1. Top of page
  2. Abstract
  3. Introduction
  4. Theoretical framework
  5. Data, variables and methods
  6. Results and analysis
  7. Discussion and conclusions
  8. References

This article asks how return migration intentions are shaped by ties to the country of residence on the one hand, and ties to the country of origin on the other. We discuss these two sets of ties in terms of immigrant integration and transnationalism, respectively. A central tenet of the study is that, at the individual level, integration and transnationalism are neither related in a predictable way nor independent of each other. In our analysis we take methodological steps that reflect this argument, and introduce an integration–transnationalism matrix. In the empirical analysis we use quantitative survey data (N = 3,053) on ten large immigrant groups in Norway, collected by Statistics Norway in 2005–06. We find that it is the relative strength of integration and transnationalism that is decisive for return migration intentions.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Theoretical framework
  5. Data, variables and methods
  6. Results and analysis
  7. Discussion and conclusions
  8. References

When migrants contemplate return, they are influenced by two sets of attachments: ties to the country of residence, and ties to the country of origin. The latter ties result from their life before migration, or from transnational practices such as return visits or long-distance communication. Attachment, in this context, refers to the totality of place-specific resources, networks, competencies and emotions that affect a person's life. In this article we ask how this pair of attachments – reflected in the concepts integration and transnationalism – together affect return migration intentions.

We contribute to the existing literature in two ways. First, we offer a conceptual framework – what we call the integration–transnationalism matrix – that facilitates analysis of migrants' multiple attachments. This framework is generally applicable to studies of migrant adaptation. It serves as a tool for empirical analysis as well as for thinking about possible ways in which migrants manage multiple attachments. Others have pointed out that integration in the country of detination can perfectly well coexist with strong transnational attachments to the country of origin (Snel et al., 2006; Erdal and Oeppen, 2013). In other words, one does not preclude the other. What we do here, by means of the integration–transnationalism matrix, is to explicate the differences between possible combinations of strong and weak attachments to the two countries. While our basic model is premised on two simplified dimensions of attachment, their intersection can inspire more nuanced analyses and policy development related to immigrant incorporation.

The second objective we pursue is this aricle is to use our framework to empirically analyse return migration intentions. We do so by means of a large quantitative survey among immigrants in Norway. The analysis contributes to understanding what makes migrants want to return to their country of origin. We build on a longer history of research on return migration intentions (e.g. Ahlburg and Brown, 1998; Dustmann, 2003; Waldorf, 1995), but seek to combine this with more recent work on the interaction between integration and transnationalism (e.g. Snel et al., 2006; Erdal and Oeppen, 2013).

Our approach to return migration intentions resembles that of Hein de Haas and Tineke Fokkema (2011), and Piotrowski and Tong (2013). The studies nevertheless differ in several ways. The other two use data on migrants who primarily migrated for economic reasons. De Haas and Fokkema study African immigrants to Italy and Spain in 1997, and Piotrowski and Tong study rural–urban migration in China. By contrast, our more diverse sample is dominated by family and asylum-related migration. In terms of theory, de Haas and Fokkema focused on evaluating the opposing claims of neoclassical economics and the new economics of labour migration. There was mixed support for both, they concluded. These theories seem less relevant in the context of our sample. Instead, we seek to contribute to the growing theoretical and empirical literature on how integration and transnationalism are related. Return migration intentions provide one window on this relationship.

Theoretical framework

  1. Top of page
  2. Abstract
  3. Introduction
  4. Theoretical framework
  5. Data, variables and methods
  6. Results and analysis
  7. Discussion and conclusions
  8. References

We are interested in how integration and transnationalism shape return intentions. All three concepts are theoretically complex in their own right. We start by examining the concept of return intentions and then turn to possible causal relationships with integration and transnationalism.

Return intentions

Migration intentions – for initial migration or return migration – are a valuable but challenging concept. Such intentions are sometimes dismissed because they are poor predictors of actual behaviour. Indeed, in the case of return migration, it is almost a rule of thumb that most migrants initially intend to return but often end up staying in the country of destination. The “myth of return” can consequently be seen as collectively held return intentions that remain unfulfilled.

Scepticism to studying intentions can be countered in two ways. First, intentions can be seen as a necessary but not sufficient criterion for action. In other words, actual returnees will be among the people who intend to return, but not all those who intend to return will actually do so. This takes us some way towards predicting behaviour. Second, and more importantly, return intentions are significant in their own right: they represent summary attitudes to the migration experience, and can affect behaviour other than return itself – for instance about investment in relationships, skills or assets.

