Balkan refugees in Sweden -
a study on labour market assimilation Bachelor thesis
Author: Anton Viklund & Nathalie Sommar Lindskog
Supervisor: Lina Aldén Examiner: Mats Hammarstedt Term: VT20
Subject: Economics
Abstract
This study focuses on annual earnings assimilation and the employment probability, described as the assimilation of annual earnings and the extent of which available workers are being used respectively (in this case workers originating from a certain country) of immigrants arriving from former Yugoslavia, i.e. Bosnia-Herzegovina, Serbia and Croatia, to Sweden during the Yugoslav wars occurring in the 1990s. Previous research made on immigrants, and in this case focusing on refugees, brings some empirical standpoints; initial annual earnings of refugees are lower than those of labour-market immigrants and natives, higher education level as well as geographic closeness and cultural similarities between source country and host country makes the transition and assimilation easier for immigrants. Immigrants incur a net-cost on public sector finances during their first years in host country, but that it diminishes as years since migration increases. These longitudinal regressions were made for two different cohorts and genders separately. The cohorts included individuals in ages 20-64 years of age from countries previously being a part of former Yugoslav that arrived in Sweden between the years of 1990 and 1995, and between 1996 and 1999. These cohorts are being studied in three cross-sections, 1990, 1995 and 1999. A brief history of the Yugoslav wars will also be presented in this thesis.
Our results show that the refugees from former Yugoslavia had a positive assimilation in to the Swedish labour market, and our results are in line with previous theory regarding labour market assimilation. Moreover, men without university education as well as women with university education assimilate faster in comparison to their corresponding opposites in regards of educational level. However, both genders, regardless of educational level, assimilated. This confirms some of the theory presented in this thesis.
Key words
Yugoslav war, refugees, assimilation, labour market, gender differences, Sweden, 1990s.
Acknowledgments
We would like to take the opportunity of sincerely thanking our supervisor, Associate Professor Lina Aldén for her guidance in writing and for supporting our work. Additional thanks go out to Professor Mats Hammarstedt for guidance through the process of writing this thesis. We would also like to thank Post-doctor Joakim Jansson for guidance in programming our statistics.
Table of contents
1 Introduction 1
1.1 Purpose 2
2 Background 3
2.1 The Yugoslav Wars 4
3 Theoretical framework 5
4 Literature review 7
5 Data 11
5.1 Descriptive statistics 13
6 Methodological framework 16
7 Results 19
7.1 Gender 19
7.1.1 Employment probability by gender 19
7.1.2 Annual earnings by gender 22
7.2 Education 24
7.2.1 Employment probability by educational level 24
7.2.2 Annual earnings by educational level 26
8 Discussion 29
9 Conclusion 33
10 Bibliography 35
Page 1(38)
1 Introduction
People have migrated through all of history, for example due to labour, family or political reasons. This study focuses on labour market assimilation, which is the development of annual earnings or employment over time, say a year. It is often done through comparing one group with their native counterparts, in this aspect Balkan refugees compared to native Swedes.
Sweden has faced immigration for several decades and reasons. In 1960’s, people migrated to Sweden as labour immigrants, from Italy, Greece, former Yugoslavia etc. Whereas in 1990’s, a large number of people migrated from former Yugoslavia due to the Yugoslav wars, altering the composition of labour immigrants arriving to mainly consisting of refugees.
At the same time, in 1995, Sweden entered the EU, which meant that it became easier for people to both migrate from and migrate to Sweden (Migrationsverket, 2020).
In comparison to labour immigrants, refugees are said to have a harder time entering the labour market and thus integrating into the host country. They often have lower initial earnings than labour immigrants, however, refugees’ earnings assimilate faster than labour immigrants’
does (Bevelander, 2016; Giri 2018; Hammarstedt and Shukur, 2006; Hirsch et al., 2014). Some researchers, such as, Bevelander (2016), Damas de Matos (2016) and Hirsch et al. (2014) shows that immigrants, despite having a slower annual wage assimilation, will catch up with native annual wages, while some researchers, such as Borjas (1985) show that it is not the case, that it takes several years for immigrants to reach the earnings of natives, but that the assimilation of annual earnings is not certain despite being in host country for a long time. These different results can be explained by cohort effects, since different groups of immigrants, i.e. cohorts, will assimilate in terms of income at different speeds and different directions.
Refugees are also said to incur a net-cost on public transfers in host country. As seen from Aldén and Hammarstedt (2019) study done on refugees’ effect on public sector finances during their first years in host country Sweden. Findings show that refugees incur a cost on public sector finances during their first years in host country, meaning a redistribution from natives to immigrants. This cost is equal regardless of education levels of immigrants, however, as time spent in host country, the net-cost of immigrants decrease faster for high educated than for low educated. The findings are in line with those of Ruist (2019) who found that immigrants incur a cost on public finances during their first years in host country, but as time spent there increases, employment rates increase and thus the net-cost of immigrants decrease. Due to these
previous findings, it is of major interest to study how refugee's annual earnings and employment rate assimilates as years in the host country increases, as it affects the countrywide finances.
The number of refugees entering Sweden as a consequence of the Yugoslav wars was large in the 1990s. Due to these closely linked series of wars in the 1990s former Yugoslavia, 138 000 people migrated to Sweden (Migrationsverket, 2020). This, along with other reasons, lead to a change of foreign-born in Sweden from consisting of 9.2 percent in 1990 to 11.3 percent in 2000 (Statistiska Centralbyrån, 2020). Subject of study will primarily focus on Serbia, Croatia and Bosnia-Herzegovina, but with a different feature in comparison to previous research, which consists of addressing possible gender differences as well as educational differences on annual earnings assimilation and employment probability. In other words, to see whether there are differences in the level of integration depending on whether the refugee is a man or a woman and if education is a deciding factor.
1.1 Purpose
The purpose of this study is to add new and relevant research that could be useful when integrating refugees on the Swedish labour market. The question of study is; How did the Balkan refugees migrating to Sweden during the 1990s assimilate into the labour market in Sweden in terms of annual earnings assimilation and employment probability? Are there any differences by gender and education? We investigate the labour market assimilation of refugees originating from Bosnia-Herzegovina, Serbia, and Croatia, with specific focus on assimilation of annual earnings as well as the employment probability. This is in order to underline a conclusion, that even though annual earnings show a sign of annual earnings assimilation, the including of employment probability might lead to the possibility of drawing a conclusion from a broader view. Previous studies show that these refugees had a good integration into the Swedish labour market (Ekberg, 2016; Sjöberg, 2018), it is of interest to investigate why this was the case. The subject of study originates from today's intense debate regarding Sweden’s migration policy, a debate that, to some degree, stems from the large amount of foreign-born individuals arriving to Sweden since around 2015 and forward. We are primarily going to study the refugees that migrated to Sweden after and during the Yugoslavian war in the 1990s.
