How can we categorise ‘nationality’ and ‘second generation’ in surveys without (re)producing stigmatisation?

Abstract

Whilst reflexive migration studies have criticised the use of categories such as ‘nationality’ and ‘second generation’ in quantitative research, several gaps on how to develop such reflexivity remain. In qualitative data, the co-construction of knowledge seems feasible during fieldwork, whereas the deductive process of quantitative research limits such interactions and is more at risk of reproducing a ‘state thought’. Through a longitudinal database, the LIVES-FORS cohort survey of the National Center of Competence in Research LIVES – Overcoming Vulnerability: Life Course Perspectives and FORS – the Swiss Centre of Expertise in Social Sciences (herafter LCS), we engage in this discussion and provide some answers. The LCS is an annual longitudinal survey that, in 2013, started following a cohort of young adults born between 1988 and 1997 who grew up in Switzerland. The underlying hypothesis of the LCS is that migrants’ descendants have access to different resources (and often a lack thereof) to Swiss natives. In this paper, we discuss both the theoretical and empirical challenges to using the categories ‘nationality’ and ‘second generation’. We show the fluidity and subjectivity of these categories. By changing the definition of the category ‘second generation’, we increased the proportion of ‘second-generation’ participants from 43 to almost 62% of the sample. Looking across the five waves of the survey, we notice a 2% unexplained variation in the first nationality mentioned by the participants and 31% missing values regarding the nationality at birth – which are both indicators that nationality is a subjective category as well as a legal one. We illustrate that the static and neutral conceptions of these categories reproduce a false and stigmatised image of migrant descendants. To avoid these pitfalls we suggest developing multilevel geographical comparisons to consider the effects of time (age and historical), to use a wider range of information in order to be more precise, to examine different nationalities instead of focusing on the traditional nationalities of labour immigrants in a given country and to explore the reasons for the lack of answers to certain questions. Thus the questionnaires should include both more flexibility in the possibilities for answers and details and more-open questions regarding sensitive issues about the definition of the self. They should be developed through a participative and bottom-up process fostering mixed methods.

Introduction

The distinctions generated by social categories always run the risk of reifying, naturalising and essentialising their object and then overlooking their process of production and the power relationships at their origin. Feminist standpoint theory has already clearly shown that scholarly work is not outside these power relationships: the knowledge production is situated and therefore always risks (re)producing stigmatisation using categories predefined by the state or by common sense and excluding marginalised voices (Haraway, 1988; Harding, 2004; Hill Collins, 1990). Categories also have a performative character, meaning that they have long-term social and political uses and implications beyond their mere production.

In migration studies, this reflexive turn in the production of knowledge also takes place, underlying some specificities. The seminal article by Wimmer and Glick Schiller (2003) on methodological nationalism reveals how migration studies are embedded in the logic of nation-states and ethnicity-centred epistemologies. Favell (2019) and Schinkel (2018) demonstrated that the notions of both migration and integration are entangled within a logic of nation-states, while Zetter (1991) criticised the bureaucratic labelling of refugees, highlighting how it depoliticises their situation. Bloch and Chimienti (2011), De Genova and Peutz (2010), Schuster (2011) and Spencer and Triandafyllidou (2020) all showed how irregularity is created by policies and associated terms. This ‘reflexive turn’ in migration studies – to use the title of a book edited by Nieswand and Drotbohm (2014) – led scholars to plead ‘for a “de-ethnicisation” (Wimmer et al., 2009) in order (…) to address the effects of hegemonic power relations in knowledge production’ (Dahinden et al., 2020, p. 3). This reflexive turn requires researchers to take ‘responsibilities’ and reflect on the categories they use and the hierarchies they might (re)produce (Martiniello & Simon, 2005). Dahinden (2016) goes a step further, suggesting some options for de-naturalising and de-ethnicising migration studies and to implement reflexivity during fieldwork in qualitative research (Dahinden et al., 2020).

Whilst qualitative research and social theories have acknowledged the problematic nature of categories – together with the essential need for them in order to make the world intelligible (Jacobs, 2018; Tajfel, 1981) – and defined alternatives to reorient research, this discussion seems rather limited in quantitative research. Quantitative social scientists have, of course, shown that statistical categories are the result of ‘conventions’. In this sense, they are not a depiction of natural phenomena but a ‘reflection’ of the social world that necessarily entails ‘transforming’ and ‘reconfiguring’ it (Desrosières, 2014, p. 39). However statistics in migration studies have not discussed how to develop statistical categories that would avoid the reproduction of ‘state thought’ and they are therefore particularly at risk of stigmatising their subject (Scholten et al., 2015; Simon, 2008).