The survey and our theoretical framework assume that migrants have the legal option of remaining in the country of residence, and that return consequently involves an element of choice. This is not always the case. For instance, undocumented immigrants might anticipate being apprehended and deported. Although this is not a concern in our sample, migrants who are not constrained by immigration regulations may still be restricted by power relations within families.

Integration, transnationalism and return

Immigrant integration and transnationalism are multi-faceted and partly contested concepts. In our analysis they represent attachment to the country of destination and country of origin, respectively. On the transnationalism side, our concept is restricted to social and economic transnational practices with a low level of institutionalization (see Itzigsohn et al., 1999; Levitt, 2001; Portes, 2003, Smith and Guarnizo, 1998). On the integration side, our use largely overlaps with the notion of “socio-cultural” integration (de Haas and Fokkema, 2011; Snel et al., 2006). The contrasting concept of “structural integration' also encompasses dimensions that may be relevant to return migration – such as employment and education –but which we treat separately.

The central relationships in our analysis are presented in Figure 1. We hypothesize that there is a negative relationship between integration and return intentions (1). In our case we expect respondents who are strongly attached to Norway to be less likely to intend to return to their country of origin. The same effect ensues when marginalization of an immigrant in Norway provides an impetus for returning. While it seems safe to assume an overall negative relationship between integration and return migration intentions, there are possible mechanisms going in the opposite direction: if integration has improved access to resources and information, return could be facilitated.


Figure 1. Theoretical framework

Download figure to PowerPoint

We also hypothesize that there is a positive relationship between transnationalism and return intentions (2). Having maintained a social presence in the country of origin, one has more to return to. Also in this case, however, opposite effects are possible: regular visits and other forms of contact with the country of origin could serve as a ‘reality check’ and rule out any illusions of what life after return might be like (see Oeppen, 2013).

There can also be causal effects in the opposite direction, from return intentions to integration (3) and transnationalism (4). If migrants expect to return, their integration efforts could be limited or highly selective. For instance, there could be little incentive to learn the destination-country's language unless it is needed for work. By contrast, migrants who intend to return have an incentive for investing in transnational relationships that facilitate reintegration.

We cannot understand how intentions to return are shaped by integration and transnationalism without also asking how these two are related to each other. A popular assumption, for instance in policy debates, is that there is a zero-sum game of attachment in which stronger integration implies weaker transnationalism and vice versa. Such a view resonates with traditional assimilationist theories. In recent social-scientific research, however, it is increasingly challenged, on both theoretical and empirical grounds (Schans, 2009; Snel et al., 2006; Tamaki, 2011).

We therefore employ a theoretical framework that sees the two as intersecting dimensions in what we call the integration–transnationalism matrix (Figure 2). This allows for different empirical possibilities: if there is a zero-sum game of attachment, quadrants B and C would be virtually empty. What we expect, however, is that these quadrants are also populated. In quadrant B there would be individuals who are strongly attached to the country of destination, and at the same time maintain strong transnational ties. Synergies exist when, for instance, resources acquired through successful integration provide the basis for extensive transnational practices. In quadrant C there would be people in a very different situation, comparatively marginalized in the country of destination and detached from the country of origin. Again, there can be a causal relationship: alienation in the country of residence could result in deprivation of the resources required for transnational practices.


Figure 2. The integration–transnationalism matrix.

Download figure to PowerPoint

The matrix enables us to examine the triangular relationship between integration, transnationalism and return migration. Instead of seeing integration and transnationalism as separate influences on return intentions, we will ask how these intentions differ by location within the matrix.

Data, variables and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Theoretical framework
  5. Data, variables and methods
  6. Results and analysis
  7. Discussion and conclusions
  8. References

We use data from the survey “Living Conditions among Immigrants in Norway”, conducted by Statistics Norway in 2005–06 (Gulløy, 2008). The survey focuses on a wide range of issues, with several questions relating to integration in Norway and ties to the country of origin, including return intentions. We are primarily interested in long-term social dynamics, which are unlikely to have changed substantially since the data were collected. However, some of the more time-specific results must be interpreted with caution. For instance, the political and economic situation in some countries of origin is likely to have changed in ways that affect the desire to return to those countries. Development in Norway may also have affected the desire to stay. The terrorist attacks of 22 July 2011, for instance, and the societal response to them, could have strengthened or weakened immigrants' sense of belonging.