This thesis has nine sections. Section one will provide information about the purpose and motivation for this particular research question. Section two will provide some background and history about previous immigration to Sweden, and a brief overview of the Yugoslav Wars.
Section three describes some economic theories about our research question. Section five includes description of our data. Section four describes previous studies and literature on this
subject. Statistical estimations and methods that are used are described in the sixth section.
Section seven provides our results, which are presented in two categories, gender and education.
Section eight is discussing of our results from previous section relating to the theories in section four, and the results from previous studies. Section nine involves our summarized results and subsequently drawing conclusions with respect to our research question.
2 Background
Sweden has since the Second World War been net-recipients of immigrants, starting during the war from neighbouring countries, continuing post-war with labour-immigrants. Immigration to Sweden increased in the 1940s until 1970s as a response to the excess demand in labour due to the industrial and economic growth (Bevelander, 2011) as well as removing some of, previously required criteria such as needs for residence and work permits for those arriving from Nordic countries. Allowing non-Nordic immigrants to enter and thereafter apply for work permits was also a new feature (Hammarstedt and Shukur, 2006). Thus, labour immigration was the cause for migration to Sweden during these years. Prior to 1960, the main source of labour-migrants was from other Nordic countries. However, during the 1960s labour-migrants from Balkans started arriving. These migrants were successful, and in the 1970s, employment rates for foreign-born women as well as males exceeded those of natives. During the 1970s, the economic growth somewhat stopped, and as a result, the demand for foreign labour stopped.
The group of labour-immigrants continuously arriving, however at a decreased rate, was from other Nordic countries. But after 1970s the stream of migrants mainly involved refugees and tied movers. During the years of 1984 to 2007, approximately 400 000 people requested asylum in Sweden, of this amount, 33 percent originated from former Yugoslavia. Former Yugoslavia, along with Iraq was the main source country of asylum-seekers through the years of 1991 to 1995 as well as 1996 to 2001. Bosnia-Herzegovina also provided a large number of asylum- seekers during the years of 2002 to 2007 (Bevelander, 2011). During the 1990s, Sweden went through a recession which caused unemployment rates to increase from approximately 3.5 percent of the labour force in 1990 to roughly 13 percent in 1993 (Ramaswamy, 1994) Previous research have showed that the initial conditions have a substantial effect on medium-term earnings and employment, with individuals entering before a recession are about 7-9 percentage points more likely to be employed, and with around 12-18 percent higher earnings compared to those entering during a time where the market is deteriorating (Åslund & Rooth, 2003). Bosnia-
Herzegovinian individuals are to a high degree Muslims, Serbians and Croatians are Christians, more precise Orthodox Christian and Catholics respectively (Magnusson, 1991).
2.1 The Yugoslav Wars
In order to understand what caused the migration of people from Serbia, Bosnia-Herzegovina and Croatia some background to the events occurring in these countries during this point in time will be presented in this section. The Yugoslav Wars is a series of wars that occurred in 1990s Serbia, Bosnia-Herzegovina, Croatia and Slovakia, and ended with a bombing in Serbia, Milosevic being removed from the power (Naimark and Case, 2003). Six countries left Yugoslavia during this decade (United Nations, n.d.). This series of wars was brutal and with a high killing streak of approximately 140000 dead victims (United Nations, 2008). In 1991, these series of wars started with the “ten-day-war” in Serbia, that started with Serbia declaring independence from former Yugoslavia. At the same time, Croatia declared independence from Yugoslavia as well, this resulted in an approximately four-year armed war that ended in 1995.
During the same period, in 1992 until 1995, Bosnia-Herzegovina declared their independence from former Yugoslavia, which resulted in the deadliest and most brutal of these six wars. This particular war took 100000 lives, women were systematically raped, and a total of two million Bosnia-Herzegovinians women and men had to flee from their homes during this war. For example, in the beginning of July in 1995, approximately eight thousand men and boys with Muslim religion in Bosnia-Herzegovina town Srebrenica were murdered during only a few days. In the period of 1998 to 1999, Kosovo was at war with former Yugoslavia. The last war was in The Former Yugoslav Republic of Macedonia and went on for a couple of months in 2001 until The Former Yugoslav Republic of Macedonia and former Yugoslavia agreed to a peace deal (United Nations, n.d.).
3 Theoretical framework
Immigrants lack host country specific knowledge upon arrival and this reflects the initial earnings differential. In other words, immigrants will have lower earnings than natives at arrival, this initial earning differential and lack of host country specific knowledge is due to, e.g., immigrants do not have the same information about the host country's customs and practices as natives does. Another factor is that, due to this lack of host country specific knowledge, in combination with being on the labour market in the host country for a shorter period of time than natives, immigrants are less attractive to the employer than their native counterpart. This host country specific knowledge could be, language skills or education, since human capital are not always perfectly transferable across countries. One possible reason for why the employer do not see recent immigrants as attractive, other than previously stated, is that they have more information about native’s productivity than the immigrants. Immigrants are expected to catch up to the wages of natives as years since migration increase, i.e., the immigrants gain host-country specific knowledge and immigrant earnings has a faster growth rate than earnings of natives. Since one additional year in the host country means one more year of host country specific knowledge, i.e. human capital, and one additional year in the labour force for the immigrant, whereas the natives only obtain one additional year of experience on the labour market (Chiswick, 1978). However, immigrants might not have higher earnings than natives, and more recent cohorts of immigrants might not even reach parity with individuals born in the host-country (Borjas, 1985). Immigrants that migrate when they are at a younger age will assimilate better and faster in the labour market than immigrants that migrate when they are at an older age. This phenomenon may be explained by the fact that those who migrate at a young age are raised and have similar lives as their native counterpart, i.e. go to the same school, learn the language, customs and culture at a young age (Giri, 2018).
The level and speed of assimilation on the labour market also depend on what type of immigrant they are, i.e., refugees, labour immigrants, etc., how old they are at arrival, their educational and ability level etc. Which means that a highly educated immigrant is expected to assimilate better in the host country compared to a low educated immigrant (Chiswick, 1982).
For example, family reunion immigrants are expected to assimilate faster than refugees, since they have an easier access to their social network (Bevelander, 2011). When comparing high educated refugees and low educated refugees to their native counterpart, high educated refugees is expected to assimilate somewhat faster than the low educated refugee (Bevelander, 2011;
Lundborg, 2013). Initial earnings of female immigrants are expected to be higher than those of
native females, and that years since migration has a negative impact on the earnings of immigrant women (Long, 1980).
