Patterns of remittances of intra-European migrants: social relations and moral obligations

ABSTRACT The importance of remittances for economic development and the maintenance of transnational social relationships have been widely discussed. Based on data from Switzerland, we analyze the roles of transnational social relations and moral obligations for the likelihood of sending remittances among intra-European migrants from Germany, Great Britain, Italy, Portugal, Serbia, and Bosnia-Herzegovina. Our data shows high levels of remitting among these groups, with migrants from South-East European countries sending remittances primarily to family and friends and migrants from Portugal and Great Britain sending remittances primarily to their own bank account. Furthermore, by using differentiated and direct measures for social relations and moral obligations, we show that strong social ties as well as moral family obligations are relevant predictors of sending remittances, beyond measures of various desires and capacities to remit usually discussed in the literature. However, these effects also vary according to social relation and remittance type. Together, the results make a strong case for the social embeddedness of remittances and the importance of including migrants from western and southern Europe in empirical research.


Introduction
Research on remittances has multiplied in the last thirty years, especially with a focus on the nexus between migration and development of the sending countries (Docquier and Rapoport 2012;Rapoport and Docquier 2006).Remittances are usually defined as monetary transfers by people with a migrant background from their country of residence (CoR) to their country of origin (CoO), often to support family and relatives (Brown et al. 2014;Carling 2008).Remittances are a type of economic behavior that is constitutively embedded within social relations, cultural scripts, non-economic motivations, and normative obligations (Carling 2008;2014;2020;Rapoport and Docquier 2006;Yang 2011).This makes remittances into a fascinating topic not only for migration research but for economic sociology, which is guided by the idea that economic transactions usually do not take place on atomistic markets, but are embedded into normative, social, and cultural relations and frameworks (Beckert 1996;Granovetter 1985).Taking into account the large and still increasing amount of international remittances in the last thirty years, research from a sociological perspective on remittances is both timely and important.
The overall aim of our study is to analyze the remittance behavior of people with a migrant background in Switzerland, focusing on the roles of social relations to family, relatives, and friends on the one hand and normative moral obligations towards the family on the other.Our empirical analysis therefore addresses two important research gaps: First, we focus on intra-European migrants in Switzerland.In general, patterns of individual remittance behavior of intra-European migrants are not well studied, apart from a small number of publications that focus on just a few countries and the older generation of labor migrants to western Europe (Bauer and Sinning 2009;Cela, Fokkema, and Abrosetti 2013;Constant and Massey 2002;Fokkema, Cela, and Ambrosetti 2013;Holst, Schäfer, and Schrooten 2008;Sinning 2011;Wolff 2019).In our study, we are able to analyze the remittance behavior of migrants from a broader set of CoOs, including also western European countries (UK, Germany), southern European countries (Italy, Portugal), and countries on the southeastern periphery of Europe (Bosnia, Serbia).This enables us to go beyond existing case studies of remittance behavior.
Second, our study includes more appropriate measures for key-concepts of the social embeddedness of remittance behavior.It is the first study to address the role of normative family obligations by directly measuring them.Previous research has usually relied on indirect indicators to measure the role of altruistic and normative motivations for remittances (e.g.Carling 2008;Docquier and Rapoport 2012;Rapoport and Docquier 2006).Additionally, our study includes measures on social relations to others in the CoO, taking into account the intensity of the social relation based on the frequency of contacts.Together, these measures allow for a more rigorous and nuanced test for the explanatory contributions of social and moral determinants for remitting.We expect both variables to have a positive effect on the level of remittances.
The paper is structured as follows.In section 2 we present the state of research on the determinants of remittances and briefly explain the context of intra-European migration.Section 3 introduces our survey data of people with migration background from six different European countries living in Switzerland.In section 4 we present the results of the statistical analyses with respect to our hypotheses and the control variables.In section 5 we summarize the results on the role of moral family obligations and social relations as additional determinants of remittances, elaborate their importance for the state of research, and discuss their limitations.