Ten countries of origin are covered in the survey, selected among the largest immigrant groups in Norway: Bosnia and Herzegovina, Serbia and Montenegro,1 Turkey, Iraq, Iran, Pakistan, Sri Lanka, Vietnam, Somalia and Chile. A total of 5,000 individuals aged 16 to 70 who had lived in Norway for at least two years were targeted, and 3,053 individuals responded to the survey. Interviews were conducted either face-to-face or by telephone, and respondents could choose between being interviewed in Norwegian and the dominant language in their country of origin. The response rate ranged from 54 per cent among Somalis to 74 per cent among Sri Lankans.

Descendants of immigrants were included in the sample and represent 9 per cent of all the respondents. For this group, “return migration” implies moving to their parents' country of origin. Among the immigrant respondents, about one quarter came to Norway as children. Because return migration takes on a different meaning for people who did not actively decide to migrate in the first place, we distinguish between two broad groups: 1) ‘adult immigrants’ who came to Norway at age 18 or above and 2) “children of immigrants” who are either born in Norway or immigrated before age 18. Note that most ‘children of immigrants’ are adults at the time of the survey: all are at least 16 years old; their mean age is 25 years. The generational differences are important to take into account, but are not the focus of our analysis.

The variables we use are either directly reproduced from the survey dataset or recoded or computed on the basis of questions in the survey or register data. Descriptive statistics of bivariate relationships with return intentions are presented in Table 1 and Table 2.

Table 1. Descriptive statistics on the categorical variables in the analysis
  Proportion who intend to return to the country of origin (%)Number of observations1
Full sampleAdult immi-grants2Children of immi-grants3Full sampleAdult immi-grants2Children of immi-grants3
  1. 1) 7 observations were deleted in the regression analysis due to missing values on one or more of the independent variables; 2) Immigrants who came to Norway as adults, i.e. aged 18 or above; 3) Children of immigrants who are either born in Norway or immigrated before age 18; R) Reference category in the regression analysis.

Total 2526223,0532,054999
Background variables
EducationLower secondary or less (and ‘other’) R222123972687285
Upper secondary2729241,284782502
University/college 1-4 years262821628461167
University/college 5 +  years2024716412242
Religiosity (Very religious)Yes2727271,5161,031485
Country of originSerbia23222628820682
Bosnia and Herz.32333033323697
Sri Lanka27282735327974
Pakistan R161516308138170
Grounds for immigrationOther222221549251298
Refuge R2426191,8121,301511
Socio-cultural integration
Not experienced discriminationYes2224191,6801,148532
Good Norwegian language skillsYes2529221,816932884
Transnational ties
Owns property in the country of originYes323232663416247
Visits to the country of originYes2829262,0531,379674
Sends remittances at least once a yearYes3131281,3871,069318
Integration–transnationalism matrix
Strongly integrated, weakly transnational R 1617151,041567474
Weakly integrated, weakly transnational 242328678537141
Strongly integrated, strongly transnational 272925866578288
Weakly integrated, strongly transnational 41404646837296
Economic resources
Employment statusNeither employment nor pension2323221,011643368
No employment, but pension R18181326223923
Insecure employment283221885579306
Secure employment262626895593302
Difficulties w/ regular expensesYes20221432326459
Difficulties w/ unex-pected expenseYes2424231,3941,032362
Table 2. Descriptive statistics on the continuous variables in the analysis. mean values by return migration intentions1
  Full sampleAdult immigrants2Children of immigrants3
Return migration intentionsYesNoYesNoYesNo
  1. 1) The number of observations is the same as for the total sample; see Table 1. 2) Immigrants who came to Norway as adults, i.e. aged 18 or above; 3) Children of immigrants who are either born in Norway or immigrated before age 18.

Age (years)373741432525
Proportion of life in Norway (%)394229286369
Sense of belonging in Norway (score)454545

Intention to return

Our dependent variable is a constructed dummy variable on return intentions. In the survey, respondents were asked “do you think you will ever move back to your country of origin, or that of your parents, to live there permanently?” A follow-up question captures the timeframe of the intended return through predefined categories (“0-4 years”, “5-10 years”, “when I am old”). A quarter of the sample replied affirmatively (Table 1). Among them, only 8 per cent intend to return within the next 5 years. For most of the prospective returnees, returning is postponed until old age (68 per cent), or within 5-10 years (24 per cent). Overall, 38 per cent responded that they did not intend to return, while 37 per cent responded “Don't know”. Interestingly, this distribution of responses is similar to that of de Haas and Fokkema's (2011) study of African immigrants in Spain and Italy.

