This is an open access article distributed under the terms of the Creative Commons
Attribution License (Global migration challenges nations that receive large numbers of immigrants. Germany is one such country. As they encounter newcomers from foreign countries, citizens and policy-makers struggle to identify what makes them different – if anything. Debates over national identity are not merely theoretical but have concrete implications for policies that regulate citizenship, such as naturalization tests and integration courses (Orgad, 2015). With few exceptions (Billig, 1995; Reicher & Hopkins, 2001) social psychology has done little to contribute to these far-reaching debates.
In the tradition of social identity theory, social psychologists thus far have treated “national identity” as one of many social identities. This implies examining identification processes rather than identity content (Ashmore, Deaux, & McLaughlin-Volpe, 2004), e.g. under what circumstances people categorize themselves in a social identity (Leach et al., 2008). Even in the rare cases when social psychologists actually examine identity content the concept remains de-contextualized and abstract (e.g., “simple” versus “complex” identity structures, Roccas & Brewer, 2002). While such abstraction allows for parsimonious generalizations across identity contexts, it also leaves a critical gap in our understanding of national identity. National identity is, of course, a social category; but it is also a socio-political construct – shaped by concrete historic contexts.
Departing from this social identity tradition, the current research inductively explores how German citizens construct their national identity. We coded nearly 1,000 open-ended responses about “what it means to be German”. Following qualitative coding, we submitted our codes to a latent class analysis. We build on a literature in the social sciences that distinguishes between civic (shared institutions) vs. ethnic (shared ancestry) models of national identity (Brubaker, 1992). Our main objective is to understand how citizens of Germany – a country that has often served as a paradigmatic case for an ethnic model of national identity but has changed recently – construct their national identity. Using a combination of qualitative and quantitative methods we aim to capture nuances in national identity as a way of talking about the self and one’s community (Billig, 1995), while at the same time identifying the structure of national identity content (i.e., investigating if an ethnic and civic type or more types emerge). To better interpret the classes we also tested how they relate to group-based guilt and degree of identification with Germany.
Past Research on National Identity at the Macro-Level
Early research developed models of national identity based on the history, politics, or laws of countries, i.e. concepts from an institutional- or macro-level. Despite differences in their nuances, most of these theories converge on two dimensions: the political structures of the state and the ethno-cultural heritage of citizens generally referred to as “civic” and “ethnic” models of nationalism (Brubaker, 1992; Kohn, 1944; Smith, 1986, 1991). According to these theories, nations with a civic identity derive from a common institutional history and the fundamental principles according to which the state is governed. Citizenship is awarded primarily for being born in the state territory (“jus soli principle”). The United States are often cited as a paradigmatic example for a civic nation.
In contrast, nations that are home to an ethnic national identity are defined by a common culture and heritage. In what is essentially an ethnic community, inclusion is restricted to those who share common descent. Citizenship is awarded primarily for being born to citizens (“jus sanguinis principle”). Germany is often cited as a paradigmatic example for an ethnic nation (e.g., Brubaker, 1992). Indeed, until 2000 having German ancestors played a large role in who could become a citizen. However, since 2000, recessive reforms have liberalized German citizenship policies. In 2016 Germany’s citizenship policies ranked among the least restrictive in Europe (Thränhardt, 2017). Of course, we do not know if German citizens’ experience of national identity parallels or deviates from these changing national identity conceptions at the macro-level.
Past Research on National Identity at the Micro-Level
How citizens define their national identity can conflict with how it is defined and regulated by political elites (Yogeeswaran & Dasgupta, 2014). It is difficult to infer one from the other. Citizens do not always agree with or even follow elite discourse and policy making, while elites are not always responsive to citizens. Thus, understanding the subjective meaning citizens assign to their national identity, their sense of belonging to an “imagined community” (Anderson, 1991) including everyday rituals and practices (Byrne, 2007) ought to be an independent exercise. We build on two research streams that have engaged in this exercise. One stream operationalizes citizens’ national identity as ratings of different membership criteria, and relies on quantitative methods. The other characterizes it as a way of talking about belonging, and relies on qualitative methods.
National Identity as Rating of Membership Criteria
Research in this stream typically asks citizens to agree or disagree with civic vs. ethnic national identity content. The most frequently used scale asks about the importance of several characteristics for being a member of a nation, including respecting a country's political institutions and laws, feelings for the nation, speaking the language, having been born in the country, living in the country for most of one’s life, having citizenship, and being Christian (e.g., Citrin, Wong, & Duff, 2001; Hjerm, 1998a; Jones & Smith, 2001; Shulman, 2002).
