Forced displacement and subsequent generations’ migration intentions: intergenerational transmission of family migration capital

ABSTRACT

A growing body of evidence for the ‘family migration capital’ hypothesis – whereby migration experience in a family leads to a greater propensity to move among migrants’ descendants – has so far relied on accounts of any migration experience, including voluntary moves. However, in the case of voluntary migration, a considerable part of the observable effect may be due to self-selection into migration and passing the migration-driving characteristics across generations rather than due to the transmission of ‘capital’ derived from migration. To minimise the influence of self-selection, we consider the effects of forced migration, where self-selection is less prevalent than in voluntary migration. Using data from the nationally representative Life in Transition Survey-III, collected in 32 countries in Eurasia (N = 41,977), in logistic regression, we show that descendants of people who experienced forced displacement as a result of World War II are more likely to report an intention to migrate than people in similar circumstances but without this kind of family experience. Our findings support the contention that migration experience leads to the accumulation of ‘family migration capital’ that is passed across generations and highlight the long-lasting consequences of forced displacement, happening on a large scale globally nowadays, for future voluntary migration flows.

KEYWORDS:

• Family migration capital
• migration intentions
• forced displacement
• intergenerational transmission

Introduction

Wars, conflicts and natural disasters displace millions of people both within countries and across international borders, the recent Russian invasion of Ukraine serving as a prime example. While the question remains open whether the forcibly displaced return to their places and countries of origin, one may also ask whether and how the experience of forced displacement may influence the propensity to migrate voluntarily among future generations, in particular among the descendants of those who were forced to move.

Migration history has been shown to have an overwhelmingly positive influence over one’s migration aspirations (understood here broadly, as in Aslany et al. 2021).¹ Apart from the role of personal experience, several works have pointed to the role of a history of family migration in predicting migration intentions (Clifton-Sprigg 2022; Ivlevs and King 2012), aspirations (Kandel and Massey 2002) and behaviour (De Jong and De Valk 2023; Francisco-Menchavez 2020; Massey and Espinosa 1997). This body of research has explained the higher propensity to migrate among those with past family migration experience through the intergenerational transmission of ‘migration capital’ accumulated in the family (see also Moret 2018). The question, however, remains to what extent pioneer migrants within those families self-selected into migration (Berlinschi and Harutyunyan 2019; Czaika and Vothknecht 2014) based on characteristics that they transmitted to their descendants, and to what extent the higher propensity to migrate among members of subsequent generations was due to the accumulation and intergenerational transmission of migration capital derived from migration (cf. Myers 1999). When the focus is on the consequences of voluntary migration, isolation of the effect that is due to the capital derived from the past move is difficult. A focus on the consequences of forced migration, where self-selection has been shown to be smaller (Ruhe and Kuhnt 2023), brings us closer to testing whether the intergenerational effect may be the result of the very experience of migration and not of the characteristics on which past migrants self-selected. While the influence of personal forced migration experience on current migration intentions has been studied before (Efendic 2016), to the best of our knowledge, existing research has not attended to the potential effect of forced migration experience on the migration intentions of members of subsequent generations (nor are we aware of any study of its effect on the subsequent generations’ migration aspirations or behaviour).

We address this hitherto unexplored question by examining the migration intentions of residents of several, mainly post-socialist, countries in Eurasia, which faced unprecedented migration flows during and in the aftermath of World War II (WWII). In this cross-sectional study, based on nationally representative household survey data from the 2016 Life in Transition Survey-III (LiTS-III), we investigate whether the descendants of those who experienced forced displacement as a result of WWII are more likely to report an intention to migrate than comparable people without this kind of family experience.

This article brings together research on the intergenerational transmission of migration behaviour (Ivlevs and King 2012; Kandel and Massey 2002; Myers 1999), long-lasting legacies of political experience, political violence in particular (e.g. Lupu and Peisakhin 2017; Nikolova, Popova, and Otrachshenko 2022; Toews and Vézina 2020; Zhukov and Talibova 2018), and the consequences of forced displacement for migrants themselves (for reviews, see Becker 2022; Becker and Ferrara 2019; Ruiz and Vargas-Silva 2013), including intergenerational effects (Bauer, Braun, and Kvasnicka 2013; Becker et al. 2020; Nakamura, Sigurdsson, and Steinsson 2022). It adds to these strands of literature in two ways. First, thanks to the more ‘exogenous’ character of forced migration, it offers stronger support for the existence of ‘family migration capital’ derived from past migration – relative to previous studies focused on the consequences of (mostly) voluntary moves. Second, it contributes to the literature on the long-lasting legacies of exposure to political violence and the intergenerational effects of forced displacement, focusing on the so far unstudied aspect of the next generation’s lives – their migration intentions.

Theoretical framework

According to the theory of planned behaviour (Ajzen 1991), intentions are shaped by perceived behavioural control (perceived ease or difficulty of performing the behaviour), attitudes towards behaviour (favourable or unfavourable evaluation of the behaviour), and subjective norms (perceived social pressure to perform or not to perform the behaviour). Family migration history may foster migration intentions by affecting its three antecedents. First, it makes migration seem less out of reach (increased perceived behavioural control). Second, a positively perceived migration experience in a family is expected to lead to more favourable evaluation of migration as a phenomenon (more positive attitudes towards migration). Third, migration experience in a family, in particular the one that was positively assessed, should facilitate understanding and support, or lessen the resistance from family members in undertaking migration (a subjective norm for migration).

