Constructions of self-identification: children of immigrants in Sweden

ABSTRACT This paper is about self-identification by a sample of young people with various migrant backgrounds in Sweden. In a survey we asked them how they present themselves to others in different contexts (in their schools and neighbourhoods or when they are out of Sweden). Our findings suggest that young people’s identification is not fixed. The paths to these various forms of identification are shaped by a variety of individual (class background, parents’ country of origin) and social factors (friendship networks and school composition). The results indicate that those who identified themselves as Swedish or hyphenated tended to be quite similar as regards their other characteristics, while those who presented themselves with the parents’ birth country or religious affiliation have very little in common.


Introduction
The external definition we are given is an inescapable part of our internal self-definition, actively constituting our identity. External labelling is more effective if the process is endowed with institutional legitimacy and governmental authority (Jenkins 2008). Classification of the population and the categorisation of people by state agencies is a clear example of such external defining. The production of 'ethnic' statistics or the statisticalisation of identity and the reductionist emphasis on differences in 'ethnic background' fails to take into account the growing proportion of young people with migrant parents who do not define themselves (at least not initially) by their migrant background. This paper is about the self-identification of a sample of young people with various migrant backgrounds in Sweden. In a survey, we asked them how they present themselves for others in different contexts (in school, the neighbourhood or outside Sweden). The results of the survey should be interpreted as a snapshot in a given time and space. As recent research emphasises 'Identities are never fully and finally "established"; instead they are seen always in process, always in a relative state of formation' (Rattansi and Phoenix 2005, 105). Despite the restrictions, the empirical material from this study contains information that enables us to examine whether these identifications are situated and changing or fixed. Moreover, we explore how the various characteristics of young people and the social environment around them are associated with a certain form of self-presentation in everyday life and in different contexts.
The process of the construction of 'otherness' and the span of identity positions in the pluralist immigrant societies of North America and European societies (with a shorter history of immigration) is not the same. Compared to North American research in the areas of racialised and ethnic identities, which has a more lengthy history, Swedish studies on this subject are rare. Studies on this issue in the Swedish context (as one of Europe's leading immigrant societies) can help us to gain a deeper knowledge of how the different institutional arrangements are experienced by the children of immigrants and how they act/react in different cultural and political climates.
The theoretical point of departure for this study is the sociological conception of identity, according to which identity is described by Hall (2006) as 'a "moveable feast"; formed and transformed continuously in relation to the ways we are represented or addressed in the cultural systems that surround us'. Identities, according to this perspective, are stories which we tell ourselves and others about who we are and who other people are. These narratives are often related to our own or to others' perception of being a member of a collective (class, national, ethnic or religious). These stories can change, shift and be contested and are often multiple.
Identities are processes of 'becoming' rather than of 'being' There are two main streams in research on ethnicized identities, as classified by Phoenix (2010). The first is the post-Eriksoniana psychological notion of identity, according to which identity is made up of 'clearly identifiable static qualities without which members of ethnic groups are likely to suffer damage' (Rattansi and Phoenix 2005, 107). This approach assumes that 'a strong, secure ethnic identity makes a positive contribution to the psychological well-being' of migrants and their descendants (Phinney et al. 2001, 502). In such an understanding, 'ethnic identity' is 'part of the "acculturation" process that takes place when immigrants come to a new society' (Phinney et al. 2001, 494). Consequently, researchers in this tradition use 'ethnic identity' and 'acculturation' interchangeably (Ibid.). In these 'essentialised, fixed understandings of ethnicities and identities', ethnic identities are assumed to be placed in a hierarchy of identity categoriesi.e. integrated, assimilated, separated and marginalised (Phoenix 2010, 314). Ethnic identity, according to this perspective, 'is ascribed and results from shared background and powerful bonds ' (2010, 300). In so doing, this research tradition fails to examine the relationality of the different components and how they influence each other. According to this essentialist understanding of ethnicity, 'One's ethnicity can be defined in terms of one's ancestral heritage, specifically the culture or cultures of origin of one's parents and grandparents' (Phinney 2005, 188). The main problem with this perspective, as Phoenix (2010) posits, is that it tends to reproduce an idea of ethnic identity that is 'static, coherent and often unitary'. We should talk about a process of identification rather than identity as a fixed essence, suggests Hall (2017), because identification is never complete but is always in process. He adds: 'Identity cannot be a fixed essence at all, as if it lay unchanged outside of history and culture, and this is so for one principal reason: identity is not given once and for all by something transmitted in the genes we carry in the colour of our skin, but is shaped and transformed historically and culturally' (Hall 2017, 127).