In the Swiss context, for instance, statistics on ‘immigration’ have thus far used the distinction between ‘national’ and ‘foreigner’ as the main axes of social differentiation, which corresponds to Swiss policy on foreigners (see Le Goff, 2005). In so doing, statistics invisibilise the citizenship and intergenerational background of the children of migrants, as they may be still recorded as foreigners after three generations due to the jus sanguinis law (see “The theoretical and methodological challenges to quantifying the category ‘nationality’…” section). The result of this statistical construction is the impossibility of being able to quantitatively study migrants’ descendants or provide an accurate understanding of their situation, even though, as mentioned by Simon (2008), they are a key issue, both politically and socially.

Whilst, with qualitative data, the co-construction of knowledge seems feasible during fieldwork and reduces the risk of reproduction of hegemonic power relationships imposing pre-defined stigmatising categories, the deductive process and the extent of the quantitative research both limit such interactions. In this context, how can we study quantitatively the descendants of migrants without reproducing stigmatisation? What challenges do the categories ‘nationality’ and ‘second generation’ raise in (longitudinal) surveys? How can we resolve these pitfalls? In this paper, we provide possible answers by opening up the survey’s ‘black box’ and looking at the empirical use, in a longitudinal survey in Switzerland, of two categories commonly employed to analyse migration processes – namely ‘second generation’ and ‘nationality’. The paper will question the scientific use of these categories and the problems they raise and sketch some possibilities for improving (longitudinal) surveys on migrants’ descendants. It draws on a longitudinal study on the transition of migrants’ descendants to adulthood, the Life Course Research Cohort Survey of the National Center of Competence in Research within the framework of the Swiss National Centre of Competence in Research – Overcoming Vulnerability: Life Course Perspectives and FORS (the Swiss Centre of Expertise in Social Sciences). Longitudinal data offer a unique way to reveal the fluidity and changing nature of these categories and enable researchers to question their neutrality and naturality.

The paper is organised in seven sections. In “The theoretical and methodological challenges to quantifying the category ‘nationality’…” section we situate the emergence of the use of the categories ‘nationality’ and ‘second generation’ in the Swiss context and discuss their specific theoretical and methodological challenges. In the “…and to quantifying the category ‘second generation’" section we describe the survey used for our analysis. The next three sections ("The LIVES-FORS COHORT survey", "Defining ‘second generation’ in a longitudinal survey" and "Defining nationality in a longitudinal survey") represent the core of the paper: first, we explore how the sample was selected based on the category ‘second generation’, highlighting the limitations and suggesting some ways to resolve them; then we discuss how nationality was recorded, its variations and how to take into consideration this fluidity. We also examine how the refusal to mention one’s own nationality reveals its subjective and fluid nature. Finally, we propose some answers to the issues raised by this paper.

The theoretical and methodological challenges to quantifying the category ‘nationality’…

As a legal category, ‘nationality’ has been considered in social research to be a basic socio-demographic datum such as ‘age’ (Anderson, 1983) and therefore a natural condition of human beings, inscribed in nation-states which are seen as ‘natural containers of social processes’ (Anderson, 2019). Nationality is used as the first marker of difference and inequality, implying a comparison between ‘nationals’ and ‘non-nationals’ and automatically a form of subordination/domination, as the latter have to align themselves with the mainstream society’s norms (Favell, 2019). This ‘methodological nationalism’ naturalises the idea of the nation-state (and consequently the nationality) as the basic unit of analysis in social research – particularly in migration research (Wimmer & Glick Schiller, 2003). It also ‘naturalises sedentarism as a norm’ and leads us to apprehend ‘mobility as a deviation’ from the norm (Dahinden & Anderson, 2021; Righard, 2021, p. 1). It posits that ‘nationals’ and ‘non-nationals’ are different groups, implying a certain homogeneity and ontological difference between them that would explain their diverse identities, behaviours and practices and prevent an approach based on migrants/foreigners’ individuality.

As emphasised by Anderson (1983), this social construction tends to omit how recent the socio-political creation of nation-states is and to think of ‘nationality’ as both a universal and a uniform issue, implying that all person have a nationality and overlooking the various pieces of legislation on it. In countries based on jus sanguinis law, citizenship is conceived as the result of the nationality of the parents; ‘citizenship’ is tied with ‘nationality’ and there is no formal difference between the two concepts. It is possible to keep registers of the origins of migrants’ children – meaning that an individual, who was born or whose parents were born, for instance, in Switzerland but whose (grand) parents had foreign origins despite living their entire life in Switzerland, can still be recorded as a foreigner, as he or she does not automatically acquire Swiss nationality even after three generations. Actually, such individuals only profit from a facilitated path when applying for Swiss citizenship.