We converted the response categories into a dummy variable: those who answered yes in some form are regarded as intending to return. The remaining two categories (“Don't know” and “No”) were combined, since “don't know” is a substantive response to the question, not a form of non-response.2

As shown in Table 1, return intentions vary considerably by country of origin. The highest proportions of prospective returnees are found among Turks (38 per cent) and Bosnians (32 per cent); the lowest proportions are found among Vietnamese (15 per cent) and Pakistanis (16 per cent). Overall, return migration intentions are more prevalent among adult immigrants (26 per cent) than among children of immigrants (22 per cent). The proportion responding “don't know” was particularly high among the refugee-dominated groups whose country of origin is marked by uncertainty. Somalis are a case in point: only 13 per cent said a clear “no” while 57 per cent reported that they did not know whether they would want to return.

Background variables

We include seven background variables in the analysis: gender, age, proportion of life spent in Norway, educational attainment, religiosity, country of origin, and grounds for immigration to Norway. Some of these require further explanations.

Proportion of life spent in Norway is calculated on the basis of age at arrival and time since arrival.3 This variable is theoretically related to integration whereas current age, as such, is not (Carling, 2008). We include square terms for these two variables to test for a non-linear relationship with the dependent variable.

Educational attainment is a constructed categorical variable capturing the educational level attained in Norway or abroad. We distinguish between the reference category “lower secondary education or less”, and three levels of higher education.

Religiosity is based on a survey question about the importance of religion in one's life. We constructed a dummy variable which distinguished “very important” from all other responses. Almost 80 per cent of the respondents in the survey adhere to religions other than Christianity.

We distinguish between three sets of judicial grounds for immigration to Norway, based on official records. The first is admission on the basis of family ties (family reunification, family formation, or accompanying family members) and the second is need for protection (as quota refugees, asylum seekers or on humanitarian grounds). Together, these two account for 82 per cent of the sample. The remaining respondents are grouped as “other”, and are mainly early labour migrants from Pakistan and Turkey or students who came to Norway before 1990. Descendants are coded with their parents' grounds for immigration, which is usually the same for both parents. Grounds for immigration do, as we will see, affect return intentions. However, these judicial categories should not be equated with motivations for migration. First, motivations can be multi-faceted while immigration rules are based on distinct categories. Second, migrants can sometimes choose their migration category for pragmatic reasons, such as if a person who migrates in order to join a spouse finds that the labour migration procedure is less cumbersome than family reunification.

The integration–transnationalism matrix

As explained in a previous section, we regard integration and transnationalism as intersecting dimensions which can be presented as a matrix. We have operationalized each dimension in the simplest possible terms, as either “weak” or “strong”, which produces a matrix with four categories (Figure 2).

Each dimension in the matrix is based on three dichotomous indicators. Respondents are identified as strongly integrated or transnational if they score positively on two of the three indicators on the respective dimension. Descriptive statistics for each indicator are provided in Table 1.

The first integration indicator, Norwegian language skills, identifies respondents who describe their skills as “good” or “very good” – the two highest levels on a five-point answer scale in the original survey. Sense of belonging in Norway is a recoded dummy variable, where those reporting the three highest levels on a seven-point answer scale of “sense of belonging in Norway” are compared to all others. The final integration indicator identifies respondents who, in their own opinion, have experienced no discrimination on the basis of their immigrant background in seven specific areas (housing, employment, workplace harassment, education, nightlife and financial services). The inclusion of this variable reflects an acknowledgement of the interactive aspects of integration: socio-cultural integration is not simply a matter of immigrants' behaviour and attitudes, but of the way in which interaction with society at large takes place.

The first indicator of transnationalism is ownership of property (a house and/or land) in the country of origin. Second, we identify respondents who have visited the country of origin during the past five years. The cost and difficulty of travel to or from Norway differs considerably between the ten countries of origin. Controlling for country background, we believe that this variable can reflect the strength of transnational relationships. Finally, we identify those sending remittances to family members in the country of origin at least once per year.

Economic resources

In line with our theoretical framework, we deliberately regard economic resources separately from integration. The relationship between economic resources and return migration intentions is ambiguous. On the one hand, economic success could inspire or be a prerequisite for return. Moreover, economic resources precipitate return when migrants are “target savers”. On the other hand, economic success could also inhibit return intentions. Migrants with secure employment and income in the country of destination may have more to lose by leaving.

In our analysis we use three measures of economic resources. The first is a four-fold classification of employment status, constructed from several questions (Carling and Hoelscher, 2013). We differentiate between “secure” and “insecure” employment, and, among those who are not employed, between people who receive other forms of long-term income, such as pensions, and those who do not.

