied is a cohort model, where the different cohorts display the difference in characteristics regarding time of arrival in the country, and is presented by equation (4) in the following way:

Y = BX + α1(Age) + α2(Age)² + δ1(YSM) + δ2(YSM)² + θi(COH) + ε (4)

(22)

Y = A natural logarithm of yearly earnings.

X = A vector of human capital variables, such as educational level, number of children living at home, civil status and residential area.

YSM = Number of years since migration for immigrants (takes the value of 0 for natives).

COH = A vector of time interval dummy variables, which divides the immigrants into groups based upon arrival in the destination country.

Females and males of the different immigrant groups are measured separately in the model, and sorted based on country of origin, with the natives as a reference group.

The coefficients that are of primary interest of this study, are the δ1, δ2 and θi. δ1 measures the assimilation effect (the yearly percentage change in immigrant earnings), and takes on a value above zero if the assimilation is positive, and a value below zero if it is negative. δ2 measures the assimilation squared effect. This coefficient will be equal to zero if the assimilation effect is linear, which means that the earnings will increase at a constant rate. However, if it takes on a positive or negative value the effect is non-linear, which means that the effect will follow a non-linear path. The coefficient θi measures the earnings gap upon arrival in the country for the different cohorts. It takes on a value below zero if there is an earnings disadvantage in comparison to natives. If the value is positive the immigrants have, on the other hand, an earnings advantage. When attaining a positive value for the δ1-coefficient and a negative value for the θi-coefficient, an estimation of the number of years it will take for the earnings to be in parity with the comparative natives can be found. To estimate the regression model, the ordinary least square methodology (OLS) is used.

To investigate the employment assimilation of the immigrant groups, the same independent variables from equation (4) will be used, but the dependent variable will be exchanged from yearly earnings to employment probability, denoted as Emp. This variable will take on the value of zero, indicating unemployment (alternatively, out of the labour force), or one, indicating employment. The model is presented by equation (5) in the following way:

Emp = BX + α1(Age) + α2(Age)² + δ1(YSM) + δ2(YSM)² + θi(COH) + ε (5) The coefficient θi measures the initial employment probability upon arrival for the different cohorts. δ1 measures the assimilation effect (the change in employment probability in comparison to natives). Finally, δ2 measures the assimilation squared effect. This is a linear probability model (LPM), which means that the outcome variable is a binary variable and the

(23)

coefficients of the explanatory variables express the change in probabilities of a one-unit change in independent variables. However, there are some problems with the LPM that needs to be considered. The first issue is that the model allows for predicted values of the outcome variable outside the bounded interval [0,1], which is unacceptable according to basic probability theory. Secondly, the error term is not normally distributed, which is an assumption that must be fulfilled to perform valid t-tests and F-tests of an econometric model and instead the LPM has a binomial distributed error term. This cause the tests of significance to be invalid.

Lastly, in contrast to the classical linear regression model (CLRM) with the assumption of the error term being homoscedastic to prove the credibility of the OLS estimates, the error term of the LPM is heteroscedastic since the variance is not constant. To handle potential heteroskedasticity, robust standard errors are used (Pedace, 2016).

Considering the descriptive statistics and previous literature there are reasons to believe that Balkan refugees face smaller initial disadvantages in employment probability and earnings compared to refugees originating from the Middle East. The assimilation process should also be faster, due to better transferability of skills and more similar culture to Swedish natives, resulting in less discrimination in the labour market. This hypothesis will be tested with estimations and calculations using the statistical program package STATA version 16.1.

7. Results

7.1 The employment assimilation of Balkan and Middle Eastern women

(24)

Table 5 and 6 present the results of the regressions comparing the employment probability of the Middle Eastern and Balkan female immigrants to the native reference group, respectively.