Geographic distance as well as cultural similarities also plays a part. Immigrants from more distant countries or countries with dissimilar culture, e.g. language and customs, is expected to have lower initial earnings and employment rates than immigrants from countries that are geographically and/or culturally closer. But, if we would hold all factors constant, the earnings and employment rates of immigrants from a country with a different language would increase more with years since migration, but by a decreasing rate, ceteris paribus. However, initial earnings and employment rate would be larger for immigrants originating from a country with similar language (Borjas, 1987; Borjas, 1994; Chiswick, 1978). Furthermore, language proficiency is expected to increase the initial earnings for immigrants in host country (Borjas, 1987; Chiswick and Miller, 2009; Dustmann and Fabbri, 2003; Dustmann and Van Soest, 2002).
The current political landscape in source country also has an effect on the assimilation rate of immigrants, whereas immigrants originating from countries with limited freedom, and especially limitations appearing more recent, is expected to have higher assimilation rates than immigrants originating from countries similar to the political landscape of the source-country.
This might be explained in the sense that the cost of returning for an immigrant originating from a political authoritarian country is very high and they might therefore be more motivated to do well on the foreign labour market (Borjas, 1987). Refugee women are expected to assimilate faster in employment rate compared to native women than refugee males does to native males (Lundborg, 2013). The initial earnings of refugees are expected to be lower than labour immigrants’, the potential success on the host-country labour market are decided by some factors in the migration process playing a part as well. Labour immigrants are expected to be more self-selected in the sense that they to a larger degree calculate the cost and benefit of migrating to another country for work, thus having a larger focus on the economic impact of the migration. Refugees have other factors to keep in mind, with the largest being their safety.
Labour immigrants are therefore expected to be more suited and prepared for the labour market when they arrive to the host-country than refugees and are therefore likely to integrate faster than the refugees (Chiswick, 1982).
4 Literature review
Chiswick (1978) showed that earnings increase at a faster pace for immigrants than for natives, the factor behind this acceleration can be ascribed to years since migration. As a year goes by in the host country, natives only obtain the extra year experience, whereas immigrants have obtained another year’s experience in the host country as well as host country specific knowledge, which increases the employment probability as well as increase their productivity and thus earnings. Another reason for the initial disadvantage is the investments in post-school training immigrants make when arriving to host country, which depresses initial earnings for them to subsequently rise. Chiswick’s (1978) results also show that the percentage increase in earnings from year since migration are roughly equal for all educational levels. In regards of other English-speaking nations, the years of schooling as well as labour market experience are more transferable between the two countries. He also found that closer cultural background and shorter distance between source country and host country has a positive effect on initial earnings. As mentioned in the theory, the initial disadvantage in earnings may be due to the fact that immigrants often have less information about the host-country’s language and customs etc.
which means that they are less attractive for the employer and to what degree cultural differences exist. Chiswick’s (1978) results can lead to the conclusion that similar cultural background with natives gives one an advantage, for example, a shared language might enable one to adapt into appreciated cultural characteristics faster than those who do not share the same cultural background nor language. After 13 years since migration, the earnings of the immigrant will equal the earnings of their native counterpart, and one additional year after that or more will result in higher earnings for immigrants than natives. 20 years after migrating, Chiswick found that immigrants had six percent higher earnings than natives (Chiswick, 1978).
Continuous study made by Chiswick (1982) yield results emphasizing the importance of country of origin for an immigrants’ assimilation process as well the reason for migrating, either being a refugee or economic migrant. The results showed that economic migrants will adjust better and experience greater success on the labour market, compared to refugees, as well as reaching parity in earnings with their native counterparts will occur at around 11 to 15 years in the US, whereas refugees never reaches parity in earnings with the native counterpart or that this takes substantially longer time than it does for economic migrants. The findings also show that refugees have lower earnings than economic migrants, as well as being disadvantaged in terms of returns to human capital, where economic migrants’ effect of schooling and country of origin labour market experience is larger on earnings than for refugees (Chiswick, 1982).
The continued study by Borjas (1985) shows different results from those of Chiswick (1978), that immigrants most likely will not catch up with the earnings of natives, and if they do so, the assimilation process is longer than the results shown in Chiswick (1978). Borjas (1985; 1994) argues that one should not use cross-sectional data when studying immigration and labour market outcomes, since these estimates are snapshots and give misleading results.
This is due to the fact that immigrants arriving today may have different skills, human capital etc. than immigrants arriving 30 years ago (Borjas, 1985; Borjas, 1994). Borjas (1985) studies how immigrants perform in the US and extended the work of Chiswick (1978) when he looked at the cohort- and ageing effects that occurs and made separate assimilation equations for natives and immigrants. He also made the assumption that the period effects are the same between natives and immigrants. He found that, when it comes to earnings assimilation, the immigrants that arrived before 1975 may catch up with the natives, but the ones arriving after 1975 are probably never catching up with the natives. This can be explained by the fact that high-skilled individuals immigrated before 1975 and after 1975 there were low-skilled immigrants migrating to the US, i.e. a bias can be seen in Chiswick’s (1978) results, seeing as positive selection occurred prior to 1975, means that earlier cohorts was better suited for the US labour market than the cohorts arriving later was (Borjas, 1985).
The economic assimilation process of immigrants in terms of earnings and employment levels are that the initial earnings at entry are lower when compared to their native counterparts (Giri, 2018; Hirsch, et al. 2014). However, this gap appears to be lower among female immigrants than among male immigrants (Lundberg, 2013). To what degree the immigrants have problems getting employed, earnings assimilate etc. will differ depending on where the immigrant migrated from. More geographically and culturally close the country is, the faster the assimilation will be (Galloway and Aaberge, 2005). Galloway and Aaberge’s (2005) results shows that immigrants’ wages increases relative to the wages of the natives, and that these results are due to multiple factors, such as, their performance on the Norwegian labour market, language proficiency, knowledge about culture in the host-country and also welfare systems, e.g. tax- and social security systems. This is also shown in the results in Hammarstedt and Shukur’s (2006) study on Sweden, that earnings differ between immigrant cohorts, especially exposed to the difference are immigrant cohorts from non-European countries which suffers from lower earnings in comparison to other immigrant cohorts and natives. Immigrants are also at a disadvantage in terms of initial earnings, and the largest disadvantage are found for immigrants originating from southern Europe. However, southern Europeans and African immigrants exhibits the highest earnings assimilation rate as time spent in host country
increases, despite this, African immigrants will still not be in parity with earnings of natives during their first 20 years in host country (Hammarstedt and Shukur, 2006). However, Damas de Matos (2016) results shows that immigrants will catch up with natives after 10 to 15 years in the host-country’s labour market.