Typically, citizens’ responses are submitted to a factor analysis that tests how many independent dimensions of national identity emerge. Most studies find two factors that map onto the civic and ethnic typology from the macro-level (e.g., Jones & Smith, 2001; Reeskens & Hooghe, 2010). However, some studies identify only one factor anchored by more “inclusive of ethnic minorities” items on one end and more “exclusive of ethnic minorities” items on the other (Meeus, Duriez, Vanbeselaere, & Boen, 2010). Others identify a third dimension, often called “culture”, grouping existing items like “speaking the language” or “being Christian”, or adding new items like “cherish traditional lifestyle”, and “dedicated to preserving a country's culture” (Eugster & Strijbis, 2011; Reijerse, Van Acker, Vanbeselaere, Phalet, & Duriez, 2013). Interestingly, the largest representative cross-country survey using this national identity content scale (ISPP) produced higher mean scores for civic than for ethnic items in Germany in 2003 (Helbling, Reeskens, & Wright, 2016). Taken at face value, this would suggest that only three years after the first liberalization of German citizenship law citizens’ understanding of their national identity was civic – a far step from its (supposedly) ethnic tradition.
The quantitative approach allows researchers to work with large representative samples, conduct cross-cultural comparisons, and link citizens’ understanding of their national identity to various macro-level outcomes, such as party positions (Helbling et al., 2016). Yet, it also has some disadvantages. Participants’ answer choices are constrained by the presented items that mainly revolve around citizenship regulations – constructs from the macro-level. While the political context certainly shapes national identities (Scuzzarello, 2015), some meanings might be detached from frames and have little connection to the macro-level. Such meanings remain unexplored.
Presenting civic and ethnic items might also communicate expectations for what are socially desirable answers, a potential alternative explanation for the high civic score among Germans. In fact, items about membership criteria may not even tap into the meaning of national identity but rather into socially acceptable conditions for restricting membership in the national community (Helbling et al., 2016).
National Identity as a Way of Talking About Belonging
Researchers in this stream see national identity as a way of talking about the self and community (Billig, 1995). According to this perspective, when prompted to think about national identity citizens actively construct explanations that locate them in a social, cultural, and historic context (Haste, 2004). Open-ended essay questions or interviews are better suited for capturing this process than closed-ended questions. Through this approach, researchers in the UK and the USA have discovered novel aspects of national identity: for example, reflections about geography (Abell, Condor, & Stevenson, 2006), or about economic opportunities (Warikoo & Bloemraad, 2018), and that citizens sometimes even question the rationality and morality of national categorization (Condor, 2006). In our past research, using open response questions we found – in sharp distinction to the closed-ended survey described above – that when asked about the value of being German, a majority of German participants evoked “heritage-based concepts”, i.e., references to cultural traditions and self-descriptive traits (Ditlmann, Purdie-Vaughns, & Eibach, 2011).
Qualitative approaches thus provide a more nuanced view on national identity content, and better capture its narrative, dynamic, and contextualized nature. At the same time, they often rely on small samples, and are better suited for capturing idiosyncrasies than commonalities in national identity content across groups in society, and across nations. To harness the benefits of both quantitative and qualitative approaches, the current research combines open-ended response coding with latent class analysis.
National Identity Content Versus Processes of Identification
Past research on national identity often confounds the meaning people assign to their national identity with the process of identification, i.e. their cognitive or affective orientation towards their national identity. For example, civic-ethnic identity items include questions about the content of national identity, e.g., “being Christian”, as well as questions about the process of identification, e.g., “feeling German.” Research on nationalism versus patriotism similarly confounds these two concepts: items ask how strongly respondents identify with different identity content: superiority over other nations vs. democratic values (Kosterman & Feshbach, 1989; Wagner, Becker, Christ, Pettigrew, & Schmidt, 2012).
Because identification has different implications depending on identity content, confounding process and content can be confusing. For example, a high degree of identification is associated with hostile behavioral intentions in Northern Ireland – but only among citizens who assign an antagonistic meaning to their national identity (Livingstone & Haslam, 2008). Similarly, a high degree of identification is associated with rejecting refugees – but only among host citizens who define their identity in essentialist terms (Pehrson, Brown, & Zagefka, 2009). Building on this insight, the current research assesses the meaning people assign to their national identity separately from the process of identification.
Similar to identification, experiencing collective emotions is also a process that is related to, but distinct from identify content. People who see past atrocities in the content of their national identity may experience collective guilt, i.e. an aversive emotion on behalf of their ingroup (Wohl, Branscombe, & Klar, 2006). Experiencing collective emotions requires some degree of identification. Too much identification, however, can lead to a glorification of the past and denial of collective guilt (Rensmann, 2004), thus altering identity content. Accordingly, one might ask what meaning people who feel collective guilt assign to their national identity. For example, Germans with their negative history associate only sport symbols with their national flag (Becker et al., 2017), is the same true for their national identity? The current research explores this bidirectional relationship between identity content and collective guilt.