In contrast, countries based on the law of jus solis automatically consider as citizens all person born within the boundaries of the nation-state (Brubaker, 1996). In other countries where jus soli also rules, the category ‘second generation’ often disappears at the administrative level when individuals are born in the host country. In some other countries with a mixed regime of citizenship, the ‘second generation’ category disappears statistically at age 13 (France) and at age 18 (the UK), when the children of migrants acquire ‘citizenship’ (Chimienti et al., 2019).

However, in both cases, ‘citizenship’ and ‘nationality’ work as categories of exclusion. They distinguish between those who are included as living on the territory of the nation-state – and are a part of it, with the right to access living conditions guaranteed by the state – and those who do not. The category ‘nationality’ automatically creates a system of domination and hierarchy based on state-controlled legislation (Simon et al., 2015), first towards ‘migrants’ and then, by extension, to the children of migrants, even though the latter were born in the country. In this sense statistics on ‘nationality’ not only provide demographic information but also become ‘a benchmark for policies and contribute to the production and reproduction of national identity’ (Simon et al., 2015, p. 2). Such an approach also tends to use cultural explanations for social and economic inequalities instead of looking at historic structural causes.

Nationality is also fluid as it can change across time both legally through naturalisation and also subjectively according to a person’s feeling of belonging (Lessard-Phillips et al., 2017; Simon & Tiberj, 2012). Nationality can be attached to different meanings and multiple feelings of belonging as it relates to a person’s subjectivity. This fluidity and plurality of meanings challenges the use, pertinence and accuracy of the category ‘nationality’ in the context of migration research and especially in surveys, given their strict framework. Before providing some ways to move beyond ‘methodological sedentarism’ and ‘nationalism’, we discuss in the next section the theoretical and methodological challenges to quantifying the category ‘second generation’.

…and to quantifying the category ‘second generation’

Use of the category ‘second generation’ is correlated with the definition of nationality; however, it also shares additional challenges. This category appeared with the first studies on the ‘assimilation’ of migrants in the USA in the 1930s and 1940s, and became popular in Europe following an increase in migrants’ descendants, their marginalised situations in the country of residence and their putative lack of integration. Population statistics became useful in investigating the situation for migrants’ offspring and developed the categories ‘of foreign background’ or ‘second generation’ (Simon & Tiberj, 2012) which, in a jus solis system such as that in France, was determined mainly by the nationality of the parents and in a jus sangunis system by the person’s own nationality.

The category ‘second generation’ highlights how difficult it is for the country of birth to consider these young people as citizens, despite the fact that they were born there, raised and lived there for most of their life. The ongoing distinction of them as ‘migrants’ descendants’ leads one to question just how much time is needed in order for someone to no longer appear as a ‘visible social problem’. It reproduces the colonial and assimilationist perspective of migrants – not only those of the first generation, who are seen as less valuable than the autochthonous population (who were thought off as the colour-blind mainstream) but also of their descendants, although they do not necessary share the same cultural or economic background as their first-generation parents. ‘Second generation’ means ‘non-belonging’, which excludes those citizens who have a migratory background (Bolzman et al., 1987). It implies ‘that only multigenerational sedentariness in a specific national territory turns a person into a true citizen’ (Dahinden et al., 2020, p. 8). It is, in this sense, also framed by a logic of ‘methodological nationalism’.

The uses and definitions of this category of ‘second generation’ can change according not only to the legal framework that rules in each country but also to the data collection process, especially the design of a quantitative survey. It is therefore not a precise expression (Bolzman et al., 2017). For example, Portes and Schauffler (1994) defined their sample of ‘second generation’ as those aged between 13 and 15 years old (at the time of their study) who were born in the United States with at least one foreign-born parent or who were born abroad but had lived in the United States for at least 5 years. For Crawley (2009), they are those born in the United Kingdom with at least one foreign-born parent. Kirszbaum et al. (2009) see them as born in France, with at least one immigrant parent while Clauss and Nauck (2009) define them as those born in Germany with two immigrant parents. These differences are even more striking as the research by Clauss and Nauck (2009), Crawley (2009) and Kirszbaum et al. (2009) is all part of the same European survey supported by UNICEF – ‘Children in Immigrant Families in Eight Affluent Countries: Their Family, National and International Context’ – and have a comparative purpose. In a nutshell, the way in which researchers operationalise this categorical definition varies, mainly depending on three aspects (Bolzman et al., 2003a):

• i) whether the person was born in the host country or arrived at a young age with his or her immigrant parents;

• ii) his or her age on arrival, which varies in the different studies between 5 and 16 years old; and

• iii) whether the individual has both parents born abroad or only one.