One possible explanatory factor for why some immigrant groups catch up with natives when it comes to annual earnings is presented by Damas de Matos (2016), which results shows that immigrants in Portugal were more mobile and changed jobs more frequently than natives, she also found that there is a link between the situation of immigrants changing jobs more frequently, e.g., often to a higher-wage paying firm and then catch up in the immigrants wage.
A positive relationship between earnings and English-language requirements of occupations were found in the study made by Chiswick and Miller (2009). More exactly, that a one standard deviation change (the marginal effect of the variable) in the required level of English for an occupation was associated with a large percentage increase in earnings for natives and refugees. This factor is also essential for job-matching, as findings show that workers lacking in language skills do well in occupations that does not require a high level of language skills but do poorly in occupations that require a high level of language skills (Chiswick and Miller, 2009). Their results are in line with findings of Dustmann and Van Soest (2002) and Dustmann and Fabbri (2003) who found that language proficiency increases the employment probability as well as increase the earnings. Moreover, Lundborg (2013) found that it is harder and more problematic for immigrants to assimilate if the immigrant migrate from Muslim countries than non-Muslim countries, and this is similar to previous studies by, for example, Chiswick (1978) and Galloway and Aaberge (2005), since there are large cultural differences between Sweden, which is a Christian country, and Muslim countries (Lundborg, 2013).
What also plays a part in the assimilation process of these are the education level obtained. Findings show that immigrants with a higher education level have a higher employment probability than those who do not (Bevelander, 2011; Bevelander, 2016). When it comes to differences in educational levels between immigrants and natives, Hirsch et al.’s (2014) results shows that, in Germany, there is not a large difference between native’s educational level and immigrant’s educational level. Due to the fact that immigrant’s education and human capital from the source country are not perfectly transferable to the host country (Lundborg 2013; Hirsch et al., 2014), the earnings that these immigrants receive is assumed to be lower in the host country than in the source country, which leads to differences in wages (Hirsch et al. 2014).
That source country culture regarding participation rates on the labour market for female immigrants play a role in their participation rate in Sweden was found by Neuman (2018). She also found that those with lower labour force participation in source country assimilates faster than those with higher participation rates in source country. Furthermore, Neuman (2018) states that some of the institutions in Sweden involving parental leave and childcare can be an explanation for this. Findings also showed that male immigrants to a larger extent participated in the labour force than female immigrants (Neuman, 2018).
That both male and female family reunion immigrants had higher employment probability than those of refugees, was found by Bevelander (2011) and as stated in the theory, this might be due to the social network already existing in the host country. Findings also showed that having children had a positive impact on employment probability for males, whereas for women, having children had a negative effect on employment probability.
However, results found by Husted et al. (2000) when studying the employment probability for refugees, non-refugees and natives in Denmark, show that after 5 to 10 years in Denmark, the employment probability of refugees are reaching the same level of employment probability rate as natives. But the geographical differences are also observed, whereas refugees from e.g. Africa enter the host country with significantly lower employment probability when compared to refugees originating from e.g. Europe and Latin-America (Husted et al., 2000).
Lundborg (2013) yield results arguing that refugees in Sweden were out of the labour force for a long time after arrival. Furthermore, no matter if the refugee resides in Sweden for a long time, the gap between native Swedes’ employment rate and refugees’ employment rate will not tangent, nor cross, regardless if years since migration is large.
Findings on the US labour market show that refugees often have lower initial earnings (Duleep and Dowhan, 2002; Giri, 2018; Husted et al., 2000) as well as employment rate than non-refugees but that they experience a larger growth in earnings than both non-refugees as well as natives (Duleep and Dowhan, 2002; Giri, 2018). The reason for this initial disadvantage in earnings and employment rate might be due to the fact that they do not migrate by choice and thus lack the preparation that a non-refugee has before migrating (Giri, 2018). Female and male refugee start off at similar shares in terms of relative employment, when compared to their native counterpart, but that female refugee assimilates those of female natives more than male refugee does to native males (Lundborg, 2013). However, the initial labour market disadvantage in earnings and employment probability tend to disappear as time in host country increases (Bevelander, 2016; Giri, 2018; Husted et al., 2000). Giri (2018) found that younger refugees’
wages assimilate faster than older refugees. The integration into the labour market in the host
country is significantly slower for refugees than for other immigrants, especially when compared to labour immigrants (Bevelander, 2016).
Studies obtaining educational- and gender specific differences in annual earnings assimilation are, from our research, somewhat restricted when focusing on these particular groups of refugees arriving to Sweden during this time period, which we hope to contribute with in this thesis. We will contribute to previous research with an insight if these, educational and gender-based differences, has any effect on the annual earning assimilation and employment probability for immigrants, more specifically, on the Balkan refugees in Sweden.
Previous studies have been made regarding Sweden, one of the main countries on the receiving end of refugees following the Yugoslav Wars. However, to the best of our knowledge, there are no collected studies on the particular subjects chosen for this thesis, which are; annual earnings assimilation and employment probability with focus on Bosnia-Herzegovina, Serbia and Croatia, during this particular time, and with respect to educational level and gender.
5 Data
This study uses LOUISE-data from Statistics Sweden and includes a random sample of foreign- born individuals’ resident in Sweden in the age of 20-64 between 1990 and 1999, as well as a sample of native Swedes in the same age group and same time period. Moreover, our data includes information about the individuals in, our three cross-sections, 1990, 1995 and 1999.
This study will only focus on individuals who were born in Serbia, Bosnia-Herzegovina, Croatia or Sweden, who were in the ages between 20-64 during the observation period. We have removed the observations with missing values on the variables used in the analysis. We then obtained two random samples, where the one used to study assimilation in employment included 107884 individuals, 97980 of the total sample are natives, i.e., from Sweden. To study annual earnings assimilation, individuals without positive annual earnings were removed from the study, this resulted in a sample group totalling 92100 individuals, of which 87326 are natives.
Immigrants are identified and grouped by region of birth, that is, one group born in Bosnia-Herzegovina and one group is a merger of individuals born in Serbia and Croatia labelled Serbia-Croatia. The reason for merging Croatia and Serbia into one group was due to our sample of individuals from Croatia not being sufficiently large, results from this group could therefore yield misleading results. Yugoslavia in our regressions is the average between
our sample groups Bosnia-Herzegovina and Serbia-Croatia. Since our data did not include Serbia as an independent country, but a part of Yugoslavia, we interpret this group as Serbia.
