Towards a precise and reflexive use of migration-related terminology in quantitative research: criticism and suggestions

Abstract

To describe migration-related phenomena, we need to reflect on the terminology and choose the most adequate one that allows us to determine whether migration is the (main) cause of a phenomenon, a consequence, or even unrelated and misattributed. We argue that the use of such terminology in quantitative and experimental research is often flawed because of its differentiated adoption in legal, political, or scientific contexts. To illustrate our argument, we focus on two commonly used terms, ‘second generation’ and ‘migration background’ to show that in many situations these terms do not accurately describe the population we study. In part, the terms imply a false homogeneity, focus on deficits, and perpetuate differences regarding national belonging where there may be structural reasons and other aspects, such as social class, that lie at the heart of observed differences. With a particular focus on quantitative research, we use survey evidence and a principled literature search, to show that both researchers and the general population often identify immigrants in terms of ethnic origin —  even though the term has its own pitfalls. We conclude that quantitative research should avoid reproducing state-created terminology and instead look beyond the strict field of immigration to consider other systems of classification like gender, ethnicity, language, or social class to reduce the negative attributes ascribed to non-citizens.

Introduction

Migration is undoubtedly a central theme in contemporary social and political discourse. It not only fuels discussions about the transformation of societies, but also brings to the fore questions of terminology regarding different immigrant groups and ways of understanding immigration (Kersting & Leuoth, 2020; Ruz, 2015; Betts, 2013). Legal terminology broadly differentiates citizens from non-citizens, formally regulating access to rights and how people can move between legal categories (Dahinden and Anderson, 2021; Schuster, 2011): to gain or lose citizenship, for instance. Demographic analysis, regularly conducted in statistical offices, reifies distinct categories and reproduces national identity (Simon, Piché, and Gagnon 2015), while political discourse and the media further enlarge the pool of terminology. Definitions tend to focus on the birthplace of the individual (or their parents) or citizenship, while other terms that circulate among these various actors include stereotyping depictions of migrants, such as ‘asylum shoppers’, or ‘welfare tourists’ (Dahinden and Anderson 2021). For a long time, research interested in social phenomena that include mobility and migration has played a part in reproducing differences between national belonging and otherness, despite existing critical stances towards the use of such terminology (Appadurai, 1993; Kertzer and Arel, 2001).

When scholars and media discuss relevant terminology, they often focus on migrants with precarious legal status, called ‘irregular’, ‘undocumented’ or ‘illegal’, and the distinction — or lack thereof — between ‘asylum seekers’ and ‘refugees’ (Hoops & Braitman, 2019). However, the literature pays less attention to other migrant groups, many of whom have lived in their country of residence for significant periods of time. Here, conceptualisations include references to the ‘second generation’ or a ‘population with migration background’, which are regularly applied in quantitative and experimental research, yet also controversially used by politicians for clamping down on ‘non-Western’ residents through discriminating policies, including the notorious Danish Ghetto laws targeting residents and citizens of ‘non-western’ descent (Burnett, 2021; Zhang, 2020). Since the way people are labelled has implications on how they are represented in laws or perceived by society, the wording of policies can further justify marginalization of those deemed undeserving (Malone, 2015; Zetter, 2007). As scholars of migration, we should thus re-think some of these most used terms — especially in quantitative and experimental research —, because they lack precision and may lead us to preconceived answers, reproducing ideals of nation states as homogenous entities against which ‘migrants of various types’ can be differentiated. We highlight four problems with current terminology (in particular ‘second generation’ and ‘migration background’), that we develop in the main analysis:

• a Many terms present immigrants as a homogeneous population, often highlighting deficiencies driven by an image of a ‘mainstream population’ contrasted with ‘immigrants’, which denies them agency in shaping their identity (yet demanding such agency when it comes to e.g. integration policies);

• b The terms used correspond to an administrative logic that is itself rooted in colonial and discriminating logics, and they do not reflect how the public perceives difference;

• c The terms perpetuate difference over time, especially in places where we want to de-emphasise difference;

• d The terms are used in various contexts, without considering structural factors or considering the consequences of categorisations, allowing for further marginalisation of those differentiated.