Resulting from the uncritical acceptance of the categories ‘nationality’ and ‘second generation’, their definitions and use in empirical research remains problematic and is the outcome of the construction by researchers (Bolzman et al., 2003b). In recent surveys there is an explicit effort made to engage with this debate and grasp the variability and subjectivity of these categories. Longitudinal surveys allow researchers to measure the changes over time, including changes regarding the categories of ‘nationality’ and ‘second generation’.

This paper investigates whether ‘nationality’ and ‘second generation’ can be categorised in surveys without (re)producing stigmatisation. How does one grasp the variability and plurality of nationality related to its subjective definition in a (longitudinal) survey? Is ‘second generation’ a relevant category in which to analyse the situation of migrants’ descendants? These two questions will lead our discussion in the following sections.

The LIVES-FORS COHORT survey

We will use the Life Course Research Cohort Survey (hereafter LCS) of the National Center of Competence in Research (NCCR) LIVES – Overcoming Vulnerability: Life Course Perspectives (hereafter NCCR LIVES)¹ and FORS (the Swiss Centre of Expertise in Social Sciences) to explore the theoretical and methodological challenges in studying migrants’ descendants. The NCCR LIVES is a research pole that investigates vulnerability over the lifecourse, where vulnerability is envisaged as a latent state related to the level (or lack) of access to available resources (economic, cognitive, social…) of individuals – at different moments of their life trajectory – which enables them to cope with the different stressors related to the social context (Spini et al., 2013; Spini et al., 2017). The underlying hypothesis of the LCS is that migrants’ descendants have access to different resources (and often a lack thereof) to Swiss natives. In this way, the social context and its supposed hostility to the condition of ‘foreigner’ or migrants’ descendants is seen as a ‘stressor’ through which vulnerability becomes manifest along the life trajectories of migrants’ descendants, particularly during their transition to adulthood. This notion of vulnerability tries to associate the possibilities determinated by the resources of individuals with the institutional and societal aspects that could influence the opportunities available to and obstacles met by migrant descendants.

The LCS is an annual longitudinal survey following a cohort of young adults born between 1988 and 1997 who grew up in Switzerland (Spini et al., 2019). The main interest of this study is ‘to describe the life paths to adulthood in Switzerland today and to compare young adults from the second generation to those whose parents have grown up in Switzerland (either born there or arriving as minors)’ (Spini et al., 2019, p. 400). A secondary objective of LCS is to compare ‘the life trajectories of children of migrants with those of Swiss natives’ (Spini et al., 2019, p. 400). To allow this comparison, LCS has been developed in order to gain an over-representation of young people from the second generation, who are under-represented in other Swiss (longitudinal) surveys.

The researchers already have data collected during five waves – from the first in 2013–2014 to the last in 2017–2018. The design of the survey aims to capture the complexity of the migration process in Switzerland, comparing migrants’ descendants from different origins and young Swiss natives. This diversity allows us to compare the experience of their transition to adulthood in the same country with that of people with a different family background (especially concerning the partner and family formation, access to the labour market and social mobility).

The LCS focuses in particular on the descendants of economic and unskilled migrants from Southern Europe, as they characterised the main migration flows to Switzerland up until the end of the 1970s; in recent waves this has included the descendants of migrants or refugees from the former Yugoslavia and Turkey, who represent a more recent and important migration into the country.

As a longitudinal study, data are collected each year in connection to the Swiss Household Panel but with a different sampling method. While the participants of the Swiss Household Panel were drawn from official registers, those for the LCS were obtained, first, by sampling with an unequal selection probability in order to create a pool of contacts and, later, by network or snowball sampling (which concerned the majority of the sample, see Spini et al., 2019 or https://forscenter.ch). Thus, migrants’ descendants were over-represented. At the same time, the sample includes a control group composed by people with no foreign origin – meaning Swiss citizens with native Swiss parents.

For this paper, we used LCS data from Wave 1 in 2013 to Wave 5 in 2017. Although at the end of the fifth wave the sample included 1961 cases from several origins we developed our analysis using only the part of the sample that answered to the ‘Social Origins’ (SO) module of the questionnaire – this covered 849 cases. This selection was motivated by the wider information this subgroup provided on the surveyed individuals (see Table 1).

Table 1 Original distribution of the sample in LCS

The sample distinguishes between ‘nationals’, ‘foreign origins’ and ‘second generation’. However, the distinction between the two latter categories is narrow in a country such as Switzerland, where jus sanguinis is the principle rule for citizenship.

In the framework of our research project entitled ‘Transition to the adulthood of migrants’ descendants’, we are precisely interested in this element of the population, who are still considered (in population statistics, legally or according to the population views) as foreigners despite the fact that they were born in Switzerland or lived there most of their lives. In the quantitative part of the project based on the LCS, one of the first challenges we faced was the definition of our sample and the ‘cleaning’ of the LCS data that we aimed to use for our study.