Remittances of intra-European migrants
There is a lack of research on current intra-European migration patterns, and this is even more true of intra-European remittances and intra-European migration that goes beyond the EU countries (Becker and Teney 2020;Schroedter and Rössel 2014).Intra-EU migration is strongly encouraged by the EU and eased by European citizenship and agreements of free movement, which are also in place between the EU and Switzerland (Bartolini, Gropas, and Triandafyllidou 2016;Schroedter and Rössel 2014;Verwiebe, Wiesböck, and Teitzer 2014).This migration is currently characterized by a relatively high skill level of migrants, although this differs to a certain degree according to the CoOs.Whereas migrants from the older EU member states are often highly skilled, this is true to a lesser degree for migrants from southeastern and eastern Europe, which make up a huge percentage of current migration flows within Europe (Becker and Teney 2020;Verwiebe, Wiesböck, and Teitzer 2014).Especially the southeastern European states lost a substantial part of their population due to migration, which was often destined to southern European countries like Italy and Spain (Cela, Fokkema, and Abrosetti 2013;King, Frykman, and Vullnetari 2013).The described pattern of contemporary migration is also true for Switzerland, as foreign citizens make up more than a quarter of its population, many of them citizens of other European countries.However, what sets migration to Switzerland apart is that it includes a higher percentage of skilled migrants from Western European countries than the overall pattern of intra-European migration (Schroedter and Rössel 2014).
Contemporary migration within Europe is driven by quite different motivations: Beyond employment and wages, career opportunities and quality of life are important motivations to move.Apart from the overall rather high level of skill, intra-European migration is usually also characterized by a comparatively high level of transnationalism, i.e. many people with a migrant background uphold their relationships to their CoO and their family and friends living there (Verwiebe, Wiesböck, and Teitzer 2014).This is also visible in the particularly high level of remittances within Europe (Carling 2008).Germany, the United Kingdom, Portugal, Bosnia-Herzegovina, and Serbia are all countries with a high aggregate level of inflowing remittances, which are also all included in our sample of countries of origin (Yang 2011).Italy, Germany, Serbia, Portugal, and Bosnia-Herzegovina are also among the most important countries that receive remittances from Switzerland (Arapovic and Brown 2009).Despite the fact that remittance outflows from Switzerland are on a remarkably high level in international comparison, there has so far only been isolated research conducted on certain cases, such as Serbian remittances from Switzerland (Lerch, Dahinden, and Wanner 2007).Therefore, research on the determinants of remittances of people with a migration background from a broad set of countries of origin in Switzerland covers an important and under-researched area.

Remittances and transnational living
In the research literature, two types of remittances are usually distinguished: compensation of employees on the one hand and workers' remittances on the other hand (IMF 2009;Taylor 1999;Yang 2011).Compensation of employees refers to the wages of border-crossing commuters.Workers' remittances refer to those monetary transfers that are undertaken by the resident migrant population of a country to their respective countries of origin (Brown et al. 2014;Carling 2008).Most empirical studies of individual remittance behavior refer to this latter type of remittances, whereas aggregate research on remittances based on balance of payment accounts often include both types (e.g.Arapovic and Brown 2009).In some contexts, the transfer of social and political ideas, behaviors, and capital has also been categorized as social remittances (Levitt 1998).In our paper, we focus exclusively on economic remittances.
Remittances are thus commonly defined, understood, and explained in the context of international migration, which embeds them within the broader concept of transnational living (Carling and Hoelscher 2013;Erdal and Carling 2020;Kushnirovich 2021).Remittances are part of the larger construct of keeping in contact and reproducing long-established relationships with family, relatives, and friends within the CoO (Zelizer 2006), thus establishing a transnational sphere of living for many contemporary migrants that links at least their CoO and their CoR.Apart from remittances, these transnational activities and relations include visits to the CoO, communication with family and friends and different forms of political and economic activity (Cela, Fokkema, and Abrosetti 2013;Erdal and Carling 2020).These forms of transnational relations and activities are not opposed to integration into the CoR, since in many cases the capacity for stable transnational relations are socioeconomic resources based on a longer stay in the CoR and a resulting higher degree of socioeconomic integration (Carling and Hoelscher 2013;Kushnirovich 2021;Portes, Guarnizo, and Landolt 1999;Roman 2013;Sana and Massey 2005).
The different forms of transnational relations and exchanges imply that remittances must be understood as being constitutively embedded within social relations, cultural scripts, non-economic motivations, and normative obligations (Carling 2008;2014;2020;Rapoport and Docquier 2006;Yang 2011).Determinants thus vary according to national, ethnic, and social contexts, like social class, which makes the explanation of remittances a very complex endeavor, not allowing for one general theoretical model (Carling 2008;2014;2020;Kushnirovich 2021;Sana and Massey 2005).In this complexity, we focus on the role of social relations to family and relatives on the one hand and moral family obligations on the other hand.However, these variables are embedded within a broader set of determinants.In structuring our presentation of these other determinants, we follow Carling and Hoelscher's (2013) suggestion to distinguish between the capacity and the desire to remit.Whereas the first term refers to migrants' social position and their ensuing ability to transfer money to their CoO, the second term focuses on the willingness and preferences to transfer money, given a certain socioeconomic ability to do so.

Capacities to remit
The clearest case of a variable measuring the capacity to remit is income.There is overwhelming support in the literature that income is a consistent predictor of the frequency and size of remittances (Antoniades et al. 2013;Carling 2008;Carling and Hoelscher 2013;Constant and Massey 2002;Holst, Schäfer, and Schrooten 2008;2011;König, Isengard, and Szydlik 2018;Roman 2013).Given the fact that world-wide and intra-European migration are increasingly characterized by migrants with high skill levels, there is a large body of literature on the relationship between skills and remittances.However, the empirical results of statistical analyses of the relationship between skill level or education and remittances are highly inconclusive (Bollard et al. 2011;Carling 2008;Carling and Hoelscher 2013;Coray 2014;Docquier and Rapoport 2012;Gentile 2019;König, Isengard, and Szydlik 2018).In general, the strong focus of empirical research on the role of income and skills highlights, that remittances are related to social class since both are indicators of the distribution of economic and cultural capital in social space (Bourdieu 1984).
The capacity to remit may additionally be based on the gender of migrants.Men often have higher earnings and have a stronger normative obligation to remit (Carling 2008).However, the empirical results with respect to gender are quite mixed, depending on the context (Carling 2008;Constant and Massey 2002;Coray 2014;Holst, Schäfer, and Schrooten 2008;2011).Therefore, in summary, income is a robust indicator of the capacity to remit and consistently statistically related to individual remitting behavior.In contrast, the effect of skill and gender varies across ethnic groups, countries of origin, and countries of residence.

