Divided Cities

Understanding Intra-urban Inequalities

Report

19 May 2018

Download PDF

Chapter 4. Spatial segregation of migrants in EU cities1

Fabrizio Natale

Marco Scipioni

Alfredo Alessandrini

Abstract

This chapter provides a broad comparison of residential distribution and segregation of immigrants in Europe, covering around 45 000 local administrative units in 8 EU member states (France, Germany, Ireland, Italy, the Netherlands, Portugal, Spain, and the United Kingdom). The analysis is based on a map of immigrant population with an unprecedented spatial resolution (i.e. cells) of 100 m by 100 m. Having discussed the importance of the local dimension for migrants’ integration, the chapter then describes the method developed to create maps and presents empirical results on, respectively, the concentration, diversity and segregation indexes across cities of destination and countries of origin. The penultimate section presents the results on possible drivers of the observed segregation indexes. The last section concludes summing up the main results and outlining possible future avenues of research.

Introduction

This chapter compares the residential distribution and segregation of immigrants in Europe in 2011, covering around 45 000 Local Administrative Units (LAU)2 in 8 EU member states (France, Germany, Ireland, Italy, the Netherlands, Portugal, Spain and the United Kingdom). The analysis is based on a map of immigrant population (defined by either country of birth or nationality) with an unprecedented spatial resolution (i.e. cells) of 100 m by 100 m. This is a significant contribution to an academic literature on segregation which so far has relied mainly on case studies or limited comparisons, with a few exceptions (Musterd 2005b; Glikman and Semyonov 2012).

Common knowledge suggests that the great majority of migrants tend to gravitate around large cities (Diaz Ramirez et al. 2018; Hardman 2008; Sanderson et al. 2015; OECD, 2018). However, this chapter nuances this assumption by better specifying the relationship between the size of the city and the concentration of migrants. The attraction of large cities becomes less evident in the case of Portugal, Spain and Italy. This might be motivated by low skilled migration working in the agriculture sectors (Okólski, 2012), and retirement migration (Betts, 2011, 133–52; Bade and Eijl, 2011).

This chapter defines diversity based on two criteria: the number of countries of origin present in a defined territory; and how evenly the different migrants and native communities are distributed in the resident population. While there has long been interest in minorities’ distribution and concentration in cities – the most recent example being the burgeoning field of research that revolves around the notion of superdiversity (Vertovec, 2007; Meissner and Vertovec, 2015) – this chapter finds that ethnic diversity is not only an attribute of large cities like Rotterdam and Berlin, but also of less-known medium and small size towns.

Furthermore, borrowing from the academic literature, in this chapter segregation is unpacked in two dimensions: clustering and isolation. The analysis on the combination of information on segregation of LAUs and countries of origin reveals that the level of clustering increases with the size of the population. Clustering is higher in general for migrants from non-EU countries, for migrants from South America and South-East Asia and for specific countries of origin which have a recent history of conflicts.

The results from the descriptive analyses are confirmed by regression models on the determinants of segregation. These models show that migrants coming from distant countries are more likely to be segregated with respect to migrants from neighbouring countries and that the diversity of the city has a positive relationship with segregation. The model also confirms a higher likelihood for segregation for groups of migrants with a high share of refugees.

Overall, our analysis tentatively corroborates empirically the observation that disadvantaged migrants tend to cluster at arrival and that their segregation often prevents them from an outwards (spatial) and upwards (socioeconomic) mobility. However, to provide a firmer claim in this regard, we would need longitudinal data which is currently missing.

There are two main limitations in the analysis. First, the data assembled are only relative to the 2011 Census and therefore do not allow to explore the evolution of segregation over time. Second, in order to provide an exhaustive picture of the complex phenomenon of residential segregation and its drivers, demographic data should be coupled with socioeconomic data, at the same level of analysis, something which is currently lacking due to data limitations.

The reminder of the chapter is structured as follows. The next section discusses the importance of the local dimension of integration from a conceptual point of view, and further elaborates on different segregation dynamics and models of integration. The third section discusses how data used in the empirical analysis was assembled and processed, and discusses methodological choices regarding the measurement of spatial segregation. The fourth section presents the results for concentration of migrants across cities, ethnic diversity in cities, segregation across cities and origins, and possible drivers of segregation. Finally, the last section concludes.

Literature review

The importance of the local dimension of integration

According to the data of the United Nations Population Division (United Nations, 2017), in 2017 in the world there were around 3.4% of migrants considering the foreign born criterion. The ratio of immigrants against the total population in EU Member States was between 1.66% in Poland and 45.19% in Luxembourg, and between 0.54% in the Slovak Republic and 13.82% in Latvia when considering only immigrants from non-EU countries.

These aggregated figures at the national level cloak the high diversity in the distribution of migrants across cities and regions within countries. It is a long acquisition of the literature that migrants tend to concentrate in cities (Sanderson et al., 2015; International Organization for Migration, 2015; Wright, Ellis, and Reibel 2008; OECD 2016a). In this light, it becomes essential to gather information on the geographical distribution of migrants at the local level, particularly when assessing both the impact of migration on the receiving societies and the outcomes of migrants’ integration into the social fabric.

This study answers to the need of better understanding local integration dynamics and challenges by providing an accurate picture of immigrants’ distribution across European cities and towns, the diversity in these geographical entities, and whether immigrants are residentially segregated therein. The literature on diversity has been extremely prolific in recent decades. For the purpose of this analysis, the focus will be on the variety of countries of origin in a given city or town and how evenly they are concentrated. While segregation has been defined in different ways and applied to different contexts (Iceland, 2014), here segregation is described as based on two dimensions, namely clustering and isolation. The former dimension captures how the ethnic groups are concentrated or evenly distributed in space. The latter dimension of isolation/exposure considers the spatial composition of the surrounding regions of each group. As a note of caution, the concept of segregation lends itself to widely different – and at times unpredictable – political uses. Further, the wide variety of methods used to measure it, as well as conceptual differences, may results in different results with diverging policy implications (Peach, 2009).

The literature on the subject has suggested that residential location has a strong influence on migrants’ integration opportunities (e.g. Musterd, 2005; Friedrichs, Galster and Musterd, 2003). The expectation emerging from the literature is to observe a “strong relationship between social process and spatial pattern”, where “highly segregated groups are unassimilated while assimilation is correlated with a high degree of residential diffusion” (Peach, 1999).

In general terms, residential segregation is negatively regarded in the literature as it is deemed to reduce the likelihood of interactions with the receiving society, which in turn may hinder the opportunities of vertical social mobility and may influence the outcomes in schooling, employment and income. Studies in the US have come to the conclusion that “high-poverty neighbourhoods are potentially stimulating negative outcomes”, as the lack of exposure to positive roles models3 negatively affects children and absence of opportunities in “such neighbourhood exacerbates the problems of having low income” (Musterd, 2005, 342). More recent studies on residential segregation list impacts as diverse as “health and deprivation effects, employment prospects”, levels of “tolerance and intolerance” and “crime and violence”, as well as the “political and civic life of minority groups” (Kaplan and Douzet, 2011). Cutler and Glaeser showed that, in segregated cities, African-Americans have considerably worse outcomes in education, employment, and higher rates of single parenthood (Cutler and Glaeser, 1997).