Perceived behavioural control: skills

Although positive attitudes towards migration are less likely to occur by forced than by voluntary migration experience, we assume that any migration experience in a family facilitates further migration, regardless of whether the previous experience was forced or voluntary. We assume that the knowledge that one’s family members migrated in the past, reinforced by the exposure to family stories and narratives about past mobility experience (see Ní Laoire 2023), even if of forced nature, may predispose migration intentions among descendants of those who were forced to move in the past by increasing their perceptions that migration is a feasible (or survivable) option (which entails higher perceived behavioural control; higher control over one’s (im)mobility, Ní Laoire 2023). This may concern overcoming psychological barriers related to cultural and language differences, which seasoned migrants know how to handle (Bernard 2022). The positive effect on subsequent generations’ migration intentions may be present even if the family narrative on the forced displacement was negatively framed (i.e. it was portrayed as a traumatic experience) or non-existent. Such a proposition is in line with the ‘mobility as capital’ approach, viewing mobility as a resource that can be accumulated through the development of specific skills, experiences and socialisation, and inherited within families, regardless of whether past experiences were positive or negative (Moret 2018; see also Cairns 2021; Ní Laoire 2023). It presupposes that the effect can be unconscious and does not even need to be accompanied by the awareness of past family experience and of the concomitant accumulation of various types of convertible capitals (Bourdieu 1986), which are transmitted intergenerationally to later serve as ‘migration-facilitating capital’ (Kim 2018) transferable to voluntary migration. This transmission may involve both skills acquired through forced displacement that can be directly used to overcome obstacles to emigration and those that are less directly useful for migration but nevertheless make the descendants of the forcibly displaced more likely to self-select into voluntary migration. Our understanding of family migration capital thus combines the two defining features of ‘migration capital’ and ‘migration-facilitating capital’ as delineated by Kim (2018) in that we see it as the resources derived from past migration experiences in a family that can be ‘put to use for’ migration (282) in the future (cf. Bernard’s 2022 definition of ‘internal migration capital’), including by subsequent generations.

While we assume that forced migration leads to the accumulation of various types of capital (including the acquisition of certain skills) that are transferable to voluntary migration, we acknowledge that some mechanisms may be weaker in the case of a forced-voluntary relative to a voluntary-voluntary dyad. We argue that, apart from the higher likelihood of occurrence of negative attitudes towards migration, this concerns above all potentially poorer access to the transnational social capital in the case of forced than in the case of voluntary migration. The latter argument gains relevance in the studied historical-geographical context, in which the intergenerational transmission of forced-displacement-induced social capital was restricted by the large proportion of temporary and internal WWII-related moves, which to a limited extent contributed to the development of social ties abroad, and by the often limited access to social capital in the country of origin, especially when migration between the origin and the destination was restricted for years. While the advantage of having social ties abroad for increasing the feasibility of migration by decreasing the associated pecuniary and non-pecuniary costs and risks was arguably smaller in the case of forced than voluntary migration, the mechanism related to the severance of social ties and networks (Bernard 2022) might have been stronger. Assuming that ‘migration skills are developed in part by the severance of social ties’ (Bernard 2022, 128), as migrants learn (and then pass this know-how to their children and grandchildren) how to ‘deal with the difficulties of leaving and entering new social contexts’, we would expect this skills-inducing effect to be stronger in case of forced migration (especially in the studied context).

Attitudes: positive long-term effects

The statement that forced migration experience cannot be accompanied by positive attitudes towards migration may, however, be questioned. Previous literature has shown that, despite constituting a disruptive event, a negative shock, usually associated with loss and disempowerment, forced – and war-time in particular – migration may in the long run prove socially empowering. By reconfiguring social and economic opportunity structures, forced migration may generate new social opportunities, compelling those affected to re-optimise their life choices (Arellano-Bover 2022; Bauer, Braun, and Kvasnicka 2013; Lubkemann 2008; Nakamura, Sigurdsson, and Steinsson 2022; Sarvimäki, Uusitalo, and Jäntti 2009; 2022). Among descendants, the awareness of the positive long-term effects of the involuntary move their family members were involved in is expected to translate into more positive attitudes towards migration, which, in turn, positively influence migration intentions.

Subjective norms: perceptions and preferences

We assume that the long-term, intergenerational effect of forced migration experience is moderated by the presence and character of family stories. On the one hand, one may expect the effect to be less likely to occur in families which did not transmit their knowledge of the forced migration experience from generation to generation as the above-discussed mechanisms related to the locus of control and positive attitudes were not at play. On the other hand, one may argue that family stories may also have the opposite effect, undermining potential future migration intentions in the family, when they convey negative attitudes towards moving. It is also possible, however, that family stories, regardless of their valence, result in greater sensitivity to the subject and contribute to the perception of mobility as a social norm and this way facilitate subsequent generations’ migration intentions.

An additional theoretical explanation for the expected higher propensity to migrate among the descendants of forced migrants may be derived from the empirical results of Becker et al.’s (2020) study. Although the authors themselves reported no relationship between forced migration experience (equated to having ancestors from Poland’s former Eastern territories, which became part of the USSR in 1945) and the intent to emigrate, they showed that the experience of forced migration shifted the preferences of those affected from investment in material possessions towards investment in portable human capital. A higher value attributed to education than physical assets, such as real estate, transmitted from generation to generation may, in turn, translate into a greater likelihood of making the decision to leave among members of subsequent generations (as they are more likely to self-select based on education and less likely to be place-bound by their non-portable assets).