The second variable is based on a survey question on whether the household had experienced difficulties with regular expenses such as the cost of food, housing and transportation during the previous twelve months. A positive score indicates that the respondent's household “often” experienced such difficulties.

The final variable identifies respondents whose household finances during the past year have mostly been in such a state that it would be difficult to cover an urgent unexpected expense of NOK 5,000, for instance for repairs or dentistry.

Results and analysis

  1. Top of page
  2. Abstract
  3. Introduction
  4. Theoretical framework
  5. Data, variables and methods
  6. Results and analysis
  7. Discussion and conclusions
  8. References

We use logistic regression to statistically investigate the explanatory power of the different variables on the intention to return to the country of origin. Logistic regression models allow us to measure the isolated effect each independent variable has on the likelihood (odds ratio) of intending to return to the country of origin, simultaneously controlling for all other variables in the model.

Before presenting the regression analysis, we show the bivariate results of combining return intentions with the integration–transnationalism matrix (Figure 3). The size of the four circles reflects the relative size of the groups defined by our integration–transnationalism matrix. This alone illustrates the basic relationship between integration and transnationalism. The weakly transnational are in the majority regardless of the level of integration, and the strongly integrated are in the majority regardless of the level of transnationalism.


Figure 3. Return expectations in the integration–transnationalism matrix.

Notes: Dark sectors represent proportion of respondents indicating that they intend to return to the country of origin. see text for details on the operationalization of integration and transnationalism.

Download figure to PowerPoint

The shaded sector in each circle represents the proportion of respondents who intend to return. Here we see clear signs of interaction between transnationalism and integration. The lowest levels of return intentions (16 per cent) are found among those who exhibit weak transnationalism combined with strong integration. At the opposite end, the highest return intentions (41 per cent) are found among respondents who exhibit strong transnationalism combined with weak socio-cultural integration. In other words, socio-cultural integration and transnationalism considered together go further towards explaining return intentions than either concept does alone.

What we observe in Figure 3 should, however, be interpreted with caution. Since we do not control for other factors, the differences between the four quadrants of the matrix could reflect differences in the composition of the four groups – for instance in terms of national origin – as much as integration and transnationalism. We now turn to the regression analysis in order to discern between different influences on return intentions.

Regression results are displayed in Table 3, which contains three models: the first regresses only the background variables; the second regression model adds the integration–transnationalism matrix, and the third model also includes the variables related to economic resources. By sequentially adding variables to the model, we are able to identify the individual effect of each block of variables on the dependent variable, and also the total effect of the model itself. For model 3, we show results for the two sub-groups – adult immigrants and children of immigrants – in addition to the full sample.

Table 3. logistic regression results for intention to return to the country of origin
 Full sampleAdult immi-grants1Children of immi-grants2
Model 1Model 2Model 3Model 3Model 3
  1. 1) Immigrants who came to Norway as adults, i.e. aged 18 or above; 2) Children of immigrants who are either born in Norway or immigrated before age 18; R) Reference category; *) p < 0.05; **) p < 0.01; ***) p < 0.001.

Background variables
Gender (male)1.437***1.360**1.331**1.281*1.292
Age (continuous)1.070**1.049*1.0441.0730.903
Age squared0.999**0.999*0.9990.999*1.002
Proportion of life in Norway (continuous)12.264**9.660**9.741**21.86993.353*
Proportion of life in Norway, squared0.130**0.177*0.172*0.0340.029*
Lower secondary or less and ‘other’ R1.0001.0001.0001.0001.000
Upper secondary1.377**1.420**1.394**1.502**1.134
University/college 1-4 years1.331*1.390*1.344*1.428**1.177
University/college 5 +  years0.8890.9660.9141.1950.296
Religiosity (very religious)1.437**1.393**1.411**1.384**1.492*
Country of origin
Bosnia and Herzegovina3.528***3.187***3.082***3.053**3.393**
Sri Lanka1.853**1.977**1.911**1.820*2.943**
Pakistan R1.0001.0001.0001.0001.000
Grounds for immigration
Protection R1.0001.0001.0001.0001.000
Integration–transnationalism matrix
Strongly integrated, weakly transnational R 1.0001.0001.0001.000
Weakly integrated, weakly transnational 1.593**1.633**1.468*2.284**
Strongly integrated, strongly transnational 1.821***1.797***1.919***1.594*
Weakly integrated, strongly transnational 3.272***3.231***3.019***4.230***
Economic resources
Employment status
Neither employment nor pension  1.3541.2331.706
No employment, but pension R  1.0001.0001.000
Insecure employment  1.498*1.596*1.375
Secure employment  1.3911.1612.406
Difficulties with regular expenses  0.7900.8910.367*
Difficulties with unexpected expense  0.9720.9261.189