The desire to remit
Turning to the desire of migrants to remit, we find an extensive discussion in the literature contrasting more altruistic motives, as in the case of unselfishly supporting family and friends in the CoO, and more self-interested motives, e.g. when transferring money to the CoO to build a house for oneself (Antoniades et al. 2013;Carling 2008;2014;Rapoport and Docquier 2006).One thing becomes clear from this discussion: the actual motives for remittances differ according to country and ethnic context, but also according to individual situations and contexts (Carling 2014;Constant and Massey 2002).A classic example is Sana and Massey's study on remittances of Mexican and Dominican migrants, were the first seem to be part of a household strategy of risk diversification and the second more a reaction to household needs (Sana and Massey 2005).Typical and consistent covariates of remittances, which indicate a more self-interested motivation to remit, are return intentions and housing in the CoO (Carling 2008;Constant and Massey 2002;Rapoport and Docquier 2006;Sinning 2011;Wolff 2019).Here, it is assumed that remittances may be targeted to prepare a return migration and therefore housing in the CoO for the migrants themselves, thus being a clear proxy for self-interested motivations.In this case remittances are not transferred to family, relatives, or friends but to the migrant's own bank account or to investments in the CoO.However, in many cases remittances are intended to take care of those left behind in the home country (Carling 2008;Carling 2014;Carling and Hoelscher 2013;Constant and Massey 2002;König, Isengard, and Szydlik 2018;Rapoport and Docquier 2006;Roman 2013;Wolff 2019).In some cases, remittances may even be a part of a kind of contractual family arrangement to send migrants to foreign countries to support the families, as suggested by the new economics of labor migration (Carling 2008;2014;Rapoport and Docquier 2006;Sana and Massey 2005).Therefore, having family, relatives, and friends in the CoO, the frequency of visits to the CoO, and other communications to people in the CoO are strong and consistent predictors of remittance behavior, which clearly shows the embeddedness of remittances within transnational social relations (Carling 2008;Carling and Hoelscher 2013;Constant and Massey 2002;Roman 2013).Other variables related to the transnational living of immigrants and their integration into the host society, as socio-cultural identification with the society of the CoR and identification with the CoO, perception of discrimination and proficiency in the language of the CoR do not show such clear results (Carling and Hoelscher 2013;Roman 2013).
Based on this discussion, we posit the following hypotheses.For each hypothesis, we test two different subhypotheses to take the difference between remittances to other people in the CoO and for the migrant's own needs into account: H1: Stronger social relations to the CoO correlate positively with sending remittances to family members and friends (H1a) and to a bank account (H1b).
Probably one of the clearest instances of an altruistic motivation for sending remittances is the feeling of a normative obligation to support others.Studies have pointed out how such remittance norms constitute a quasi-contractual agreement among family members (Sana and Massey 2005).Sending remittances becomes part of how a morally good family member is defined, thereby turning financial payments into symbols of commitment, care, and compassion (Hannaford and Foley 2015;Zelizer 2006).They are internalized by family members, describe a sense of duty and are upheld by sanctions (Hannaford and Foley 2015;Zelizer 2006).Moral obligations therefore provide a strong motivation to support others abroad by sending remittances.Yet, they might also be a driver for sending money to a personal bank account, since they also increase the importance of various other aspects of social life in the CoO, e.g.meeting family and friends in person and caring for them.Having funds available in the CoO facilitates these other modes of family support.
So far, it is not clear to what extent such moral norms shape remittance behavior, however.The role of altruistic and egoistic motivations in research on remittances was usually studied with indirect indicators, e.g.having real estate in the home country as an indicator of more self-interested motivations and having family and friends in the CoO being indicators of more altruistic motivations (e.g.Carling 2008;Docquier and Rapoport 2012;Rapoport and Docquier 2006).Yet, these indicators are embedded within quite different normative settings, making it difficult to draw inferences from statistical results of the role of these indirect indicators to the actual relevance of altruistic and normative considerations (Carling 2008;Simoni and Voirol 2020).Therefore, in this paper, we measure normative obligations to family directly and statistically analyze its covariation with the frequency of remittances.Based on this discussion we derive the following hypotheses: H2: Stronger normative family obligations correlate positively with sending remittances to family members and friends (H2a) and to a bank account (H2b).