Prior criticism on methodological nationalism shares some of our concerns (Wimmer and Glick Schiller, 2002). We agree with not treating nation-states as single social and political entities and hence natural units of analysis (Kalir, 2013) and broaden the criticism about rigid perspectives to address the fixation in migration and mobility studies on one aspect of peoples’ biography: Border crossing determines identity and becomes inherited, irrespective of self-identification (Kalir, 2013). This form of classifying the population can reduce people to ‘others’ within the territory of the nation-state. While similar arguments could be drawn for the legal or public use of terms (Dahinden and Anderson, 2021), we focus on official statistics and survey questions used in academic research that can encourage an understanding and description of population groups difficult to defend when examined closely, as they might risk reinforcing stereotypes (Anderson & Blinder, 2019). Our contribution lies in developing these arguments and applying them to quantitative and experimental research and non-marginal immigrant groups where notions of ‘objective’ categories and classifications face little reflection by researchers.

Conceptual limits of current approaches

In public and academic discourse, we find many ways to describe individuals who are perceived as non-citizens. Individuals seeking refuge are defined as (failed) asylum seekers, refugees, detainees, or deportees (Hamlin, 2021) and are differentiated according to their deservingness (Karakoulaki, Southgate, and Steiner 2018; Sanchez & Romero, 2010). Laws and policies deepen or even demand a differentiation between citizens and non-citizens (Dahinden et al., 2020). Yet, even within the category of citizens we find that governmental institutions, statistical offices, or international organisations sort between ‘Western’ and ‘non-Western’ people or ‘descendants’ (Clante Bendixen, 2022). Emphasising difference, the terms hide or ignore differences within groups that their members may find important (Brubaker, 2002; Chimienti et al., 2021a, 2021b). Critical postcolonial research argues that official (ac)counting often targets people at the margins, such as ‘the poor, the sexually profligate, and the criminal’ at home, while clearly contrasting them against the majority population ‘as “different” in problematic ways’ (Appadurai 1993: 317). Such categorisations can negatively affect immigrant groups, which is why they are sometimes challenged by non-official actors. For instance, the news network Al Jazeera decided to no longer refer to ‘migrants’ given the violent situation in the Mediterranean (Malone, 2015). However, we find little reflection in quantitative and experimental research.

Instead, the way researchers or public administrators use terms can prevent them from seeing how the definition of ‘migrants’ — often conflating migration status, race, ethnicity, and asylum — affects the interpretation of results and our “understanding of the impacts of migration” (Anderson & Blinder, 2019:2). Public statistics may define immigrants and their descendants as having a ‘migration background’ based on citizenship and place of birth of individuals and their parents (sometimes grandparents), as well as legal residence or citizenship status (Simon, 2017). Self-identification or whether members of the mainstream society regard them as ‘migrants’, plays no role in these definitions (Chimienti et al., 2021a, 2021b; Jenkins, 1997). Yet administrative, legal categorisation defines access to various rights and media or political discourse can increase the perception of ‘foreignness’ and ‘otherness’ by symbolically singling out certain groups (Hoops & Braitman, 2019). Consequently, members of the ‘second plus’ generation may not be perceived as members of the ‘native’ mainstream society, for instance if skin colour or outward religious markers allow a continuous differentiation (Shavers & Moore, 2019).

We are not the first to contest terminology; see for example critical interrogations of ‘migrants’ (Dahinden, 2016; Elrick & Schwartzman, 2015; Favell, 2019), ‘refugees’ (Carling, 2017; Sigona, 2018), ‘integration’ (Korteweg, 2017; Schinkel, 2018), or the description of how ‘migration background’ affects pupils at school (Horvath, 2019). Work on ‘methodological nationalism’ by Wimmer and Glick Schiller (2002), going back to Martins (1974) and Smith (1983), highlights that lived groups do not necessarily correspond to citizenship and nationality. In the tradition of methodological nationalism, terminology such as ‘refugee’, ‘second generation migrant’ or ‘tourist’, adds an “empirically oriented social science practice [that] is taking national discourse, agendas, loyalties and histories for granted” (Wimmer and Glick Schiller, 2002:301). Our aim is not to repeat existing critiques, nor to banish certain words, but to look at how past criticism is applied (or not) in current quantitative and experimental studies, adding to the often theoretical (cf. Dahinden and Anderson, 2021) or qualitative studies (Glick Schiller, Çaglar, and Guldbrandsen, 2006) that discuss migration terminology. We also seek steps towards a practical way forward within research done on mobility and migration: here ‘groups’ are not abolished in research — because we need ‘groups’ to study social realities —, but we conscientiously choose the way we approach these social realities.¹ In this sense, our call for more carefully reflected terminology includes a call for a partial resistance to official categorisations and we argue for a plurality and diversity in approaches — especially if they are based on thoughtful consideration. As such, we wish to explore productive ways to use terms such as ‘second generation’ and ‘migration background’ that can be valuable when used in a de-stigmatising way.