A cohort is a group that share characteristics, for example year of birth. Here, the cohorts are groups that share time of arrival. The first cohort is made up of the refugees that arrived between 1990 and 1995, and the second cohort is made up of the refugees that arrived between 1996 and 1999. The frequency of refugees from, e.g., Bosnia-Herzegovina be larger in one cohort than another, which, in a cross-section data, would give misleading results. As shown in the literature review, Borjas (1985) argues that one should not use cross-section data in studies like ours, since this could lead to bias results if there are differences between the cohorts. The difference could be in, e.g., skills, so, e.g., earlier cohorts may have lower skills than more recent immigrants, and this would give misleading results (Borjas, 1985). Therefore, our study is based on longitudinal data, which is collected data over several years, 1990, 1995 and 1999, this allows us to follow the individuals over time. There might be other types of immigrants, besides refugees, in these groups. However, since there was a pending war during the years of migration, we assume that most of the immigrants are refugees.
Individuals might leave or enter our sample for different reasons, such as death, turning older than 64, turning 20, an immigrant might enter through immigration during the years, individuals could also leave the sample through moving to another country, or return migrate to their source country. These factors are possible and not taken into account in this thesis, nor is the fact that our data is limited to a ten year period, from 1990 to 1999. This means that we are unable to observe the refugees’ long-term assimilation, and might lead to misleading results if, for example, the most able refugees return migrate to source country.
Our empirical analysis includes the variables; age, educational attainment, number of children in the household, gender, if the individual resides in a metropolitan area or not, marital status, employed or not and their annual earnings. Annual earnings are defined as income from wage-employment, self-employment and sickness pay. Employment is defined as a person being employed. According to Statistics Sweden, an individual can be characterized as being employed if the individual has been working during the month of November. A negative aspect with this way of characterization is that everyone, despite those not actually working a full year, will be included as employed. This can inhibit the average annual earning of the groups with fewer months worked in a year.
We also need to be cautious when drawing conclusions regarding the impact that education, i.e. the variables; primary school, high school and university, has on the labour market assimilation. Since we lack information regarding the individual’s educational level for
a large fraction of our sample. As in our summary statistics, an improbable share of low educated people in Sweden are displayed, the sample groups from the former Yugoslavian countries also consists of a small fraction of low educated individuals.
5.1 Descriptive statistics
Table 1 shows the descriptive statistics of the males in the samples included in the years 1990, 1995 and 1999. As seen in the table, natives were on average older than immigrants with an average age of 42 compared to the Bosnia-Herzegovinians average age of 37 and the youngest group of Serbia-Croatian with an average age of 36.
Regarding educational level, a very small difference can be seen between the two immigrant groups. 0.0293 percent of the males from Bosnia-Herzegovina had a primary education as their highest achieved education, as 17.3 percent of the native Swedes had primary education as their highest achieved educational level. Whereas the highest share of university graduates were found in the Serbia-Croatian group, with 36.4 percent having a university education, the average for Swedish males was 28.4 percent. However, a large fraction of the sample lacked information regarding their educational levels, which might lead to misleading information regarding this group of immigrants.
Refugee males were to a higher degree married than native Swedes, with 70.5 percent of the refugees from Serbia-Croatia being married while 48.3 percent of the natives were married. Refugees, on average had more children living at home than natives. Natives had, on average, 0.879 children living at home, in comparison to the refugees from Serbia-Croatia which had the largest number of children at home, more precisely, an average of 1.703 children.
There are no significant differences between the refugees and natives when it comes to frequency of residing in a metropolitan area, all groups were around 33 percent.
The average value of females in the samples included in the years 1990, 1995 and 1999 are shown in table 2. As for males, native women were on average older than the refugees. Where the refugees were around 35 years old when they enter Sweden and, on average, natives were 42 years old.
Looking at education, natives had a higher share of people with primary school education as their highest achieved education, meanwhile Bosnian-Herzegovinian refugees does not have any women with primary school as their highest achieved education. At the same time, a higher degree of the refugees has a high school education as their highest achieved education, in comparison to natives. The results are similar within our group of university educated, where Bosnian-Herzegovinian women are to a higher degree university educated, 36.7 percent, than natives, that are least university educated, by 33.1 percent. However, as mentioned earlier, results regarding education might be misleading for these particular immigrant groups since a large fraction of the sample lacked information of educational level.
Native women were to a lower degree married than the refugees, as 52.9 percent of native women are married, and 73.5 percent of the Serbia-Croatia women were married.
Refugee women have more children living at home than natives, with Serbian-Croatians have the largest number of children living at home. Native reside in a metropolitan area to a lower degree than the refugees, but the results do not differ significantly, since 33.5 percent of the native women and 40 percent of the Serbia-Croatian women reside in metropolitan areas.
Results in table 3 shows the percentage difference in the share of employed between refugees from former Yugoslavia and natives, and the differences over the three cross-sections. This measure presented in table 3 is calculated separately for each gender and cohort in our three cross sections 1990, 1995 and 1999. All values in the table are negative for the immigrant groups, except for Bosnian-Herzegovinian women in the 1990 cohort. This shows that the employment rate for refugees were lower compared to their native counterpart. We also see a pattern of male refugees having a lower differential in share of employed, with the exception for Bosnian-Herzegovinian females in the 1990 cohort. And lastly, that immigrant groups do not converge to the share of employment to those of natives during our researched years.
The initial difference in share of employed for Bosnia-Herzegovinian males in 1990 was 43.4 percent lower than their native counterpart, whereas the Serbia-Croatian men had a share of employed that was 39.6 percent lower. In 1995, the disadvantage in share of employed increased to roughly 84.3 percent lower for Bosnia-Herzegovinian men and 75.3 percent lower for Serbia-Croatian men. In 1999, the differentials in share of employed for both groups had decreased. When we look at the different cohorts and their disadvantage in share of employed, we can see that, in 1999, refugees in the 1996-1999 cohort had a larger disadvantage than those in the 1990-1995 cohort. While women in both cohorts had a larger disadvantage in share of employed than the males.
Results in table 4 yields the percentage difference in the annual earnings between refugees and natives, and the differences over the three cross-sections. This measure is calculated separately
for each gender and cohort in our three cross sections 1990, 1995 and 1999. The majority of the values in this table are negative, which means that these immigrants faces a disadvantage to their native counterparts in annual earnings, in all three cross sections and in both cohorts.
As the table shows, for Bosnian-Herzegovinian men in 1990, annual earnings were 23.8 percent lower than their native counterpart, and in 1995 this was reduced to be 22 percent lower.
For the females from Bosnia-Herzegovina, the annual earnings were 31.6 percent lower, but in 1995, this disadvantage was reduced to 23.8 percent. Serbia-Croatia women face lower annual earnings than the corresponding males. However, Bosnian-Herzegovinian refugees had a larger differential than Serbia-Croatian refugees.
In 1999, the gap in annual earnings had been reduced for all groups, but do not converge with natives. Moreover, these results show that males in the year 1999 in the 1990-1995 cohort had 7.2 percent lower annual earnings, while women had 8.6 percent disadvantage in annual earnings, in 1999. Males in the 1996-1999 cohort had in 1999, 13.1 percent lower annual earnings and women 14.7 percent lower annual earnings. Furthermore, we see that males have smaller differentials in annual earnings than women in both cohorts in our sample, except in 1990.