Comparative studies of urban segregation in Europe are few, and tend to highlight the methodological difficulties in simultaneously account for “the varying impact of the welfare state, via the specific historical paths that have been followed in different cities, to differences in the cultural realm” (Musterd, 2005, 345). In a pioneering study comparing residential segregation in the UK and the US, Peach (1999) tracked different integration trajectories for different minorities in London and New York, with London's Afro-Caribbean population4 and New York’s Latino communities advancing towards assimilation into the mainstream (“melting pot”5), while South Asian population in London converging towards a “structural pluralistic model”6, and African-Americans in New York remaining segregated. Differently put, the comparison of residential patterns in these two cities coupled with a socioeconomic survey of these communities show that ethnic segregation7 is not necessarily associated with income segregation. In any case, the very fact that Censuses started to track in a more accurate manner minorities – such as the UK census in 1991 (Glazer, 1999) – reveals how salient their integration had become in the eyes of policy-makers. More broadly, studies on segregation have now moved far beyond an exclusive residential focus, to include several other aspects such as workplace segregation, or the role played by social media (van Ham and Tammaru, 2016). Indeed, scholars have become increasingly interested in understanding whether residential segregation dynamics are replicated in other social spaces, which in turn may exacerbate or moderate the potentially negative effects of residential segregation (see the case of ethnic enclaves in Zhou, 2006).

The policy response to debates on segregation has recently focused on dispersal policies, particularly in the case of asylum seekers and refugees. Countries as diverse as Germany, Italy, Sweden, and the United Kingdom activated such policies at some point in the past to deal with what were perceived as large arrivals of asylum seekers (Boswell, 2003; Stewart, 2012; Bloch and Schuster, 2005; Bolt, Phillips, and Van Kempen, 2010). Public administrations have enacted such policies based on an assumption of a negative impact of an excessive concentration of migrants both in time and space on cities’ capabilities to cope with varying patterns of immigrant settlement (be it temporary or permanent). As a corollary, dispersal policies are regarded as offering better chances of effective migrant integration. In parallel, countries have highlighted that dispersal policies are also a form of solidarity among local administrative units in their efforts of receiving asylum seekers and immigrants.

Box 4.1. Social and political consequences of immigration at the local level

Insights from the academic literature

In sociology, the relationships between natives and migrant communities have been frequently framed under either “intergroup contact” or “group threat theory”. Intergroup contact theory expects opposition to migration to wane as diversity increases. This is because “interpersonal interactions with [migrant communities] will decrease prejudice, as positive experiences with [immigrants] reduce both stereotypical thinking and anxiety about an out‐group and enhances empathy towards its members” (Eger and Bohman, 2016, 879). Group threat theory posits that opposition between natives and immigrant communities is a result of a perceived or real threat “due to intergroup competition for scarce resources, such as jobs, welfare benefits, or political power” (Eger and Bohman, 2016, 878). Eger and Bohman run a simple correlation between migrant stocks (OECD data) and hostility towards migrants (measured through a set of questions in the European Social Survey), and find only very weak relationships (R2=0.03) between the two variables (Eger and Bohman, 2016). However, the level of aggregation might be an important factor here. In other words, it is possible that the relationship between migrant stocks and anti-immigrant attitudes is diluted to a considerable extent at country level. This is why it becomes essential to analyse this relationship a lower aggregation level. Weber (2015) approaches the problem of level of analysis from a different angle. He posits that group threat theory might work at the national level where subjective perceptions of volumes of migration may be at work, whereas at regional and local level contact theory should hold as this is the space where exchange between groups actually occur. In other words, according to the level of analysis, different result might emerge.

The assumption of much of the literature investigating the relationship of migrant communities and natives is that either the pace of settlement or the sheer size of these communities matter. A too rapid rate of immigration, or a too concentrated local presence might trigger a perception of migrants as a threat to the societal security and capacity. Pottie-Sherman and Wilkes (2017) reviewed the literature on the association between group sizes and attitudes towards immigration (55 studies) and noticed a lack of consensus on the magnitude and direction of this relationship. The authors boil down the likely reasons of such disagreement to three factors: 1) measurement issues (who is an immigrant? what is opposition?); 2) methodological differences in measuring the relationship; 3) geographical level upon which the relationship is measured.

Source: Pottie-Sherman, Y. and R. Wilkes (2017), “Does size really matter? On the relationship between immigrant group size and anti-immigrant prejudice”, International Migration Review, 51 (1), pp. 218–50, https://doi.org/10.1111/imre.12191.

Dynamics of segregation and models of integration

The most often cited integration models, in the specialist literature and public debate alike, are the assimilations and multicultural (or pluralist) ones. The concept of spatial residential segregation has a correspondence in both, in that the assimilation (or melting‑pot) is spatially translated into a process of gradual dispersion towards a more uniform distribution of the population and in a lower level of clustering (Peach, 1999), whereas multiculturalism could translate migrant communities that maintain patterns, often paired by visible and distinct sociocultural traits. In this latter hypothesis, the urban landscape would appear as a mosaic of closed communities. These insights are interesting for us as they provide a way of understanding settlement beyond the simple description of residential distribution.

Conventional narratives, confirmed by empirical evidence, hold that the process of integration can be regarded as a series of subsequent stages. After being admitted in a country, migrants tend to choose to (or are forced into) settle in areas where pre-existing contacts can reduce the costs and limit the challenges of moving into a new country. The key element is that the location choice for migrants is not random but affected even more than in the case of the domestic population by network effects, housing conditions and a starting position of disadvantage. This initial clustering of the migrants in first areas of arrival resonates with multiculturalist policies of recognition of minority groups’ rights to difference (Iceland, 2014).

The expectation of assimilationist models is that as immigrants move up on the socioeconomic ladder, they also gradually detach from their “ethnic enclaves” and settle throughout the city – what Iceland called spatial assimilation (2014). These “up-ward and out-ward trajectories” (Zelinsky and Lee, 1998) result in an assimilation process that is both socio-economic and spatial. Theories of assimilation and their spatial counterparts often refer to the case of immigrants in the US before 1965, but similar outcomes have been registered also more recently for Canada (Fong and Hou, 2009). The academic literature tends to attribute the immigrants’ relatively even distribution across cities to their socio-cultural proximity to the destination country and the vertical social mobility (Zelinsky and Lee, 1998). The important element to retain here is that proximity might be an important factor in determining segregation. The empirical analysis of this chapter proxies this with distance of each migrant community to the country of origin.