Analysis: homogenous groups with a deficiency logic

Defining immigrants is the least difficult part once we accept that settlement does not need to be permanent and that some people effectively live in two different countries (Geurts & Lubbers, 2017; Zufferey et al., 2021). Turning to descendants of immigrants, we can insist on two immigrant parents to count a child as ‘second generation’ or having a ‘migration background’, but often we use a single immigrant parent (e.g. Reeskens & van Oorschot, 2017). Not only does this attest to an incoherence between individuals who might have a vastly different residence status (e.g., citizenship versus permanent residence), but it also contributes to a deficit logic, where the experience of crossing the border is a deficit that is passed on to one’s descendants. Once we reach the so-called third generation (see Jiménez et al., 2017), complexity increases: Should a single grandparent from a different country be counted as ‘migration background’, or should we insist on four immigrant grandparents? The further we go down the generations, the less likely the perception of a homogeneous group applies, and the more likely researchers with a fixed ‘systematic’ approach introduce problematic classifications. Consider for instance a study indicating that “[b]ecause we are interested in opinions towards the influx of immigrants, we exclude respondents who have been born abroad or have at least one parent who was born abroad” (Reeskens & van Oorschot, 2017:184). This is clearly systematic, but also fixed in that all children of immigrants are denied the possibility of being part of the majority population. By contrast, one step to open such pre-defined and researcher-driven exclusionary criteria is to change the operationalisation of migration background away from e.g., place of birth and to a linguistic operationalisation, e.g. via the first language learned (Hamel, 2022).

Through a principled review of published research (cf. Munn et al., 2018), we can show that there are many different approaches to how researchers define, measure, and grapple with differences between what they mean and what they measure (Appendix 1). In this review, we searched for the terms we criticise and coded how they are applied in quantitative and experimental research. We find many articles referring to ‘migration background’ and ‘second generation’ immigrants, but the way the population under question is measured varies a great deal: foreign birth, a specific country or region of birth, official definitions like the migration background defined by the German government, or ethnic understandings that treat some immigrants as ‘closed’ minority groups are all commonplace. It is difficult to discern a coherent mapping between the categories used (e.g. migration background) and the measurement (e.g. foreign birth). For example, one study refers to ‘migration background’ and measures whether both parents were born abroad (Bächle et al., 2013), while another uses the same concept but uses an algorithm to identify ‘Turkish’, ‘Arab’ and ‘Greek’ names (Rüdiger et al., 2021). Sometimes authors are constrained by the variables available in existing datasets, but we can discern attempts to capture ‘cultural difference’ or ‘marginalisation’ without saying so, or instances where readers are left in the dark how ‘migration background’ was measured (see Appendix 1).

Leaving measurement aside, when we differentiate between people with ‘migration background’ and those without, for example as predictor variables in a quantitative analysis, we treat a heterogeneous population as homogeneous (Brubaker, 2002; Lucassen, 2022; Wimmer, 2009): A refugee fleeing persecution is captured alongside an expat arriving through policies addressing the highly skilled, neglecting differentiated selection procedures and thus access to rights. Put differently, knowing whether a person meets the definition of a ‘migration background’ is unlikely to tell us anything useful about life chances, unless we take it as a — poor — proxy for something else. Indeed, in the reviewed studies, we typically encounter these terms in the context of ‘deficiencies’ like insufficient language skills, ‘incompatible’ values, unwillingness to integrate, low levels of education, or poverty (see Zhang, 2020).

In our search, we find a differentiation between how terms such as ‘second generation’ or ‘migration background’ are used. On one hand — and ideally — they can demarcate how marginalisation and poverty are inherited, often because a migration background is linked to lower incomes, etc. On the other hand, the use of second generation refers to demarcating (a subgroup of) ‘foreign nationals’ living ‘somewhere else’, e.g., second generation Turks in Austria, France and Sweden (Schnell, 2015). This approach “flattens idiosyncrasies and creates boundaries around these homogeneous bodies, since it performatively limits their extent […] statistics are to bodies and social types what maps are to territories: they flatten and enclose” (Appadurai 1993: 333f.).