In the next section, we present the results of our analyses on the categories ‘nationality’ and ‘second generation’, and how the epistemological and empirical aspects mentioned can influence the collection and management of data in migration studies.

Defining ‘second generation’ in a longitudinal survey

The first issue we faced when trying to define our sample based on the LCS was the low number of ‘second-generation’ respondents, which is striking if we consider that the LCS aims to reach a large number of migrants’ descendants living in Switzerland. It was hypothesised that the networking among people with the same nationality would allow each young migrant descendant (during the recruitment phase by snowballing, see Spini et al., 2019) to connect the LCS interviewers with other young adults with the same nationality and lead to an increased sample. This idea was built on the assumption that the second generation will spend time more with others of the second generation from the same origin than those with Swiss-born parentage, probably based on Putnam’s (2007) or similar theories that ethnic diversity decreases social capital. It did not turn out to be the case in the LCS.

As the operationalisation of the category ‘second generation’ was enacted prior to our research, the percentage of individuals belonging to the ‘second generation’ in our sample was limited a priori by the definition given by LIVES-FORS. The definition of the category ‘second generation’ is more a methodological than a legal problem. In the LCS the second generation is defined as those:

• with both parents who immigrated to Switzerland after the parents’ eighteenth birthday;

• who were born in Switzerland between 1988 and 1997 and were thus between 16 and 25 years old at the beginning of the survey, and between 23 and 32 by 2020; or

• who had lived in Switzerland since 01 January 2013 and/or had attended school in Switzerland before the age of 10 (Spini et al., 2019).

The LCS used a broad definition of ‘second generation’ through including people who were born in Switzerland, as well as people who arrived before their tenth birthday. However, it uses a restrictive criterion for the parents (as both parents had to be born abroad, have foreign citizenship and to have arrived in Switzerland before their eighteenth birthday).

Using these criteria, the LCS identified 364 individuals out of 849 (43%) as ‘second generation’ (Table 2). In addition, 34% of the sample were categorised in the LCS as of ‘foreign origin’, because they did not mention any Swiss origins but did not correspond to the criteria for the second generation defined by LCS. A further 23% of the sample were classified as of ‘no foreign origin’ – thus representing those of Swiss origin. In order to increase our sample of ‘second-generation’ individuals, we chose to modify its definition, which emphasises even more the plasticity of this category.

Table 2 Recovered cases in the new ‘second-generation’ category

We considered that starting school in Switzerland at the age of 9 or 10 might not create sufficiently important differences because both are very close to the selection age for secondary schools (at 11 years old), and this does not leave (in both cases) much time to redress inequalities. Thus, we included people who had arrived and enrolled in school in Switzerland before the age of 11.

Moreover, we revised the ‘foreign origin’ category to represent a residual category, including individuals who had two foreign parents but whose age on arrival in Switzerland could not be identified. This also includes individuals whose parents arrived before their eighteenth birthday. We assume that the distinction in the age of arrival of the parents (after 18 years old) is a methodological issue that might not have a significant impact on the ‘second generation’ trajectory. Therefore, our revised and working category of ‘second generation’ excludes the variable ‘arrival age’ when both parents are of immigrant origin.

Whilst we included more cases, we also made some distinctions between the ‘second generation’ sample who were born in Switzerland and those who arrived at a later age as, according to the literature, being born in a country compared to being a later arrival has a different impact on the life trajectory (Bolzman et al., 2003a; Portes & Rumbaut, 1990). In addition, we distinguish between the ‘second generation’ and the ‘2.5 generation’, who are those with one parent of foreign origin and another of Swiss origin (Gomensoro, 2014). Our working sample of ‘second generation’ thus includes (see Table 2, figures in bold):

• Second-generation born in Switzerland: individuals who were born in Switzerland, both of whose parents were immigrants and of foreign nationality;

• Second-generation, not born in Switzerland, but who arrived before the age of 10: individuals who arrived in Switzerland before the age of 10 and whose parents were both immigrants of foreign nationality.

We categorised the other migrants’ descendants as the 2.5 generation, which includes individuals who were born in Switzerland and who had one native Swiss parent and those who arrived at or after the age of 11 in Switzerland; being of Swiss origin includes those whose parents were both native Swiss.

With this new categorisation, we ‘recovered’ 176 individuals (in bold in Table 2) from the residual category ‘foreign origins’ that we included in our new ‘second-generation’ category. We will use this new sample of 849 cases as our basis. At the same time, the new categorisation allows us to refine the category of ‘second generation’ and to distinguish between the ‘second generation’ who were born in Switzerland and those who were not. In addition, we increased the proportion of ‘second-generation’ participants from 43 to almost 62% of the sample. The same principle about the place of birth was used with the 2.5 generation in order to keep the specificity of the place of birth for further analysis. Finally, the residual category of ‘foreign origins’ almost disappeared through the use of this more accurate categorisation.

