Theoretical work in psychology (e.g., Anant, 1966; Hagerty et al., 1992) portrays belonging as the satisfaction of an individual’s need to be personally involved with their environment and to feel part of a larger social entity—socially embedded. Thus, belonging is often seen as part of a healthy state of being and defined in positive terms as a central feature of a person’s well-being (Hagerty et al., 1992). When belonging is challenged or eroded, it has the potential to become a severe psychological stressor, threatening mental health (Nibbs, 2014). Beyond these basic premises, psychological research has operationalized belonging mainly from a personality and individual differences research perspective, framing the tendency to experience a sense of belonging (or lack thereof) as a trait or fundamental mode of relating (e.g., Hagerty & Patusky, 1995; Lee & Robbins, 1995; Malone et al., 2012). This emphasis on individuals’ psychological makeup determines these constructs’ applicability and, in particular, limits their pertinence to enquiries which are interested in the sense of belonging as context-dependent and time-variant.

Sociological studies of belonging have focused on processes of exclusion and inclusion, often in the form of legal membership, and on a general feeling of “being at home.” Both accounts tend to neglect the emotional dimensions of the sense of belonging. On the one hand, studies focusing on legal membership as markers of belonging (Krzyzanowski & Wodak, 2007; Sicakkan & Lithman, 2005) as well as on identity politics and place-related aspects of exclusion and inclusion (Castles & Davidson, 2000; Yuval-Davis, Anthias, & Kofman, 2005) hinge heavily on pre-defined categories of belonging to account for personally and culturally mediated emotional processes, or focus only on certain aspects of belonging such as ethnic identity (e.g., Maehler et al., 2019). References to “feelings of being at home,” on the other hand, have been criticized for romanticizing belonging (Antonsich, 2010; Yuval-Davis, 2006) and for mainly focusing on potential objects of belonging, instead of defining what these feelings of being “at home” actually entail.

A comprehensive concept of belonging that accounts for some of these limitations is provided by Pfaff-Czarnecka (2013), who defines belonging as “an emotionally charged, ever-dynamic social location” (p. 13). Based on this understanding, we conceive of a sense of belonging as an emotionally charged experience that is dynamic, constantly in flux, and frequently challenged rather than being stable and fixed. We also regard the sense of belonging as context-sensitive and primarily related to a social location, rather than to specific objects or “modes” of belonging. Building on Pfaff-Czarnecka’s (2013) definition and drawing on a broad theoretical literature, we emphasize four central elements of belonging (or lack thereof): (1) connectedness, (2) participation, (3) identification, and (4) congruence.

The connectedness element represents the feeling of having a place within a social system, a national space, or geographical space (Antonsich, 2010; Dixon & Durrheim, 2004; Pfaff-Czarnecka, 2013). Lacking this sense of place and being disconnected from social networks promotes feelings of loneliness and insecurity (Newcomb, 1990; Pieloch, McCullough, & Marks, 2016). Previous studies highlight the importance of connections to, for example, occupational or educational institutions (Gibson, Bejínez, Hidalgo, & Rolón, 2004; van Houtte & van Maele, 2012). Any socially defined structure can satisfy or frustrate what Kohut (1984) labeled “connectedness”—a central element of belonging.

The second defining element of SoB goes beyond having a place within an environment or system and focuses on a reciprocal feeling of acceptance and being valued within a system, arising from socially participating in this system. According to Baumeister and Leary (1995), humans need “frequent, non-aversive interactions” (p. 497) as well as people they trust (Pearce, 2008) to feel like they participate in the lives of others, to feel indispensable. A lack of participation can result in feelings of being left out and unrecognized, frustrating what Hagerty (1992, p. 173) labeled the “valued involvement” element of belonging.

Third, the identificational element of belonging essentially captures the feeling of personal “fit” or “non-fit” within one’s environment (Hagerty et al., 1992; Pfaff-Czarnecka, 2013). Identificational feelings arise when individuals’ roles within their environment resonate with their identity or the location they envisage for themselves within that environment (Antonsich, 2010; Probyn, 1996; Yuval-Davis, 2006). An individual’s identification with their place in their environment arises out of emotional acceptance of their involvement and their social integration (Orton, 2012). A strong “fit” or identification typically leads to feelings of safety and stability, which are essential for a sense of belonging; lack of identification, in turn, erodes this sense.