Methods and data
The analyses are based on our own survey called TRANSSWISS.It was conducted in 2017 and relies on a stratified random sample of the total population of Switzerland.We sampled for firstand second-generation persons with migrant background from Italy, Germany, the UK, Portugal, Serbia, and Bosnia-Herzegovina, thus covering migrants from a diverse set of western, southern and southeastern European countries.The second generation describes people who were born in Switzerland with foreign citizenship or former foreign citizenship (i.e. who have been naturalized).To have sufficient data for different generations of migrants, we additionally sampled for second-generation Italians and Germans who had been naturalized.The samples were drawn from two different data sources.The first one is provided by the federal statistical office and contains data from the population registers of the Swiss communes and cantons.Additionally, we used the Migration Information System (ZEMIS), which collects data on Switzerland's foreign population.Among other things, it records whether a person has been naturalized.
The survey was aimed at all people who have a valid residence status in Switzerland, belong to one of the stated groups, had lived in Switzerland for at least 24 months, and were aged between 18 and 62 (a description of the sample is in the Appendix).This age cut-off was chosen because the study was aimed at the economically active population before retirement and a possible remigration.The data was collected online and by mail.We achieved an overall response rate of 35%.The questionnaire was available in seven different languages and covered topics such as social relations, economic interactions, political participation, identification, and integration.
Dependent variables: We use two dependent variables for the analysis of sending remittances, differentiating two types of recipients: friends and family (F&F) and the respondent's own bank account (Table 1).To measure these two types of remitting, respondents indicated on a four-point scale ranging from 'not at all' to 'about once a month' how often they sent money abroad to various recipients in the past 12 months, excluding donations or the purchase of commodities and services.In case of family and friends, we used the highest frequency among several categories of recipients, namely spouse, parents, siblings, children, other relatives, and friends.The respondent's bank account was a category on its own.For the analysis, we dichotomized these variables (0 = no remittances, 1 = at least once a year).
Independent variables: A first set of independent variables relates to the respondent's migration background.We categorized respondents according to their CoO, namely Germany (GER), Italy (ITA), Great Britain (GB), Portugal (PRT), Serbia (SRB), and Bosnia-Herzegovina (BIH).This covers firstand second-generation migrants.In case of first-generation migrants, it is simply their birth country, and in the case of secondgeneration migrants with one parent born abroad, it is the birth country of said parent.If both parents were born abroad, we categorized according to the mother's birth country.Additionally, we computed a variable separating firstfrom second-generation migrants.Second, we operationalized a series of variables for the capacity to remit (Carling 2008;Carling and Hoelscher 2013).These encompass measures related to capital endowment and social class (Bourdieu 1984).We included net equivalent household income (as defined by the OECD 2015) and occupational status, separating workers, lower-level employees (e.g.technician or secretary), higher-level employees (e.g.teacher or manager), self-employed, and others (e.g.students, etc.).We additionally measured the highest level of education (compulsory, higher education, university degree).Finally, we include gender (0 = women; 1 = men) as an additional variable related to the capacity to remit.
To capture the desire to remit, we included a measure for ownership of real estate in the CoO (Carling 2008;Rapoport and Docquier 2006).We differentiated between real estate owned by the respondents themselves and real estate owned by the spouse or other family members.Furthermore, we operationalized identification with the CoO and identification with the CoR, to measure socio-cultural integration and transnational identification.The measurement for Switzerland was based on four items, each measured on five-point scales, indicating the extent to which the respondents identified with Switzerland, the municipality, the canton, and the Swiss national team at sport events.To assess the reliability of the index, we used Cronbach's Alphaa measure of internal consistency based on the intercorrelations between the variables.With a Cronbach's alpha of 0.79 this can be considered as good.The measurement for the CoO was based on three similar items.Cronbach's alpha was also good (0.72).In both cases, we computed the mean, with higher values reflecting a stronger identification with Switzerland or the CoO, respectively.Furthermore, we took proficiency in the language of the CoR, i.e.Switzerland into account.The latter represents an index based on the mean of four variables, each measured on a five-point scale, with higher values reflecting higher levels of language proficiency (Cronbach's alpha: 0.96).Additionally, we include a measure for perceived discrimination.In five questions, we asked respondents how often they have had the impression of being disadvantaged due to their origin in various contexts, such as on the labor market or encounters with the police.Each question had a fivepoint scale.We computed the mean of these variables, with higher values indicating a higher frequency of perceived discrimination (Cronbach's alpha: 0.82).
Coming finally to the key variables of our hypotheses, we distinguish three types of social relations: (i) contacts to family members (spouse, siblings, parents, or children), (ii) other relatives, and (iii) friends.We chose these three types of social relations on theoretical and substantial grounds (Zelizer 2006) as they relate to the core family, the extended family, and contacts outside the family, respectively.We asked respondents how often they were in contact with others in their CoO.Values ranged from 'never' to 'daily' on a six-point scale.In each case, we took the highest frequency among the contacts within a relationship type.For example, if a respondent was in contact with siblings in the birth country once a year, but daily with parents, the respondent received the value 'daily' for the family relation.Hence, the variables also take the intensity of the social relation based on the frequency of contacts into account, thus giving an indication of the strength of the ties, instead of a simple binary measure (Granovetter 1985).We supplemented these measurements with a more subjective component of social embeddedness.We asked what country respondents perceived as their center of life.This variable took a value of zero if the respondents considered Switzerland as their center of life and a value of one if they considered another country, including the CoO, as their life's center.
Finally, we developed a measure for the moral obligation to support family members.In contrast to previous studies (e.g.Carling 2008;Docquier and Rapoport 2012;Rapoport and Docquier 2006), we included direct measures of the relevant normative beliefs.It consisted of three statements referring to a sense of duty to support parents or children, even in the face of potentially high costs.A sample item reads: 'When a parent is seriously ill or frail, it is primarily the duty of adult children to care for him or her.' Again, we computed the mean of the items, each measured on a five-point scale, with higher values representing a stronger moral obligation.Cronbach's alpha was acceptable with 0.64, being below the commonly used threshold for satisfactory internal reliability of 0.7 but far above the threshold for unacceptable measurements of 0.5.However, Cronbach's alpha also needs to be interpreted in context.It is generally lower for indices with a small number of items, i.e. three items in this case (Graham 2006).
Control variables: All models included age and employment status (unemployed, parttime, full-time) as control variables.
Additional variables: Additionally, we analyze questions on the reasons for sending remittances abroad in the last twelve months.Respondents were able to indicate for a selection of reasons, such as health expenditures or children's education (see Table 3), whether they apply fully, partially, or not at all.Respondents were able to agree with several reasons.To simplify, we dichotomized these variables, indicating whether a reason applies at all or not.
For the multivariate analysis of sending remittances to family and friends (F&F) or to a bank account, we use logistic regressions, since these two dependent variables were measured on a binary level (Long 2011).Unfortunately, when using logistic regression, the size of coefficients from different models cannot be easily compared (see Mood 2010).Therefore, we also report average marginal effects (AME) at the predictor's means to compare effect sizes across models.AME depict the change in the probability of the occurrence of the outcome (in our case sending remittances) when a dependent variable increases by one unit, holding constant all other dependent variables at their means.
Due to the extensiveness of the questionnaire, we employed a planned missing data design (Enders 2010).Most of the respondents received a random selection of thematic question blocks to improve the efficiency of the data collection and reduce respondent fatigue.This results in missing values which were completely at random, meaning they were unrelated to the study variables and hence estimable by multiple imputation (MI) without bias.MI produces multiple copies of a dataset, each containing estimates for the missing values.These values vary between datasets due to a stochastic element in the estimation process, depicting the uncertainty of the estimation.The analysis model is then computed for each dataset separately and results are combined (Enders 2010).For the subsequent analysis, we only included respondents which were administered questions on remittances, resulting in a net sample of 3416 cases.We used 40 imputations for the estimation process, corresponding to the highest proportion of missing values.To check the validity of the procedure, we inspected the distributions of the variables in each imputed dataset.No anomalies were detected.All computations were conducted in Stata 17.