These different ways of categorising those of the ‘second generation’ show how this category is highly malleable depending upon the exact parameters of the definitions, also related to the frameworks for citizenship. In the Swiss context, redefining this category allows us to increase the number of respondents considered as ‘second generation’ in our sample. In addition, this redefinition allows us to identify a subgroup within the ‘second generation’ who were not considered at the beginning of the survey (176 ‘second generation’ respondents and 81 ‘2.5-generation’ ones).

These first considerations clearly highlight the fact that the category ‘second generation’ is not only imprecise but also discriminatory and therefore controversial, as though these young people were different to or even lesser citizens than, those born of native Swiss parents. The lack of precision prevents a proper image of migrants’ descendants being given and is a barrier to comparisons either internally (between groups) or cross-nationally. With the purpose of providing a picture of social and economic inclusion of the whole population and their changes over time, a (longitudinal) survey should therefore not hypothesise categories of distinction in their design; rather these should be the result of the analysis of the different variables of inequality such as class, gender, race, ethnicity, place of birth etc. and their intersectional analysis. (Longitudinal) surveys should also implement self-categorisation by their respondents, enabling the latter to provide a self-definition of what is relevant for them, which would prevent the formation of stereotypical assumptions on inequality. In order to avoid using the mainstream as the norm of comparison, differences between but also within the categories which arose from the results should be developed. In so doing, such analysis will provide configurations of both people and inequalities. The category ‘nationality’ raises further issues, as we discuss in the next section.

Defining nationality in a longitudinal survey

The analysis of nationality in our subsample of migrants’ descendants immediately threw up some inconsistencies. We found differences in the distribution of nationalities across the five waves, with fewer people from a given nationality in a latter wave than in a former one. Some changed their nationality during the survey (mentioning, for instance, in latter waves, that Swiss was their first nationality) without, however, mentioning that they had undergone naturalisation or a change in citizenship status. What is more, an important number of people did not answer the question about whether or not they had Swiss citizenship from birth (see “Investigating the ‘No answer’ response in a longitudinal survey” section, Table 6). These inconsistencies have both an epistemological and a methodological impact, which we discuss here.

As already above mentioned, the definition of nationality for a survey is as much related to the legal framework that defines nationality and citizenship as it is to a subjective question which raises methodological issues (as in the case of self-identification, cf. Lessard-Phillips et al., 2017; Simon & Tiberj, 2012). The definition of nationality can vary not only according to geographical context and across time – as citizenship law can change and one person can acquire another nationality – but also from the subjective perspective during the lifecourse of an individual according, for instance, to his/her economic inclusion in the local labour market. This variability challenges how we accurately record the nationality of an interviewee in the framework of a survey – and even more of a longitudinal survey – which has to keep the same definition/question whilst the interpretation may vary over time.

In the LCS, ‘nationality’ was asked as an attribute that could have multiple meanings – someone could indicate three nationalities without ordering them in any consecutive way. Comparing the different waves, we see that there are some variations in what people declare as their first nationality (see Table 3). We put the first and the second nationalities mentioned in the LCS into five categories that regroup the most important migrant origins in Switzerland in recent decades. These groups were Southern Europeans, who were part of the first massive migrations into Switzerland after WW2, mostly coming from Italy and Spain (Bolzman et al., 2003b); Eastern Europeans, who were part of the more-recent waves, arriving after the 1990s and including migrants from Kosovo, Turkey, Macedonia, Croatia, Serbia, Bosnia-Herzegovina, Slovenia and Albania; and the Portuguese, who are currently the largest group of migrants in Western Switzerland, arriving at the same time as Eastern European migrants (OFS, 2019).

Table 3 Nationality in LCS waves

On average, we found slight variations in the declared nationality for all groups: there were, on average, seven individuals by nationality group who mentioned a different nationality from one wave to the next. Although minimal (on average they represent 1% within each group), these variations are interesting because we can neither explain them by attrition, nor by replacing them in the survey sample (attrition and replacements are quantified as ‘no dataset’ and excluded from computations), nor even by respondents’ acquisition of Swiss nationality during the survey (naturalisation). Only 39 cases for all five waves indicated that they went through a naturalisation process and only 34 individuals in the sample (4%) said they had received it after 2013 (Table 4 framed). Therefore, the 1% variation which we observed in the first declared nationality, from one wave to another, cannot be explained by naturalisation during the survey. Whilst this percentage seems insignificant, it shows the fluidity of the category ‘nationality’ despite all the investigative efforts by the promoters of the survey. We believe that this percentage might be higher when considering the whole LCS sample of 1691 cases.