Nevertheless, past migratory trends have shown that this trajectory “is neither inevitable nor unidirectional” for immigrant communities (Peach, 1996). Immigrants might remain clustered into ethnic enclaves or neighbourhoods. In an ideal application of the multicultural model, this residential scenario involves ethnic minorities’ full participation in the economic and social life of the country, while keeping their own culture, language, and values. Empirical studies, though, highlighted a possible negative side of multiculturalism, named pillarisation. Pillarisation is the outcome of multicultural policies, when immigrant groups result instead isolated from the host society and spatially segregated. When this assimilationist dispersal across the urban landscape or “multicultural” interaction between ethnic enclaves and the surrounding urban territory do not happen, segregation becomes especially problematic, as it might signal persistent integration difficulties and inequalities across generations (Iceland, 2014).

Arbaci and Malheiros (2010) challenged the assumption that dispersal across cities and peripheries is associated with upward mobility by focusing on immigrant experiences in Southern Europe, where de-segregation processes were accompanied by marginalisation (2010). Recent contributions regard the idea of spatial assimilation as outdated and incapable to fully grasp the increasing complexities of immigrant integration in European cities (Crul, 2016). Others question even more explicitly the links between immigrant integration and spatial assimilation (Bolt, Özüekren and Phillips, 2009).

These diversified integration trajectories suggest that integration models and their spatial counterparts may follow very different dynamics, depending on a variety of factors. With a specific focus on segregation, the academic literature singles out several determinants of segregation. The characteristics of the receiving society or city may be important, as well as the socio-cultural distance between the migrant community and the receiving society (mentioned above). Socio-economic characteristics of immigrants are also relevant in determining segregation (Iceland and Wilkes, 2006), but to a limited extent, according to Peach (1999). In his analysis of segregation in Germany, Sager observed that “differences in income, education, language skills and village/city size have the potential to account for 29%–84% of the residential isolation” for four migrant groups, i.e. immigrants from the Balkans, Italy, Turkey and Eastern Europe (2012). In a comparative study of segregation in Australia, Canada, New Zealand, the United Kingdom and the United States, other factors weighted in shaping segregation outcomes, namely “size of the group being considered as a percentage of the urban total, but also urban size and urban ethnic diversity” (Johnston, Poulsen, and Forrest, 2007).

As a general hypothesis, it should be expected that the higher the gaps, the more difficult will be for migrants to achieve out-ward and up-ward mobility. In addition, at the contextual level, housing markets might force minorities into relatively deprived neighbourhoods, sheer discrimination from landlords might make it difficult to find accommodation, and natives might flight from areas where immigrants start to settle (“white flight”). An agent based model developed in 1969 by the Nobel prize winner Thomas Schelling (1971) shows how the patterns of intra-urban mobility may be explained on the basis of the individual preferences to live close to persons of the same race or homophily. The model shows how individual preferences can produce striking collective results and bring to the collapse of mixed neighbourhoods and high levels of segregation. Further, minorities can willingly isolate themselves, and here cultural and historical factors play a role (Musterd, 2005).

Box 4.2. Segregation in geographical and social spaces

Besides the classical assimilation versus pluralist models of integration a third model introduced by Zelinsky (Zelinsky and Lee, 1998) and named heterolocalism considers how the increasing availability of means of communication allow establishing and preserving strong socio cultural ties independently from residential locations. Although formulated in 1998 this model anticipated the importance of social media in maintaining relations and defining a social space which overcomes geographical proximity constrains. With this model a pluralist society may coexist with a spatial pattern of dispersion.

The heterolocalism model has a correspondence at international level in the idea of transnationalism, whereby identity and cohesive communities are formed and maintained across national boundaries. Both heterolocalism and transationalism show the importance of considering a more encompassing dimension of social space than just focusing on where people live. In fact, the research frontier about segregation is moving ahead from measures of residential segregation based on census data to consider how people with migrant background interact with the receiving society also through work, education and leisure activities (Wong and Shaw, 2011; van Ham and Tammaru, 2016).

One important dimension of segregation relates to the level of exposure between the migrant and the host population. This exposure which allows confrontation between different culture and opinions can be seen as an essential condition to maintain democratic participation process and avoid polarisation and radicalism. Sunstein (2017) points out how virtual communities emerging in social media are increasingly characterised by phenomena of self-insulation. Social media like Facebook and Twitter have the undisputed merit of greatly enhancing the possibilities to access information and communicate with person all over the globe, and in this sense they are contributing to improve exposure to diverse opinions and to connect with others, overcoming the limits of the geographical proximity. However, social media have an explicit objective of tailoring the provision of information on an individual basis and are designed to favour the formation of communities of like-minded. This filtering of information and contacts results in a self-insulation from those who don’t share the same interest and set of values. In this way social media create echo chambers in which the more extreme ideas are amplified, public opinion is polarised.

Data processing and measurement

Assembling and processing the data

Despite the advancement of methods and wealth of indexes and tools to measure segregation there have been few empirical applications to compare the patterns of spatial segregation between cities in different countries (for recent examples see: Peach, 2009; Musterd, 2005). In most empirical applications the calculation of the segregation indexes is limited to few large cities and considers aggregate ethnic groups. The main reasons for the few cross-country comparisons reside in: the difficulty of assembling data from several national statistical institutes; the lack of standardisation in the aggregation levels, geometries, definitions; and confidentiality requirements.

To address these challenges, this study uses a data set that provides for the first time the possibility of mapping migrant communities in several EU Member States, at high spatial resolution. The uniqueness of the data set consists both in the high level of spatial resolution and the large geographical coverage which includes almost 45 000 LAU in France, Germany, Ireland, Italy, the Netherlands, Portugal, Spain and the United Kingdom. The original data used to produce the high resolution map was obtained through ad hoc extraction of 2011 national census statistics. These ad hoc extractions provided data on the resident population broken down by country of birth and/or citizenships at the lowest possible level of geographical detail.

The geometries used to represent the data were polygons corresponding to census sampling areas in the case of France (TRIRIS and IRIS), Italy, Portugal and Spain, a spatial grid with cells of 100 by 100 m in the case of Germany, postal codes in the case of the Netherlands and so-called output areas in the case of Ireland and the United Kingdom. Differences in geometries and resolution were harmonised through a spatial processing method called dasymetric mapping. With this method the population by origin from the original census data was redistributed and spatially disaggregated into a uniform grid using as ancillary information the land cover classes (CORINE land cover) and the presence of built up areas (European Settlement Map) in each cell. The basic idea behind the process of dasymetric mapping is that the total population of the census area is proportionally allocated to the cells included in the polygon if these are characterised by the presence of built up areas and a residential land cover, rather than for example green areas and agricultural land.