Empirical results
Table 2 presents the frequency of remitting at least once a year by CoO and type of remittances (friends and family, respondent's bank account, and a combination of the two).Overall, about one-fifth of respondents indicated that they had sent remittances in the past twelve months.The prevalence of sending remittances was slightly higher for friends and family compared to one's own bank account.Sending remittances was thus a rather widespread practice in the population of intra-European migrants studied, confirming its possible economic importance (IMF 2009) and significance for transnational societies (Verwiebe, Wiesböck, and Teitzer 2014).
However, we also find strong country differences.First, remitting was especially frequent among respondents from Portugal, with 46% of the respondents doing so, followed by Serbia, Bosnia-Herzegovina, and Great Britain.Sending money abroad was much less common for migrants with German and Italian backgrounds.A Chi 2 test, which allows to statistically assess whether there is an association between sending remittances and countries, yields a highly significant result (p < 0.001).
Second, when looking at the recipient of the remittances, we find additional country differences.Respondents from southeastern European countries (Serbia and Bosnia-Herzegovina) led when it came to friends and family as recipients of remittances.They very rarely remitted to their own bank accounts.Indeed, no respondent did so in the case of Bosnia-Herzegovina.Italian immigrants exhibit a similar pattern, but on a much lower overall level.In contrast, Portuguese immigrants lead in terms of remitting to one's own bank account among the six countries, whereas sending remittances to family and friends is less common among them.Immigrants from Great Britain are second in terms of remitting to a personal bank account.However, the composition of the two types of remittances is much more balanced there.Similarly, German immigrants show a rather balanced pattern in types of remittances.Again, for both types of sending remittances, the Chi 2 tests yield highly significant differences between countries.
Overall, these findings suggest that especially migrants from southeastern European countries remit more often to support others.This may be related to the peripheral economic status of southeastern European countries within Europe and the ensuing necessity of migration and remittances to support the living standards of family and relatives left in the home country (King, Frykman, and Vullnetari 2013).Furthermore, Portugal and Great Britain exhibit the highest proportions of migrants sending remittances to their own bank account.
Such an interpretation receives further backing when we look at the specific purposes for sending remittances, summarized in Table 3.First of all, across all countries, the main reason for sending remittances was personal expenses, followed by various types of social support, such as helping those in distress, for example.Business investments and children's education were less important.Second, we find substantial and mostly significant country differences (based on Chi 2 -tests)with a similar countryspecific pattern as in Table 2. Social support, namely helping those in distress, supporting living standards, and health expenditures, were much more common reasons to remit for respondents in the Serbian and Bosnian groups.In contrast, personal expenses and business investments were most prevalent in Great Britain and Portugal, and to a lesser extent in Germany and Italy.Third, despite these country differences, we find that reasons to remit were quite heterogenous within each country as well, with every goal of remitting receiving at least some consent in most countries.In sum, these results show that the functions of remitting vary substantially between countries and between individuals within countries (Carling 2008;Carling and Hoelscher 2013;Rapoport and Docquier 2006).
To explain which capacities and desires explain different levels of remitting, we turn to the results of the logistic regression models with remitting to family and friends as the dependent variable (Table 4).Model 1 only contains the country dummies and the variable for being born in Switzerland next to the controls.Looking at the countries of origin,  we can confirm the results from the previous bivariate analysis.With the exception of Italy, all CoOs differed from Germany in their level of remitting to family and friends on a statistically significant level.Looking at the negative and statistically significant coefficient for being born in Switzerland, we see that second-generation migrants had a lower likelihood of remitting to family and friends, as could be expected from the literature (Carling and Hoelscher 2013;Kushnirovich 2021;Portes, Guarnizo, and Landolt 1999;Roman 2013;Sana and Massey 2005).
Model 2 introduces variables from established explanations focusing on capacities and desires to remit.Household income was, as expected, a strong and highly significant covariate of remitting to friends and family (Antoniades et al. 2013;Carling 2008;Carling and Hoelscher 2013;Constant and Massey 2002;Holst, Schäfer, and Schrooten 2008;2011;König, Isengard, and Szydlik 2018;Roman 2013).In terms of occupational status, only being self-employed reached statistical significance.Educational level was statistically insignificant.Hence, only economic capital, especially in terms of income and to a minor extent occupational status, is relevant for remitting.Finally, males were more likely to remit (Sana and Massey 2005).
Turning to desires for sending remittances, we find statistical evidence for real estate investment in the CoO as a driving motivation.Both types of real estate, that is, real estate owned by the respondent and real estate owned by family members, were predictors of the likelihood to remit to family and friends.We furthermore find that identification with the CoO increased the likelihood of remitting, whereas a strong identification with Switzerland had the opposite effect.Perceived discrimination and language proficiency do not reach any level of statistical significance.
Having a look at the difference in effect sizes of the country-related variables between models 1 and 2, we find that the average marginal effects (AME) for Great Britain became markedly smaller, which was also true for Portugal's AME to a more modest extent.However, the AME for the southeastern European countries did not diminish at all.Hence, previous explanations of remitting are helpful for explaining group-specific differences only for a subset of cases.
Model 3 introduces the variables measuring social relations and moral obligations.Regarding the former, the model confirmed that frequent contacts with family members and the respondents' view of their center of life were both positive and statistically significant covariates.Thus, respondents with strong social ties to family members in their CoO and respondents who considered a foreign country their center of life were more likely to remit to family and friends.Yet, other types of social relations, namely those with relatives and friends, did not exhibit the same statistical effects.These results show how the embeddedness of remittances into transnational relations is confined to the core family.Therefore, our first hypothesis H1a is corroborated by the data only for family relations.We turn next to our second hypothesis H2a regarding family obligations.Respondents with a strong conviction that there is an obligation to support family members in need, even in the light of high costs, were more likely to send remittances to family and friends.