Current Research
We asked nearly 1000 German citizens to provide open-ended responses to the question “what does it mean for you to be German?” We then content coded these responses and submitted the codes to a latent class analysis (LCA). Latent class analysis allows us to identify groups of participants who share a common understanding of national identity, and to make comparisons between groups. It also provides insights into the structure of national identity (i.e., if an ethnic and civic type or more types emerge), and the complexity of each type (i.e., how many different categories characterize the responses of participants in each group).
We selected Germany as a research case for two reasons. The main reason is that Germany – often cited as a paradigmatic case for an ethnic model of nationalism – underwent substantial changes at the macro-level in recent years. We want to know if and how these changes are reflected in the voices of citizens.
A secondary reason is the difficult relationship of many Germans with ethnic-based nationalism because of World War II (Joppke & Rosenhek, 2002). On the one hand, historically jus sanguinis principles were introduced to German citizenship laws partly as special provisions for ethnic Germans who were expelled from areas in Eastern Europe (so called “Aussiedler”) after World War II. On the other hand, many Germans today feel conflicted about the idea of ethnic nationalism because of the central role it played in Hitler’s rise to power. In light of German history it is interesting to examine how strongly identified Germans construct their national identity. Do they resort to the traditional jus sangunis ideas or can they strongly identify with a more inclusive meaning of being German? It is further interesting what identity content promotes the experience of collective guilt, and, relatedly, if associating history with national identity always implies more collective guilt. To understand how the national identity types that emerge in the LCA relate to process of identification and collective guilt, we include these as covariates in our model.
Methods
Participants
Using purposive sampling (Battaglia, 2008) we successfully recruited nine-hundred-eighty-seven German citizens to participate in our study. Our recruiting strategy targeted different societal groups (young and elderly, supporters of different political parties, ethno-religious minorities, low SES groups). Table 1 displays successfully contacted organizations. About one third (n = 328) of the participants were female (n = 18 missing), mean age of the sample was 44.53 years (SD = 18.45), the education level of the sample was rather high with 76% of participants reporting a high level of education, 18% a medium, and 6% a basic educational level (n = 20 missing). 15% of the participants had a migration background, i.e. at least one parent was not born in Germany (n = 54 missing). Political orientation was M = 5.05 (SD = 2.40) on an 11-point Likert-scale (1 = left; 11 = right). A clear minority (8% of participants) lived in a city with a population above 1 million, 25% in a city with more than 100,000 citizens, 36% in a medium size city between 20,000 and 100,000 residents, 15% in smaller cities (less than 20, 000 residents), and 17% lived in villages (n = 17 missing). We used random imputations for missing values in all analyses involving missing demographics (R package random.imp; Gelman & Hill, 2006).
Table 1
Number of Successfully Contacted Organizations
| Political activist groupsa | 7 |
| Major political parties | 16 |
| Youth organizations of major political parties | 7 |
| Turkish cultural organizations | 6 |
| Christian organizations | 2 |
| Welfare institutions (e.g., Agency for Labor, retirement homes) | 6 |
| University organizations | 6 |
| Companies | 1 |
| Total | 51 |
aFrom the left and the right end of the political spectrum.
Compared to information about the German population from the German statistical office (https://www.destatis.de/EN/Homepage.html), participants in our sample were more left-leaning, a bit younger, more educated, and more often male than the German average. However, our sample includes a sufficient number of members from all groups to make meaningful comparisons between demographic groups. To ensure that any new results are not merely a function of the specific composition of our sample, we furthermore replicated the main results from past research using closed-ended national identity content scales with our data (Appendix B).
Procedure
We programmed and administered our survey online using the program Unipark. Participants were recruited through advertisements on listserves or newsletters of different political, social, and educational organizations. To gain access to these listserves and newsletters, members of our research team first identified appropriate organizations online, and then mailed letters and sent emails to the contacts listed on organizations’ websites. We selected organizations that represent different political, educational, age, and income groups, as well as cultural associations targeting first and second generation immigrants. The letters and emails sent to organizations described the topic of the survey, and asked to forward the link to their members. In the retirement homes we also offered a paper-pencil version of the survey. After participants provided informed consent and completed the survey, they were debriefed and entered into a raffle for 350 Euro. All data were collected between September 2011 and March 2012.
Materials
Talking About National Identity
To assess how participants construct their national identity, we asked the following open-ended question:
What does it mean to be German?
Please think about what it means for you to be German carefully. Please take your time for that – about 3 minutes. Then please write freely in the box below what it means for you to be German. Write as much or little as you wish. Spelling and grammar are not important. There are no right or wrong answers. Let us know your thoughts and feelings!