Table 4 Declared naturalisation and observed change of nationality across waves

Table 4 also shows further evidence of inconsistences if we look at the difference between the declared acquisition of nationality (naturalisation) and the change of nationality registered in the data across waves: 29 individuals (3%) declared that they had acquired Swiss nationality after 2013 but that they did not change it during the survey (in bold and italics). Secondly, 10 individuals who changed their declared nationality during the data collection did not say that they had received Swiss nationality. This variation between the declared nationality and the changes of nationality, as observed across the five waves, corresponds to 2% (n = 18).

In order to explain these variations in the first nationality mentioned by the interviewees, we explore whether they were related to the place of birth of the persons. We distinguished between people who were born in Switzerland and have Swiss citizenship and those who were born either in Switzerland or abroad but who do not have it (see Table 5). This cross-tabulation shows another puzzling result: about a third of the sample (31%, n = 264) did not mention whether or not they had Swiss citizenship from birth. This absence of an answer is difficult to interpret (see next section). Table 5 shows that the ‘no response’ rate is higher among people who were not born in Switzerland (38%). Thus, data about birthplace and citizenship at birth provide no further explanation for the variation in first-declared citizenship in the LCS.

Table 5 Cross-tabulation, place of and nationality at birth

In other words, we cannot explain the variations in declared nationalities across the waves of the LCS with any objective criteria. This inability demonstrates that nationality is more than just legal data; it is closely related to a subjective feeling of belonging. This is not, of course, a new result as several qualitative studies on belonging have already provided evidence of it (cf. cf. Frauenfelder, 2007; Lessard-Phillips et al., 2017; Poglia Mileti, 2019; Simon & Tiberj, 2012). However, our analysis does allow us to quantify this variability and the subjective nature of nationality.

So how to take into consideration this variability in a longitudinal survey which is related to nation-state and ethnicity-centred pitfalls? As highlighted by Lessard-Phillips et al. (2017), a first step would be to include the different geographical levels at which the comparison is made: comparing the situation of the same profile of children or young adults at both local and national levels – and also transnationally – examining their situation in different local and national contexts as well as in the country of origin. This is what Green calls ‘divergent models’ which, according to her, allow one to ‘locate the explanation of difference at the point of arrival and not at the point of departure’ (Green, 1994, p. 15).

Such multilevel comparisons provide a more-nuanced understanding of what is happening at the sub-national level. Cities have been described as the key sites of settlement and inclusion and therefore the comparison of migrants and their descendants at this local level would be more accurate and would better distinguish the contextual causes of differences compared to individual ones (Crul & Heering, 2008). Yet, comparing national contexts is also necessary – given the national root of entry and settlement legislations (Guiraudon & Lahav, 2000) – as is taking into consideration the non-urban context that might be forgotten in comparisons to urban contexts (Lessard-Phillips et al., 2017). Besides, a multilevel comparison avoids the need to reduce the comparison to a minority/majority dichotomy (of native parentage) – including comparison to other youth with similar backgrounds in other contexts – and also to compare them to their parents (intergenerational comparison).

To this measure we add that comparison should also take into consideration the whole possible spectrum of diversity by also examining cases of non-traditional national origin among migrants and their descendants; studies often tend to focus on those national origins which are presumed to be more vulnerable. Taking also into account those from privileged origins might shed light on the unequal access granted to immigrants and their treatment by the ‘host’ society – which, in this sense, would not be related only to race, gender, culture, disability etc. but would increasingly concern from which part of the world an immigrant is coming.

An additional measure highlighted by Lessard-Phillips et al. (2017) is to take into consideration in the comparison the time, age, period of arrival and duration of stay, as well as people’s mobility over time. Time has an important role in the outcome of inclusion that put into perspective the importance of the geographical context. Taking time into consideration allows the researcher to record the variability and plurality of meaning which nationality may entail for an individual and also the geographical mobility of people.

In this sense longitudinal surveys are better suited to providing more accurate comparisons because time is considered both in the data collection and in the analysis; this allows examination of the processes instead of the outcomes and offers a further way to understand the complexity of the mechanisms at play in the lives of the children of migrants (Lessard-Phillips et al., 2017).

In the next section we discuss respondents’ refusal or inability to answer – another aspect often neglected in surveys even though it may reveal interesting phenomena.

Investigating the ‘no answer’ response in a longitudinal survey

As mentioned above, we found an important number of ‘No answer’ responses to the question ‘Are you Swiss since birth?’ – 31% of our sample gave no answer (264 individuals, see Table 6). This is striking, as other similar questions such as ‘Were you born in Switzerland?’ (place of birth), do not present any missing data. How should we interpret this lack of response? The LCS does not allow analysis of the statistical relationship between the variations in declared nationality and the non-response to nationality at birth. Indeed, there is no a direct relationship between the two variables in the dataset.