Research hypotheses

Given these theoretical premises, our study aims to verify the following research hypothesis: Experience of forced displacement among members of the preceding generations is positively related to subsequent generations’ migration intentions (H1). We assume that the effect of the experience of voluntary migration in a family includes the self-selection effect on top of the accumulation and transmission of family migration capital. In the case of forced migration, this self-selection-induced part of the observable effect is minimised, which brings us closer to isolating the true effect of family migration capital. We also formulate a hypothesis regarding the role of education: The effect of forced displacement on the migration intentions of the descendants is partly mediated by education acquired by subsequent generations after the war (H2). This is based on the assumption that forced displacement-induced preference for investment in education rather than physical assets facilitates migration intentions.

WWII-related forced population movements

World War II and the resultant border changes drove millions of people into exile, with Central, Eastern and South-Eastern European populations particularly strongly affected (Kosiński 1982). WWII-related forced displacement encompassed different types of movements (see Stola 1992) both during and after the war and differed in terms of their scope (international vs internal) and permanency (possibility of return).

Forced war-time population movements included, first, mass exodus due to direct military actions – evacuation and flight from hostilities on the part of advancing or retreating troops in the front zone and transfers of prisoners of war. Second, they included forced migrations due to war-time revisions (many of them temporary) of borders and expanding foreign rule. The evictions and expulsions from the incorporated lands involved mostly Polish but also Czechoslovak, Romanian, Hungarian, Bulgarian and Yugoslav territories (Kosiński 1982), and included, among others: several million people involuntarily sent to Germany as forced labour; Poles and Jews sent from the Polish lands incorporated into the Third Reich to the German-administered General Government; people sent to concentration camps; and the Soviet deportations from the newly annexed territories (first the Polish Eastern Provinces, then the Baltic States and Bessarabia). The intra-USSR war-time eastward flows also involved displacements from territories occupied or threatened with occupation by enemy troops to the Soviet interior, which included the evacuation of enterprises from the west of Russia (Lieberman 1983), as well as the 1941–1944 violent deportations of entire ethnic groups branded as ‘traitors to the motherland’ to Siberia and Central Asia (see Kreindler 1986), accompanied by compensatory forced migrations to their abandoned territories (Polian 2004). Many of those who (were) resettled during the war (and survived), including those freed from the German labour camps and forced labour in Nazi Germany as well as evacuees and those displaced due to temporary war-time territorial changes, returned to their homelands after the war (Kosiński 1969). Some – entire nationalities displaced within the USSR – were rehabilitated and allowed to return to their ethnic homelands only years after the war (Polian 2004).

Most of the late-war and post-war forced movements were ethnically driven and permanent – in many cases return was impossible due to the redrawing of borders and restrictions imposed on emigration from the countries of the communist bloc throughout the Cold War period, which essentially froze their inhabitants’ migration intentions for several decades. The biggest WWII-related late-war and post-war forced population movement was the mass exodus of Germans from Eastern and Central Europe (mostly from the former eastern territories of the German Reich, which Germany lost after the war) to West Germany in the wake of the Nazi defeat.² The second largest WWII-related late-war and post-war forced population movement was the displacement of over 1.5 million Poles from Poland’s former Eastern territories to the new Poland in the aftermath of WWII. Their resettlement to the newly acquired Western Territories – Silesia, Pomerania and East Prussia – started in 1944 and was accompanied by the forced displacement of almost 0.5 million Ukrainians from Poland’s southeastern territories to the Ukrainian Soviet Socialist Republic.³ Later, in 1947, under the so-called Operation Vistula, a remaining 140,000 Ukrainians (or rather Eastern Slavs) and people from mixed families were forcibly dispersed throughout the Western Territories of northern and western Poland.

Data and methods

The study draws on individual-level data from the 2016 Life in Transition Survey-III (LiTS-III), conducted by the European Bank for Reconstruction and Development in collaboration with the World Bank.⁴ The data were collected face-to-face through Computer-Assisted Personal Interviewing. The survey polled over 51,000 respondents in 34 countries (approximately 1,500 in each country), including 29 post-socialist countries of Central and Eastern Europe and Central Asia (including Mongolia but excluding Turkmenistan), as well as Turkey, Greece, Cyprus, Germany, and Italy. LiTS-III elicited information about respondents’ current migration intentions and their past family experience of forced displacement, which makes it a suitable dataset for our study. The question on WWII-related forced displacement was not asked in Italy and Cyprus and thus they were excluded from our analyses.

Current migration intentions were measured by the question on intentions to move abroad within the coming year (Do you intend to move abroad in the next 12 months?), with possible answers ‘Yes’ (4%) and ‘No’, coded by us as 1 and 0 respectively. Nearly 2.5% of respondents provided a ‘don’t know’ answer, which we also coded as 0. In two robustness checks, we (1) excluded the ‘don’t know’ answers from the analysis and (2) included them as a separate outcome of the dependent variable. The results of the corresponding binary and multinomial logit regressions, which can be found in Supplemental Tables S4 and S5, respectively, are consistent with our main results.

Forced displacement experience was measured by the retrospective question on the necessity to move as a result of World War II and was asked in relation to the respondent, his/her parents and grandparents (Did you, your parents or any of your grandparents have to move as a result of the Second World War?). Since we are interested in intergenerational effects, we limit our sample to respondents born after WWII. 10% of them provided a positive answer. Importantly, this could encompass a range of migratory experiences – both internal and international, driven directly by war as well as by governments in peacetime (post-war population exchanges and compulsions). Although we acknowledge that a distinction between voluntary and forced migration is by no means binary (Becker and Ferrara 2019; see also Bartram 2015), we rely on respondents’ self-perception of the move. 12% of respondents provided a ‘don’t know’ answer and a further 2% refused to answer; we create two dummy variables for these categories.