The second theoretical stream in the field (that serves as the point of departure for this study), claims that 'identities are socially constructed, multiple, potentially contradictory and situationally variable' (Phoenix 2010, 298). This view focuses on 'ethnicized rather than ethnic identities' and regards ethnicities as 'part of the process of boundary maintenance and so of border negotiations and contestations ' (2010, 298). Identities, in this interpretation, are processes of becoming rather than of being (Hall 2006). Key features of this perspective are set out below.
First, the relational character of any identity. The identity of an individual only acquires meaning and significance in relation to what it is noti.e. in relation to other identities (Rattansi and Phoenix 2005). So we could say 'Identity is never unilateral; what other people around me think about who I am is no less important than what I think about myself'. As Jenkins suggests, 'Others don't just perceive our identity, they actively constitute it ' (2008, 96). People who observe that they are lumped together by others in a group may accept the content and implications of that categorisation and internalise their subordination. However, being stigmatised by others through negative labels may also evoke the resistance of the group and could lead to a political process of mobilisation.
Second, individuals are structurally positioned within hierarchic economic, political, social and cultural systems. These structural positions shape a person's life opportunities through locating him/her within certain networks of power relations with various resources. When a society is divided by class, gender and ethic distinctions, individuals find themselves already positioned within a discursive field that is never entirely of their own choosing. They experience their shifting and intersecting positions through dominant discourses that tend to reflect the configuration of power relations in a given historical moment in a particular society.
Third, and related to the previous point, the logic of categorisation/ classification of a population by state agencies and civic institutions (through the assignment of identificatory labels) is powerfully constitutive and can have serious consequences for subordinated groups. Statistical reports produced by, inter alia, government bodies and census categories (which box the population into different 'ethnic' groups) are significant sources of such classification. In certain social contexts, like the Jews in Germany during the 1930s, identities are brutally forced on people when the work of boundary maintenance separates a population into 'us' and 'them'.
Fourth, identities are never fully established but, rather, always in process and a state of formation. Put another way, identity narratives are 'not fixed, immutable or primordial' (Jenkins 2008, 19); they can change, shift and be contested. As Walzer (1997) writes, when immigrants move across state frontiers, they carry their old values, attitudes and religions but they are subject to constant transformation by adjustment to the new environment. Consequently, their self-identification is precarious.
Fifth, self is not rational and conscious each time but de-centred by the operation of the unconscious (desires, motivations, projections and anxieties). The 'conscious' self is neither in complete control nor in possession of complete self-knowledge (Hall 2006).
In line with these theoretical frames, the assumption behind the construction of our questionnaire is that the labels which young people choose for their self-designation may change over time and settingsi.e. respondents may label themselves variously in different ways depending on the spatial context (in Sweden or abroad, at school or in the neighbourhood). Furthermore, the social climate (what is happening at a specific time in society and the surrounding milieu), can promote this or that kind of selfcategorisation. The construction of belonging, as Yuval-Davis (2006) reminds us, 'can be an act of self-identification or identification by others'.
As Rumbaut (1994), puts it, while ethnic self-labelling for some groups 'becomes optional and recedes into the social twilight', for others (in particular the stigmatised groups) it may become like 'a resilient resource or an engulfing master status'. Such self-labelling may be stronger among groups who experience more discrimination.