The final element of belonging, congruence, acknowledges the potential existence of multiple, overlapping sources of belonging that can come into conflict with each other. For Savage, Bagnall, and Longhurst (2004), congruence refers to the concurrence of an individual’s current place of residence and their life story. In our understanding, congruence indicates a non-conflicting, cohesive relationship between coexisting sources of belonging. This means that congruence is potentially challenged in situations of re-orientation toward new social and cultural environments or when multiple environments are in conflict. These situations can result in unsettling feelings of being torn between different realms of belonging or a far-reaching loss of belonging.

Item generation and selection

We searched the literature for scales measuring “sense of belonging,” “belonging,” and “belongingness” to construct a brief Challenged Sense of Belonging Scale (CSBS) based on items from one or multiple existing scales. Specifically, we looked for items that correspond to the four previously defined elements of belonging: participation, connectedness, identification, and congruence. Our motivation for selecting items from existing scales was to draw on established wordings to improve our chances of constructing a valid and reliable scale with apt, answerable items. We identified two scales from the psychological literature that include some items that are relevant to our approach: the Social Connectedness Scale (SCS) and the Sense of Belonging Instrument (SOBI). Importantly, as addressed above, these scales were designed to capture a sense of belonging as a psychological trait or mode of relating, while we conceive of a sense of belonging as a situationally contingent experience that is subject to changes and challenges across contexts and time. In other words, these scales and many of their items address related but different constructs (e.g., the SOBI-items: “I could disappear for days and it wouldn’t matter to my family”, “I would describe myself as a misfit in most social situations”, “I feel like a piece of a jig-saw puzzle that doesn’t fit into the puzzle”; e.g., the SCS-item: “Even among my friends, there is no sense of brother/sisterhood”). We therefore selected questions from these scales that work as situationally sensitive items and not primarily as dispositional measures, especially when combined with other items (see Table 1). Our brief scale has the advantage that it can be included in large, multi-purpose surveys such as the one used in this study.

The Social Connectedness Scale (SCS) was designed to capture the “connectedness” element of Kohut’s (1984) “belongingness need” (need for a sense of “being a part of” in avoidance of loneliness and alienation) (see also Lee & Robbins, 1995). Connectedness “allows people to maintain feelings of being ‘being human among humans’ and to identify with those who may be perceived as different from themselves” (Lee & Robbins, 1995, p. 233, paraphrasing Kohut, 1984, p. 200). Although the authors justify their scale by referring to an increased lack of societal belongingness, also among immigrants, the SCS was created in the context of personality and individual differences psychology. We nonetheless found items that reflect our participation and connectedness elements in the SCS. We changed the wording of the connectedness item from “disconnected from the world around me” to “disconnected from those around me” to reduce phrasing redundancy (our Identification and Congruence items aim at “worlds”), as well as to emphasize challenged social connectedness in this item.

The Sense of Belonging Instrument (SOBI) was created by Hagerty and Patusky (1995) using Hagerty and colleagues (1992) concept of belonging. One of the two central elements of belonging captured by this scale is “fit” or, as we call it here, identification. While the SOBI, like the SCS, is designed to capture a psychological trait, with other items capturing a sense of being a misfit or an outsider in social situations, one item expresses lacking identification in a way that is both psychologically and sociologically applicable.

Finally, our concept of belonging includes congruence. We did not find an item that reflects this element in the existing literature. Therefore, referencing transnational theory’s notion of being “torn between worlds” (Levitt & Glick-Schiller, 2004), we formulated the congruence item of the scale ourselves. This item was devised with the migration context in mind, but, as we argue, it is suitable for capturing a multitude of incongruences.

As noted above, the objective of our scale is to identify instances of challenged or lacking belonging with respect to a particular context. All of our scale items are negatively worded, ascertaining the extent to which the sense of belonging is challenged rather than the extent to which it is intact. We agree with the authors of the SCS that it is the “frustrations of belongingness” (Lee & Robbins, 1995, p. 235) that are of interest to research, especially research on marginalized or disadvantaged groups and groups vulnerable to mental health problems. Additionally, we worded all scale items consistently in the negative based on the consideration that reverse wording, rather than lowering response bias, could lead to erroneous response patterns owing to inattention and confusion (Van Sonderen, Sanderman, & Coyne, 2013).