6 Methodological framework
We are going to use quantitative methods and more specifically, a series of regression analysis, i.e., longitudinal data analysis. Longitudinal data analysis has multiple benefits, such as the ability to see whether or not there are any longitudinal- or cross-sectional correlation between our dependent variables and our explanatory variables, it is also useful in distinguishing between the ageing effect and the cohort effect (Diggle et al., 2013). As being described in the theory part, Borjas (1985) criticizes Chiswick's (1978) method where he observes snapshots and not the same individuals over time, which leads to cohort effects. Borjas (1994) states that avoiding cohort effects is done through following the same group of individuals over time. We are therefore going to follow these individuals, our cohorts 1990-1995 and 1996-1999, over time, i.e. during the 1990s, in three cross-sections 1990, 1995 and 1999. Natives in the ages 20 to 64 are used as a control group, in this thesis. We ran two regression, for employment probability and annual earnings, as following:
Prob(Y! = 1) = ß𝑋 + 𝛾"𝑌𝑆𝑀#+ 𝛾$𝑌𝑆𝑀#$+ 𝛿"(𝑎𝑔𝑒) + 𝛿$(𝑎𝑔𝑒)$+ ∑%#&"𝜃#(𝐶𝑂𝐻)+ 𝜀 (1) 𝑙𝑜𝑔(𝑊#) = ß𝑋 + 𝛾"𝑌𝑆𝑀# + 𝛾$𝑌𝑆𝑀#$+ 𝛿"(𝑎𝑔𝑒) + 𝛿$(𝑎𝑔𝑒)$+ ∑%#&"𝜃#(𝐶𝑂𝐻)+ 𝜀 (2)
The probability of being employed is the variable to be explained in equation (1) and is a dummy variable which will yield the value 1 if the individual is employed and 0 otherwise.
YSM is years since migration and will yield a value of zero if the individual is a native, and a positive number between 1 and 9 if the individual is a refugee, which shows the number of years a refugee has been in the country. COH is a dummy variable for time of immigration, i.e., cohort, here being either refugees entering Sweden sometime between 1990 and 1995 or between 1996 and 1999. If an individual entered Sweden between 1990 and 1995, cohort for 1990-1995 will have the value 1, and 0 otherwise. If the individual entered Sweden between 1996 and 1999, 1996-1999 cohort will yield a value of 1, and zero otherwise. Swedes will yield value 0 in both cohorts. Age is calculated as the year of 1990, 1995 and 1999 minus year of birth. YSM and age both have an additional variable which is squared, to allow these variables to have a non-linear effect on the outcome variable, where the squared age is divided by 100.
X is a vector of control variables including educational level, number of children living at home, marital status, if the individualresident in a metropolitan area.
𝛾" is the assimilation effect i.e., the yearly percentage change for refugees, in relation
to natives, in employment probability, and annual earnings. If this coefficient yields a positive value, there is positive assimilation, which means that the gap between refugees and natives in terms of employment probability and/or annual earnings decreases, as years spent in Sweden increases. If this coefficient yields a negative value there is a negative assimilation, meaning that the gap in employment probability and/or annual earnings between refugees and natives are increasing as time spent in Sweden increases. A value of zero means that there is no assimilation takes place and the gap remains constant as time passes. Meanwhile, 𝛾$ is the squared assimilation effect, which means that the assimilation effect will diminish over time, i.e. is non-linear, so if the value is negative, and vice versa if positive. But if this coefficient yields a value of zero, the assimilation is linear. ẟ is the age effect which, according to Altman (2014) occurs when one gathers experience from chronologically getting older. It is added as an attempt of distinguishing years since migration (YSM) from the effect that ageing has on experience (Altman, 2015), and is assumed to have a positive value. The cohort effect, 𝜃, can take any value, both positive and negative, but is expected from our theoretical standpoint to yield a negative value, since immigrants are expected to have a lower employment probability than natives at arrival. 𝜃 illustrates the initial employment probability differential, and initial annual earnings differential, between the refugees and their native counterparts, 𝜃" shows the initial employment probability differential, and initial annual earnings differential for the first
cohort, entry in 1990–1995, and 𝜃$ shows the initial employment probability differential and initial annual earnings differential for the second cohort, entry in 1996–1999.
The results for equation 1, employment probability, will be in percentage points. The same regression is run for equation (2) as for equation (1), but here log(W), the log annual earnings, is the variable to be explained. In this case the sample is smaller since we only include individuals with positive annual earnings when running the regression. The results on annual earnings, equation 2, will be in log points, so we will need to antilogarithm them after running the regression.
We included a vector containing educational level, year of arrival, age, and if the individual resides in a metropolitan area, in our regression to control for other factors that can influence our variable to be explained, besides years since migration. The cohorts are used to group and distinguish between foreign-born individuals immigrating to Sweden between years 1990-1995 and 1996-1999. However, we do not take all relevant factors into account, such as language proficiency and differences in culture. The educational levels are Primary School, High School and University in our estimates and are computed as dummy variables. Where the dummy variable for “Primary School” will yield a value of 1 if the individual has attended schooling for less than ten years, and 0 otherwise. Our dummy variable for “High School” will yield a value of 1 if the individual has attended schooling for 10-12 years of schooling, 0 otherwise, and our dummy variable for “University” will yield a value of 1 if the individual has more than 12 years of schooling, and 0 otherwise. We then used primary school as reference group in our regressions for differences between educational levels.
The geographical areas will be presented as a dummy variable labelled “metropolitan”, which will be equal to one if the individual reside in a metropolitan area, this area represents Stockholm, Gothenburg and Malmö, and will be equal to zero otherwise. The metropolitan area in our data is Stockholm, Gothenburg, and Malmö, which is defined as the definition by Statistics Sweden (Statistiska Centralbyrån, n.d.).The variable “number of children living at home” includes all children, both underage, i.e. younger than 18, and adult, i.e. older than 18 or is the age of 18. These models cover the possible assimilation effects, initial differences between natives and refugees as well as the change in differences over time
This model will be repeated for both genders separately, to see gender differences in assimilation, and for education to study differences in assimilation based on educational attainment. The explanatory variables chosen in this regression are based on earlier mentioned theories regarding immigrants’ employment probability and annual earnings and the effect on the assimilation of the variable to be explained. This model is a simplification of the reality,
and due to our many explanatory factors, can give us more information about what causes the assimilation on the labour market, depending on which gender and educational level one has.