Remembering that there is no essential ethnic identity, we should acknowledge that ethnic self-labels are not necessarily related to the cultural attitudes of young people, nor to their everyday behaviour. According to Rattansi and Phoenix (2005, 107), knowing how young people label themselves 'does not indicate how they live their lives or what their cultural practices are ' (2005, 107) or, as Brubaker (2004) reminds us, labels for belonging to certain categories can be quite disconnected from actual social and cultural practices. Portes and Rumbaut (2001) demonstrated that the experience of the children of immigrants, their feeling of belonging to and attitudes towards the larger society differs substantially depending on which label they choose. In other words, these labels indicate the variety of structural constraints and social climates which affect these different forms of ethnic self-definition. The paths to various self-categorisations are shaped by the complex ways in which individuals' characteristics (their class background and other resources and experiences) interact with how society at large (the state, the school system and the media), treats them and affects their sense of who they are and where their place is.

The Swedish context
Having lost about a quarter of its population in nineteenth century, Sweden turned from a country of emigration to one of immigration in the twentieth. Based on data from Statistics Sweden (SCB 2017), of the roughly10 million population of Sweden, 18.5 per cent were foreign-born. The largest foreignborn groups were immigrants from Syria (170,000), Finland (150,000), Iraq (140,000) and Poland (91,000). Among the Swedish-born population, 5.5 per cent were native-born with two foreign-born parents and 7.5 per cent were native-born with one foreign-born parent.
The same percentages for individuals between 0 and 19 years were 11.2 (foreign-born), 13.8 (native-born with two foreign-born parents) and 12.0 (native-born with one foreign-born parent). Thus, about 37 per cent of the Swedish population in this age category had some form of migrant background at the time. Sweden, in this way, like most of our contemporary Western nation-states, is a cultural hybrid, 'mongrelized and diasporized beyond repair' (Hall 2017, 143). As previous empirical research has shown, immigrants and their children tend to be over-represented in the lower echelons of the labour market, with lower wages, poorer working conditions and less employment security (Behtoui and Neergaard 2010). Sweden's urban landscapes are marked by overlapping patterns of ethnic and economic segregation. The poor neighbourhoods of large cities are inhabited by a population whose backgrounds are predominantly in non-Western countries (Östh, Clark, and Malmberg 2015).
The xenophobic and racist underpinnings of the current anti-immigrant environment in Sweden have also had an impact on the identity formation of young people with migrant background. Violent radical nationalist racism has increased in recent years in Sweden. The aggressive extreme-right nationalist organisations propagate retreating to an ethnic absolutism. As a latecomer to the growing European extreme-right nationalism, the Swedish Democrat Party (Sverigedemokraterna) scored about 18 per cent of the total vote in the 2018 parliamentary elections. This party, with its historical roots in the old Swedish National Socialist movement of the 1930s, now has the potential to hold the balance of power between the main political blocks of centre-right and left-green political parties. Such a harsh and polarised socio-political climate may be matched by a retreat to more 'defensive identities' among young people with migrant background in response to the ensuing cultural racism (cf. Hall 2006). They may, construct a stronger identification with their parents' origins. The revival of religious orthodoxy or cultural traditionalism is another alternative in such a situation.

Previous research
Swedish survey-based empirical research in this field has the first theoretical streamnamely the psychological notion of identity and acculturationas its theoretical frame. The outcomes of the different acculturation types on the psychology, educational achievements and labour-market position of young people of migrant background have been the focus of these studies. Nekby and Rödin (2010) found that there are no significant differences in employment probabilities between the descendants of immigrants who identify only with the majority culture (assimilated) and those who identify with both the majority culture and the ethnic group (integrated). On the other hand, those who have a weak attachment to the majority culture (separated and marginalised identity categories) have fewer employment probabilities than the two categories mentioned above. Nekby, Rödin, and Özcan (2009) found that integrated males have a higher probablility of completing their tertiary education than those in the assimilated, marginalised and separated categories (no systematic differences were found for women). Furthermore, the authors showed that the frequency of acculturation types was different for the various immigrant groups. Virta, Sam, and Westin (2004) have shown that, among the descendants of immigrants from Turkey in Sweden and Norway, those who identify themselves as Turkish and those who have the 'integration' type of identification tend to have higher levels of adaptation. Hällsten, Edling, and Rydgren (2018) found that those offspring of immigrants who are oriented towards both the majority and their parents' culture (the integration category) have the highest level of adaptation and a more privileged background.