Translation procedure

The original English language items were translated into German by some of the authors through a forward-translation-based process involving independent translations, comparisons, and discussion. All three authors involved are German native speakers and native or fluent in English and culturally competent regarding German and major English-speaking communities as well as competent regarding the content of the items and sociological testing practices (Muniz et al., 2013). The translation of the German or English questionnaire into Arabic and Farsi/Dari was provided by professional Arabic or Farsi/Dari mother tongue translators in the translation agency that works with Kantar, the company that conducts the IAB-BAMF-SOEP Survey of Refugees. The translation process within the translation agency used a TRAPD-like procedure (Harkness, 2003): First, two translators separately produced a translation. Next, a third translator discussed the translations with the others and resolved any discrepancies. Finally, trained interviewers working with Kantar, who were native speakers, tested the translated questionnaire in the field and reported back to the translation agency in case of problems (Britzke & Schupp, 2019, p. 49; Jacobsen, 2018).

Sample

We tested the scale using the IAB-BAMF-SOEP Survey of Refugees (Goebel et al., 2019; IAB-BAMF-SOEP Survey of Refugees 2018, 2020). The IAB-BAMF-SOEP Survey of Refugees is a random sample of asylum seekers and refugees who migrated to Germany between 2013 and 2016 (Kühne, Jacobsen, & Kroh, 2019). It is a joint project of the German Socio-economic Panel (SOEP), the Research Centre of the Federal Office for Migration and Refugees (BAMF-FZ), and the Institute for Employment Research (IAB). The sampling frame is the German Central Register of Foreign Nationals (AZR), which comprises all individuals living in Germany for longer than 3 months without holding a German passport (Babka von Gostomski & Pupeter, 2008). The CSBS was implemented in the third wave of the panel containing 4376 individuals living in 3061 households (for more information on sampling and response rates, see Kühne et al., 2019).

At the beginning of each CAPI interview, respondents chose their preferred interview language—English (N = 341, 8.7%), Arabic (N = 2957, 75%), Farsi/Dari (N = 485, 12.4%), Pashto (N = 30, 0.8%), Urdu (N = 54, 1.4%), or Kurmanji (N = 44, 1.1%). All questions were then presented in German and the chosen language side-by-side (Jacobsen, 2018).

For the present study, we analyzed data from the 3783 third wave respondents who received a questionnaire including the CSBS and who responded in English, Standard Arabic (hereafter just “Arabic”), or Farsi/Dari. These language groups had a sufficient number of participants for separate analyses.

Beyond language, our sample is heterogeneous in various. Most Arabic-speaking respondents are from Syria (73%), and most Farsi/Dari-speaking respondents are from Afghanistan (79%). English is linked to a more diverse range of countries of origin. Many English-speaking respondents are from Eritrea (32%), but the majority is distributed in small proportions across a vast array of countries. Across all three languages, most respondents have a medium educational background (Levels 2–4 of ISCED11, United Nations Educational, Scientific and Cultural Organization (UNESCO, 2012). The Arabic-speaking group had the highest proportion of respondents who completed tertiary education, with around 20% having attained tertiary qualifications. In terms of religion, our sample is also diverse. Most respondents report an Islamic denomination. However, those who answered the CSBS in English most often reported a Christian denomination. The sample is rather young, with an average age of around 33 years across all languages. Most respondents have only been in Germany for around 3 years.

Table 2

Key characteristics of the sample across languages

Note. ^aNA refers to missing values or insufficient information in order to generate a variable (e.g., ISCED11)

^bWe reduced the original ISCED 2011 classification (International Standard Classification of Education) into 3 categories (0, 1 = 1; 2, 3, 4 = 2; 6, 7, 8 = 3; 5 = not applicable), reflecting primary, secondary, and tertiary education (see UNESCO Institute for Statistics)

Item analyses

To investigate the validity of the scale in several languages and to account for potential measurement errors introduced by our translation procedures (Harkness et al., 2010), we carried out all analyses separately for the three language groups (English, Arabic, and Farsi/Dari). Throughout our analyses, we treated the CSBS Likert scale item responses as metric. All analyses were carried out using Stata 14.S.E. All four variables follow a similar distribution in all three languages (see Table 3 for means and SD, and Figures 1, 2, and 3). What stands out is that the highest value of the scale, indicating no challenge to one’s sense of belonging, is the modal value for all four items for all languages. The other four responses, indicating various degrees of challenged belonging, display an almost normal distribution. The proportions of missing items vary between 2.7 and 14.4% (see Table 6 in the “Appendix” section), depending on the survey language. Respondents who answered the scale in English skipped more questions than those who responded in Arabic or Farsi/Dari across all items. While 92.8% of the Arabic-speaking respondents and 89.1% of the Farsi-speaking respondents answered all four scale items, only 78.0% of the English-speaking respondents did. Among English-speaking respondents, average item nonresponse across all four items was 13.2%, compared to only 3.5% among Arabic-speaking and 5.8% among Farsi-speaking respondents. There does not seem to be a strong systematic pattern as to who answered all CSBS questions and who did not. There is some indication that respondents with higher levels of education were more likely to answer all scale questions and that those who did not report their legal status were less likely to answer all scale questions. However, the findings are not consistent across languages and the confidence intervals are large. Thus, for further analysis, we treated the missing values as “missing completely at random” (Little & Rubin, 1987), refraining from imputation or weighting adjustments.