In line with the existing literature on the determinants of migration intentions, we controlled for: gender, age, education level (primary, secondary, tertiary), being married, employment status (proportion of time employed in the 12 months before the interview), self-perceived position on a 10-step wealth ladder, ethnic minority status, self-assessed health level, and living in an urban area. In addition, we controlled for respondents’ past international migration experience – an important factor enabling future migration (Bernard and Perales 2021; Bernard and Vidal 2020; De Jong 2000). The survey contains information on whether respondents had moved from abroad to their current place of residence, completed secondary education abroad or were born abroad.⁵ We combined these pieces of information to create a categorical variable that is equal to one if at least one of the aforementioned experiences was mentioned (we also include a dummy for missing answers).⁶ The definitions and summary statistics of all variables included in the analysis are provided in Supplemental Table S1.

We used logistic regression with country-fixed effects to model the relationship between the preceding generations’ experience of forced displacement and the respondent’s current migration intentions in our baseline results. Country-fixed effects enable us to account for all observed and unobserved country-level influences that might affect people’s migration intentions. As a robustness check, we additionally estimated mixed-effects (random intercept) models with 2-level clustering (individuals nested within countries) and 3-level clustering (individuals nested in primary sampling units, which are nested within countries, assuming that the experience of forced displacement sometimes concerned whole localities or some specific regions more than others). Additionally, we conducted a drop-one test to make sure that our full-sample results are not driven by a particular country. We also performed two further robustness checks, by excluding from the sample: (1) respondents with prior international migration experience (likely to be subject to various selection processes) and (2) ethnic minorities (whose ancestors might have been targeted with forced displacement based on their ethnic belonging). These robustness checks (see Tables S6 and S7, Figure S3⁷ and Table S8 in the supplement) yielded results consistent with our baseline results.

To address hypothesis H2 and test for the mediating role of education, we estimated a parallel–serial mediation model, with father’s and mother’s education as primary (parallel first-step) mediators and respondent’s own education as a secondary (serial second-step) mediator (see Figure S1 in the supplement), assuming that the experience of WWII-related forced displacement might have influenced not only respondents’ but also their parents’ educational choices.⁸ Additionally, for the oldest respondents (born before 1960), we estimated the indirect effect using a simple mediation model with respondent’s own education level as a single mediator (see Figure S2), assuming that the educational choices of the parents (due to their age when the war ended) were less likely to have been influenced by the experience of forced displacement. The significance of indirect effects was tested through a bootstrapping procedure (Hayes 2022), including 5,000 bootstrap samples and 3 as the random seed number.

Results

Table 1 reports our baseline results. Specifications (1.1)-(1.2) include only the regressor of interest – the set of indicator variables capturing the experience of WWII-related forced displacement in the family – and country-fixed effects. The results suggest that, relative to having no family history of WWII-related forced displacement, reporting such history increases the likelihood of having migration intentions by 1.6 percentage points (p < 0.01) for the no-control full sample and 1.6 percentage points (p < 0.01) for the no-control main estimation sample, and of providing a ‘don’t know’ answer by 1.2 percentage points (p < 0.01) and by 1.3 percentage points (p < 0.01), respectively.

Table 1. WWII-related forced displacement in the family and migration intentions.

	No controls, full sample		No controls, main estimation sample		Controls, excluding international migration experience, main estimation sample		Controls, including international migration experience, main estimation sample
	(1.1)		(1.2)		(1.3)		(1.4)
	Odds ratio	Marginal effect	Odds ratio	Marginal effect	Odds ratio	Marginal effect	Odds ratio	Marginal effect
WWII-related forced displacement in the family
No	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
Yes	1.466***	0.016**	1.470***	0.016**	1.544***	0.019***	1.542***	0.019***
	(0.161)	(0.005)	(0.163)	(0.005)	(0.172)	(0.005)	(0.172)	(0.005)
Don’t know	1.338**	0.012**	1.368***	0.013**	1.143	0.005	1.140	0.005
	(0.119)	(0.004)	(0.123)	(0.004)	(0.105)	(0.004)	(0.105)	(0.004)
Refusal to answer	1.152	0.005	1.254	0.009	1.177	0.006	1.162	0.006
	(0.234)	(0.008)	(0.261)	(0.009)	(0.256)	(0.009)	(0.252)	(0.009)
Age					0.959***	−0.002***	0.959***	−0.002***
					(0.003)	(0.000)	(0.003)	(0.000)
Female					0.726***	−0.012***	0.724***	−0.012***
					(0.048)	(0.003)	(0.048)	(0.003)
Married					0.809**	−0.008**	0.808**	−0.008**
					(0.057)	(0.003)	(0.057)	(0.003)
Secondary education					0.920	−0.003	0.946	−0.002
					(0.127)	(0.005)	(0.135)	(0.005)
Tertiary education					1.301	0.011	1.341	0.012*
					(0.197)	(0.006)	(0.209)	(0.006)
Higher on wealth ladder					0.875***	−0.005***	0.876***	−0.005***
					(0.021)	(0.001)	(0.021)	(0.001)
Employed					0.876	−0.005	0.877	−0.005
					(0.066)	(0.003)	(0.066)	(0.003)
Lives in urban area					1.376***	0.012***	1.368***	0.012***
					(0.096)	(0.003)	(0.095)	(0.003)
Ethnic minority					1.328**	0.011**	1.288**	0.010**
					(0.120)	(0.003)	(0.122)	(0.004)
Better health					1.079	0.003	1.082	0.003
					(0.048)	(0.002)	(0.048)	(0.002)
International migration experience
No							Ref.	Ref.
Yes							1.274	0.010
							(0.175)	(0.006)
Missing answer							1.186	0.007
							(0.142)	(0.005)
Country-fixed effects
Constant	0.054***		0.054***		0.384**		0.365**
	(0.008)		(0.008)		(0.122)		(0.120)
Pseudo R^2	0.064		0.065		0.118		0.118
BIC	14,261.3		14,051.7		13,376.3		13,390.6
N	41,977		40,966		40,966		40,966