Degree of Identification
To assess to what degree participants identified with being German we adapted eight items from previous research (Doosje, Branscombe, Spears, & Manstead, 1998; Hofmann, Baumert, & Schmitt, 2005; Karasawa, 2002; Mummendey, Kessler, Klink, & Mielke, 1999). We selected items that cover the cognitive (self-categorization) and affective (attachment) process of identifying as German, independent of the content of this identity. After writing their identity essays, participants indicated their agreement with each of eight items (e.g., “I identify with other Germans”, “I am proud to be German”) on a five-point Likert scale (1 = strongly disagree; 5 = strongly agree). Items were averaged to form a single score of “degree of identification” with higher values indicating higher identification (Cronbach’s α = .91; M = 3.49; SD = 0.97). Twenty-six participants failed to complete this scale (see Table 2 for a complete list of items).i
Table 2
Degree of Identification Scale: Wording of Items
|
Note. Participants indicated the importance on a five-point-Likert scale ranging from 1 (= strongly disagree) to 5 (= strongly agree).
Group-Based Guilt
In order to capture emotions of group-based guilt, we adapted five items of a standard measure (Doosje et al., 1998) to the context of historical wrongdoings of Germans towards Jews during the Holocaust. Participants indicated their agreement with each of five items (e.g. “I feel guilty about the negative things Germans have done to Jews”, “I feel regret for Germanys harmful past action to Jews”) on a five-point Likert scale (1 = strongly disagree; 5 = strongly agree). Items were averaged to form a single score of “group-based guilt” with higher values indicating more guilt (Cronbach’s α = .88; M = 2.46; SD = 1.04, 73 missing).
Demographics
To obtain information about our sample and to investigate variation in the content of national identity within Germany, we asked participants a series of demographic questions, including gender, age, education, place of residence, migration background, and political orientation. For demographics, degree of identification and group-based guilt covariates we used random imputation to replace missing values.
Content Coding
Development of Coding Manual
To content code participants' essays about what it means to be German, we modified a coding system that was developed in past research with similar text material (Ditlmann et al., 2011) using an empirical-inductive approach (Berg, 1998; Mayring, 2000; Smith, 2000). We – again inductively – revised and expanded the original coding manual in light of the current, more extensive text material: One author identified additional or different themes while reading all of the current essays and conducting the coder training described below. Both authors then discussed each proposed modification with the goal of capturing the themes from the data as best as possible while honoring the principle of parsimony (Berg, 1998). This procedure resulted in 12 final categories that we define and describe in Table 3.
Table 3
Variables of the Content Analysis of Essays and Intercoder Reliabilities
| Text example | Percent Agreement | Cohen’s Kappa |
|---|---|---|
| Biology | 98.20% | .91 |
|
||
| Culture | 94.24% | .88 |
|
||
| History | 97.48% | .94 |
|
||
| Language | 99.28% | .98 |
|
||
| Nativism | 99.64% | .99 |
|
||
| Personality Traits | 98.65% | .95 |
|
||
| Polyethnicity | 98.20% | .94 |
|
||
| Democracy | 98.92% | .97 |
|
||
| Freedom | 99.64% | .98 |
|
||
| Economy & Safety | 96.04% | .89 |
|
||
| Welfare | 98.20% | .92 |
|
||
| Formality | 98.92% | .97 |
|
||
| Total | 98.20% | .93 |
Note. Percentage agreement and Cohen’s κ are based on the coding of a sub-sample of n = 278 essays out of the overall sample (corresponds to 28.16%) by two independent coders.
Coding Procedure
Two male coders used these 12 categories to analyze the content of the national identity essays. Before starting the actual coding process to ensure a high level of objectivity in the coding, both coders first practiced coding with four training modules, 20 texts each from a prior research project with a similar open-ended question (Ditlmann et al., 2011). After finishing a single module, the coders discussed difficulties and clarified questions, until they reached a satisfying level of independent coding agreement (Cohen’s kappa .80 or above; Cohen, 1960; Neuendorf, 2002).
Coders were instructed to divide all essays into theme units that express single ideas and aspects of national identity. They then went through the essays unit by unit. Double coding of a theme unit in two different categories was not allowed (i.e., the codes were mutually exclusive). Additionally, to achieve high reliability, categories could only be coded once per essay. In other words, coders coded whether a given category was present or absent in each essay. People used on average 9.58 categories in their essays (range: 3 – 12). One coder coded all essays (N = 987), while a second coder coded 278 randomly selected essays. For a summary of the coefficients of the intercoder reliability of the two coders see Table 3. Percent agreement and Cohen’s κ (Cohen, 1960) were computed using the “irr” package (Gamer, Lemon, Fellows, & Singh, 2010) for the statistical software R (R Development Core Team, 2011). Both reliability estimators were very high across all categories (Mpercentage= 98.20%, SDpercentage= 1.46; Mκ = 0.93; SDκ = 0.08, Neuendorf, 2002). The number of double-coded essays (N = 278) exceeded the minimum number required to infer an estimated percentage-agreement of .90 for the entire sample from the reliabilities reported in Table 3 (α-error of 5%; Riffe, Lacy, & Fico, 2005).