By viewing the tables, we come across a positive employment assimilation of 3.1 percentage points for the Balkan females and of 1.6 percentage points for the Middle Eastern females, at the 1% level of significance. The assimilation is non-linear for both groups, as the coefficients for the squared years since migration variable are negative. By comparing the two values we find that the Balkan females have a faster employment assimilation than the females originating from the Middle East. Further, by viewing the different cohorts it is visible that both groups face a disadvantage in employment probability at arrival in the country, when comparing to the natives. For the Balkan females it decreases from 52.1 to 44.3 percentage points in the earliest cohorts, and then grows to 67.2 and 69.2 percentage points in the 1991-1995 and 1996-1999 cohorts. For the Middle Eastern women, it is seen to continuously increase throughout the time intervals. The disadvantage ranges between 29.2 to 53.5 percentage points for the earliest cohorts, and then further grows to 72.9 and 78.9 percentage points for the 1991-1995 and 1996- 1999 cohorts. When comparing the groups, we find that the Balkan females initially face a larger disadvantage than the females from the Middle East. This trend, however, turns and the disadvantage grows greater for the Middle Eastern females in the more recent cohorts.

Furthermore, marriage increases the employment probability with about 6.1 percentage points for both groups. Residence in a metropolitan area positively affects the probability with 1.4 percentage points for the Middle Eastern females, but is, however, statistically insignificant for the Balkan females. An additional child living at home decreases the employment probability of the immigrants with about 2.7 percentage points. In addition, Balkan and Middle Eastern females with a degree from upper secondary school (Edu 2) are 13.4 respectively 14.0 percentage points more likely to be employed than those with a lower education (Edu 1). If they attain a university degree (Edu 3) the likelihood increases further to 16.6 percentage points. Finally, age increases employment probability for both groups, with about 5.7 percentage points. The employment probability increases with a decreasing rate as the coefficients for the squared age variable are negative. In this context, the age variable can be interpreted as years of experience.

(25)

7.2 The employment assimilation of Balkan and Middle Eastern men

Table 7 and 8 show the regressions comparing employment probability of the Middle Eastern and Balkan male immigrant groups, to the Swedish reference group. Since the years since migration variable has a positive coefficient of 0.24 percentage points for the Middle East and 4.7 percentage points for Balkan, while the squared years since migration is negative for both groups, this can be interpreted as a positive, increasing employment assimilation at a decreasing speed. However, the YSM variable for the Middle Eastern group is statistically insignificant and conclusions cannot be drawn from the value of this coefficient. The Middle Eastern cohorts show increasing negative values over the time intervals, ranging from 23.7 for the pre-1980 cohort to 69.8 percentage points disadvantage for the 1996-1999 cohort, which implies that earlier cohorts face a higher initial employment probability compared to latter ones. The Balkan immigrant group shows a different pattern over the cohorts, where the smallest disadvantage of 57.0 percentage points is connected to the 1986-1990 cohort and the pre-1980-cohort face the greatest disadvantage of 75.4 percentage points.

The age variable, which could partly be interpreted as years of experience, is 4.7 percentage points for Middle Eastern male immigrants and the corresponding number for Balkan males is 5.15 percentage points. This means that the employment probability increases with age. This increase is however decreasing, since age squared takes negative values for both immigrant groups and the decrease is slightly faster for Balkan males. Higher educational attainment has

(26)

a positive effect on earnings for both groups. Completed upper secondary school studies and university studies increases the employment probability by 4.5 and 4.7 percentage points, respectively for Balkan males and the corresponding values for the Middle-Eastern group are slightly higher at 5.1 percentage points for education group 2 and 6.1 percentage points for education group 3. An individual who is married runs about 12 percentage points higher probability of being employed compared to those who are not married, and this applies for both immigrant groups of measure. Residence in a metropolitan area has a 0.85 percentage points negative effect on earnings for Balkan males and this value is statistically significant on the 5% level. The coefficient for Middle Eastern men living in metropolitan areas is positive but statistically insignificant. Having children living at home decreases the employment probability by 0.75 percentage points for Balkan immigrants and 0.56 percentage points for the Middle Eastern male immigrant group.