There are some qualitative Swedish studies that have used the decentred and de-essentialised notion of young people's identities (See e.g. Moinian 2009;Runfors 2016). Based on interviews with children (12-16 years old) with migrant backgrounds, Moinian (2009, 40) writes that 'Children's constructions of their own identity were fluid and varied, depending upon the various situations and contexts'. They believed that they were ascribed an identity by the majority society when they were denied the possibility to negotiate or challenge this categorisation. Runfors (2016Runfors ( , 1849 reports how her informants (young people with migrant backgrounds who had grown up in low-status urban spaces located on the outskirts of Stockholm) used 'the dominant national categorisation mode . . . the word invandrare (immigrant)', for self-labelling, which positions them in racialised and ethnified scales of judgement. Consequently, they developed a mode of selfdesignation that 'was not oriented towards ethnicity. Nor were they oriented towards the hyphenated identity positions . . ., they seemed to regard hyphenated identity positions as unavailable'. As an alternative, 'they spoke of themselves as immigrants, but no so much as an act of resistance. Rather they used the language of domination' (2016, 1856).

Data and measurements
Young people participating in this study are, at first hand, those who participated in the first wave of an earlier research survey and who were 15-6 years old in Spring 2016. To enlarge the number of descendants of immigrants, we added a random sample of children of immigrants from Asia (Iran + Iraq = 1,200), Africa (Somalia = 600) and another European country (Bosnia = 600) to the second wave of the survey. These specific groups were chosen since they have been the largest groups of immigrants from these regions to Sweden over the last 20 years. In collaboration with Statistics Sweden (SCB), we sent a questionnaire survey to these young people who were now 17-18 years old. Incentives for the young people to complete the questionnaire (cinema tickets) were included to maximise the response rate. After three reminder letters were sent out, 51.4 per cent of young people (52.5 from the initial group and 50.3 from the complementary group) completed questionnaires. All respondents from the second group and only those from the first group who had a migrant background and answered the questions are included in this study (total 1,619 individuals).
Alongside the results of the questionnaire, the following information was also obtained from the longitudinal register databases of Statistics Sweden: respondents' age, gender, own and parents' birthplace (grouped) and parents' highest educational attainments and job status based on the Swedish Standard Classification of Occupations (SSYK).
In line with our theoretical point of departure, we have used selfcategorisation questions with a wide range of choices. We assumed that adolescents might label themselves in different ways depending on the context. Furthermore, we did not expect that the labels they used in answering these questions would necessarily be related to their everyday cultural behaviour and attitudes.
We asked the respondents to what extent do they use the different designations when they present themselves for others. There was wide range of choices that the young people could make in categorising themselves, and the labels they chose could change over different settings (in school, the neighbourhood or outside Sweden). Such self-categorisation could include labels such as: 1) a plain national Swedish label (national); 2) a hyphenated Swedish label (Swedish-Iranian, Somali-Swedish); 3) an ethnic-minority group or national label (ethnic) connected to the parents' country of origin (e.g. Somalia or Iraq), and 4) a religious affiliation (e.g. Muslim, Christian, Jew). Among the proposed answers, we had an open-ended option (other) where the respondents could write an answer in their own words. These written selfdesignations were also coded either as one of the abovementioned four categories, or as missing (we had about 5 per cent missing after the final coding. For all labels in the school, neighbourhood and abroad we had 5.3, 5.1 and 4.8 per cent missing answers). 1 If the first two of these types of identification (national and hyphenated Swedish) are mainly in relative harmony with mainstream identities, the last two variants (ethnic and religious) indicate a greater endorsement of some form of otherness. The statement was measured on a five-point scale ranging from 'Never = 1ʹ to 'To a great extent = 5ʹ. Because the different types of self-categorisation were measured independently on continuous scales, the respondents did not need to choose one label over another. Instead, they expressed their sense of belonging to each label along a continuum.