The result of the spatial processing was a grid covering the entire territory of the 8 Member States included in the study, where each cell reports the residential population by origin at three different levels of aggregation (country, continent, EU vs. third country origin). From such data it was possible to calculate the concentration of migrants for LAU, FUA, at the different levels of the Nomenclature of territorial units for statistics classification (NUTS 1, 2 and 3) or for hoc grouping of specific cells.

The main limitation in the final data set is that while it provides information at high spatial resolution it can only reproduce partially the aggregated figures at national and regional level. The high level of detail in the data implies that for many cells the data falls below confidentiality thresholds and therefore either the data is completely missing or it is presented in aggregated form in respect of the country of origin dimension.

Figure 4.1 exemplifies a calculation of the share of migrants from China, Ecuador and Senegal in each cell falling in the boundary of the LAU of Genoa in Italy.

Figure 4.1. Example of a calculation of the share of migrants from Ecuador, China and Senegal in each cell falling in the boundary of the LAU of Genoa in Italy

The simple visual inspection of the figure indicates that: the Ecuadorian community - the largest in the city - is both clustered in at least two areas as well as marginally present in other parts of the city; the Chinese community is much smaller and tends to spread rather evenly along the coast, with a very narrow displaying of peaks of concentration; and the Senegalese community is the most segregated, with a presence almost exclusively delimited in a narrow area of the city.

Measuring spatial segregation

A large body of research on residential spatial segregation has contributed to develop several indexes to measure spatial residential segregation. These indexes range from very simple and a-spatial measures of diversity to more advanced indexes which consider the spatial structure of the distribution of populations in census tracts or in continuous spatial distributions. See Annex 3.A for technical details on the measures used in this chapter.

Segregation indexes introduced since the 1950s and 1960s (dissimilarity, exposure index, variance ratio index, entropy-based information theory index) typically measure the dissimilarity in the ethnic composition of the residential areas, without considering the spatial relations between the residential areas. This leads to two flaws which have been described as checkerboard problem and modifiable areal unit problem.

Figure 4.2 exemplifies the checkerboard problem with a toy example representing three cities and two ethnic groups of “black” and “white”. The “black” group has the same ratio to the total population in ‘a’ and ‘b’. A-spatial dissimilarity indexes and concentration profiles calculated for the three cities would show identical values while it is self-evident that the “black” group has very different spatial patterns of clustering.

A seminal paper by Massey and Denton (1988) identifies five main dimensions to measure spatial segregation. Further to this paper Reardon and O’Sullivan (2004) collapse the segregation measure along two main dimensions of isolation/exposure and clustering/evenness.8 The clustering/evenness dimension captures how the ethnic groups are concentrated or evenly distributed in space. In the examples in Figure 4.2, the clustering is highest in ‘a’ and decreasing progressively in the examples in the ‘b’ and ‘c’. The second dimension of isolation/exposure considers the spatial composition of the surrounding regions of each group. The black group in ‘b’ although preserving a certain level of clustering has a higher number of “whites” in its surrounding and is therefore less isolated in respect of ‘a’. Intuitively, exposure can be seen as a measure of the probability of the group of entering in contact with other groups, which is influenced by the amount of shared boundaries of the regions.9

The modifiable areal unit problem is determined by the use of aggregated data by census areas. The anomaly produced in this case consists in the fact that distant groups within the same census area are considered more clustered than geographically closer groups falling in two distinct census areas. Ideally this anomaly could be addressed by using point data on the exact residential locations; however, this data is normally not available in the census statistics, since it would violate data protection rules. Another solution is to redistribute the population of the census area into a continuous density surface through a spatial smoothing process based on an equal distribution or kernel density. This is similar to the method of dasymetric mapping which was adopted in this study to harmonise the different census geometries into a uniform grid. This grid, considering its high spatial resolution, approximates a continuous spatial distribution of population and at least partly addressees the modifiable areal unit problem.

Figure 4.2. Toy example of segregation measures calculated for three ideal cities

Results

Concentration of migrants across cities

A very simple measure of concentration of migrants can be calculated by dividing the migrant population by the total population in each city. Considering this ratio at the level of single LAU gives a considerable refinement in respect of the statistics aggregated at higher administrative levels of provinces, regions and nations. All the results refer to the year 2011.

The distribution of the concentration of migrants across LAU shows an extremely variegated picture (Figure 3.3). In the case of LAUs with a population of more than 1 000 inhabitants, the median value of concentration of migrants, considering both intra‑EU and third countries origins, for all countries considered, is of 7%, and the upper quartile of 23%. Values of concentration above the median are recorded in France, Germany, Ireland and Spain. Some examples of LAUs with the highest values of concentration are San Fulgencio in Spain (70%), Wembley in the UK (68%), Dublin North in Ireland (65%), Büsingen am Hochrhein in Germany (48%) and Aubervilliers in France (37%).

These high values of concentration are less evident in the case of FUA (Figure 4.4) where the median values range between 16% in Germany and 7% in France and Portugal. A high level of concentration of migrants is present in Achen in the Netherlands (73%), Fuengirola in Spain (67%), Leverkusen in Germany (57%) and London in the UK (57%). Box 4.3 discusses the link between the spatial concentration of migrants and urban poverty for the case of five Dutch cities.

Figure 4.3. Concentration of migrants in LAUs with a total population of more than 1000 persons

Figure 4.4. Concentration of migrants by FUA

Box 4.3. Is there a link between the spatial concentration of migrants and urban poverty? Evidence from the Netherlands

While there is growing evidence of how, across Europe, migrants are on average more likely to be at risk of poverty (Eurostat, 2018), fewer studies have explored the spatial dimension of such relationship in contemporary European cities.

Magante and Luca (2018) explore the case of the Netherlands, and ask: is there a link between the residential concentration of immigrants and urban poverty? They provide robust exploratory analysis on the country’s five biggest cities, namely Amsterdam, Eindhoven, The Hague, Rotterdam and Utrecht. Their analysis combines novel fine-grained data on the residential distribution of immigrants in urban neighbourhoods with data on income distribution. The results confirm how areas characterised by a large share of migrants show significantly higher levels of poverty, measured as the share of persons in the bottom income quintile. Neighbourhoods with a higher concentration of migrants also feature higher income inequality, possibly reflecting the heterogeneous socioeconomic structure of areas inhabited by both natives and foreign-born people.

Figure 4.5. Migrants’ residential concentration and poverty rates in Dutch urban neighbourhoods

Econometric results confirmed the positive relationship between migrants’ concentration and poverty, which is robust to controlling for city fixed-effects, neighbourhood population, income inequality (a proxy for gentrifying areas), and level of ethnic diversity. Even according to the most conservative estimates, a 1% increase in the share of migrants is correlated to a 0.32% increase in the share of poverty. With the exception of The Hague, the link between migrants and poverty is stronger outside of the capital city and increases inversely to city size. Besides, the intensity and the sign of the relationship is greatly affected by the composition of immigrant communities. Controlling for covariates, the relationship is insignificant for migrants from ‘old’ EU member states (EU15 countries). By contrast, it is significant for migrants from both the ‘new’ member states which have joined the EU in 2004 (EU13 countries) and non-EU countries. EU13 natives, in particular, are the ones for whom the link is strongest.