Note: * p < 0.05, ** p < 0.01, *** p < 0.001. The table reports binary logit odds ratios and (average) marginal effects. Standard errors are in parentheses. Dependent variable: Intentions to migrate (0 = ‘No’ or ‘Don’t know’, 1 = ‘Yes’). Survey weights that add up to the number of interviews conducted are applied in all regressions. For complete econometric output, see Table S3 in the online supplement.

Specification (1.3) adds control variables. The magnitude of the forced-displacement-in-the-family dummy now amounts to 1.9 percentage points (p < 0.001), while the marginal effect of the ‘don’t know’ category drops in magnitude and becomes statistically non-significant. The latter could be explained by the greater likelihood of specific groups of respondents (e.g. the younger and less educated) to provide a ‘don’t know’ answer:⁹ once we include these characteristics as controls, the ‘don’t know’ category loses significance.

It is useful to assess the magnitude of the effect of the key regressor relative to that of other controls and to the dependent variable. Thus, the increase in the likelihood of reporting migration intentions by 1.9 percentage points from having a history of forced displacement in the family, is equivalent to the effect of reducing the respondent’s age from 37 to 27 or reducing their self-perceived wealth from 7 to 3 (on a 10 step ladder). Evaluated against the sample mean of the dependent variable, a history of forced displacement in the family is associated with a 49% (0.019/0.039) increase in the likelihood of reporting migration intentions. These comparisons highlight the substantive significance of the association between a family history of forced displacement and respondents’ migration intentions.

Next, we want to control for respondents’ previous international migration experience (specification (1.4)). The results suggest that the association between international migration experience and migration intentions is statistically non-significant.¹⁰ The marginal effect of the regressor of interest remains unchanged (both in terms of magnitude and statistical significance).

Table 2 reports the results of a mediation analysis. It demonstrates that overall the history of forced displacement in the family had a positive indirect effect on migration intentions through parental and respondent’s education operating in sequence (specification (2.1)). In particular, the analysis shows that there are five pathways to migration intentions (cf. Figure S1): a direct one from forced migration experience in a family (b = 0.340, SE = 0.088; 95% CI [0.167; 0.514]); an indirect one through father’s education (b = 0.022, SE_boot = 0.009, 95% CI [0.005; 0.039]); an indirect one through respondent’s education (b = 0.007, SE_boot = 0.004, 95% CI [0.001; 0.015]); an indirect one through mother’s and respondent’s education (b = 0.004, SE_boot = 0.002, 95% CI [0.000; 0.008]) and an indirect one through father’s and respondent’s education (b = 0.005, SE_boot = 0.002, 95% CI [0.000; 0.009]). We did not, however, find a significant indirect effect of forced migration experience via mother’s education (b = 0.004, SE_boot = 0.010, 95% CI [−0.015; 0.023]). For older respondents (born before 1960; specification (2.2); cf. Figure S2), for whom the serial paths with parental education might not be relevant due to their parents’ age, we observed a positive indirect effect of WWII-related forced displacement in the family on migration intentions via education (b = 0.057, SE_boot = 0.022, 95% CI [0.016; 0.103]).

Table 2. Direct and indirect effects of WWII-related forced displacement in the family on migration intentions through education.

WWII-related forced displacement in the family = Yes^a	Effect	SE/Bootstrap SE	95% CI/bootstrap CI
(2.1): whole sample (N = 39,138)
Relative direct effect	0.340	0.088	0.167–0.514
Relative indirect effects:
IV → mother’s education → DV	0.004	0.010	−0.015––0.023
IV → father’s education → DV	0.022	0.009	0.005–0.039
IV → respondent’s education → DV	0.007	0.004	0.001–0.015
IV → mother’s education → respondent’s education → DV	0.004	0.002	0.000–0.008
IV → father’s education → respondent’s education → DV	0.005	0.002	0.000–0.009
(2.2): born before 1960 (N = 11,560)
Relative direct effect	0.056	0.254	−0.442–0.554
Relative indirect effect:
IV → respondent’s education → DV	0.057	0.022	0.016–0.103

^a Comparison group: WWII-related forced displacement in the family = No.

Note: IV: WWII-related forced displacement in the family; DV: migration intentions; both parental and respondent's own education level are measured on a continuous (1-8) scale; age, gender, marital status, subjective wealth, employment status, living in an urban area, ethnic minority status, self-perceived health and country are controlled in relation to the dependent variable but not the mediators; direct and indirect effects are on a log-odds metric. We run mediation models with the PROCESS macro (v4.3; Hayes 2022) in SPSS 29. For specification (2.1), we estimated PROCESS model 80 (see Figure S1 for a conceptual diagram) and for specification (2.2) – PROCESS model 4 (see Figure S2). Bootstrap confidence intervals for the specific indirect effects were generated using the percentile method. Indicator coding was used to represent the multicategorical DV.