Leaving homogeneity aside, classifying ‘immigrants’ into the first and second (and subsequent) ‘generations’ is simple only at first sight. The ‘first generation’ describes immigrants: people who have crossed an international border to live in another country. The ‘second generation’ describes children of the ‘first generation’ who were born and grew up in the country where their parents came to live. The ‘third generation’ describes children of the ‘second generation’, whose grandparents came to live in the country the ‘third generation’ was born and grew up. We often approach immigrants this way because it constitutes an (apparently) straightforward way to count and is typically done systematically — implying scientific rigour. When we look more carefully, however, counting immigrants is no longer so easy since marriages between immigrants and the majority population do not fit neatly with the notion of ‘generations’ (Song, 2015), making the classification a pragmatic but unsatisfying choice.

Typically, when we refer to ‘second generation’ immigrants — even when referring to a specific subset — we take an external perspective: It is us researchers who define the supposedly relevant boundaries through the definition we use. An unintended consequence of this is that we treat immigrants as a constructed, externally defined category in which group members have a limited role in shaping membership (Chimienti et al., 2021a, 2021b; Jenkins, 1997). For instance, when we construct a variable ‘migration background’ in a quantitative dataset, we do not consider identification or assimilation; typically, we only consider nationality and place of birth. This means that we deny the social construction of the category (‘immigrant’, ‘second generation’) and imply internal homogeneity by using an ‘objective’ approach. If we use these terms, we ignore the fact that specific groups and subgroups of immigrants may be integrated or separated to different degrees (Will, 2019), ignoring the otherwise well-established literature on segmented assimilation where patterns of integration can vary between different immigrant groups (Portes & Zhou, 1993), acculturation, integration, and intermarriage (Schmitter Heisler, 2008; Song, 2009), in addition to denying individual agency. While we would not argue that it lies entirely within the scope of individual agency to determine whether a person is considered part of the majority population, we have no doubt that individual actions and behaviour can contribute (Choi et al., 2019; or so-called ‘CV-Whitening’ where individuals hide parts of their CV such as languages spoken or change their name to ‘blend in’, Kang et al., 2016; Ruedin and van Belle, 2023).

Administrative logic, not how people see the world

Oddly enough, integration policies across the Western world demand individual efforts — only to deny this by the way immigrants are counted in (statistical) analysis. Here, research reproduces the official language and discloses the dependency of researchers on data gathered by public offices (Heckathorn, 2006), thus complicating the conceptualisation of empirical studies. The terminology we criticise follows this administrative logic — a neat classification of the population according to rights and duties (see Connelly et al., 2016 on how these data are constructed; Will, 2019). By contrast, the public often has an ethnic understanding of immigration, something that needs to be addressed through a separate method of data-gathering. In 2016, we asked a sample of the Swiss population “When you think about immigrants coming to and living in Switzerland, which of these groups do you normally think about?”.² While 36 per cent of respondents consider citizens from European countries immigrants, it is 65 per cent who imagine non-European countries. At the same time, only 14 per cent identified Swiss-born children of non-Swiss citizens as immigrants (Panel A in Fig. 1). These major differences clearly indicate that the citizenship-based approaches behind ‘migration background’ and ‘second generation’ do not correspond to everyday perceptions of different immigrant groups. Instead, the respondents seem to focus on cultural distance and the (implied) motive of migration. We could recognise a similar understanding in the literature analysis, such as when researchers refer to ‘second generation’ but empirically focus on Turkish immigrants and their descendants (e.g. Barwick & Beaman, 2019; Appendix 1).

Fig. 1

Whom the population think about when they think about immigrants. All categories in Panel A fall under the official definition of ‘migration background’ in Switzerland: “citizens from European countries”, “citizens from non-European countries”, and “Swiss-born children of non-Swiss citizens” (‘second generation’). The wording in panel B is slightly ambiguous (“a person from [Germany]”), but this cannot explain differences between ‘Western’ and ‘non-Western’ countries of origin. Higher scores in panel B capture a stronger association with ‘immigrant’. Switzerland, 2016, N = 368 respondents

When we provided a list of people in the survey (see Fig. 1), the majority indicated that people from Germany or the US are considered ‘not an immigrant’ (median answer 1 for both, on a scale from 0 ‘clearly not an immigrant’ to 4 ‘clearly an immigrant’, Panel B in Fig. 1; see also Ruedin, 2020), whereas people from Eritrea or Syria are perceived as ‘immigrants’ (median 3.4 for Eritrea, 3.8 for Syria). People tend to think of asylum seekers (69%) and people from poor countries (54%), especially countries with armed conflicts (66%), despite asylum constituting the motive for immigration for only 4.3 per cent of the immigrants in 2014. In addition, return migrants are commonly treated as ‘natives’, despite being new to the place, potentially having few contacts, and facing similar challenges of integration as immigrants do (Vathi et al., 2016).