7 Results
7.1 Gender
7.1.1 Employment probability by gender
Since our subject of study is to find out differences in assimilation of employment probability and annual earnings, two regressions were computed as described in the methodology part of this thesis, one for employment probability and one for annual earnings. The values of table 5- 6 is the average of our refugee individuals compared to their native counterparts.
We begin by looking at table 5 and male refugees, the employment probability of males from Bosnia-Herzegovina in the 1990-1995 cohort had an initial employment probability that was 94.5 percentage points lower. Corresponding group in the 1996-1999 cohort had a 76.6 percentage point lower initial employment probability. Bosnian-Herzegovinian males suffered the largest initial disadvantage of the two male groups in the 1990-1995 cohort, while Serbian-
Croatian males suffered the largest initial disadvantage in employment probability compared to their native counterpart in the 1996-1999 cohort.
Moving to the assimilation effect of the two male refugee groups, which is shown in the
“years since migration” variable, Serbian-Croatian males had the highest assimilation effect from an additional year in the country. Increasing their employment probability by 16 percentage points, however by a diminishing rate as seen from years since migration raised to a power of 2. One can also see that the assimilation effect on employment probability is larger for males than for females, regardless of country of origin, but males’ assimilation effect will diminish faster than the females.
Looking at personal characteristics and the change in employment probability for refugee males, age shows a positive correlation, and equal value of increasing the probability of being employed by 5.2 percentage points for both male groups. However by a diminishing rate as seen from age raised to a power of 2. Being married also shows to have a positive impact on the probability of being employed, however, Serbian-Croatian males see a larger increase than Bosnian-Herzegovinian males, with an increase of 11.7 percentage points. Regarding education, a high school or university education increases the probability of being employed by similar values for the two male groups once again. However, Serbian-Croatian males increased their employment probability to a larger extent than Bosnian-Herzegovinian males, with Serbian-Croatian males increasing their employment probability by 4.5 percentage points for high school graduates and 4 percentage points for university graduates. Number of children living at home are shown to inhibit the probability of employment. For Serbian-Croatian males an additional child living at home reduce their probability of employment by 1 percentage point, while Bosnia-Herzegovina males only saw a 0.4 percentage point decrease. Residing in a metropolitan area did not yield any statistically significant values for males, which means that we cannot statistically conclude that these values differ from zero.
If we then look at table 5 and females, similar patterns found for males were found for female’s initial employment probability compared to their native counterpart. Bosnian- Herzegovinian females have the largest initial disadvantage in employment probability by 88.1 percentage points. We also see that, as for males, Serbian-Croatian females has the largest assimilation effect, increasing their probability of being employed by 10.5 percentage points, however by a diminishing rate. Meanwhile, females from Bosnia-Herzegovina assimilate at an increasing rate. We also see that, as for males, the females in the 1990-1995 cohort suffers a larger initial disadvantage in employment probability than their corresponding group in 1996- 1999.
Females from Serbia-Croatia possess a larger assimilation effect in comparison to Bosnia-Herzegovina females. But this assimilation in employment probability will increase with a diminishing rate for Serbia-Croatian females, which goes in line with the results for their corresponding males. Meanwhile, the results show that Bosnia-Herzegovina females employment probability assimilation increase with an increasing rate, at a 10-percentage level of significance. One observable pattern is that males have a larger assimilation effect from years since migration than females, which is shown in the “years since migration” and “years since migration raised to a power of 2” variables.
Completing table 5 with female personal characteristics, we see that age is correlated with higher employment probability, however by a diminishing rate. The diminishing rate shows a trend of diminishing even further as the effect of age on employment probability increases. Such as Serbia-Croatian males having the largest increase, but also the largest diminishing rate of the refugee groups. As well as Serbian-Croatian females having the largest increase in employment probability from being one year older, but also the largest diminishing rate of the refugee groups. Being married is positively correlated with employment probability, however more so for male refugees than female refugees. Ranging from the highest of 11.7 percentage points increase in employment probability from being married for Serbia-Croatian males, to the lowest which are females from Serbia-Croatia who see an increase of 5.9 percentage points in employment probability. Attending high school as highest education has a positive effect on the assimilation for all. A university education leads to an even higher increase in employment probability than high school for females. This holds true for both genders but shows to be more important for females. The only variable that was included and shows to have a negative impact on the assimilation of employment probability for refugees are number of children living at home. Additional children will slightly inhibit the assimilation process, especially for female refugees. Lastly, residing in a metropolitan area is beneficial for both female groups.
7.1.2 Annual earnings by gender
Looking at table 6 and starting with the results showing average annual earnings for refugee males with their native counterparts as reference group. The results show that immigrant males from Bosnia-Herzegovina suffers the largest disadvantage in terms of initial annual earnings, with a 94.6 1percentage deficit in initial annual earnings in 1990-1995 cohort. The 1990-1995 cohort experience a larger initial disadvantage in annual earnings, than the initial disadvantage the 1996-1999 cohort group experiences to their.
The assimilation effect of both male groups shows to be increasing their annual earnings, however by a diminishing rate. Where the assimilation effect is larger for males originating from Bosnia-Herzegovina, than their corresponding group, and the diminishing rate of the assimilation effect is larger for the Bosnia-Herzegovina males as well.
Age also raise the annual earnings of both male refugee groups, around 17.6 percent for both male groups. However, by a diminishing rate for both male groups as seen from age raised to a power of 2. Being married also results in a larger annual earnings assimilation, an increase of around 23.5 percent, but a bit larger for males from Serbia-Croatia. Attending high school or university shows to have a positive effect on the annual earnings assimilation, whereas males
1 Percentages for annual earnings are calculated by; 𝑒!− 1
with high school as their highest achieved education increase their annual earnings equally, with 32 percent. The effect on annual earnings are larger if males have a university education.
However, Serbia-Croatian males benefit the most from a university education, as their annual earnings increase by 49.5 percentage compared to Bosnian-Herzegovinian males with a 48.7 percentage increase. The only variable in this regression that decrease the annual earnings is
“number of children”, which means that one additional child living at home will decrease the annual earnings for both male groups. This negative effect will be larger for males from Serbia- Croatia, than Bosnia-Herzegovina, with a 3.2 percent decrease in annual earnings. Residing in a metropolitan area will increase the annual earnings more for males from Serbia-Croatia than Bosnia-Herzegovina, however increasing annual earnings for both male groups. The increase in annual earnings that comes from living in a metropolitan area ranges between 11.6 percent and 13.1 percent.
Viewing the results for female refugees’ initial annual earnings compared to their native counterpart, the results shows, as for males, that the 1990-1995 cohort suffers a larger initial annual earnings disadvantage than the 1996-1999 cohort does. Where the initial disadvantage in 1990-1995 cohort ranges from 81.5 percent to 94.6 percent and 1996-1999 cohorts initial disadvantage in terms of annual earnings range between 74.3 percentages to 87.7 percentage.