The model Chi-square is significant for all regression models, which indicates that the overall models are significant (not shown). The AIC, a measure of the relative goodness of fit of a statistical model, suggests that the model's explanatory power increases when we introduce the integration–transnationalism matrix, but not when we add the variables on economic resources.4

The odds ratio indicates the effect of the independent variable on the intention to return, holding all other variables in the model constant. For categorical variables, an odds ratio greater than 1.0 indicates an increased likelihood of intending to return compared with the reference category (a positive effect). An odds ratio of less than 1.0 indicates that the individual is less likely to intend to return than those in the reference category (a negative effect). In the case of continuous variables, the odds ratio represents the expected change that results from a one-unit change in the independent variable, e.g. ageing one year.

We now turn to discussing individual results (Table 3). Model 1 shows the effects of our six background variables. Men are found to be significantly more likely than women to have return intentions. Age and proportion of life spent in Norway both have an inverted U-shaped curvilinear relationship with return intentions. This indicates that return intentions increase with age to a certain point (age 42) and subsequently declines. Similarly, the odds ratio for proportion of life spent in Norway peaks at 61 per cent (43 per cent for adult immigrants). Migrants who have spent a smaller or larger proportion of their lives in Norway are less likely to have return intentions. The odds ratio for proportion of life spent in Norway is a large number since it measures the effect of going from 0 (just having arrived) to 1 (having spent one's entire life in Norway).

The effect of educational attainment also follows an inverted U-curve. Respondents with upper secondary education are significantly more likely to have return intentions than those with less education or with graduate-level university education. Those who are very religious are also more likely than others to intend to return.

There are large differences by country of origin: the odds of having return intentions are more than three times larger among Turkish respondents than among otherwise similar Pakistani or Vietnamese respondents. In other words, national origin contributes strongly to explaining differences in return intentions also after controlling for other background variables. For instance, the low level of return intentions among Vietnamese and Pakistanis do not disappear when we consider their long migration history compared with other groups.

Finally, those who have immigrated to Norway for family reasons have 1.5 times higher odds of intending to return than those who were granted residence on the grounds of protection or other reasons. This finding could imply different return intentions within families, whereby those who have immigrated for family reunification or marriage are more intent on return than the partner who arrived for protection, work or studies. Our results contrast with those of de Haas and Fokkema (2011) who find no effect of main motive for migration. As noted earlier, however, there is a difference between judicial grounds for immigration and self-reported migration motives. Furthermore, there seems to be a greater diversity of migration forms in our sample. It is not surprising that this is reflected in the results.

In summary, we find in Model 1 that all the background variables are important in predicting return intentions, with particularly large and significant isolated effects for country background. As we shall see, these effects hold across all three models.

In Model 2 we introduce our main independent variable of interest, the integration–transnationalism matrix. The regression confirms the conclusion from the descriptive statistics: those who are strongly integrated and weakly transnational are significantly less likely to intend to return to the country of origin than all others. The differences are substantial; respondents at the other end of the matrix – weakly integrated and strongly transnational – have 3.3 times higher odds of intending to return. As in the bivariate analysis (Figure 3), the two remaining categories – weakly integrated and weakly transnational, and strongly integrated and strongly transnational – are similar to each other and occupy an intermediate position.

When we introduce the matrix in Model 2 the background variables remain remarkably stable, both in the size of the individual effects and in significance levels. At the same time, the matrix adds considerable explanatory power in predicting return intentions, demonstrated by the lower AIC. Return intentions can thus be partly explained by the integration–transnationalism matrix, separately from, for instance, country effects. We do find, however, that the positive and significant effects of coming from Somalia, Sri Lanka or Iraq are slightly reinforced with the introduction of the matrix. Turks and Bosnians, who according to the descriptive statistics are the most positive about returning, have slightly lower odds when we control for integration and transnationalism. In other words, the bivariate results were influenced by particularly strong transnationalism and/or weak integration among these groups. In Model 2 we also see that the effect of educational attainment becomes stronger, with a more pronounced inverted U-curve across educational levels.

In the final model, Model 3, we introduce the measures of economic resources. The effects on return intentions are small; only those with insecure employment status are more likely than the reference group to have return intentions. The modest effect of employment status resonates with de Haas and Fokkema's (2011) finding that structural integration is not significant for return intentions.