Data Analysis Strategy
After coding we conducted an LCA. We used LCA to identify meaningful profiles of national identity based on our 12 coding categories. LCA models profiles or typologies in discrete data (Goodman, 2002; McCutcheon, 2002). It assigns each participant to a class with a specific probability based on the unique combination of codes in his or her response. The classes, in turn, are defined by how probable it is that a class member mentioned each category. Latent Class Analysis is similar to confirmatory factor analysis in that a number of clusters/factors is specified based on fit criteria. It is different in that observations (in our case people) get grouped into classes, as opposed to variables into factors. We used the package “PoLCA” (Linzer & Lewis, 2011, 2013) for the statistical software R and selected the number of latent classes based on fit. We then replicated the model, this time including the covariates degree of identification and group-based guilt in the LCA. To assess model fit, we used the Bayesian Information Criterion (BIC; Schwarz, 1978). Smaller values of the BIC indicate better fit, and the fit index includes a penalty for number of classes in the model (see Appendix A).
In the first and main part of the results section, we present the findings from the LCA. After selecting the model with the best fit, we describe each of the emerged classes and then proceed with interpreting the relationships between covariates and latent classes, i.e. whether having a higher or lower value on degree of identification and group-based guilt is predictive of membership in specific classes. We conclude the section with an analysis of demographic variables as predictors of class membership in a multinomial logistic regression. The combination of inductive content coding and LCA can harness the benefits of qualitative and quantitative approaches. It captures nuances in national identity, yet allows for the identification of specific identity-types and comparisons between them.
Results
Latent Class Analysis
To identify classes of national identity content, we compared models with 1 to 8 latent classes using the criteria described above (see Appendix A for the BIC value for each model). We selected the four-class model because the BIC is slightly lower than in the three- and five-class models. When moving from a three- to four-class model, the evidence against the higher BIC is Δ24.86; that is strong evidence against the higher BIC (Kass & Wasserman, 1995). When moving from a four- to a five-class model, the evidence against the higher BIC is weak, Δ2.02 (Kass & Wasserman, 1995).
Figure 1 shows the relative frequencies of coding categories conditional on latent class of national identity content in the selected four-class solution. For example, participants in the heritage-based class have a .82 probability to write something that would be coded as culture as compared to a .48 probability in the entire sample.
Figure 1
Four latent classes of national identity compared to the overall distribution.
Inspecting the relative frequencies for each of the classes, we labeled the largest class (39% of participants) heritage-based national identity because culture and language have the highest probabilities in this class. We labeled a smaller class (19% of participants) ideology-based national identity because broad-based, abstract values like democracy and freedom have relatively high probabilities, which is consistent with Ditlmann et al.'s (2011) original definition for ideology-based national identities. We labeled the second most frequent class (26% of participants) legal-formalistic national identity because the category formalities, which primarily describes the legal rules and obligations of holding or obtaining citizenship, has a relatively high probability. Finally, we labeled the fourth and smallest class (16% of participants) traits-based national identity because high probability categories were traits, such as being punctual. We did not label classes cultural, ethnic, or civic to highlight that they emerged from open responses that were broader and more nuanced than common ethnic and civic identity content scales. In what follows we describe the four classes in more detail.
Heritage-Based National Identity Class (39% of Participants)
Being German means to have a common identity and history. To live according to certain values and virtues […]. To identify with my culture. Pluralism is enriching. (Study Participant)
Being German means to me to belong to the nation of “poets and philosophers.” This obliges me to think about things and use my intellect. (Study Participant)
The topics most frequently mentioned by participants in this class are culture – for example the reference to poets and philosophers above – and language, for example, “I associate my personal being German with the language I speak.” The heritage-based class thus includes a wider range of expressions of ethno-cultural heritage than typical ethnic identity scales. In addition, categories traditionally associated with the ethnic identity concept (e.g., biology: .25; and nativism: .30) – while more frequent in the heritage-based class than in the overall distribution – play a subordinate role relative to language (.56) and culture (.82). As illustrated by the reference to “pluralism” in the second quote above, another category, poly-ethnicity, appears here (.18) more often than in the overall distribution (.13). An endorsement of one’s own culture thus may be associated with a respect for all cultures in some cases.