We argue that in many cases when researchers and administrators use terms like ‘second generation’ or ‘migration background’, they mean ethnic minority groups but try to avoid naming them as such. An ethnic group is defined by common descent, or a common national or cultural tradition, which may involve a shared language, culture, a common narrative of origin and a common claim to history, intermarriage, and a shared identity (Ben-Rafael & Sternberg, 2015). The established scientific literature not only clarifies the meaning of ethnicity, but also highlights that ethnic groups and their fuzzy boundaries are socially constructed (Brubaker, 2009; Wimmer, 2009), malleable and not static (Bochsler et al., 2021; Picker, 2017). While ethnicity can in some cases correspond to nationality (Parameshwaran & Engzell, 2015), it can also span different countries of origin (e.g., Albanians from Albania, Kosovo, or North Macedonia), or even have differences within countries (e.g., Bosniaks, Croats, and Serbs from Bosnia and Herzegovina).

In our view, perceiving some immigrants and their descendants as ethnic minority groups can have advantages over terms like ‘migration background’: Ethnic groups are about societal groups living together, boundary-making and identities that are explicitly socially constructed (Strijbis, 2019), and their grouping can be critically studied, or we can measure when they are discriminated against. We can uncover racist logics (see discourse on non-Western immigrants in Denmark) that otherwise remain hidden when using broad terms such as ‘migration background’. Yet, ethnic groups and their boundaries are also defined from the bottom up, by members of the group — although the view of outsiders can heavily influence this process by pre-defining boundaries and showing groups how they are perceived from the outside (Elrick & Schwartzman, 2015; Jenkins, 1997). With this, ethnicity allows a flexible description of relevant groups and multi-identification of individuals, something that inflexible concepts like citizenship, nationality, or ‘migration background’ cannot offer. Ethnicity, while also used to oppress distinct groups, offers access to multiplicity and certain ranges of agency. Self-declared ethnicity can be measured and used in statistical analyses, should this be desired (Bochsler et al., 2021). Consequently, there is nothing surprising or ‘wrong’ about the observation that some people change their ethnic identity when governments change or that ethnic identity is related to power structures (Bochsler et al., 2021; Posner, 2017).

That said, the survey results cited do not allow us to conclude that the public always thinks in terms of ethnicity. The motive of migration plays a role (compare Blinder, 2015): Seeking asylum (66%) and working (46%) are associated with being perceived as an ‘immigrant’, while studying in Switzerland is not (17%) — possibly because of the often-temporary nature of the relocation (although international students fall under the UN definition that considers residence of 1 year as the minimum to talk of migration). These differences are best explained with a non-ethnic approach, which highlights that the ‘correct’ approach to understanding population groups depends on the context and social phenomenon under consideration (Glick Horvath, 2019; Schiller et al., 2006).

We are aware of the critique of treating ‘migrants’ as ethnic groups, terms that are conceptually different (Glick Schiller and Çağlar, 2013). In the administrative context, ethnicity and race are often unclear to individuals asked to fill out forms (Kertzer, 2017). At the same time, surveys and census data often reify ethnicity as a group to which one belongs (Jenkins, 1997), rather than being about the feeling of belonging. In much of continental Europe the reluctance to regard immigrants and their descendants as ethnic groups is explained by historical legacies such as the Holocaust and colonialism, and strong anti-racist norms that often translate into attempts to adhere to so-called colour-blindness (Ivarsflaten et al., 2010; Song, 2018). In Western Europe, we often observe a rejection of data on ethnicity with reference to Nazi Germany (Simon, 2017) or more recently Rwanda. These cases highlight that in the wrong hands, data on ethnic groups can be fatal. However, new questions in the European Social Survey reveal that individuals have no problems identifying their ethnicity in Western Europe (Schneider & Heath, 2019). Similarly, qualitative research in Switzerland describes how pupils with Albanian parents are considered non-Swiss by their peers regardless of their passport (Duemmler, 2015), pointing to the inadequacy of studying such positioning through ‘migration’. A perspective of ethnicity could grant us tools to study the reproduction of exclusive boundaries and focus on deficiencies that effectively construct children of some immigrants as ‘ineligible’ to become ‘Swiss’ or ‘Dutch’. Indeed, the public seems to differentiate between various kinds of immigrants (Ruedin, 2020). In line with these studies, we argue for a differentiated and situation-specific analysis of such social phenomena.