The integration of migrants is among the top priorities for both national and local governments in many OECD countries. There is significant research on neighbourhood effects, i.e. on how the spatial concentration of disadvantage may perpetuate social exclusion and negatively influence the integration of migrants (Bolt and van Kempen, 2003; Friedrichs, Galster, and Musterd, 2003; Musterd, 2005). The importance of the local level as a locus where integration can occur is also embodied in the EU Action Plan on the integration of third country nationals (EC, 2016). While results do not find systematic cases of spatial concentration of specific communities and poverty as in the contexts of American and Asian cities (cf. Massey and Denton, 1993; Garcia-Lopez and Moreno-Monroy, 2016), they nevertheless underline the importance of policy solutions that ensure the full integration of migrant communities into host societies (cf. Musterd, 2005; OECD, 2008).

Source: Bolt, G. and R. van Kempen (2003), “Escaping poverty neighbourhoods in the Netherlands”, Housing, Theory and Society, 20, pp. 209-222; EC, (2016), Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions, COM 2016, 377 FINAL, Brussels; Eurostat (2018), Migration integration statistics – at risk of poverty and social exclusion, http://ec.europa.eu/eurostat, (accessed on 15 March 2018); Friedrichs, J., G. Galster and S. Musterd (2003), “Neighbourhood effects on social opportunities: The European and American research and policy context”, Housing Studies, 18(6), p. 797–806, https://doi.org/10.1080/0267303032000156291; Garcia-Lopez, M.A. and A. Moreno-Monroy (2016), “Income segregation and urban spatial structure: evidence from Brazil”, CAF Working papers, 08(2016); Magante, C. and D. Luca (2018), “Testing the link between migrants’ spatial concentration, poverty, and inequality: new micro-geographical evidence from the Netherlands”, Unpublished working paper, Gran Sasso Science Institute, L’Aquila; Massey, D. and N. Denton (1993), American Apartheid. Segregation and the Making of the Underclass, Harvard University Press, Cambridge, MA; Musterd, S. (2005b), “Social and ethnic segregation in Europe: Levels, causes, and effects”, Journal of Urban Affairs, 27(3), pp. 331–48, https://doi.org/10.1111/j.0735-2166.2005.00239.x; OECD (2008), Jobs for Immigrants (Vol. 2): Labour Market Integration in Belgium, France, the Netherlands and Portugal, OECD Publishing, Paris, http://dx.doi.org/10.1787/9789264055605-en.

By considering the data of the eight countries together there is no apparent relation between the size of LAUs, measured in terms of total population, and the concentration of migrants. However, Figure 4.6 shows that differences in correlations between population size of the administrative unit and migrant concentration emerge when analysing each country separately. In particular, in decreasing order, the Netherlands, the United Kingdom, Portugal, show moderate to strong positive correlations (r approximately 0.45 and above) between LAUs population size and concentration of Third Country Nationals (TC), while lower values are recorded in the cases of France, Germany Ireland, Italy and Spain. This relationship between size and concentration is weaker for EU nationals, if not negative (in Portugal and Spain).

As shown in Section 3.3, past studies have tested for associations between cities’ sizes and migrant concentration and segregation (Sager, 2012; Johnston, Poulsen, and Forrest, 2007). The high correlations for some of the countries in the sample support the hypothesis that that migrants preferably settle in large cities. That said, the lower values in the case of France, Germany, Italy and Spain, as well as the negative results for Portugal and Spain, alert us that the relationship is not straightforward, and other intervening factors might be at play, as migrants in those countries tend to be more dispersed in smaller cities.

To explain these contrasting results, one has to consider that national characteristics of the receiving countries and the type and origin of migration may affect the likelihood that migrants will settle in large rather than medium and small size cities. For example, the negative correlation for intra-EU migration in Portugal and Spain may be explained by the fact that among migrants for these two countries there are many retirees in particular coming from the United Kingdom which privilege touristic destinations along the coastline rather than living in Madrid or Lisbon (Betts, 2011; Bade and Eijl, 2011). Similarly, the lower correlation for extra-EU migrants in Italy and Spain in respect of the other countries in the sample may be related to the relatively high share of unskilled labour force employed in farming activities.

Here labour market dynamics are at play. According to the dual market theory (Piore, 1986), the labour market of developed industrial economies is segmented in two types of jobs. Jobs in the secondary sector are characterised by low wages, lower social status and employment instability. Migrants tend to fill these jobs since they typically view their migration as temporary. Empirical analyses of the EU labour market identified three distinct segments on the EU labour market describing the coexistence of “good” jobs on one side and “bad” jobs on another. In such labour market structure, non-EU immigrants have higher probability than natives of being employed in “bad” jobs, although the immigrants-natives gap varies significantly among MS (Grubanov-Boskovic and Natale, 2017). This chapter postulates that the observed patterns of geographical substitution between nationals and migrants in some countries may have a relation to the substitution which is taking place at the level of the labour market.

To sum up, the geographical patterns of migration and the level of concentration in small versus large cities and rural areas versus urban areas may reveal a more nuanced view of the mobility transition theory (Zelinsky, 1971) and the general view of a higher likelihood for migrant to move towards large cities.

The somehow surprising evidence that in some countries migrants are also likely to settle in small cities gives a new perspective on the territorial aspects of migration. The prevailing narrative on migration indicates that migrants are attracted by the large and global cities, which in turn witness a constant increase in their diversity and level of multiculturalism (Meissner and Vertovec, 2015; Vertovec, 2007). In fact, this chapter suggests that this process may be happening also in smaller cities.

Although the data set does not record age structures and population changes over time, it is plausible to imagine that in some cities the increasing share of migrants is substituting the native population and compensating for ongoing trends in depopulation and population aging. At lower geographical scales, a similar effect is frequently observed in the mobility dynamics in neighbourhoods, where migrants settle in urban peripheries and more deprived parts of the city replacing the national population. In fact, the idea of migrants compensating for the depopulation of small towns in eastern Germany and southern Italy emerged during public discussions about the reallocation of the large inflows of asylum seekers in 2015 in Germany and Italy.

Figure 4.6. Correlations between city size and the concentration of migrants from third countries and intra-EU

Ethnic diversity of cities

The diversity of cities can be measured by considering both the variety of migrants’ origins (Vertovec, 2007) and the evenness in the distribution of population across different countries of origin. Among the several diversity indexes available in the literature (for instance: Simpson, Rao - Stirling, Gini, Blau) one of the most commonly used is the Shannon entropy index (Shannon, 1948). Similarly to other indexes, the value of diversity measured through the Shannon entropy index increases with the increase in the number of countries of origin and when there are equal population shares across the different migrants’ origins. In other words, on one side of the spectrum, the index approaches zero if the majority of the population is represented in one dominant group (most likely the domestic population) and shares in the other groups are extremely small. At the other side of the spectrum, the index reaches the maximum when all groups composing the population are of equal size.