However, Bosnian-Herzegovinian females suffers the largest initial annual earnings disadvantage of the two female groups, and the assimilation effect are higher for Bosnian- Herzegovinian females than for Serbia-Croatian females. As also shown for their corresponding males. However, all groups show a positive assimilation effect from additional years in the host country, a pattern showing in the two separate genders is that those experiencing a larger initial disadvantage in annual earnings has a larger assimilation rate.
Remaining variables for female refugees shows that age is beneficial for the annual earnings assimilation, increasing the initial annual earnings by 14.8 percent for both female groups, by a diminishing rate. Being married shows to increase the annual earnings, however more so for Bosnian-Herzegovinian females, leading to an increase of 1.5 percent, while no statistically significant values for Serbian-Croatian females are displayed. Attending high school or university shows to have similar effect in comparison to the corresponding native groups, and has a slightly larger impact on Bosnian-Herzegovinian female’s earnings assimilation. Increasing Bosnia-Herzegovinian female’s annual earnings by 35.3 percent from having a high school education and 59.7 percent from having a university education, in comparison to Serbian-Croatian females increasing their annual earnings by 34.6 percent for those with a high school education and 59.5 percent for those with a university education.
Residing in a metropolitan area also shows to increase the annual earnings of female refugees, similar values of 10.5 percent increase for Serbian-Croatian females and 10 percent increase for Bosnian-Herzegovinian females, similar to the results for males. Yet again the only variable that shows to inhibit the assimilation is having children living at home, displaying negative values for the earnings assimilation. Depressing the earnings assimilation by 7.2 percent for Serbian-Croatian females and 7 percent for Bosnian-Herzegovinian females.
7.2 Education
7.2.1 Employment probability by educational level
The question of study is partly if there are any differences in employment probability between males with and without a university education, which can be studied in table 7. The values displayed are the average of our refugee individuals with their native counterparts as reference group. The results show that men without a university degree will face larger initial disadvantage in employment probability in both cohorts, however the difference is not large.
The results show that the assimilation effect is larger for men without a university education than men with a university education, once again not large but still to be seen. The diminishing rate does not differ greatly between the two either. This means that the assimilation rate of refugee males from Serbia-Croatia without a university education will increase by a faster
diminishing rate than their corresponding group of Serbia-Croatian males with a university education, however not by much. As for Bosnian-Herzegovinian males, the results are not statistically significant which means that we cannot say anything about the diminishing rate of these individuals.
There are differences between males from Serbia-Croatia and Bosnia-Herzegovina and their assimilation effect on employment probability, the results shows that men from Serbia- Croatia will, regardless of educational level, have lower initial disadvantages in employment probability in comparison to men from Bosnia-Herzegovina in the 1990-1995 cohort. When viewing the 1996-1999 cohort, one can see the reverse results, that males from Bosnia- Herzegovina, regardless of educational level, face lower initial disadvantage in terms of employment probability than their Serbian-Croatian counterpart. However, the “years since migration” variable shows that Serbian-Croatian males’ assimilation effect is larger than the assimilation effect for males from Bosnia-Herzegovina, regardless of education level. Both male groups assimilate at a diminishing rate, although Serbia-Croatia males more so than their corresponding group of Bosnian-Herzegovinian males. But these results regarding the diminishing rate of the assimilation is not statistically significant for the Bosnia-Herzegovinian males.
As in table 7, it is of interest to study this for the females as well. Table 8 display results that are the average of employment probability for females with and without a university education, with their native counterparts as reference group. The results for females, in the 1990-1995 cohorts, goes against the trend found for male refugees, as females in the 1990-1995 cohort is found to have higher initial disadvantage in employment probability if they have a university education. In cohort 1996-1999, females from Bosnia-Herzegovina also follow this reverse trend, i.e. have higher initial disadvantage in employment probability if they have a university education. However, Serbian-Croatian females in the 1996-1999 cohort face a larger initial disadvantage in employment probability if they do not have a university education.
Females will, according to these results, have a larger assimilation effect if they have a university education, with Serbian-Croatian women experiencing a larger assimilation effect in comparison to their Bosnia-Herzegovina counterparts. But the results of the assimilation effect for females from Bosnia-Herzegovina with a university education are not statistically significant. Bosnian-Herzegovinian women with no university education have an assimilation effect that increases by an increasing rate, however this result is only statistically significant at a 10% significance level.
7.2.2 Annual earnings by educational level
Furthermore, the question of study also includes if there are any differences in annual earnings between male refugees, with and without a university education, which can be studied in table 9. These results are the average values of the group, with the reference group being their native counterpart. As mentioned earlier, one need to take the antilogarithm on the values presented in the tables representing the annual earnings results to receive the values in percentage.
According to these results, refugee males in the 1990-1995 cohort, face a higher initial annual earnings disadvantage if they have a university education, however not by much. The same trend can be found in cohort 1996-1999 for Bosnia-Herzegovina men. The opposite is found for refugee males from Serbia-Croatia in cohort 1996-1999, i.e. the initial annual earnings disadvantage is larger for those with a university education.
The assimilation effect for male refugees without a university education is larger than the assimilation effect for male refugees with a university education. This assimilation effect will increase the annual earnings as years since migration increases, however, by a diminishing rate.
The results shows that Bosnian-Herzegovinian males face a larger initial annual earnings disadvantage in comparison to their Serbia-Croatian counterparts, regardless of educational level. However, the assimilation of Bosnian-Herzegovinian males showed to be faster, but also diminish at a faster rate as seen from years since migration raised to a power of 2.
Table 10 displays the average results for annual earnings of female refugees with and without a university education compared to their native counterpart. As in table 9, one need to take the antilogarithm on the annual earnings values in table 10 as well to receive the values in percentage. A pattern that follows throughout this table is that the initial annual earnings disadvantage for university educated refugee females are higher compared to their corresponding groups of female refugees without a university education. This trend can be seen in both cohorts. As was the case for males, females from Bosnia-Herzegovina face lower initial annual earnings compared to Serbia-Croatia women. However, the assimilation effect is higher for Bosnian-Herzegovinian females, and we can also see that those with a higher assimilation rate has a higher diminishing rate. These findings goes in line with the findings for their corresponding males.
The assimilation effect for university educated females is larger than for those females without a university education, meaning that the annual earnings will increase at a faster pace for university educated than for those without a university education, however by a larger diminishing rate. These results are the reverse of the results found for their corresponding male group. The results shows a trend of increasing assimilation effects as initial disadvantage increases, for both females and males.
The regressions run in these 6 tables has an r-squared value between 0,1 and 0,3.
This means that between 10 percent and 30 percent of the dependent variables are explained by