In general, the heritage-based class appears to be a broader and more complex version of the recently introduced cultural identity type (Eugster & Strijbis, 2011; Reijerse et al., 2013). It could also be seen as a modern version of what Meinecke (1908) called “Kulturnationen”, nations that are primarily based on a common culture. Interestingly, history is also mentioned more frequently in the heritage-based class (.36) than in the general distribution (.18). In many instances mentioning history goes along with highlighting a special responsibility because of the Holocaust, for example in this response by one participant: “It also means to be to bear the historical burden of the ‘Third Reich.’ A certain responsibility for those who are persecuted arises from that […].”
Ideology-Based National Identity Class (19% of Participants)
Being German means to live in a free constitutional state, to live in peace and economic security, to travel worldwide. It means great social welfare and health insurance. (Study Participant)
As illustrated by the opening quote, the ideology-based national identity class is dominated by democracy (.47), welfare (.52), and economy & safety (.49). Freedom, e.g., “a free constitutional state” or “living in a free country” also occurs more often here (.34) than in the overall distribution. Like the heritage-based national identity vis-à-vis the ethnic identity concept, the ideology-based identity includes a wider range of expressions of civic values than typical civic identity scales. In particular, the categories welfare, economy, and safety that are frequent in the ideology-based class go beyond common civic identity scales. They are reminiscent of findings by Warikoo and Bloemraad (2018) that identified economic opportunities as central narratives in interviews about US national identity with immigrant-origin youth.ii
Legal-Formalistic National Identity Class (26% of Participants)
For me, being German means that I was born in Germany by accident and have a German passport. Nothing more!! (Study Participant)
I happen to be a resident of this country. (Study Participant)
The most frequent and clearly dominant topic in the legal-formalistic class is formalities (.40), which describes the legal requirements for obtaining and holding citizenship. The only other above-average category in this class is nativism (.29) – the formal requirement of having been born in the country. The emphasis on the randomness of place of birth (“accident”, “happen to be”) in the quotes above illustrates that at least some participants list formalities to distance themselves from the notion of a meaningful German national identity. Perhaps they provide politico-legal definitions of citizenships and borders instead of narratives about meaning to express resistance to any form of nationalism.
Traits-Based National Identity Class (16% of Participants)
Typical positive traits of a German for me: honest, industrious, punctual, neat […]. Typical negative traits of a German for me: narrow-minded, interested in politics but not active enough, envious. (Study Participant)
The most frequent and only above-average category in the traits-based class is personality traits (.62), which includes both positive and negative traits that are supposed to characterize Germans. Despite its low frequency, this is an intriguing class, because it is completely distinct from national identity at the macro-level. Top-down national identity scales do not include them, but national stereotypes, often traits ascribed to a group, have a long tradition in psychological research (Peabody, 1985). Although citizens are unable to accurately name common personality traits of their co-nationals as measured by a Big Five questionnaire (Terracciano et al., 2005), self-ascribed traits can demarcate boundaries. People sometimes prefer that their group is characterized by distinct traits, including negative ones, over being indistinguishable from another group (Mlicki & Ellemers, 1996). Traits can also define prototypical group members, and then exclude those who are less prototypical (Huynh, Devos, & Altman, 2015). Accordingly, describing Germans as punctual or neat can be a way of reinforcing group boundaries and negotiating membership in one’s national group.
National Identification and Group-Based Guilt
To check if membership in one or more of the four identity content classes is associated with higher or lower levels of identification and/or group-based guilt, we included identification and guilt indicators as continuous covariates in the LCA. Table 4 shows the coefficients from the multinomial regression and Figures 2a and 2b depict the pattern graphically.iii
Table 4
Multinomial Regressions Predicting Membership in Each Identity Class Relative to Ideology-Based Class (Using R/poLCA)
| Variable | Legal-formalistic
|
Traits-based
|
Heritage-based
|
|||
|---|---|---|---|---|---|---|
| Coef. | SE | Coef. | SE | Coef. | SE | |
| National identification | -.87** | .16 | .14 | .17 | .33* | .15 |
| Collective guilt | -.13 | .15 | -.39* | .15 | -.17 | .13 |
| Constant | .06 | .21 | -.21 | .21 | .65** | .18 |
Note. N = 987; df = 930; BIC = 10945.
*p < .05. **p < .01.
Figure 2a shows that as people identify more with being German, they are less likely to belong to the legal-formalistic class and more likely to belong to the heritage-based class. We find no evidence for a similar relationship between national identification and membership in the ideology or the traits-based class. This pattern confirms what we also deduced from the analysis of class content: the legal-formalistic class characterizes people who reject the notion of national identification and thus reduce its meaning to administrative formalities. Meanwhile the largest, heritage-based class characterizes people who identify with and feel especially attached to Germany.