Although the countries of origin of migrants do not necessarily correspond to different socio-cultural characteristics, it is common practice for quantitative measures of diversity to be based on the distribution of population shares across countries of origin, which can be taken as proxy of the sociological concept of diversity within cities. Similarly to the results regarding the concentration of migrants, the values of the diversity index across LAU also show a highly variegated picture (Figure 4.7).10

The figure confirms the common perception of some European cities as “superdiverse”, like Berlin and Rotterdam. However, what is interesting to notice is that also medium and small size cities like for example Baranzate in Italy, Forest Gate South in the United Kingdom, Monaghan in Ireland and Teulada in Spain exhibit high values of diversity. The relation between diversity and the total population of the LAU follows a similar pattern encountered in the case of the concentration of migrants. In general, there is a positive correlation between the diversity and total population of the LAU. High correlation are present in the case of the Netherlands (0.6), the United Kingdom (0.4), small correlation in the case of France (0.2), Germany (0.2), Ireland (0.2), and almost no correlation in the case of Italy (0.08) and Spain (0.06).

The positive correlation between the size of the city and values of diversity holds also if diversity is calculated using FUAs as geographical units, instead of the LAUs. FUAs have across all countries higher median values of diversity in respects of LAUs (Figure 4.7). As in the case of LAUs (Figure 4.6), values of diversity are higher than the overall median in the case of France, Germany and Ireland, and lower in the case of the Netherlands, Spain and the United Kingdom.

Looking at specific countries, Germany nearly doubles its median diversity, driven by the relatively high diversity recorded in large cities such as Frankfurt and Munich. The role of capitals and very large cities also emerges more clearly when considering FUAs (e.g. Amsterdam, London, Milan, Paris), supporting insights in the literature about superdiversity in Europe. These results are, however, also conditioned by the definition of LAUs, which in some countries is fragmented in specific neighbourhoods within the larger boundaries of the FUA. This is clearest in the case of London and Paris, which in Figure 4.6 are divided in several units, while they appear as single entities in in Figure 4.7.

Segregation across cities and origins

Besides calculating simple concentrations of migrants and diversity indexes for each LAU, the spatial segregation measures of spatial information theory index and general spatial exposure/isolation proposed by Reardon and O’Sullivan (2004) were calculated for each LAU with more than 1 000 inhabitants and each country of origins of the migrants.

Segregation indexes were calculated for a total of 267 280 combinations of 41 532 unique LAU and 186 unique countries of origins.11 For confidentiality reasons, the data for Spain, UK and Ireland only includes 21, 20 and 5 major countries of origin.

Figure 4.8 shows the two measures of segregation for a sample of large cities (top 20 by size of the population), and for the origin countries which recorded the highest level of clustering in each of city. Although the sample represents only a small part of the entire data set, it shows that it is difficult to identify a consistent pattern of segregation by origin and cities. Origins which may appear segregated in one city do not feature among the most segregated in other cases. The only exception in the sample is the Chinese community, which appears as the most clustered in Berlin, Munich, Naples, Stuttgart, Toulouse and Turin.

In general, there is not a uniform behaviour between the two measures of isolation and clustering with the exception of migrants from the Philippines in Hamburg and from Bulgaria in Madrid. This is to be expected since the two dimensions proposed by Reardon and O’Sullivan are designed to capture two distinct aspects of segregation.

Figure 4.7. Diversity (Shannon entropy) by LAU

Figure 4.8. Diversity (Shannon entropy) by FUA

Figure 4.9. Values of segregation (clustering and isolation) for selected cities

Figure 4.9 and 4.10 provide an overview of descriptive statistics on clustering and isolation. The graphs respectively represent the values of segregation and isolation for all combinations of cities and origins considering a breakdown by classes of cities on the basis of their population and by continents and world regions. Figure 4.9 suggests a negative relationship between the size of the city and clustering dimension. In other words, the bigger the city, the lower immigrant communities tend to cluster. On the other hand, there is no clear pattern of association in the case of isolation.

Differences in clustering also emerge considering the intra-EU versus third country origin (data not shown for brevity). In particular, the median clustering is lower in the case of intra-EU migrants (0.09) in respect of the third-country origin (0.14). This difference may be indicative of the fact that in general intra-EU migrants are facing fewer obstacles in settling in and therefore are more likely to spread spatially. No meaningful difference is recorded when it comes to isolation.

Figure 4.10. Clustering and isolation by city size classes

Figure 4.11. Clustering and isolation by continent and regions of origin

From Figure 4.10 it emerges that there are also specific differences in the level of segregation by continents and sub-regions of origin. Migrants from Latin America have in general higher median levels of segregation, according to both the isolation and clustering dimensions, followed by migrants from Asia and Europe. Lower clustering is found in the case of migrants originating from Northern America and Oceania, as well as for migrants coming from Africa. That said, there is substantial within-continent variation which can be observed at the sub-regional level. For instance, migrants from Eastern Europe have a higher tendency to be clustered in respect of migrants from other regions in the same continents, whereas migrants from South-Eastern Asia show the same tendency with regards to clustering, but have the lowest isolation median values for the continent. While migrants from Latin America have the highest median clustering values, their isolation values are not radically different from other sub-regions in other continents.

Table 3.1 shows more in details the five top countries recording the highest level of clustering in each continent. An interesting finding which emerges from this table is that some of the countries in Africa and Asia which have the highest median values of clustering are also countries which produced relatively large flows of refugees. This is the case of, inter alia, Afghanistan, Eritrea, Myanmar, Somalia, and Sudan (UNHCR, 2015). A possible explanation may be that migrants from these origins are more likely to be part of a fragile group escaping from wars and violence, thus relying even more from the support of existing diaspora besides host-states support. As a consequence, they could have a tendency to concentrate in areas where their communities are already rooted, hence resulting in higher clustering values. This would be in line with the idea that segregation, particularly at first, is a way for recently arrived immigrants – in this case, refugees – to cope with what might be a difficult transition and integration process (Peach, 1996). In addition, this trend could also be fuelled by internal relocation policies for asylum seekers by some countries (Boswell, 2003). On the other hand, these countries’ clustering values are on a par with states that are not major source countries of refugees, such as Argentina, Indonesia or Israel, again suggesting heterogeneity in the factors driving segregation of migrants in Europe. To better understanding these dynamics, the following section turns to the likely determinants of segregation in the eight countries analysed in this section.

Drivers of segregation

The determinants of segregation described in the literature can be broadly classified in three groups: drivers at individual level, structural characteristics of the receiving society, and structural characteristics of the migrant group. At the individual, socioeconomic characteristics such as education, income, occupation may play a role. Group characteristics include cultural and ethnic features such as religion, language proximity and visibility of the minority.