The details of the respondents' immigrant background in this study were obtained from Swedish register data and sorted into the following regional categories, either based on the country of birth of the respondent (the so-called 'first generation', with subscript suffix 1 for each group) or country of birth of the parents (the so-called 'second generation', with subscript suffix 2). As a result, we have the following groups: 1) Mixeddesignated those respondents who had one Swedish and one foreign-born parent, 2) North-West or 'the Global North' countries contains the EU 15, Norway, Switzerland, Canada, Australia, New Zealand, the United States and is labelled as NW group, 3) Other European countries, 4) Asia and 5) Africa. 2,3 Swedish citizenship and living in a 'big city' (= 1 and otherwise = 0) are dummy variables. Respondents who answered that more than 70 per cent of students in their schools have an immigrant background were coded 1 (attending an ethnically segregated school) and zero otherwise. Those who estimated that more than half of their friends have the same ethnic or cultural background were coded as 1 (mostly immigrant friends), otherwise zero. Duration of residence of the family is a variable from register data (time of migration of parents). Respondents' socioeconomic background included their parents' highest educational attainment, job status (in both cases, the parent with the higher-status job and higher level of education determined the family's collective status) and total family income (grouped into 10 classes). An index was then constructed with these three factors (education, job status and income)named the family SES.

Results
As the descriptive data in Table 1 indicate, the duration of residence of the family of respondents who were foreign-born is shorter than for those who were born in Sweden. In our sample, those born in 'NWgroup' and 'other European countries' were, to a greater extent, residents of big cities compared to those from the same region who were born in Sweden. On the contrary, those born in 'Asia' and 'Africa' were less likely to live in big cities. The Swedish citizenship rate was very low for those who were born in the 'Global North' countries 'NW1 and). Regarding the family background of our respondents, we can observe that young people with Mixed and NW backgrounds had more resources (parents' education, income and job status) in their family. We can also see that those pupils with origins in Asia and Africa were more likely to attend segregated schools. However, we could see no systematic differences relating to our respondents' homogenous friendship networks. When we investigated the class background of those young people who attended segregated schools, we saw that they had statistically significant lower family resources (education, income and job status of parents) compared with those who not attend such schools. However, similar class disparity is not evident when it comes to the question of homogenous friendship networksi.e. even if respondent's friends had a migrant background, they had roughly the same class background. Figure 1 depicts four different types of self-categorisation. In our sample, the ethnic label (compared to other labels) is the dominant category, followed by hyphenated, plain national Swedish identity and finally religion. As Phinney et al. (2006, 77) state, in almost all American studies in this field, the ethnic label scores higher than all other types of self-identification, 'although there is wide individual variation'. Swedish data were also included (815 young 15-year-old Swedes with immigrant backgrounds) in the 2006 ICSEY, the International Comparative Study of Ethno-Cultural Youth Project (with 13 nations) on self-identification. As Phinney et al. (2006, 89) write, in all these countries, the ethnic label was 'stronger than national identity' (which, in the case of Sweden, means Swedish). While, according to results of this comparative study, 'ethnic identity'with a mean of 4.3 on a 5-point scalewas stronger than plain 'national identitywhich had a mean of 3.3 for all countriesthe corresponding figures for Sweden were 4.4 and 3.

2.
To what extent the various labels are chosen by our respondents may differ depending on the context. As mentioned above, we asked respondents about three different contexts (school, neighbourhood and abroad).