Discussion

This study has focused on the potential role of past family experience related to migration and mobility, in a specific situation when this experience had an involuntary nature. Specifically, we seek to answer the question of whether an experience of forced displacement in a family may influence the current migration intentions of its members. Building on research on the potential of the intergenerational transmission of migration behaviour (Ivlevs and King 2012; Myers 1999), we investigate whether the descendants of people who experienced forced displacement as a result of WWII are more likely to report an intention to migrate than people in similar circumstances but without this kind of family experience. Our results provide evidence that – in line with our first hypothesis – a family history of forced displacement is indeed positively linked with respondents’ current migration intentions.

While our data do not allow us to test the underlying mechanisms empirically, in theoretical terms, we interpret this finding as the intergenerational transmission of ‘family migration capital’ that accumulates within a family regardless of whether past moves were voluntary or forced. We argue that family members involved do not need to be aware of the accumulation and transmission of this capital for an effect to occur, but it is expected to be stronger in families, in which this awareness is present and reinforced by family narratives. In the latter case, we explain it through the higher likelihood that descendants of forced migrants, familiar with their family experience and exposed to family stories and narratives about migration, perceive migration as a feasible (or survivable) option. We also argue that this perception may be reinforced by the awareness that, when viewed in the long run, an involuntary move has positively influenced the life chances of one’s family members. Last but not least, we contend that the higher migration propensity among the descendants of forced migrants may be linked to the intergenerationally transmitted forced-migration-induced higher preference for investment in portable human capital rather than in physical assets (Becker et al. 2020). Importantly, all these should be viewed as complementary rather than competing mechanisms. We also found some tentative support for our second hypothesis, stating that the effect of forced displacement on migration intentions is (partly) mediated by education, which suggests that the mechanism connected to investment in portable human capital may indeed be at play (complementary to other mechanisms we describe, though not decisive).

Since the forced character of past family WWII-related moves limits self-selection into migration, our findings provide stronger evidence for the existence of ‘family migration capital’ than found in previous studies predicting current migration intentions from any family migration experience, and hence to a large extent from the experience of voluntary migration. Thanks to a more exogenous character of a forced move, we can be more confident that the effect is due to the past migration experience and not pre-existing differences between individuals (e.g. in terms of social, cultural or economic capital) that have been transmitted intergenerationally.

Our study has some limitations. First, the LiTS-III question that we used to measure the experience of WWII-related forced displacement left respondents room for different interpretations. Firstly, the formulation ‘have to’ may refer, in people’s view, to mass forced displacement, when their ancestors were deprived of any alternatives, as well as to more voluntary decisions, which however might have been interpreted as taken under a narrow opportunity structure. Secondly, respondents might have accounted for both international and internal moves. Although theoretically the survey provides information on the location of respondents’ mother’s and father’s families in 1939 (a country and a region), which we hoped to be able to use to distinguish between international and internal moves, the quality of these data does not allow doing so.¹¹ What the comparison of these limited data with respondents’ place of birth/current location suggests, is that most of the reported forced moves occurred within country borders. One may thus ask why the experience of internal forced displacement should translate into a higher propensity to migrate internationally. Existing literature provides some clues why such a proposition is plausible: internal migration experience has been shown to be positively associated with the odds of subsequent onward international migration (Bernard and Perales 2021; see also King and Skeldon 2010 and Bernard and Vidal 2023), in particular past personal experience of forced internal displacement proved to be positively related to current intentions to migrate internationally (Efendic 2016). Moreover, as Myers (1999) argues, the socialisation model does not presuppose that the similarity between parents and children in terms of their mobility behaviour depends on the scale of the movement (internal, international, interstate). The transferability from internal to international moves in the case of internal moves of a forced nature may be additionally motivated by the fact that they are more likely to entail a feeling of uprootedness, which often accompanies an international move, and by the fact that people forced to move might have ended up in locations that might not be optimal for them (Becker and Ferrara 2019). Besides, the distinction between international and internal movements with regard to WWII-related flows is not particularly useful, preventing the assessment which of the two is expected to bear a larger effect. A substantial share of the intra-USSR WWII-related forced migration flows, although formally taking place within state borders, involved movements over several thousand kilometres, to distant territories characterised by different socio-environmental, including climatic, conditions. Other WWII-related population movements in Central, Eastern and South-Eastern Europe were, in turn, formally international in scope at the time when they occurred, even though they were confined to the territory of a single state from before territorial changes (e.g. in the case of Yugoslavia, Kosiński 1982). All this explains why we build on past research concerning both internal and international migration as well as residential mobility in our study.

Second, although self-selection into mobility has been shown to be smaller in war-induced than in voluntary mobility (Ruhe and Kuhnt 2023), the mechanism described by Myers (1999) as a status-inheritance model – assuming that the similarity in mobility behaviours between parents and their adult children is due to common characteristics that are associated with mobility – cannot be completely excluded. Even though it is generally acknowledged that agency by forced migration decisions tends to be more limited than by voluntary moves, recent research highlights that self-selection also occurs in conflict-induced migration (Birgier et al. 2018; Guichard 2020; Oruč 2009), and individuals forced to move should not be viewed as passive victims without agency (Bohra-Mishra and Massey 2011; Seven 2022; see also Bartram 2015). Some of the characteristics based on which forced movers self-select may be transmitted intergenerationally and consequently affect the migration intentions of their descendants.¹² Large part of the WWII-related forced migration flows had, however, a universal character, which minimises self-selection into migration. Moreover, we rely on family members’ self-perception concerning the involuntary character of the move.