As mentioned above, introducing the economic measures does not increase the explanatory power of the overall model. The effects of the integration–transnationalism matrix remain stable. However, we see that the significance of age, which was reduced in Model 2, now disappears altogether. This implies that in the previous models, the age variable was partly reflecting differences in economic resources between age groups. Age alone, we now see, has no significant impact on return intentions.

By contrast, the significance of proportion of life in Norway remains in Model 3. This indicates that the intention to return is a function of relative time spent in Norway independently of integration levels, transnational ties and select economic measures.

The last two columns in Table 3 show the results of Model 3 on the two subdivisions of the sample, adult immigrants and children of immigrants. The effects mostly go in the same direction for both groups, but vary in size and significance.

Figure 4 visualizes key results of the regression analysis: how position in the integration–transnationalism matrix affects return intentions, controlling for all other variables in Model 3. As in the table, we use the category “strong socio-cultural integration; weak transnationalism” as the reference group, and see how much higher the likelihood of return intentions is in the three other groups. The figure visualizes the uncertainty of the statistical analysis by means of shaded areas representing the 95 per cent confidence interval around the estimates.


Figure 4. Return expectations in the integration–transnationalism matrix.

Notes: Black lines represent odds ratios relative to the reference category (strong socio-cultural integration; weak transnationalism). shaded areas represent 95% confidence intervals. Notes: 1) immigrants who came to norway as adults, i.e. aged 18 or above. 2) children of immigrants who are either born in norway or immigrated before age 18.

Download figure to PowerPoint

Intentions to return are most prevalent among migrants with strong transnational ties and weak socio-cultural integration. There is no discernible difference between the two middle groups. The pattern that emerges thus suggests that it is the relative strength of integration and transnationalism that is important for return migration intentions.

The overall picture is the same for the two subsets of the sample, adult immigrants and children of immigrants. There are slight differences between the generations, however. Among the children of immigrants, integration appears to be more important than transnationalism in explaining differences in intentions to return: among the weakly integrated, the strength of transnationalism has no statistically significant effect. It is challenging to compare the groups, however, since the smaller sample of children of immigrants results in larger margins of error.

Discussion and conclusions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Theoretical framework
  5. Data, variables and methods
  6. Results and analysis
  7. Discussion and conclusions
  8. References

We have sought to find out how integration and transnationalism interact in shaping return migration intentions. The findings of the regression analysis can be summarized as follows.

First, return intentions vary systematically by gender, age and migration history. We find lower odds of return intentions among women, the young, the elderly, and the people who have either spent very little or very much of their lives in Norway. The most likely prospective returnee appears to be a man in his forties who has spent about half his life in Norway. He is more likely to see return as an option for retirement than an immediate plan.

Second, return intentions vary systematically by educational attainment. The highest likelihood of indending to return is found among people with a medium-level educational attainment; people with very low or very high levels of education are least likely to intend to return. Not surprisingly, perhaps, highly educated children of immigrants have the lowest odds of intending to return.

Third, religiosity has a significant, independent effect on return intentions. Respondents who say that religion is very important in their lives are more likely to intend to return to their country of origin. This finding could be directly related to the greater possibilities for letting one's faith shape daily life in the country of origin. Alternatively, it could reflect a more diffuse orientation towards traditional values that correlate with religiosity and stand in contrast to the relative liberal and secular environment in Norway.

Fourth, return intentions vary substantially between migrants from different countries. This reflects two overlapping effects: (1) differences between the groups' composition and migration history, and (2) differences between the countries to which respondents would return. Looking at the descriptive statistics, one could think than disparities between the groups were due to differences in characteristics such as age, length of stay or grounds for immigration, which have independent effects on return intentions. The multivariate regression analysis, however, shows that this is not the case. In other words, being Bosnian, for instance, is associated with very high odds of return intentions that must be attributed to either the specific characteristics of Bosnia as a potential destination, the unmeasurable qualities of “Bosnianness”, or variables that are not included in the model.

Fifth, immigrants who were admitted on the basis of family ties are significantly more likely to have return intentions than those who entered Norway as workers, students or for protection. In most cases, the family ties underlying migration are with other immigrants, not with native Norwegians. For family migrants, the source of the attraction of moving to Norway – the family members – are potentially mobile. Especially if they are immigrants from the same country of origin, joint return might be seen as feasible and desirable.

Sixth, integration and transnationalism – the way we have conceptualized these dimensions – play important roles for return intentions. More specifically, return intentions are shaped by the relative strength of integration and transnationalism. The highest likelihood of return intentions is found among people who are both weakly integrated and strongly transnational. What is remarkable here, is how integration and transnationalism can cancel each other out and produce intermediate odds of intending to return. This is the case with people who are both strongly integrated and strongly transnational as well as with people who neither strongly integrated nor strongly transnational. Moreover, these two groups have strikingly similar likelihood of return intentions. Hence, we conclude that it is the relative strength of integration and transnationalism that matters.