IDENTITIES: GLOBAL STUDIES IN CULTURE AND POWER
The most conspicuous was the tendency of young people with a migrant background to present themselves with the national label (Swedish) when they were away from Sweden compared to when they were there. Information provided by respondents of previous Swedish qualitative studies (Moinian 2009) explains this phenomenon as follows: either how people outside Sweden treat them -'Nobody counted me as a real Iranian there [in Iran] . . . They thought I was strange' (2009, 43)or respondents' own feelings outside Sweden -'When I visit my relatives in Iran . . . I think then that I am not like them, Iranians, but that I'm closer to Swedes ' (2009, 44).
A factor analysis (PCA) to find the patterns of correlation between the different types of identification (not shown here but available upon request), demonstrated that those belonging to national and hyphenated categories congregate on one axis and those who chose ethnic and religious labels on the other. In addition, young people in the Mixed group and those from Africa 1 are in reverse positions regarding their place on the first and second axes. Young people in the Mixed group tended to choose Swedish and hyphenated identity labels to a greater extent than the ethnic and religious ones. The opposite is the case for respondents belonging to the Africa 1 group.
Compared to American studies of self-identification (Portes and Rumbaut 2001), our rates of hyphenated-identification categories are lower. Phinney et al. (2006, 90) also report that hyphenated identification is pertinent for respondents within 'settler societies' such as Canada, New Zealand, Australia and the United States. As Walzer (1997,33) writes, many people in immigrant societies like the United States prefer to present themselves by a hyphenated identityfor example, that of an Italian-American. Such selflabelling 'symbolises the acceptance of "Italianness" by other Americans, the recognition that "American" is a political identity without strong or specific cultural claims'. However, hyphenated identification, as a discursive possibility, does not yet exist in Sweden, despite the large-scale labour-market migration of workers after the Second World War and the recent arrival of refugees.
It is worth noticing that, because of the lack of hyphenated identification as a discursive possibility, some young people with immigrant backgrounds in Sweden try to find other substitutes. One example is the Swedish 'inbetweenship' association (mellanförskapet) which was initiated to create a debate on issues related to belonging and identity. Members of this association are those who define themselves as 'born in Sweden but in some way outside the white norm, those who have access to neither Western, nor non-Western cultures, white nor non-white forms of identification'. Another example is the group of children of immigrants in Sweden who wrote the anthology Third Culture Kids, a manifesto of identity (Babak Azarmi 2017).
In the next section, we examine how the paths to various selfcategorisations are shaped through the individual respondents' characteristics (their class background and other resources) and the social environment around them.

Factors that shape the various self-categorisations
As the identity labels in our study are on a continuous scale, linear regression models are used to examine the impact of the various individual and social factors.
Regarding migrant background and self-identification, those respondents who were born in an African country (Africa 1, mostly in Somalia) are our reference group. As the results of Tables 2 and 3 show, all other groups (including Africa 2) tended to identify themselves as Swedish or having a hyphenated identity to a greater extent than did our reference group (Africa 1). The only exception was those born in the 'Global North' countries (NW1). As Sayad (2004, 162) suggests, the main feature of immigration as a system is 'the relation of domination that prevails at the international level'. When we now have a 'hierarchical nation-state system', with a 'similar hierarchy of rights and freedoms of their people', which Castles (2005, 42)  refers to as a 'hierarchical citizenship', those who are born in the powerful and prosperous NW nations do not run the risk of experiencing exclusion and stigmatisation, and thus have no need to choose Swedish or hyphenated identities instead of that of their parents' homeland (they can say, for example, 'I am French'). Compared with the Africa 1 group, all others (including Africa 2) tend to choose the ethnic label to a much lesser extent. When it comes to the religious label, we see a similar pattern but, here, there is no significant difference between the Africa 1 and Africa 2 groups. Thus, even those who were born in Sweden with parents from African countries (mainly Somalia) tended to choose the religious label more often than other groups. This is probably because, as Yuval-Davis (2010) suggests, of the dissimilar meaning of religious identity for the different groups. As she found in her study, 'Muslimness' for people from the Balkans was 'one of the markers of ethnic/national identity boundaries and difference' while, for Kurds, it was 'one of the common (and therefore, for most, almost irrelevant) collective characteristics that they shared with their national hegemonic Other, namely, the Turks'. Nonetheless, 'in the Somali case, it [Muslimness] was for many a central cultural and religious cluster of collective and individual identity narratives and practices'.