Third, political violence is often not randomly targeted (Czaika and Kis-Katos 2009; Ibáñez and Vélez 2008; Lupu and Peisakhin 2017). Victims may be targeted, for instance, based on their ethnicity, religious denomination, political views, wealth or education level. Some individuals within the population are thus more likely to be targeted and consequently to experience forced displacement. While a considerable part of the WWII-related forced population movements have a ‘total character’ (e.g. evacuations), some were more selective, for instance, Soviet deportations of former Polish citizens to the USSR’s interior during the first Soviet occupation of 1939–1941 (Eberhardt 2011; Gross 2002 (1988)) or deportations from the Baltic States and Moldavia, which were to a large extent targeted at the old local elites labelled as ‘class enemies’ (Polian 2004). If the characteristics based on which people are targeted are generally regarded as conducive to migration and at the same time are subject to intergenerational transmission, these inherited characteristics may be partly responsible for a higher likelihood of declaring migration intentions among victims’ descendants.

Forth, the type (e.g. evacuations, expulsions, deportations) and character of WWII-related forced movements differed not only in terms of selectivity and randomness of being targeted but also other circumstances surrounding them that through different feelings that they evoked in retrospect (e.g. shame of being ‘on the wrong side of history’ or trauma) might have influenced the existence and the character of post-mobility family stories. The latter might have also been influenced by the fact whether forced migrants managed to return to their pre-displacement places of residence. In the case of families whose members have not managed to return to their pre-WWII places of residence despite wanting to do so, family narratives can be shaped by the ‘myth of return’, potentially facilitating subsequent generations’ – return – migration aspirations. Although LiTS-III provides information on the location of respondents’ mother’s and father’s families in 1939 and on respondents’ intended country of destination, these data have many missing values, which along with other flaws of data on pre-1939 locations (discussed in endnote 11) prevents more fine-grained analyses. While cross-tabulation matrices of WWII-related forced displacement in a family by intended destination country for each country yield very small entries, no patterns suggesting that the mechanism connected to the ‘myth of return’ is at play are visible. It should also be borne in mind that many forcibly displaced as a result of WWII returned after the war, while the involuntarily immobile of the communist period had over a quarter of a century after emigration restrictions had been lifted to realise their ‘return’ migration aspirations (until 2016, when the survey was conducted). Moreover, even among those who or whose ancestors have not returned and who intend to migrate to their families’ pre-war places of residence, not all consider their intended migration in terms of a ‘return’. They often use their ‘roots’ pragmatically by destination choice, e.g. as a means of simplified access to residence and the labour market (especially relevant by the current non-EU–EU origin–destination dyads), which blurs the boundary between ‘return’ and ‘onward’ migration. Unfortunately, our data do not allow us to distinguish between different types of WWII-related forced movements nor account for their temporal character. While we expect temporary forced moves to be less consequential for subsequent generations’ migration intentions than long-term or permanent forced moves (cf. Efendic 2016), further research is required to establish this. To address the above limitations, future studies would need to acquire more detailed data on family migration history, including the trajectory of the war-induced movement and the circumstances that accompanied it as well as any migration experience following it.

Fifth, we rely on subsequent generations’ accounts of forced displacement experienced by their ancestors, which are determined by the awareness of the experience and largely dependent on the existence of family stories, and may be subject to recall bias. To be able to fully capture the effect of past forced displacement on subsequent generations’ voluntary migration intentions, including the potential ‘unconscious transmission’ of migration capital, one would need to have access to objective data on people’s past family experience of forced displacement.

Finally, LiTS-III data do not enable us to reliably distinguish between the second- and third-generation forced migrants, which would allow more informed inferences about the persistence of family migration capital. To address this question, we used age categories, adopting the 1970 year of birth threshold (for the results, see Supplemental Table S9). The marginal effect proved to be smaller for the older than the younger subsample of respondents (0.9 vs 2.6 percentage points), which is linked to generally lower migration intentions among older individuals. When evaluated against subsample means of the dependent variable (0.020 vs 0.063), the increase in the likelihood of reporting migration intentions associated with a history of forced displacement in the family seems to be comparable for respondents born before 1970 (45%) and those born after 1970 (41%). This suggests that intergenerational transmission of migration capital spans more than two generations. Nevertheless to be able to test this in a more rigorous manner (as well as to more accurately test the mediating role of education, see endnote 8), one would need more detailed data on respondent’s parents and grandparents (regarding their year of birth, education and experience of forced displacement).