These results clearly support the view that remittances for family members and friends are intricately linked to social relations and normative obligations (Rapoport and Docquier 2006;Carling 2008;2014;2020;Yang 2011).They exhibit a statistically robust covariation controlling for established explanations based on capacities and desires to remit.
Once we take these variables into account, identification with the CoO and the CoR even turn insignificant (Carling and Hoelscher 2013;Roman 2013).Moreover, based on the changes in the AME of the country-specific variables, we see how social relations and moral obligation are helpful to explain group-specific differences in the cases of Portugal, Serbia, and second-generation migrants since all of the respective effects further diminish in size.Hence, these determinants add to and go beyond the power of usual explanations in various contexts of migration.
To gain more detailed insights into the importance of particular capacities and desires for explaining differences in sending remittances to family and friends we computed a series of additional models.These models allowed us to assess the relative reduction in the AME of the country and generational effects when we add a particular capacity or desire (Baron and Kenny 1986; results not shown but available on request). 1Thereby, we find that income is especially relevant to explain the country effects of Great Britain and Italy.The center of a respondent's life is especially relevant for Portugal.For both SSE countries (Serbia and Bosnia-Herzegovina) we see the biggest reduction in effect size when entering family obligations.This is in line with the profiles of these migration groups.Looking at Table 5, which provides an overview of the distribution of the independent variables to explain remittance behavior, we see that respondents from Great Britain are characterized by high-income, high-skilled employees, respondents from SSE countries by strong family obligations, and respondents from Portugal by an especially strong orientation towards life in a country abroad (i.e.real estate abroad, identification with the CoO and perceiving a country abroad as the center of one's live).Turning finally to the generational effect, we find that social relations are of special importance to explain the difference between firstand second-generation migrants (Fokkema, Cela, and Ambrosetti 2013).Table 6 presents the logistic regressions for the second type of remittances, namely sending money to the respondent's own bank account.Since none of the respondents in Bosnia-Herzegovina exhibited this type of remitting, we combined the southeastern European (SSE) countries, Serbia and Bosnia-Herzegovina, to prevent non-identification of the logistic regression models (Long 2011).Starting with model 4, we find clear and statistically significant differences between all CoOs, with Italy and the SSE cases having a lower propensity to remit and Great Britain and Portugal having a higher propensity to remit compared to the reference category Germany.As before, second-generation migrants were also less likely to send money to a bank account abroad than those born in the CoO.
Turning to model 5, we find that income, gender, identification with the CoR and with Switzerland, and owning real estate abroad are significant determinants of remitting to a bank account.Again, perceived discrimination is not significantly correlated.However, in contrast to remittances to family and friends, education and language proficiency are significant covariates while occupational status and ownership of real estate by family members are only significant on the 10% level.The coefficient of language proficiency is negative.This shows that familiarity with the language of the CoR is mainly a proxy for socio-cultural integration, diminishing the desire to send money abroad.These determinants are most successful in explaining the remittance behavior of respondents from Great Britain, second-generation migrants, and to a lesser extent those from southeast European countries, as can be seen from the reductions of the AMEs from model 4 to model 5.
In model 6, we add the key variables of interest.Only the variables for social relations again limited to family networks and the respondent's center of life -, turn out to be significant covariates.In contrast to remittances for family members and friends, family obligations do not exhibit a statistically significant covariation.Taken together, we therefore consider H1b to be confirmed by the data, while H2b is not supported.Additionally, once we introduced these variables, the statistical significance of educational differences and language proficiency drops to the 10% level.Interestingly, when it comes to sending money to a personal bank account, both variables for identification remained significant.Hence, socio-cultural integration and transnational identification are of relevance (Carling and Hoelscher 2013;Roman 2013), but only for this type of remitting.Looking finally at the changes in the AME of the country-related variables, we see how the effects diminish for the second-generation migrants and to a lesser extent for Portugal from model 5 to model 6.Yet, there was no reduction in the effect sizes for Italy and the SSE countries.
As before, we conducted additional analysis on the importance of particular capacities and desires for explaining differences in sending remittances to a personal bank account.Here, identification with the CoO and the CoR turns out to be the major driver behind the country differences for the Italian and the southeastern European groups.Similarly, personal ownership of real estate is most important in the Portuguese case, reflecting their profile as migrants with a stronger orientation towards life abroad (see Table 5).Language proficiency yields the most substantial reduction for Great Britain.Swiss language proficiency is lowest in this group (see Table 5).In general, therefore, we find that determinants related to socio-cultural integration are most important in all cases for explaining the country differences in sending remittances to a personal bank account.Finally, as with sending remittances to family and friends, social relations lead to the biggest reduction of the generational effect.
Overall, we find many commonalities in the determinants of remittances for family members and friends and the determinants for remittances to a personal bank account.Yet, one important pattern emerges.Sending money to family members and friends is associated with moral obligations but not with the identification with the CoO or CoR, while the opposite is true for sending money to a personal bank account.Irrespective of socio-cultural integration, moral obligations compel migrants to directly support distant others in their CoO.Sending remittances to a personal bank account, in contrast, depends on the desire of belonging to the CoO or the CoR.Strong ties to core family members are relevant in both cases.Social capital in terms of network embeddedness thus possibly works as a kind of social lubricant, facilitating the motivating forces of obligations and identifications alike (Granovetter 1985).