To test whether the CSBS showed the expected single-factor structure, we employed a confirmatory factor analysis for all language groups separately. This analysis was based on a Pearson correlation matrix and used maximum likelihood estimation methods. We defined sufficient fit as follows: a CFI of at least 0.90 (Bentler, 1990), RMSEA below 0.08 (Kline, 2010), and SRMR below 0.08 (Hu & Bentler, 1999).

To assess construct validity, we first employed a confirmatory factor analysis for all language groups separately. As indicated by the factor loadings (see Table 3), all items map onto the same latent construct. Additionally, fit-indices (CFI, RMSEA, SRMR) suggest that the proposed structure fits the data well. 

Table 3

Factor loading (standardized) for the three language groups

Item parameters

Table 4 shows the mean and standard deviation of each of the four items for CSBS separately for the three language groups.

Table 4

Means and Standard Deviations

Note. Scale ranging from 1, “strongly agree”, to 5, “strongly disagree”, English: N = 292; Arabic: N = 2838, Farsi/Dari: N = 454

Objectivity

By their very nature, the constructs of sense of belonging and challenged sense of belonging capture a very subjective experience. It is therefore desired that respondents answering the scale tap into their own subjective interpretation. However, the CSBS represents a standardized instrument that comprises individual items that capture clearly defined and demarcated aspects of sense of belonging. The scale thus enables an objective and comparable assessment of challenged sense of belonging as defined here and is comparable across different national and language contexts.

Reliability

Evidence for internal reliability

Reliability was measured using pairwise Pearson correlations, Cronbach’s alpha, and McDonald’s omega. As shown in Table 5 Cronbach’s alphas and McDonald’s omegas for the different sub-samples are all adequate to good. Additionally, the item-test correlations (ITC) and the item-rest correlations (IRC) are all of a sufficient magnitude, indicating that the items are highly correlated with the overall score.

Table 5

Cronbach’s alpha, item-test correlations, and item-rest correlations for each factor over languages

Note. CSBS: English: N = 266; Arabic: N = 2745, Farsi/Dari: N = 432

Pairwise correlations shown in Table 6 section reveal that some items are only moderately correlated with each other. Nevertheless, all effects are strongly significant, and no correlation is smaller than 0.42.

Table 6

Pairwise correlations between items

Validity

Evidence for construct validity

Due to the good fit of the single-factor model, construct validity can be assumed (see Item analyses).

Evidence for convergent validity

We calculated partial correlations controlling for duration of stay in Germany separately for all three language groups in order to assess the association between the CSBS score and measures linked to belonging as evidence for convergent validity (see Table 7). Across all languages, CSBS scores are positively associated with greater mental health, that is: less challenged sense of belonging is associated with greater mental health. Out of all associations tested, this is the strongest one, with lower and upper confidence interval bounds for Pearson’s r values of a minimum of 0.32 and a maximum of 0.55 across languages. The well-being measure, life satisfaction, is positively, albeit less strongly, associated with the CSBS total score. The Pearson’s r confidence intervals show that there is an association across languages: the weakest among Farsi-speaking respondents (0.11–0.29) and the highest among Arabic-speaking respondents (0.25–0.32). Regarding the social embeddedness variables, we find that the number of people with whom respondents can share private thoughts and feelings is modestly positively associated with CSBS among Arabic-speaking respondents (0.03–0.11) and modestly or not at all associated in Farsi-speaking respondents (− 0.05 to 0.14). Frequency of contact to Germans in general and to Germans in the respondents’ own friend groups, are both positively associated with the CSBS score with confidence interval bounds for Pearson’s r between 0.01 and 0.34 across languages. A higher frequency of contact to people from the country of origin who are not relatives is modestly or not at all associated with higher CSBS scores in Farsi-speaking respondents (− 0.08 to 0.10) and modestly negatively associated with CSBS scores among English-speaking respondents (− 0.20 to 0.03), indicating that a higher frequency of contact could actually be linked to a more challenged sense of belonging.