7.3 The earnings assimilation of Balkan and Middle Eastern women

Table 9 and 10 portray the results of the regressions comparing yearly earnings of the Middle Eastern and Balkan female immigrants to the native reference group, respectively. By viewing the tables, we come across a positive earnings assimilation of 8.4%² for the Balkan females and of 5.2% for the Middle Eastern females, at the 1% level of significance. The assimilation

2 coefficient

(27)

is non-linear for both groups, as the coefficients for the squared years since migration variable are negative and slightly higher for the Balkan group. This implies that the female earnings increase, but at a decreasing speed. A comparison of the two values shows a faster earnings assimilation for the Balkan females than for the females originating from the Middle East.

Furthermore, by viewing the different cohorts, it is visual that both groups face an earnings disadvantage at arrival in Sweden, compared to natives. For the Balkan females, the disadvantage the largest of around 67.1% for the earliest cohort, as well as for the two most recent ones. In the 1980-1985 and 1986-1990 cohorts the disadvantage decreases from 58.0%

to 49.8%. For the Middle Eastern women, the disadvantage attains a different trend, being the lowest for the earliest cohort and then continuously increase throughout the time intervals. The disadvantage ranges between 52.8% to 62.2% for the earliest cohorts, and then further grows to 73,0% and 76.5% in the 1991-1995 and the 1996-1999 cohorts.

By looking at civil status, marriage increases the earnings by 2.2% for the Balkan females and 1.8% for the Middle Eastern. Residence in a metropolitan area positively affects the earnings with about 10.6% for both groups. An additional child living at home negatively affects the earnings of the female immigrants with around 6.7%. Further, Balkan and Middle Eastern females that have a degree from upper secondary school (Edu 2) earn about 30% more than those with a lower education (Edu 1). While the females with a university degree (Edu 3) earn about 53.1% more than the lower educated if Balkan, and about 55.9% if Middle Eastern.

Finally, age increases earnings for both groups, with 14.1%. The earnings are, however, increase with a decreasing speed as the coefficients for the squared age variable are negative.

(28)

7.4 The earnings assimilation of Balkan and Middle Eastern men

Table 11 and 12 display the results of regressions comparing yearly earnings of the Middle Eastern and Balkan male immigrant groups, to the Swedish reference group, respectively. The coefficients for years since migration is positive at the 1% level of significance for both groups and has a higher value for Balkan males. This indicates a positive earnings assimilation, consisting of 13.1% yearly increase in earnings for Balkan males and 1.7% for Middle Eastern men. Years since migration squared is negative for the Balkan group, which means that the earnings increase with decreasing speed. This coefficient is positive but statistically insignificant for the Middle Eastern group. The three earliest observed cohorts, pre-1980, 1980- 1985 and 1986-1990, from Balkan have higher initial earnings disadvantage, 87.0%, 80.5%

and 71.6%, compared to the corresponding Middle Eastern cohorts with the values 64.5%, 61.6% and 66.3%. Immigrants arriving before 1990 from the Middle East seem to face smaller obstacles in the labour market, but the Balkan males assimilate faster, on average. For the two most recent cohorts of measure, 1991-1995 and 1996-1999, the opposite pattern is discovered.

The Balkan immigrants face a 74.8% and 68.5% entry disadvantage, respectively, while the same cohorts from the Middle East has a disadvantage of 76.2% for cohort 1991-1995 and 77.6% for the latest arriving cohort. The coefficients for the age variable are positive and close to equal in both regressions, 16.4% in the Middle Eastern case and 16.6% for the Balkan immigrants. This is interpreted as an additional year of age will increase the yearly earnings by 16.4% versus 16.6%, but this effect is waning by just over zero percent each year for both

(29)

groups. Having completed upper secondary school studies has approximately a 25.0% positive effect on earnings for both groups and university studies would increase earnings by over 40.0%. Being married increases the earnings by 24.2% for Balkan males and 22.5% for Middle Eastern males, compared to not being married. There is a positive effect of living in a metropolitan area, consisting of 11.0% for Balkan and 12.4% for the Middle East. An additional child living at home reduces the earnings by 2.7% for the Balkan immigrant group, compared to 2.0% for Middle Eastern men.