The length of family's residence in Sweden was significant only when it came to religious identity labels. Thus, adolescents whose family had spent more time in Sweden reported a weaker sense of attachment to a religious identification. Bilingualism was the most common among those who identified themselves with ethnic and hyphenated labels. The impact of gender was mixed and inconsistent for identity-label choosing in our sample.
Regarding environmental/external factors, we can observe that the socialisation effects of having many peers with a migrant background is negative for identification as Swedish. On the contrary, those who have a segregated friendship network are more strongly linked to an ethnic and religious identification than others.
Approximately the same pattern is observable for those young people who attend ethnic segregated schools and who tend to identify themselves more often with an ethnic label and rarely present themselves with the Swedish label. In other words, students in less-segregated schools are more likely to adopt a plain Swedish identity than those in the fairly segregated ones. According to Portes and Rumbaut (2001), attending segregated and poor American inner-city schools has another impact. It leads the descendants of immigrants away from identification with their parents' country of birth and towards stigmatised labels such as Black, Latino or Hispanic.
Those who feel that teachers are discriminating against students with a migrant background are more likely to reject a Swedish identity and adopt an ethnic or religious one.
Respondents who perceive that the migrant background of a student makes her/him less popular with other students tend to adopt a Swedish or a hyphenated identity more frequently than others who are not of the same opinion. The results for those choosing ethnic and religious identity labels are not statistically significant.
Living in big cities (compared with smaller ones) decreases the probability of Swedish identification. The impact of obtaining Swedish citizenship is contradictory. On the one hand, it heightens the probability of a person presenting him/herself as Swedish and, on the other, it prompts citizens to report higher scores on the religious identity label. This is perhaps because a significant number of NW1 and OtherEurope1 groups have preserved their original citizenship whereas the majority of young people of African descent in our sample have been able to obtain Swedish citizenship.

Summary and discussion
State agencies (e.g. Statistics Sweden, The National Agency for Education), teachers and other adults in day-care centres and schools, as well as the media, often label the offspring of immigrants in Sweden as invandrare (immigrants). This label is constructed as a negation of everything 'Swedish'. This dominant discourse of Swedishness, as Swedish ethnologist Daun (1998) writes, 'has a blood component' and is related to the principles of 'jus sanguinis' (right of the blood), as opposed to 'jus soli' (birth) or 'jus domicili' (residence). The underlying assumption for this understanding of Swedishness is that not all of those who were born and brought up in the country are Swedes. A Swede is born of parents who are native-born 'Swedish'; he or she has a 'Swedish' appearance and name, and speaks Swedish without a foreign accent. Thus, all others became invandrare (immigrants) 3 . What Daun does not say is that the imagined 'problematic' immigrants (invandrare) do not include all immigrants but certain groups of immigrants. Those whose roots are in the rich and powerful nations of the Global North are not invandrare, thus not stigmatised.
As Scott (1992, 15) puts it, it not the case that some groups of people have a subordinated position and are discriminated against because they are already differentquite the contrary. 'Difference and the salience of different identities are produced by discrimination, a process that establishes the superiority or the typicality or the universality of some in terms of the inferiority or atypicality or particularity of others'. It means that all identities are produced within a system of inequality. We should take into consideration the historical, political and social circumstances by which the different identity categories are formed.
Nevertheless, young people with a migrant background (who were either born in Sweden or grew up there) challenge this forced categorisation/classification. They actively negotiate various forms of identity and belonging; they construct a variety of self-identifications and present themselves in diverse ways in various contexts. They construct and reconstruct forms of identification in their daily lives, in school, the neighbourhood and outside the country. The question of choosing an identity label -'who I am'is intimately, as our results indicate, connected with where they are and who the others are. The findings of our study suggest that young people's identification is not fixed and bounded but a relatively fluid, situational and dynamic process.