Table 7

Partial correlation coefficients

Note. All correlation coefficients display partial correlation coefficients; we control for the time since arrival in Germany (in years)

Our exploration of the associations between belonging and concepts we predicted to be linked to belonging revealed significant associations between CSBS scores and several indicators of well-being and social embeddedness. The negative relationship between a challenged sense of belonging and overall mental health and life satisfaction across language groups is in line with the emphasis on the importance of a sense of belonging for well-being throughout the belonging literature (e.g., Hagerty et al., 1992; Nibbs, 2014). Moreover, the negative associations between a challenged sense of belonging and indicators of general social embeddedness and social embeddedness in Germany in particular (having people with whom one shares feelings, as well as contact with Germans and having Germans in one’s group of friends) are as expected. While some of these associations were of a small magnitude for some language groups, the results align with the emphasis on social embeddedness and connectedness throughout the belonging literature (see Baumeister & Leary, 1995). Interestingly, a challenged sense of belonging is not related to the frequency of contact with individuals from the country of origin, suggesting that contact with people from the host society may be more important for fending off challenges to one’s sense of belonging. The findings are robust with respect to the time since respondents arrived in Germany. Overall, these findings provide evidence for the validity of the CSBS based on associations with external variables (AERA, APA, & NCME, 2014). It should be noted that differences regarding significant effects in these associations between language groups could be due to sample size differences.

Descriptive statistics (scaling)

The manifest scale scores of the English-language CSBS shows a mean of M = 3.6 and a standard deviation of SD = 1.2 (N = 266) with a skewness of -0.42 and a kurtosis of 2.15. The manifest scale scores of the Arabic-language CSBS shows a mean of M = 3.5 and a standard deviation of SD = 1.2 (N = 2745) with a skewness of -0.24 and a kurtosis of 2.05. The manifest scale scores of the Farsi/Dari-language CSBS shows a mean of M = 3.2 and a standard deviation of SD = 1.2 (N = 432) with a skewness of 0.05 and a kurtosis of 1.95.

Cross-language comparability

Measurement invariance testing across language groups

Based on this confirmatory factor analysis, we tested for measurement invariance to assess whether the model estimates the same latent construct in all language groups. If this is the case, the means of latent indicators can be reliably compared across groups. We employed a bottom-up estimation approach for measurement invariance testing (Saris, Pirralha, & Zavala-Rojas, 2018). This means that we started with the least restrictive model (configural invariance, loadings, and intercepts are not set to be equal across groups), applying the following rules to determine further steps: In the case that this model shows an adequate fit, loadings are restricted to be equal across groups in order to test metric invariance. If this model fits the data well and the deterioration of the CFI does not exceed 0.01, intercepts are set to be equal as well in order to determine full scalar invariance (see Chen, 2007). If this model shows appropriate fit as well and the deterioration of the CFI is 0.01 or smaller, full measurement invariance is given.

The test for measurement invariance indicates full scalar invariance as all steps of the bottom-up procedure show adequate fit (configural: CFI = 0.98, SRMR = 0.03; metric: CFI = 0.98, SRMR = 0.06; scalar: CFI = 0.97, SRMR = 0.06). With full scalar invariance supported, we assume that the latent construct is the same across language groups.

Further literature

The scale was first published in the journal Measurement Instruments for the Social Sciences.

Acknowledgement

Open Access funding enabled and organized by Projekt DEAL.

• Lukas Marian Fuchs, Institute for Sociology - Freie Universität Berlin, Garystraße 55, 14195 Berlin, Germany; E-Mail: lukas.marian.fuchs@gmail.com.
• Lena Walther; Department of Psychiatry - Charité University Medicine Berlin, Hindenburgdamm 30, 12203 Berlin, Germany; E-Mail:lenawalther526@gmail.com.
• Jannes Jacobsen, Institute for Sociology - Freie Universität Berlin, Garystraße 55, 14195 Berlin, Germany, E-Mail: j.jacobsen@fu-berlin.de

Data stems from IAB-BAMF-SOEP Survey of Refugees, a joint project of the German Socio-economic Panel (SOEP), the Re-search Centre of the Federal Office for Migration and Refugees (BAMF-FZ), and the Institute for Employment Research (IAB). All data is available free of charge through the Research Data Centers (Forschungsdatenzentren) of the involved institutions.