Table 6 Second-generation response to ‘Swiss since birth’

Whilst the total number of individuals in the categories ‘Swiss origins’ and ‘2.5 generation’ answered the question as to whether they were born with Swiss citizenship, 49% of those in the second-generation category did not answer (Table 6). The relationship between being second-generation and not having answered this question is statistically significant, with a medium degree of association (sig. 0.00 and R Pearson/ Phi/ Cramer 0.47).

If we inquire further into this relationship, we find that there is no great difference between nationalities in the non-response to this question whilst the explicative variable seems related to the level of education: the lower the level of education, the higher the non-response to the question ‘Swiss since birth’ (Table 7).

Table 7 Response to ‘Swiss since birth’ by education

We see that, according to whether or not they have foreign origins and, for some, a condition of vulnerability (here a low level of education), questions seem to have different meanings: whilst those born in Switzerland answered the question as to whether or not they were Swiss since birth, those categorised as second generation did not. Is it because the former felt it was more legitimate to answer? Or is it just because they were sure of their nationality at birth? Indeed some respondents might have had information about when they were naturalised. Or maybe it was because some migrants’ descendants felt that the question was offensive as they just indicated in a previous answer that they were Swiss.

In any case, the difficulty in interpreting the absence of answers calls for researchers to take respondents’ feelings more into consideration during the survey process by adding an open question, for instance, allowing interviewees to explain their reasons, both in the survey and during the analysis.

Conclusions

At the beginning of this article we mentioned the reflexive turn in migration studies based on qualitative research and we asked whether similar theoretical and methodological change could take place in quantitative research and under what conditions. In this paper we have provided some form of answer by exploring the use of the categories ‘nationality’ and ‘second generation’ in a longitudinal survey. Our interest in ‘nationality’ focused on how interviewees answered and interpreted questions related to the concept whilst, for the ‘second generation’ category, we looked at how researchers defined it.

We first showed the bias of these categories towards methodological nationalism and the issues which such bias raises in terms of interpretation. The analysis based on the LIVES-FORS COHORT survey demonstrated and quantified the variability and subjectivity of these categories. Birth nationality can be different from actual or current nationality; equally, a person’s nationality at the beginning of a survey can be different to that in a later wave of the study. Moreover, nationality is subjective, even if the law frames it. Thus, we observed a 2% variation in the declared ‘nationality’ and the change of nationality observed across waves that is not related to naturalisation. We also found 31% of non-responses to the question on nationality at birth, which could indicate either that this topic is sensitive or, at least, that this is not objective data but subjective, which could lead to different interpretations.

The redefinition of the category ‘second generation’ allowed us to increase the number of ‘second-generation’ cases from 43 to 62% of the sample. The variability in the definition of the category ‘second generation’ empirically expresses two situations. First, there is no convention about who the ‘second generation’ actually are because the ‘second generation’ do not exist in reality as a defined group. This category is conceived for individuals with a ‘recent’ migrant background as, in reality – and particularly in the Swiss case – most of the population have ancestors who migrated from another country at some point in time. Second, this category represents, somehow, a form of violence as it creates a distinction and homogenises migrants’ descendants, who are classified in one group despite their differences and only according to their parents’ country of origin. The one-third of our sample who did not answer the question ‘Are you Swiss since birth?’ could represent an important clue in understanding these feelings and should be investigated further.

In order to avoid the reproduction of structures of domination and exclusion in the study of people of multiple origins, research should develop multilevel geographical comparisons (between cities and between nations) and reflect on the effects of time, considering the different mechanisms at play. Data collection should allow a more flexible approach that considers a wider range of information – such as, for instance, place of birth, nationality/ies at birth, later nationalities and year of acquisition – in order to be more precise and to examine different nationalities instead of focusing on the traditional nationalities of labour immigrants in a given country. We also need more information on the reasons behind the lack of answers to certain questions, as they might be another indicator of the stigmatisation of people of a certain nationality or an additional indicator of the variability of nationality, which might not be known by everyone at birth. To do so, the questionnaires should include both more flexibility in the possibilities for answers and details and more open questions regarding sensitive issues about the definition of the self. They should be developed through a more-participative and bottom-up process which will foster mixed -methods. Finally, data analysis should enable researchers to question common-sense categories and those that arise from the results and are not subsumed. Longitudinal surveys which allow the analysis of the processes instead of the outcomes offer more ways to understand the complexity of the mechanisms at play in the lives of the children of migrants.