The role of the different drivers for segregation is analysed through two multivariate regression models in which the observed segregation indexes of isolation and clustering are put in relation with the following explanatory variables: size of the city, relative size of the migrant community in respect of the population of the city, diversity of the city, geographical distance and contiguity between country of origin and destination, share of refugees from the country of origin to the total migrants population in the country of destinations.

The inclusion of the bilateral distance and contiguity variables is intended to capture socio-cultural differences between the migrants and the receiving society. In particular, the hypothesis to be tested is whether geographical distance, which may be considered a proxy also of socio-cultural differences, may favour a high level of spatial segregation. For instance, if migrants from African or Asian countries are more spatially segregated than migrants coming from within the EU and in particular from neighbouring countries.

As copiously argued in the academic literature (Vertovec, 2007; Meissner and Vertovec, 2015; OECD, 2016b; Sanderson et al., 2015), and confirmed in the previous sections, immigration is gradually changing the character of most EU cities by increasing their diversity. One key question is if more diversity is associated with an increase of segregation and the formation of separate clusters in the urban landscape, or if diversity is evolving along the line of the assimilation model in which migrants tend to be dispersed in the city. We address this question by including among the explanatory variables the overall diversity of the city measured through the Shannon index described in the previous section.

The last explanatory variable is represented by the share of refugees in respect of the total migrant population. The variable is aligned to the same combinations of the LAU and country of origin pairs for which segregation measures were computed. The reference year for this variable is 2010 and therefore it provides lagged values in respect of the 2011 reference year of the census data. The variable can be interpreted as the likelihood that a migrant included in the high resolution map is a refugee.

Figure 4.12 shows the results of the two models in terms of standardised regression coefficients (more summary statistics are shown in Annex 4.B).

Continent	Country of origin	Clustering	Isolation
Africa	Uganda	0.5985	0.0539
	Somalia	0.5288	0.0905
	Sierra Leone	0.4599	0.0073
	Ethiopia	0.4292	0.0209
	Eritrea	0.4071	0.0207
	Sudan	0.3536	0.0394
	Burundi	0.312	0.0057
	Djibouti	0.2385	0.0027
	Liberia	0.2384	0.0314
	Rwanda	0.2289	0.0041
Asia	Myanmar	0.6407	0.0891
	Nepal	0.5007	0.0274
	Brunei	0.4808	0.1477
	Jordan	0.4404	0.0047
	Indonesia	0.4318	0.0484
	Afghanistan	0.4107	0.0402
	Iraq	0.4044	0.0419
	Israel	0.4032	0.0338
	Uzbekistan	0.3897	0.0044
	Syrian Arab Republic	0.3758	0.0185
Europe	Montenegro	0.6003	0.0624
	Belarus	0.3512	0.0044
	Norway	0.2553	0.0041
	Bosnia and Herzegovina	0.249	0.019
	Ukraine	0.2464	0.0194
	Russia	0.1956	0.0383
	Macedonia	0.1858	0.0392
	Moldova	0.1731	0.0191
	Albania	0.1359	0.0336
	Serbia	0.1109	0.0099
Latin America and the Caribbean	Paraguay	0.4512	0.08
	Dominica	0.4377	0.0615
	Argentina	0.42	0.083
	Cuba	0.4061	0.0535
	Aruba	0.398	0.042
	Dominican Republic	0.3801	0.0201
	Bolivia	0.3367	0.0571
	El Salvador	0.3252	0.0218
	Suriname	0.3195	0.0361
	Colombia	0.3177	0.055
Northern America	United States	0.1139	0.025
Northern America	Canada	0.1071	0.0021
Oceania	Kiribati	0.6678	0.1152
Oceania	Australia	0.0991	0.0212

Source: Elaborations based on data sources detailed in section “Data processing and measurement”.

Figure 4.12. Drivers of isolation and clustering

What emerges from the figure is that on the one hand, the relative size of the community drives to more isolation of the migrant communities. On the other, the bigger the immigrant community, the more it tends to be scattered throughout the city. The negative relationship of this variable on clustering is consistent with the idea that a large migrant community may occupy several areas of the city and exhibit therefore a pattern of higher spatial dispersion than a small community. At the same time, the bigger this community, the less there is a chance of exposure with other communities in the area, simply because of its sheer size. At least from a geographical perspective, being member of a large community is related to a lower probability of encounters with the national population also if the community is not clustered. The fact that the relative size of the community affects the clustering and isolation in opposite directions confirms the usefulness of considering two distinct dimensions when assessing segregation as proposed by Reardon and O’Sullivan (2004).

The variable which has the strongest positive relationship with clustering is the size of the city. The positive sign indicates that communities in large cities tend to remain more geographically circumscribed. This is in line with the literature on ethnic enclaves, where immigrants recreate part of the social and economic fabric of the countries of origin, but are not necessarily isolated from the outside.

The coefficients for the distance variables have all the expected signs (negative for contiguity and positive for distance) to confirm the hypothesis that the likelihood for segregation is increasing for migrants coming from countries of origin which are geographically far from the receiving country. The negative relationship is particularly strong in the case of the contiguity variable. This is telling that migrants coming from neighbouring countries will have a higher geographical dispersion and are not necessarily confined in few specific neighbourhoods.

The diversity of the city has a negative relationship both on the clustering and isolation. This is indicating that cities which have both large migrant communities and migrants coming from several origins drive immigrants to spread across the urban territory and not to isolate themselves.

The coefficients of the last variable measuring the level of forced migration from a specific country of origin have a positive sign both for clustering and isolation. This confirms the hypothesis that fragility and disadvantage may determine at least in initial phase of immigration the conditions for agglomeration and spatial segregation. These conditions may be linked to the reception policies for asylum seekers which force people to reside in assigned reception centres or to integration policies where public administration provides housing in specific geographical areas.

Another factor at play may be represented by the tendency to settle closer to pre-existing diasporas to benefit of the supports of the network of migrants. The evidence provided in this study albeit at a much lower geographical scale would be in line with the effect of diasporas as a pull factor for migration which is well documented for aggregated migration flows at national level. Whatever the reasons for agglomeration, one conclusion is that the outward and upward or assimilation trajectories which are described for other type of migrants are probably less applicable for migrants who given their countries of origin are more likely to be in a condition of disadvantage.

A highly skilled migrant from the US to Rome will have freedom of choice about where to settle while an Eritrean escaping war will initially go close to his/her community or where public administration would provide housing opportunities. The high values of segregation encountered indicate that initial places were refugees are more or less forcibly landing creates a focus of attraction and concentration also in the long term.

Conclusions

This chapter has illustrated an analysis of a new data set mapping the concentration of migrants in EU cities at high spatial resolution. The analysis showed that diversity is not only a characteristic of well-known large cities, but also of less-studied medium and small size towns. Furthermore, clustering is higher in general for migrants from third countries, for migrants from South America and South-East Asia and for specific countries of origin which have a recent history of conflicts.