8. Discussion

With the use of OLS- and LPM regressions, we have examined the earnings and employment probability of Balkan and Middle Eastern immigrants in Sweden. By referring to historical events of wars in former Yugoslavia and the Middle East and the descriptive statistics showing an increased inflow of individuals originating from these regions during the 1990s, the assumption has been made that the data consists primarily of refugee-migrants. The focus is to study the labour market assimilation of refugees from mentioned countries, but since the data set includes no information about the cause of migration, we cannot be sure that only refugees are included in the study, which needs to be considered. Based on prior research we predicted a positive employment and earnings assimilation for both groups attaining an initial disadvantage at arrival in Sweden. The hypothesis to be tested, stated that Balkan immigrants have a faster assimilation rate and face lower initial disadvantages in the labour market. Most studies investigate the assimilation process of immigrants in general, while in this study men and women are given separate analysis.

Departing from the assimilation model, constructed by Chiswick, and developed by Borjas, gives the first intention that the assimilation process will be positive, but leaves the question of which speed different cohorts from different countries assimilate into the labour market. The incitement behind refugee migration is safety and not motivated by economic reasons. This could potentially make the regression results differ from reports on labour-migrants, whose priority are to increase earnings. As predicted in the hypothesis, which was based on previous research (see Adsera & Chiswick, 2007 and Husted et al, 2001) Balkan refugees have a faster assimilation process, regardless of gender or outcome variable, compared to Middle Eastern counterparts. This could be due to European immigrants having more transferable skills and more linguistic and cultural similarities. However, the earlier cohorts in the study, pre-1980

(30)

and 1980-1985, originating from Balkan face greater disadvantages compared to the cohorts arriving from the Middle East. Latter Balkan cohorts do better than earlier ones and face smaller disadvantages than the Middle Eastern, which is in line with the findings of Hammarstedt and Shukur (2006). Previous studies argue that discrimination against non-European immigrants is a potential explanation of why some group assimilate faster than others (see Ahmed, Andersson

& Hammarstedt, 2010, Hammarstedt & Shukur, 2006 and Carlsson & Rooth, 2007). In addition, our findings show a very large variation in earnings assimilation when comparing the male immigrant groups. This trend is seen to follow in the assimilation of employment as well, but in a lesser way. Worth noting is that the estimation of employment assimilation of Middle Eastern men is statistically insignificant in our study, which makes the interpretation uncertain.

The women show smaller variation, in general, when comparing the immigrant groups. This indicates that gender is a strong component that needs to be taken into consideration when examining labour market performances of immigrants.

No group or cohort of measure, however, shows tendencies to catch up with natives during the first ten to fifteen years of residence in Sweden, according to Table 3 and 4, since the employment and earnings differentials, relative natives, are constantly negative. This is consistent with previous literature (see Aldén & Hammarstedt, 2019, Adsera & Chiswick, 2007 and Lundborg, 2013). The large disadvantages in employment and earnings for the most recent cohorts of the study (the 1991-1995 and 1996-1999) can be explained by the recession in the economy in the 1990's. This resulted in a large decrease in labour demand, which caused the probability of attaining employment for immigrants to decrease. During these years, the Yugoslav wars and the Middle Eastern conflicts contributed to a large influx of refugee- migrants to the Sweden. As previously mentioned, refugees are seen to have a harder time assimilating than labour-migrants, which can be viewed as another explanatory factor to the results (see Brell et al. 2020). It could also be problematic that our data does not inform about number of working hours, only if an individual is employed or not and yearly earnings. This could potentially make us comparing individuals with full time jobs to people working part time. If that is the case, the earnings differentials presented could potentially be larger due to differences in working hours and not only great earnings gaps between immigrants and natives.

The educational attainment, area of residence, having children living at home and civil status has approximately the same influence on employment probability and earnings when comparing the immigrant groups separately for gender. In this sense, the country of origin does not seem to matter, regarding treatment in the labour market. The estimate for marriage tends

(31)

to be higher for men, while the coefficients for children living at home have a greater negative effect on females’ outcome variables. This could be connected to different gender roles for men and women, where men tend to work to support the family financially, while women take responsibility for childcare and household.