Our position favours capturing self-declared ethnicity in anonymous studies to allow diagnostics of social problems. This is not the same as arguing for recording ethnic groups in official data and censuses, which tends to take a static view on group membership and is influenced by politics on ‘which box to tick’ (Cooley, 2019).

Perpetuation of difference over time

Ostensibly, the terms ‘second generation’ and ‘migration background’ cater for the fact that the boundary between ‘natives’ and ‘immigrants’ can change and therefore address the critique of methodological nationalism. The terms do so by differentiating between ‘immigrants’ and their descendants who have not moved across the border, but for several reasons cannot or are not considered part of the ‘native’ population.³ The fact that we use these terms signals that we consider them to be different in some way, a particularity used for boundary-making. When we designate people having a ‘migration background’ and define them by the citizenship and birthplace of their parents, we highlight an immutable characteristic. There is no room for changes over time or across contexts of the everyday: A person keeps the ‘migration background’ irrespective of naturalisation, residence time in the country of destination, language ability, or identity and behaviour.

By differentiating between ‘generations’, individuals are placed in the ‘immigrant’ category, not as part of the majority population, despite the view that their assumed ‘difference’ has diminished (Klarenbeek, 2019; Lessard-Phillips et al., 2017). Like the ‘one drop’ definition of Black populations used in the United States in the 19th and 20th Century (compare Citrin et al., 2014) migration remains a deficit that is inherited and that can potentially be lessened over time — while at the same time, as recent policies show, can remain a sturdy reason for discrimination. It expands beyond nationality and allows for a constant definition that is often racialised (Zhang, 2020). In the established Danish context, the Ghetto laws allow for the forced eviction of ‘non-Western immigrants’ from areas where their number is higher than 30%, including both (foreign born) immigrants and their (Danish-born) descendants, based on the parents being born outside Denmark (often targeting Muslim minorities and thus disclosing discrimination towards a specific religion). No obvious point exists at which descendants of immigrants cease to be ‘immigrants’; they inherit foreignness, independent of time spent at a place. Existing definitions of ‘migration background’ have no room for individuals becoming part of the majority population within one’s life course (Lessard-Phillips et al., 2017). They are caught between groups and are seen as partially (non)belonging to both, their relatives, and citizens.

Being ‘a migrant’ trumps the actual situation of individuals and becomes relevant not only legally but also as an ascribed identity. Yet, the emphasis on differences between people with a ‘migration background’ and members of the ‘mainstream population’ often happens where we recognise diminishing differences: among children and grandchildren of immigrants. While they are not considered part of ‘the majority’, the categorisation into generations often describes an increased similarity to the majority population. This is apparent in terms like ‘1.5 generation’ — but also finer distinction like ‘1.25’ and ‘1.75 generation’ (e.g. Tran, 2017). The term ‘generation’ expresses similarity to the majority population without measuring this. We end up lumping everyone together: All immigrants are regarded as (inherently) ‘different’, where the second generation is still more 'different’ than similar, while for the third (and subsequent) generation we ostensibly discern some lingering difference. However, we do not measure anything, neither socio-economic background, education, language, and socialisation in the country of current residence, nor the potentially structural factors that can disadvantage these groups. Consider for instance the observation that “[t]he only exception is among ‘3 + generation’ Asian-Americans, who are consistently more likely to vote in states with high proportions of co-ethnics” (Ramakrishnan & Espenshade, 2001:891). While we can readily identify this ‘group’ in quantitative data, this exception may describe a statistical artefact where random differences in small groups appear relevant, may point to missing variables, or indeed suggest that we are looking at ethnic groups as the authors imply in this instance: Although the group is identified by one of their ancestors having migrated, the political behaviour at present is unlikely to be shaped by this fact. Migration may have played a (large) role in the genesis of the groups in question, but we do not expect that this characteristic defines the group on its own. Hence, if necessary for our specific research context, we need to be precise in referring to children of immigrants, children with one immigrant parent and one non-immigrant parent, and immigrants who have completed much of their education and socialisation in the current country of residence rather than using obfuscating terminology.