Figure 2b shows that as people report more group-based guilt, they are more likely to belong to the ideology-based class relative to the traits-based class. The traits-based class thus seems to allow members to draw boundaries without worrying about a responsibility for the past. Meanwhile, writing about history in the ideology-based class seems to be connected to experiencing group-based guilt. Perhaps the ideology-based identity content helps people cope with collective guilt (i.e., through highlighting democratic achievements). Or perhaps it makes people especially aware of Germany’s atrocities in the past because they contrast so strongly with its democratic values today.
Figure 2
a. National identification as a predictor of class membership (controlling for collective guilt mean-centered). b. Collective guilt as a predictor of class membership (controlling for national identification mean-centered).
In contrast, while members of the heritage-based class are highly likely to mention history, there is no relationship between membership in this class and feelings of group-based guilt. This suggests that references to history in the heritage-based class take a variety of forms, and only some are related to feeling guilty for the past. Finally, membership in the legal-formalistic class is not connected with collective guilt. This is unsurprising, since people only experience group-based emotion on behalf of groups that have at least some meaning for them (Branscombe et al., 1999).
Demographics That Are Predictive of Class Membership
Finally, we conducted a multinomial regression with participant demographics as predictors and class membership as outcomes to explore what distinct demographic groups can predict membership in different classes. We included political orientation, gender, age, education, urbanity of region, and parents’ migration background as demographic predictors in our analysis (Table 5). These are covariates that are frequently assessed in research on national identity and immigration attitudes (O’Rourke & Sinnott, 2006). Table 5 suggests that older people are more likely to be in the ideology-based than in the legal-formalistic or heritage-based class, politically leftist people in the legal-formalistic class, people with more education in the heritage-based class, and men in the heritage and traits-based class.
Table 5
Multinomial Regressions Predicting Membership in Each Identity Class Relative to Ideology-Based Class (Using Stata/ mlogit)
| Variable | Legal-formalistic
|
Traits-based
|
Heritage-based
|
|||
|---|---|---|---|---|---|---|
| Coef. | SE | Coef. | SE | Coef. | SE | |
| Politics: left-right | -.35** | .05 | .05 | .05 | .00 | .04 |
| Age | -.01* | .00 | -.00 | .00 | -.01* | .00 |
| Basic degree | .55 | .41 | -.51 | .48 | -1.05* | .42 |
| Middle degree | .09 | .28 | .04 | .29 | -.26 | .25 |
| Urban-rural | -.04 | .07 | .01 | .07 | -.06 | .06 |
| Migration background | .16 | .27 | -.28 | .32 | -.38 | .26 |
| Male | .78 | .21 | .99** | .23 | .90** | .19 |
| Constant | -.24 | .18 | -.58* | .21 | .45** | .17 |
Note. N = 987; (LR) χ2 = 176.37; p > .001.
*p < .05. **p < .01.
Discussion
Our latent class analysis unraveled four national identity classes: a heritage-based identity class with a strong focus on language and culture (39% of participants), an ideology-based identity class that revolves around democracy, welfare, freedom, and economy and safety (19% of participants), a legal formalistic identity class that is mostly concerned with the legal requirements for citizenship (26% of participants) and a traits-based identity class (16% of participants). The first two of the emerged identity classes resemble the broad dimensions that macro-level theories converge on: ethnic versus civic nations.
The heritage-based identity class characterizes the greatest number of responses. This finding is consistent with Germany's strong ethno-cultural tradition (Joppke & Rosenhek, 2002). Interestingly, while higher identification with the nation increases the probability of membership in the heritage-based class, lower levels of group-based guilt do not. A heritage-based national identity thus appears to be an expression of nationalistic sentiments that comes with an enhanced feeling of responsibility for the past – at least for some class members. Furthermore, poly-ethnicity is more frequent here than in general, suggesting that a strong sense of one's culture does not necessarily lead to the exclusion of other cultures. Thus, while feeling attached to an essentialist national identity is associated with exclusionary attitudes towards immigrants (Pehrson et al., 2009), feeling attached to a more nuanced, less essentialist heritage-based identity might not be exclusionary in the same way. Yet, the large size of the heritage-based class contrasts with findings of higher mean scores for civic than for ethnic subscales in countries with strong ethno-cultural traditions (Hjerm, 1998b; Reijerse et al., 2013).
The ideology-based identity is popular among older participants, although this is a rather small effect. Perhaps the emphasis on democratic institutions is reminiscent of post-WW2 education about national identity.
Importantly, the heritage- and ideology-based classes are more heterogeneous than the ethnic and civic identity typology. Participants mentioned several concepts that go beyond typical closed-ended questionnaires (e.g., welfare, economy and safety, history). Also, neither the heritage- nor ideology-based identity class are “pure”: poly-ethnicity, traditionally associated with a civic understanding of national identity, appears most frequently in the heritage-based class. While culture is extremely frequent in the heritage-based class (.85), it is also the fourth most frequent category in the ideology-based class (.43). Finally, language, the second most frequent category in the heritage-based class, can also be interpreted as civic if it is understood as an important precondition for democratic participation (Hjerm, 1998b). This mix of ethnic and civic elements evokes scholarship at the macro-level that proposes that jus sanguinis and jus soli elements are often mixed in different manners and proportions in states and political movements (Smith, 1991).