Seventh, economic resources have no clear effect on return intentions. Including these variables hardly improves the explanatory power of the model. This finding is perhaps not surprising, since there are plausible reasons why economic resources could either strengthen or weaken return intentions, depending on the circumstances. Economic theory in this area is of limited relevance to our sample, since traditional labour migrants constitute a small minority. Our finding is nevertheless important as a corrective, since economic variables play a big role in the literature on return migration.

Eighth, the determinants of return migration intentions appear to be relatively similar among adult immigrants and children of immigrants. As shown by the descriptive statistics, return intentions are remarkably prevalent among people who were born in Norway or spent at least some of their childhood in the country: 22 per cent intend to return; compared to 26 per cent among adult immigrants. This is a notable finding. Even if children of immigrants are unlikely to actually move to their parents' country of birth, signalling a wish to do so can be central to identifying with one's origins.

Beyond the results of our regression analyses, we have made more general empirical and theoretical contributions to the study of integration and transnationalism. Our proposed integration–transnationalism matrix reflects the argument that integration and transnationalism are neither related in a predetermined way nor independent of each other. The distribution of respondents across all four quadrants of the matrix attests to its value (Figure 3). The integration–transnationalism matrix is premised on particular conceptualizations of the two dimensions: we have used generally applicable indicators of individuals' attachment to the country of residence and country of origin, respectively. By separating economic resources from integration we avoid a conflation of being integrated and being successful.

There are several limitations to our study that must be noted. First, the sample is limited to countries outside Western Europe and North America, with living standards that are substantially lower than in Norway. It is valuable to study integration and transnationalism across a greater diversity of migrant groups, as in the seminal study by Snel et al. (2006). Paradoxically, the diversity of national origins nevertheless presents a methodological challenge: it is likely that determinants differ between the groups, but the often small numbers make it challenging to run separate analyses. Not only is there a limited sample within each group, but the prospective returnees often make up a small minority. There are, for instance, fewer than 50 Pakistani and Vietnamese respondents in our study who intend to return.

Second, survey data about intentions is tenuous. We cannot assume that all the respondents had an unequivocal intention that they could present to the interviewer upon being asked the relevant question. In our case, data collection was also vulnerable to nuances being changed in translation. Interviews were conducted in ten languages, and slight differences in the tone of the question could have affected the results.

Extensions from our study could go in several directions. What we are clearly missing out on in an individual survey like ours, are the complex intra-family dynamics of return migration decision-making. For instance, we have seen that male individuals are more positive about return than female individuals, but we don't know how the possibility of return is negotiated between spouses, for instance. More generally, survey data of the type we present here can be successfully used in conjunction with ethnographic data, providing that the populations are reasonably homogenous. For instance, Carling et al. (2012) used the same survey to study remittance-sending among Pakistanis and Somalis, and used ethnographic interview data from the same groups to provide interpretation of regression results.

It remains to be seen how many of the migrants in our survey actually return to their countries of origin. Their return intentions are nevertheless interesting in their own right. With respect to our central research question, we have shown that these intentions are shaped by people's attachment to their country of residence and country of origin, relative to each other. These attachments do not represent a zero-sum game: indeed, more than half of our sample have either weak attachments in both directions or strong attachments in both directions. What we have shown, however, is that when attachments are stronger in one direction than the other, return intentions are systematically affected.

  1. 1

    The survey was conducted before the dissolution of the Republic of Serbia and Montenegro in June 2006.

  2. 2

    By this we mean that “don't know” describes an attitude to return, comparable to other response categories. By contrast, “don't know” responses to factual survey questions such as “what is your highest level of education?” must be treated as non-response. We repeated the analysis with different treatments of the “don't know” category: merging it with the “yes” answers and keeping at a separate outcome in a multinomial regression. These alternative approaches produced similar outcomes with respect to our independent variables of interest.

  3. 3

    Estimated on the basis of the first immigration to Norway.

  4. 4

    The Akaike Information Criterion is calculated as AIC = 2 Log + 2((k–1) + s) where k is the number of levels of the dependent variable and s is the number of independent variables in the model. The preferred model is the one with the lowest value of the AIC.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Theoretical framework
  5. Data, variables and methods
  6. Results and analysis
  7. Discussion and conclusions
  8. References