Summary and discussion
This study has focused on the explanation of remittance behavior of firstand secondgeneration migrants from Germany, Italy, Great Britain, Portugal, Serbia, and Bosnia-Herzegovina living in Switzerland by transnational social relations and normative family obligations.In doing so, we have addressed two important research gaps: First, we focused on intra-European migrants in Switzerland, since patterns of their individual remittance behavior are not well studied (Bauer and Sinning 2009;Cela, Fokkema, and Abrosetti 2013;Constant and Massey 2002;Fokkema, Cela, and Ambrosetti 2013;Holst, Schäfer, and Schrooten 2008;Sinning 2011;Wolff 2019).Second, our study analyzed the roles of social relations to the CoO and normative family obligations.We used data from the TRANSSWISS survey to test hypotheses derived from the theoretical discussion on social relations and normative family obligations, with two types of remitting behavior as dependent variables: remitting to family and friends and to the respondent's own bank account.All the statistical models included additional variables that are usually referred to in explanations of remitting (Carling and Hoelscher 2013).
The analysis mainly supports our hypotheses.Frequent social contacts with one's family in the CoO and perceiving a country abroad as one's center of life contribute to both types of remitting, namely to support family and friends but also to move the funds into a personal bank account.Interestingly, the effect of social relations is restricted to the core family.Social relations with relatives and friends in the CoO are irrelevant for sending remittances.In contrast, normative family obligations are only statistically relevant for remittances to family and friends.They do not explain remittances to one's own bank account.This underscores the altruistic motivation to remit for the former type of remittances based on internalized family obligationsa moral norm to support one's family even in the face of personal costs, but it does not have indirect effects on remitting on one's own bank account.Besides, we found that income, male gender, and owning real estate are consistent predictors, underscoring the importance of social inequalities and social class differences for remittances.
Generally, we observed a surprisingly high level of remitting among intra-European migrants living in Switzerland.However, we also discovered striking differences in the types of remittances, the motives for remitting, and the capacities and desires explaining differences between migrants from Germany, Italy, Great Britain, Portugal, Serbia, and Bosnia-Herzegovina.Table 7 provides a typological overview of this variation.We found that migrants from Serbia and Bosnia-Herzegovina adhere very closely to the traditional model of remittances, assisting family and relatives back home, primarily to help in distress and support living standards.The higher level of remitting to family and friends is best explained by the desire to support one's family and not capacities to remit.This should be seen in relation to the socioeconomic situation of the western Balkan states, where remittances play an important role for sustaining living standards (King, Frykman, and Vullnetari 2013).In contrast, the higher level of remitting of migrants from Great Britain is primarily explainable by better economic opportunities, i.e. higher income.This relates to their profile of high-skill and high-income transmigrants, reflecting their class position.Finally, we find that the most important reason for the difference in remitting between firstand second-generation migrants lies in the generally weaker social ties to the CoO.
Coming to the limitations of our study, first, it is quite evident that we can only speak of covariations but not of causal relations.For nearly all statistical correlations presented in our paper, the direction of causality is not fully clear.Even the level of income could be based on the desire to remit, since migrants may decide to work more and harder so that they are able to remit.Therefore, this study leaves us with a pattern of interesting correlations.In the future, studies of remittance behavior should turn more to longitudinal designs.Second, we have focused on individual-level variables, mostly leaving out the socioeconomic contexts of the CoR and the CoOs.We fully acknowledge, however, the relevance of macro-level explanatory factors, for example economic development, infrastructure, or stability of financial institutions such as national currencies (e.g.Sana and Massey 2005).We encourage future research to extend our approach by including more countries in order to simultaneously test micro-and macro-level explanations with multi-level models.Another strategy would be to take subnational variations within countries into account, since most of the countries show a strong regional variation in important macro-level variables like GDP per capita, unemployment, and outmigration thus overcoming methodological nationalism.In sum, our findings clearly show how social relations to the CoO and normative family obligations shape remitting behavior beyond the usual determinants of capacities and desires to remit.This makes a strong case for the embeddedness of transnational economic behavior prominently argued for in economic sociology (Beckert 1996;Granovetter 1985;Zelizer 2006).Beyond the mere focus on the determinants of remittances, our study underscores that intra-European migrants deserve more attention in contemporary migration research.It is especially important to focus also on migrants from western and southern Europe, which are as yet understudied.Their patterns of migration and transnational living remain a terra incognita in several areas.Therefore, research on migration should be broadened to include a more heterogenous set of CoOs.Note 1. Basically, this is a type of mediation analysis, where the countries and the generational variables are the independent and sending remittances the dependent variables, with the particular capacity or desire to remit as mediator.(Partial) mediation by the desire or capacity occurs when the effect of an independent variable is reduced (Baron and Kenny 1986).More concretely: The baseline model is always model 3.Then, we estimated model 3 without a capacity or desire, for example income, which we call here model 3'.We then computed the relative change in the AMEs for all countries and the generational effect between model 3 and model 3'.Finally, we compared the size of the relative changes for all mediators, that is, all capacities and desires.We report those capacities and desires with the strongest reduction for each country and generational effect.

