Inter-regional migration among international migrants in Sweden: Patterns, drivers and destination

Louisa Vogiazides, Stockholm University

A growing literature examines the mobility of international migrants within their host country. In Sweden, just as in other migrant-receiving countries, the majority of studies focus on immigrants’ short-distance residential mobility within urban areas, assessing its role in shaping residential segregation. The long-distance mobility of the foreign-born population is less prominent in empirical research. To address this gap, the present study explores the patterns, determinants and destination of inter-regional moves among recent immigrant cohorts in Sweden. More specifically, it examines immigrants’ mobility away from their first region of settlement, so-called ‘secondary migration’. The study uses longitudinal geo-coded individual-level register data to follow the residential patterns of newly arrived immigrants over a twelve-year period, from 2004 to 2015. The study proceeds in two steps. First, it applies discrete-time event history analysis to analyze the propensity and timing of secondary migration. In a second step, the study will apply discrete choice modelling to study the destination choices of secondary migrants. The study employs an innovative method to measure the neighbourhood context based on individualized neighbourhoods calculated with the newly developed software EquiPop. EquiPop uses the geo-coded data on individuals’ residence to compute their k-nearest neighbours, where k represents the researcher-defined count of surrounding individuals. The great advantage of this method is that it bypasses the so-called Modifiable Areal Unit Problem (MAUP), which is a source of statistical bias affecting any geographical analysis based on arbitrarily defined aggregations of geographical areas.

Introduction
A growing literature examines the mobility of international migrants within their host country. In Sweden, just as in other migrant-receiving countries, the majority of studies focus on immigrants’ short-distance residential mobility within urban areas, assessing its role in shaping residential segregation. The long-distance mobility of the foreign-born population is less prominent in empirical research. To address this gap, the present study explores the patterns, determinants and destination of inter-regional moves among recent immigrant cohorts in Sweden. More specifically, it examines immigrants’ mobility away from their first region of settlement, so-called ‘secondary migration’. Applying event history analysis and discrete choice modelling on Swedish longitudinal register data, the paper analyses the propensity, timing and direction of secondary migration flows.

Prior research
Research on the determinants and effects of internal migration patterns among international migrants in Sweden is rather scarce. A study of internal migration flows during the period 2005-2008 reveals that foreign-born individuals are more likely to undertake long-distance moves compared to natives (Andersson 2012). This trend remains even when controlling for a range of mobility-related socioeconomic variables, which raises the question of whether differences in internal migration propensities could be due to cultural characteristics. Another explanation can be the fact that migrants lack so-called ‘location-specific insiders’ advantages’, such as work and family ties, which constitute disincentives to move (Fischer & Malmberg 2001).

Consistent with life-course literature, Andersson found that migrants who are young (in their mid-twenties) and single are more prone to relocate internally. Female migrants and recipients of social allowances are also more mobile. Migrants’ propensity to move over long distances also decreases with the duration of their stay in Sweden (Andersson 2012).
Beside personal characteristics, the characteristics of the neighbourhood of residence also influence migrants’ likelihood to move internally. Residents of immigrant-dense neighbourhoods are indeed less likely to move over long distances compared to residents of native-dominated neighbourhoods (Ibid.). A recent study has also revealed that foreign-born individuals are less likely to relocate from municipalities where there is a large presence of other immigrants from the same region of origin (Aradhya et al. 2017).

Empirical design, data and methods
This paper employs register data from GEOSTAR, a full-population database compiled by Statistics Sweden containing longitudinal, geo-coded, annually updated and individual-level data including a wide range of demographic, socioeconomic and residential variables. Secondary migration is defined as a move away from the Labour Market Area (LMA) of first settlement. LMAs are annually-constructed by Statistics Sweden based on commuting zones .
The study uses longitudinal geo-coded individual-level register data to follow the residential patterns of newly arrived immigrants over a twelve-year period, from 2004 to 2015. The risk set consists of all foreign-born individuals who immigrated (for the first time) to Sweden between 2003 and 2013 and are at working age (between 25 and 64 years-old) during the follow-up period (N=407,967). The study proceeds in two steps. First, it applies discrete-time event history analysis to analyze the propensity and timing of secondary migration. In a second step, the study will apply discrete choice modelling to study the destination choices of secondary migrants.
The study employs an innovative method to measure the neighbourhood context based on individualized neighbourhoods calculated with the newly developed software EquiPop. EquiPop uses the geo-coded data on individuals’ residence to compute their k-nearest neighbours, where k represents the researcher-defined count of surrounding individuals. The great advantage of this method is that it bypasses the so-called Modifiable Areal Unit Problem (MAUP), which is a source of statistical bias affecting any geographical analysis based on arbitrarily defined aggregations of geographical areas (Östh et al 2014).

Preliminary results
Long-distance mobility is generally a rare phenomenon, as the vast majority of people tend to stay in their region of residence. The case of the secondary migration of foreign-born individuals is no exception. The hazard rate of experiencing secondary migration is very low. On average, only 2.57% of the foreign-born individuals in the risk set move away from their first region of settlement every year. The hazard rates of secondary migration are significantly different according to the type region of initial settlement. It is more than four times higher among individuals who first settled in the countryside compared to those who settled in a metropolitan region. Rates of secondary migration also differ according to individuals’ type of immigration permit and region of origin. Students and refugees are more likely to leave their first region of settlement compared to individuals who immigrated for work or through family ties. Individuals from the Middle East and Africa also have a higher tendency to relocate compared to other immigrant groups.

The longitudinal dataset used in this study allows examining not only the occurrence of secondary migration but also its timing. Ten years after settlement, individuals have a survival rate in their first region of settlement of 0.8.

Presented in Session 1221: Internal Migration and Urbanization