9. Conclusion

The aim of this study was to examine if Balkan refugees would have an easier and faster assimilation than refugees from the Middle East. This assumption was based upon the economic theory that human capital is more easily transferred between countries that are alike (similarities regarding culture, institutions, language). In accordance with our hypothesis, the Balkan refugees were seen to transfer their human capital more easily to the Swedish labour market than the Middle Eastern refugees, but face higher initial disadvantages in earlier cohorts. The female assimilation rates were found to be similar between the immigrant groups, with employment assimilation rates consisting of 1.6 percentage points for Middle Eastern women, contra 3.6 percentage points for Balkan women. The female earnings assimilation rates amounted to 5.2% for the Middle Eastern group and 8.4% for Balkan. Regarding the employment assimilation of the male immigrant groups, the Middle Eastern men were given a positive, but statistically insignificant value of 0.2 percentage points, while the Balkan males faced an assimilation rate of 4.8 percentage points. The Balkan males exhibited a much higher earnings assimilation rate, 13.1%, than the Middle Eastern males, 1.7%, which may be an indicator of discrimination. This explanation is strengthened further, as the initial disadvantages in employment and earnings are seen to continuously grow higher for the Middle Eastern immigrants as more recent cohorts arrive in the country. These results provide some evidence of how the Middle Eastern refugees are struggling more than their European counterparts to integrate into the Swedish labour market, as well as highlight how the performance of immigrants differs between gender. Nonetheless, since both immigrant groups experienced high earnings and employment disadvantages at arrival in Sweden, reaching parity in earnings and employment rate with natives seems distant, regardless of gender and origin.

Similar results for the immigrants originating from Southern Europe and Asia are found in the study by Hammarstedt & Shukur (2006). Limitations of our study can be expressed as the attainment of small samples, as well as using data from an earlier period. These factors may result in conclusions not applicable to recent circumstances. To fully underline the struggles of the Middle Eastern immigrant group, further research needs to be conducted. This can be done by examining data regarding the larger immigrant flows in the 21st century.

(32)

10. References

Adsera, A. & Chiswick, B.R. (2007) “Are there gender and country of origin differences in immigrant labour market outcomes across European destinations?”, Journal of Population Economics, 20(3): 495-526.

Ahmed, A., Andersson, L. & Hammarstedt, M. (2010) “Can Discrimination in the Housing Market Be Reduced by Increasing the Information about the Applicants?”, Land Economics, 86(1): 79-90.

Aldén, L. & Hammarstedt, M. (2019) “Refugee Immigration and Public-Sector Finances:

Evidence from Sweden”, FinanzArchiv Public Finance Analysis, 75(3): 296-322.

Andersson, L. & Hammarstedt, M. (2011) “Transmission of self-employment across immigrant generations: the importance of ethnic background and gender”, Review of Economics of the Household, 9(4): 555-577.

Berman, E., Lang, K. & Siniver, E. (2003) “Language-skill complementarity: returns to immigrant language acquisition”, Labour Economics, 10(3): 265-290.

Bevelander, P., (2001) “Getting a foothold: Male immigrant employment, integration and structural change in Sweden, 1970-1995”, Journal of International Migration and Integration, 2(4): 531-559.

Borjas, G.J. (1985) “Assimilation, changes in cohort quality and the earnings of immigrants”, Journal of Labour Economics, 3(4): 463-489.

Borjas, G.J. (1994) “The Economics of Immigration”, Journal of Economic Literature, 32(4):

1667-1717.

Borjas, G.J. (2020) Labor Economics, Eighth edition. McGraw-Hill, New York.

Brell, C., Dustmann, C & Preston, I. (2020) “The Labour Market Integration of Refugee Migrants in High-Income Countries”, CReAM Discussion Paper Series, 2020-01/20.

Carliner, G. (1980) “Wages, earnings and hours of 1st, 2nd, and 3rd generation American males”, Economic Inquiry, 18(1): 87-102.

The Assimilation of Balkan and Middle Eastern Refugees in Sweden