Statement on informed consent
All participants were informed about the goals and aims of the research, they were informed that and how anonymity is assured and how their information is used.Participants were also informed that participation in the survey is voluntary.

Table 1 .
Overview of the variables for the regression analysis (CoO = CoO; CoR = CoR, i.e.Switzerland).

Table 2 .
Relative frequency (in percentage) of remitting at least once per year per CoO and type of remittance, i.e. respondent's bank account or friends and family (F&F), in percent.
number of respondents per country.The two highest values for each type of remittances are highlighted in bold.N is the total number of valid cases per remittance type.† p < 0.1.*p < 0.05.**p < 0.01.***p < 0.001.

Table 3 .
Relative frequency (in percentage) of purposes for remitting per CoO in percent.Relative frequencies are computed by dividing the absolute frequency of the respective purpose by the number of respondents per country.The two highest values for each purpose are highlighted in bold.N is the total number of valid cases per reason.Respondents were able to choose several purposes.

Table 4 .
Logistic regressions with remitting to friends and family as dependent variable.

Table 5 .
Descriptive statistics of the independent variables per CoO.
Note: The highest value for each variable is highlighted in bold.

Table 6 .
Logistic regressions with remitting to respondent's bank account as dependent variable.

Table 7 .
Overview of country differences for sending remittances to family and friends (F&F) and to the respondent's bank account.