The way we count generations in migration research contrasts with how we use it in other contexts, such as technology or in professional contexts: A ‘second’ or ‘third generation’ craftsperson is understood to stand out, due to accumulated experience and thus their competences. In migration research, the use of generations implies a lesser connection to the mobility of prior family members (e.g. parents and grandparents). As we have seen in the survey, public understanding does not correspond to how generations are used by researchers: children of immigrants are by many not actually perceived as an immigrant. The only way out of this conundrum is to differentiate between immigrants and everyone else — which means giving up on the logic of generations. This is one possibility that we suggest: If there are relevant differences among descendants of immigrants, this may suggest a matter of ethnic boundaries and discrimination that do not consider whether a person is an immigrant or a descendant of an immigrant (as is often done in field experiments on discrimination, Zschirnt & Ruedin, 2016).

Concluding discussion

In sum, we criticised the construction of ‘migrant’ terminology and broadened the discussion on differentiation beyond debates on ‘asylum’ and ‘illegality’ with a focus on quantitative and experimental research. We demonstrated that the common terms ‘migration background’ and ‘second generation’ lack precision and imply false homogeneity, focus on deficits, and perpetuate difference exactly in situations where we assume increasing similarity because of social and civic integration — in fact, these terms deny ‘immigrants’ the possibility of demonstrating the integration demanded of them. Both terms essentialise groups despite extensive research demonstrating that immigrant groups are internally heterogeneous, and that stereotyping can lead to discrimination, irrespective of what we call the affected populations (Zschirnt & Ruedin, 2016). Their use often distracts from understanding relevant social phenomena and the mechanisms that cause them. Yet, our review showed that despite the rigidity of the terms, we still find an apparent acceptance of their social construction in the literature (Appendix 1; see also Sen & Wasow, 2016). Despite critical work on the use of migration terminology, quantitative and experimental work often uses a standardised set that can prevent us from properly understanding the relation between migration background and poverty, or reasons for discrimination, rather than reproducing difference.

What we criticise is however not necessarily the variables in these datasets, but how existing data are employed and presented in quantitative research specifically. Hence, we do not want to discard specific labels as such, for instance references to ‘origin’ or ‘background’ in studies that are interested in how ‘migrants’ are discriminated or supported. Instead, we insist on precise measurement of social phenomena and call for more careful descriptions and circumscriptions of the (sub-)populations we study.

Our analysis on the shortcomings of above-discussed terminology within quantitative works also adds to work done within qualitative research (Dahinden, Fischer and Menet, 2020). Here we tap into the debate on de-migranticisation (Dahinden, 2016) and re-migranticisation (Dahinden and Anderson, 2021). This should help us social scientists not to be overly influenced by the politicisation of immigration and assume that migration should a priori be the right perspective. Indeed, we encourage all researchers to routinely consider systems of classification beyond migration and citizenship in our quest to describe inequalities that often arise at the intersection of the historical, social, and political (Crenshaw, 1989). Categories are perspectival and performative and for quantitative and experimental analysis (which is often data driven), we should therefore find ways to precisely measure whether the phenomena we study really correspond to existing variables. By using large datasets, however, quantitative analysis is a priori in an advantageous position to reflect the heterogeneity within groups studied (Chimienti et al., 2021a, 2021b).

For quantitative and experimental research, we have further suggested several steps for a more careful and thoughtful analysis. For the design stage, we suggested a thought experiment to consider alternative perspectives, and we highlighted that self-declared ethnicity should be included in a larger number of surveys if we want to avoid quasi-ethnicity via variables such as ‘migration background’. That way we can stop trying to draw inferences based on country of birth and citizenship which then act as quasi-ethnicity (Simon, 2017) — an approach that leads to the problematic concepts of ‘migration background’ and ‘second generation’. Specifically, when designing quantitative and experimental research we should ask ourselves how we would interpret the results if gender or educational differences were the relevant categories (rather than control variables). Indeed, identifying and naming relevant social mechanisms directly should be preferred. Such a focus on mechanisms and social ‘problems’ may lead us to recognise that many phenomena attributed to migration are not really about immigration but about discriminating policies, communication, or economic disparities (Horvath, 2019).