Despite these limitations, the paper adds to the current body of literature on family influences on migration intentions and behaviour (Brunarska and Ivlevs 2022; Clifton-Sprigg 2022; De Jong and De Valk 2023; Francisco-Menchavez 2020; Ivlevs and King 2012; Massey and Espinosa 1997; Myers 1999), by showing that the positive relationship between past family mobility experience and current migration intentions also holds when the past mobility experience was involuntary and thus providing stronger support for the existence of ‘family migration capital’. It also contributes to the scholarship on the long-term consequences of forced displacement for families affected, in particular the long-lasting legacies of WWII-related forced migrations (Bauer, Braun, and Kvasnicka 2013; Becker et al. 2020; Sarvimäki, Uusitalo, and Jäntti 2022), by demonstrating that forced displacement may have long-lasting consequences for future, voluntary migration flows, determining subsequent generations’ migration intentions. Since migration intentions are not a perfect predictor of an actual move (Bradley et al. 2008; De Jong 2000), a promising avenue for further research is the identification of the potential role of past family experience of forced displacement in driving subsequent generations’ migration behaviour. Even though not all intentions become realised – either because they turn into a preference to stay or due to the lack of an ability to realise them (Carling 2002; Carling and Schewel 2018) – they are, along with planning, considered the closest antecedent of migration behaviour (De Jong 2000; cf. the nature of the mindset, Carling and Mjelva 2021). This is why we expect the effect of family experience of forced displacement to be also observable for subsequent generations’ migration behaviour, which warrants further investigation. Although past studies demonstrated that determinants of migration intentions might diverge from determinants of migration behaviour (Brzozowski and Coniglio 2021; Carling 2002), family migration history ‘pull[s] in the same direction’ (Carling 2002). Since past personal migration experience proved to be positively related to both the formation and realisation of migration intentions (Bernard, Kalemba, and Nguyen 2022; De Jong 2000), we expect that similarly intergenerational transmission of family migration capital increases migration aspirations, which lead to migration intentions, which precede migration behaviour. Finally, future research may also find it worthwhile studying whether the effect of past migration is indeed stronger in the case of voluntary than in the case of forced migration as predicted by our theoretical framework. Importantly, our findings suggest that any type of migration, voluntary or forced, can generate an intergenerational snowball effect, when past migration facilitates further migration through the transmission of family migration capital. This mechanism contributes to the self-perpetuating character of migration (De Haas 2010) and gains special relevance in light of the growing forced migration flows globally.

Acknowledgments

We wish to thank the editors and two anonymous reviewers for their valuable feedback on an earlier draft of this paper.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The survey data that support the findings of this study are available from the EBRD website (exact link provided in the paper). Replication code for the analyses conducted in Stata, allowing the recreation of all the operations on the original survey data, is available at the osf.io platform (doi: 10.17605/OSF.IO/4K9ZH). The paper provides all the necessary details to recreate the mediation models estimated with the SPSS PROCESS macro.

Additional information

Funding

This work was supported by the Polish National Agency for Academic Exchange under the Bekker NAWA Programme [BPN/BEK/2021/1/00226].

Notes

1 Aslany et al. (2021) treat migration aspirations as an umbrella category encompassing terms such as desires, intentions, plans and expectations.

2 Earlier, in 1939-1941, 0.5 million ethnic Germans from all over Europe, many of them ‘socially-pressured’ or fear-driven, were resettled to Nazi Germany as part of the ‘Heim ins Reich’ programme (Ahonen et al. 2008).

3 Post-war population exchanges also involved agreements between the USSR and Czechoslovakia and Czechoslovakia and Hungary.

4 https://www.ebrd.com/what-we-do/economic-research-and-data/data/lits.html.

5 Respondents were asked how long they had lived in their current place of residence. Those who provided an answer other than ‘the whole life’ were asked three subsequent questions: where they had moved from, where they were born, and where they had completed secondary school.

6 For additional insights, we also run models in which we account for each of the three international migration experiences separately and for all of them at once but incorporating them as three dummy variables (see Table S10). Our results remain unchanged in these four additional specifications.

7 No single country proved to fully drive the positive, statistically significant association between the variables of interest found in the full sample. Having said this, Belarus, Ukraine and Poland appear to push up the magnitude of the association. When they are excluded all at once, the coefficient drops by half but is still significant at the 95% level.

8 Unfortunately, our data contain neither information on the year of birth of respondent’s parents and grandparents nor the time when they acquired their highest education level. We also do not have information on the level of education of respondent’s grandparents, which along with parental and respondent’s education level might be used to capture the potential educational social mobility. Another problem is that we do not have information on whether forced displacement was experienced by the family of respondent’s father or mother, while the parallel-serial mediation model contains both paths. Thus, the obtained results should be treated with caution.

9 The multinomial logit regression, in which the outcomes of the WWII mobility variable are regressed on the set of controls, confirms that younger and less educated respondents are more likely to provide a ‘don’t know’ answer.

10 This result may seem surprising. We think it may be attributed to the quality of the international-migration-experience dummy, which is based on fragmentary information on personal migration histories. In the case of respondents who were born and completed their secondary education in their current country of residence, this measure unfortunately does not account for international moves, if they happened before the last internal move.

11 First, there were many missing values on these variables. Second, in the case of some of the younger respondents, the respective questions: In which country (in 2015 borders) and geographical region did your mother's/father’s family live in the beginning of 1939? If you do not know the region, specify the closest city to the place where your mother's/father's family lived in the beginning of 1939. could not be answered unequivocally, as their mother’s/father’s families at that time might have involved families of a respective grandfather and a grandmother living in two different countries. Third, apparently some respondents referred to 1939 and not 2015 borders when replying to these questions (as revealed by answers involving pairs such as: Poland and Vilnius/Trakai/Ignalina) and the regional variables contain many more missing values than the country-related ones so they cannot be effectively used to verify whether a respondent referred to pre-WWII rather than 2015 boundaries.

12 In our sample, respondents who reported a family history of WWII-related forced displacement are more likely to belong to an ethnic minority group and to have higher education levels – both their own and parental – than people without such history (see Supplemental Table S2). There could be several reasons for the latter: investment in education by the forcibly displaced after displacement (Becker et al. 2020) but also self-selection into forced displacement on the basis of education (Birgier et al. 2018; Guichard 2020) and a greater likelihood of falling victim to violence on the basis of education (Eberhardt 2011; Polian 2004). By bivariate comparisons, these differences with regard to education level can also be driven by age.