Two additional classes emerged: the legal-formalistic and the traits-based national identity. Citizens in the traits-based class use self-descriptive stereotypes, presumably to demarcate boundaries (Brewer, 1991; Mlicki & Ellemers, 1996). This interpretation is further supported by the finding that experiencing group-based guilt reduces Germans’ chance to fall into the traits-based class. An essentialist national identity expressed through self-ascribed traits seems to preclude the aversive emotion of group-based guilt.
Participants in the legal-formalistic identity class consider national identity a merely bureaucratic construct. This raises the question if they hold a “post-national mindset”. If the answer is yes, then almost one quarter of the participants has no personal relationship with Germany. Indeed, identification with Germany reduces people’s chance to fall into this class. Meanwhile – perhaps reflecting the cosmopolitan position of center-left parties in Germany – being politically progressive increases their chance.
Limitations
Our sample is more educated, left-leaning, and male than what would be representative of Germany as a whole. It does, however, include a sufficient number of members from each of these groups to predict class membership with demographics as we did above. This analysis suggests that our data might overestimate the size of the heritage-based, legal-formalistic, and traits-based classes relative to the ideology-based class. That is because, as we describe and interpret above, being educated and being male (both groups that are overrepresented in our sample) increases the odds of falling into the heritage-based class. Being male also increases the odds of falling into the traits-based class. Finally, being left/center-left (another overrepresented group) increases the odds of falling into the legal-formalistic class. Yet, given how small the ideology-based class is currently (19%), it is unlikely that it would be the most frequent class even with a representative sample.
As a robustness check, we replicated the commonly found two factor solution with a civic and an ethnic scale, and many of the common associations between different demographics and endorsement of civic versus ethnic national identity content in Appendix B. These replication results give us confidence than the pattern we detected in our LCA meaningfully deviates from patterns identified in past research using civic and ethnic identity content scales, rather than merely being a function of our sample.
Nevertheless, future research should replicate our study with a sample of less educated German citizens and a broader set of political and other minority groups. It should also apply the same approach to other countries. It will be interesting to see if and to what extent similar categories and a four-class solution emerge in countries that were traditionally classified as civic at the macro-level (e.g., USA) or that have an entirely different history and policy landscape (e.g., Singapore).
Implications and Conclusion
What can our results contribute to public debates about national identity? In Germany, as in many Western democracies, the public debate about national identity is closely linked to questions of immigrant access and integration. While past research often confounds the questions of who we are and who can join us, the current research deliberately focuses on the former question. In doing so, we find that ideology-based concepts play a subordinate role in how Germans see their collective self. This is in stark contrast to citizenship tests in Germany and elsewhere that require primarily civic knowledge from immigrants (Michalowski, 2011). Of course, ideology-based categories, such as democracy, characterize all liberal democracies, and as a result German citizens may not see them as distinctly defining their own nation. Yet, if asked what immigrants should know, they might emphasize such categories more than heritage-based ones. If true, this means that what we see subjectively as most defining to who we are is different from what we expect from newcomers. Future research could ask citizens what they think immigrants should learn about being German.
Another reason for the disconnect between Germans' subjective definition of being German, and citizenship tests and policy debate, might be that while requiring civic knowledge for membership in a democracy can be justified on normative grounds, requiring and testing knowledge about culture is much more problematic and difficult (Orgad, 2015). Our findings underscore some of these difficulties. Even though the heritage-based class was largest, less than half of the participants were classified there. What is more, the heritage-based class itself is rather heterogeneous, featuring many categories with different response probabilities. And about a third of the participants in our sample denies that there is anything distinctly German at all. If there is so little consensus among Germans about what it means to be German, it is difficult to teach newcomers about it.
Perhaps the lack of consensus is a result of the competing narratives at the macro-level in Germany that sometimes conflict with people's personal experiences and histories. Civic education that does not merely teach civic knowledge but deliberately links it to national identity might lead to a greater consensus among Germans, and between Germans and immigrants who take immigrant education classes and citizenship tests. Future research could investigate this hypothesis.
In conclusion, being German can have a heritage-based, ideology-based, legal-formalistic, or traits-based meaning. The heritage- and ideology-based identity classes are broader and more multifaceted than the civic and ethnic types from past research. Combining qualitative and quantitative methods, we refine and expand our understanding of how citizens construct their national identity, and thus make a social psychological contribution to some of the most heated debates of our time.