Some of our respondents identified themselves with the mainstream and shifted to a Swedish or hyphenated (e.g. Iranian-Swede) self-image. Many others hold on to their parents' country of origin when they present themselves as, for example, Iraqi or Bosnian (the ethnic label). A smaller group chose the religious label. As our results show, compared with other forms, identification with the parents' country of birth was the most prominent. According to comparative studies in this field, the prevalence of ethnic labels in countries like Germany and Sweden, with their shorter histories of immigration, is due to the lesser availability of hyphenated identification for children of immigrants in these contexts compared with settler societies (like Canada or Australia) with a long history of immigration (Phinney et al. 2006). At the same time, we should take into account the time factor. Internalising the dominant discourse by young people with an immigrant background appears to be the main explanation for the prevalence of ethnic labels. However, the present harsh anti-immigrant climate may have intensified reactive ethnic and religious identity formations as a response to this widespread cultural racism. However, we should remember Yuval-Davis (2010) remarks, that religious allegiance in the case of some groups as young people with Somali background may be 'a central cultural and religious cluster of collective and individual identity narrative and practices'.
The paths to these various forms of identificationas our results have demonstratedare shaped by a variety of individual and social factors. The general pattern of the results indicates that those who identified themselves as Swedish or hyphenated tended to be quite similar as regards their other characteristics, while those who presented themselves with the parents' birth country or religious affiliation tended to have very little in common. Those born in Sweden (2nd G.) are more likely to be in the first group (who chose the hyphenated or the Swedish labels) than those born abroad (First G.); quite the reverse for ethnic and religious identification.
When it comes to the migration background, our results demonstrated the following. Compared to those who were born in an African country and those who were born or had their roots in the Global North (NW 1 and 2), all other groups (including those who had African parents) identified themselves to a greater extent as Swedish or with a hyphenated identity. Compared with the Africa 1 group, all others (including Africa 2) tended to choose the ethnic label to a lesser degree. As for the religious label, we see a similar pattern but with no significant difference between the Africa 1 and the Africa 2 groups.
A higher socioeconomic background was associated with a greater orientation towards the national identity (Swedish) and less to religious affiliation. As Portes and MacLeod (1996) remark, since higher education and more consumer power are associated with 'whiteness' in Global North countries, then young people from the better established and economically stronger families have easer to choose a plain national identity (e.g. American or Swedish). On the other side, religious identificationaccording to (Castells 1997, 66)-is 'generated by actors that are in positions/conditions devalued and/or stigmatised by the logic of domination, thus building trenches of resistance and survival on the basis of principles different from, or opposed to, those permeating the institutions of society'.
Adolescents who had lived longer in Sweden were less likely to identify themselves through religion, and bilingualism was the most common among those who identified themselves with ethnic and hyphenated labels. Attending segregated schools and fewer contacts with peers of non-migrant background was negatively associated with identification as Swedish and positively associated with ethnic and religious identification. Perceived discrimination was associated with young people's distancing from a Swedish identity and had a positive association with ethnic or religious identity.
Although our findings are based on young people with an immigrant background perceiving varying degrees of stigmatisation and different class backgrounds, the cross-sectional nature of this study, the specific time of the survey and the age of the respondents were important limitations. More longitudinal researchand the use of ethnographic methods with 'thick descriptions'needs to be conducted in order to explore the numerous ways in which the various identities are constructed and revised in depending on the social context. This kind of research will help us to obtain a deeper understanding of how the descendants of immigrants try to balance the desire to root themselves in their parents' communities of origin while, at the same time, using every available opportunity in the Swedish context to extend their freedom from the same parental influence. How their family's position in the social hierarchy of the new country, their own individual characteristics and other environmental factors interplay in this process is an intriguing issue for future research in this field.