While we provided survey evidence that the majority population often understand social divisions in terms of ethnicity, we certainly do not call for an ethnicisation of society. Using statistical interaction terms between (self-declared) group membership can address intersectionality (Hancock, 2014). However, we note that interaction effects and subgroup analysis quickly outrun even larger datasets, so ‘big data’ alone will not be the answer and cannot replace disciplined theoretical thinking. Therefore we urge researchers to use well-reflected terminology — independent of or critically reflecting political discourse, and free from pre-conceptions to the extent that this is possible. If we argue that often self-declared ethnicity may be more appropriate a term, we do so in the spirit of Brubaker (2002) and Wimmer (2008), recognising that these groups are neither fixed, nor mutually exclusive — especially as immigrant integration, adaptation, and assimilation advance. It is not a question of replacing the terms ‘migration background’ and ‘second generation’ but a call for a more flexible and situation-specific approach. In this sense, we note that migration is not the problem, but there are problems around migration, and that their causes need to be studied beyond a purely migration-focused lens (Horvath, 2019).

For the analysis and interpretation of results, we hence encourage researchers to consider alternative perspectives, a precise description of the population under study — especially when subsets are studied — rather than referring to imprecise terms like ‘immigrants’ or ‘migration background’. We suspect that a more precise vocabulary is useful to capture various phenomena: using ‘language skills’ if this is what matters in education, or ‘ethnic groups’ or ‘stereotypes’ when it comes to discrimination, or when appropriate ‘social class’, socio-economic exclusion, marginalisation, economic vulnerability, or social segregation to name just a few possibilities.⁴ We also highlighted how interaction effects allow quantitative researchers a precise analysis that emphasises heterogeneity within the population under study.

There is no single terminology or concept that fits all purposes, but with a focus on social phenomena and mechanisms we can at least strive for a precise description rather than rush into preconceived conclusions simply because a concept like ‘second generation’ is widespread in administration, politics, and academic literature: Let us pause for a moment to reflect on why we consider a particular social cleavage important.

Appendix

To better characterise the literature and support our argument, we have undertaken a principled review of articles. We searched in the migration journal with the highest impact factor (Comparative Migration Studies), as well as in the articles included in the Directory of Open Access Journals (DOAJ). With the DOAJ we aim for a broad coverage across disciplines and specialities; with the CMS we want to check if the language use in the specialist literature is notably different. We searched for: “second generation”, “migration background”, “migration origin”, “self-declared” ethnicity, “language use”, “language skills”, “language proficiency”, “unwillingness to integrate”, “incompatible values”, and “low level of education”. For the DOAJ, we added migra* in the case of “second generation”, “language skills”, “language proficiency”, and “low level of education” to exclude excessive numbers of false positives unrelated to human migration. The keywords reflect the terms we criticise for the way they are used in the literature, as well as a range of specific concerns that we listed as possible specific and precise terms (Fig. 2).

Fig. 2

PRISMA Table for the principled literature review on second generation, migration background, migration origin, and related terms, 2023

In a second step we looked at how each of the articles made use of the terms identified by the search (Table 1). We can see that there are many more relevant articles referring to generic terminology (‘migration background’, ‘migration origin’, ‘second generation’), the use of which we criticise. By contrast, more specific references to language, education, and self-declared ethnicity are clearly less common in the literature. The keywords not included in the table yielded very few articles, none of which turned out to be relevant.

Table 1 Relevant articles found in CMS and DOAJ for different keywords, 2023

Looking at how the articles actually operationalised immigrants (Table 2), we find that the most common form reflects an ethnic understanding, followed by the country of birth being different (‘foreign born’) and specific countries or regions of origin. There are some differences between the keywords used as entry point: ‘migration background’ yields many studies that use a German-style official definition, but also many that rely on being born abroad. Ethnic perspectives are common. Using ‘second generation’ as entry point, ethnicity is most common, followed by foreign birth, German-style migration background, and country of origin. Using ‘ethnicity’ as keyword, we get almost only studies using an ethnic understanding; for ‘language’ as a keyword, we also get many ‘ethnic’ perspectives and references to being born abroad.

Table 2 How immigrants are operationalised in relevant studies, overall, and by keyword, multiple classifications possible, 2023