The large size of the migrant community reduces the clustering, but it increases its isolation. Migrants coming from distant countries are more likely to be isolated compared to migrants from neighbouring countries, which are more evenly dispersed. In fact, the diversity of the city has a positive effect in contrasting isolation. All in all, groups of migrants with a high share of refugees/asylum seekers are more likely to be isolated.

To further expand this research, one of the most pressing needs is to combine the data on geographical distribution with other socioeconomic data at the same level of analysis. While this exercise has granted us a wealth of information on immigrants’ distribution across Europe, it has also shown that to answer more elaborated academic- and policy-related questions on immigrants housing, education, residential patterns, we need to combine such information with other data which could provide a more refined picture as well as control for other intervening factors. So far, this has proven to be a challenge, and only further access to administrative data at the local level can fully realise the potential of this mapping exercise. As a second avenue of research, there is the opportunity of complementing the large amount of information provided by this mapping exercise for the year 2011 with results from past Census rounds. This will offer the possibility of having a time perspective that so far has been missing, and hence enable us to start talking about trends and not only provide a snapshot of immigrant settlement patterns.

Bibliography

Arbaci, S. and J. Malheiros (2010), “De-Segregation, peripheralisation and the social exclusion of immigrants: Southern European cities in the 1990s”, Journal of Ethnic and Migration Studies, 36 (2), pp. 227–55, https://doi.org/10.1080/13691830903387378.

Bade, K.J. and C. van Eijl (eds) (2011), The Encyclopedia of Migration and Minorities in Europe: From the 17th Century to the Present, Cambridge Univ. Press, Cambridge.

Betts, A. (ed) (2011), Global Migration Governance, Oxford University Press, Oxford.

Bloch, A. and L. Schuster (2005), “At the extremes of exclusion: Deportation, detention and dispersal”, Ethnic and Racial Studies, 28(3), pp. 491–512, https://doi.org/10.1080/0141987042000337858.

Bolt, G., A.S. Özüekren and D. Phillips (2009), “Linking Integration and residential segregation”, Journal of Ethnic and Migration Studies, 36(Nov), pp. 169–86, https://doi.org/10.1080/13691830903387238.

Bolt, G., D. Phillips and R. Van Kempen (2010), “Housing policy, (de)segregation and social mixing: An international perspective”, Housing Studies, 25(2), pp. 129–35, https://doi.org/10.1080/02673030903564838.

Bolt, G. and R. van Kempen (2003), “Escaping poverty neighbourhoods in the Netherlands”, Housing, Theory and Society, 20, pp. 209-222.

Boswell, C. (2003), “Burden-sharing in the European Union: Lessons from the German and UK experience”, Journal of Refugee Studies, 16(3), pp. 316–35.

Crul, M. (2016), “Super-diversity vs. assimilation: How complex diversity in majority–minority cities challenges the assumptions of assimilation”, Journal of Ethnic and Migration Studies, 42 (1), pp. 54‑68, https://doi.org/10.1080/1369183X.2015.1061425.

Cutler, D.M. and E.L. Glaeser (1997), “Are ghettos good or bad?”, The Quarterly Journal of Economics, 112(3), pp. 827–872.

Ramirez, M. D., Liebig, T., Thoreau, C., and P. Veneri (2018), “The integration of migrants in OECD regions: A first assessment”, OECD Regional Development Working Papers, No. 2018/01, OECD Publishing, Paris, http://dx.doi.org/10.1787/fb089d9a-en.

EC (2016), Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions, COM 2016, 377 FINAL, European Commission, Brussels.

Eger, M.A. and A. Bohman (2016), “The political consequences of contemporary immigration”, Sociology Compass, 10(10), pp. 877–92, https://doi.org/10.1111/soc4.12409.

Eurostat (2018), Migration integration statistics – at risk of poverty and social exclusion, http://ec.europa.eu/eurostat, (accessed on 15 March 2018).

Eurostat (2015), Glossary: Local administrative Unit (LAU) - Statistics explained, http://ec.europa.eu/eurostat/statistics-explained/index.php/Glossary:Local_administrative_unit_(LAU).

Fong, E. and F. Hou (2009), “Residential patterns across generations of new immigrant groups”, Sociological Perspectives, 52(3), pp. 409–28, https://doi.org/10.1525/sop.2009.52.3.409.

Friedrichs, J., G. Galster and S. Musterd (2003), “Neighbourhood effects on social opportunities: The European and American research and policy context”, Housing Studies, 18(6), p. 797–806, https://doi.org/10.1080/0267303032000156291.

Garcia-Lopez, M.A. and A. Moreno-Monroy (2016), “Income segregation and urban spatial structure: evidence from Brazil”, CAF Working papers, 08(2016).

Glazer, N. (1999), “Comment on ‘London and New York: Contrasts in British and American models of segregation’ by Ceri Peach”, International Journal of Population Geography, 5(5), pp. 347–51, https://doi.org/10.1002/(SICI)1099-1220(199909/10)5:5<347::AID-IJPG149>3.0.CO;2-G.

Glikman, A. and M. Semyonov (2012), “Ethnic origin and residential attainment of immigrants in European countries”, City & Community, 11(2), pp. 198–219, https://doi.org/10.1111/j.1540-6040.2012.01398.x.

Grubanov-Boskovic, S. and F. Natale (2017), “Migration in a segmented labour market”, JRC (Joint Research Centre) Technical Report, Publications Office of the European Union, Luxembourg, https://doi.org/10.2760/213154.

Hardman, A. (2008), “Immigration and the city”, in S.N. Durlauf and L.E. Blume (eds), The New Palgrave Dictionary of Economics: Volume 1 – 8, Palgrave Macmillan UK, Palgrave, pp. 2863–65., https://doi.org/10.1007/978-1-349-58802-2_762.

Iceland, J. (2014), Residential Segregation: A Transatlantic Analysis, Migration Policy Institute, Washington, DC, www.migrationpolicy.org/research/residential-segregation-transatlantic-analysis.

Iceland, J. and R. Wilkes (2006), “Does socioeconomic status matter? Race, class, and residential segregation”, Social Problems, 53(2), pp. 248–73, https://doi.org/10.1525/sp.2006.53.2.248.

International Organization for Migration (ed) (2015), Migrants and Cities: New Partnerships to Manage Mobility, World Migration Report 2015, International Organization for Migration (IOM), Geneva.

Johnston, R., M. Poulsen and J. Forrest (2007), “The geography of ethnic residential segregation: A comparative study of five countries”, Annals of the Association of American Geographers, 97 (4), pp. 713–38, https://doi.org/10.1111/j.1467-8306.2007.00579.x.

Kaplan, D.H. and F. Douzet (2011), “Research in ethnic segregation III: Segregation outcomes”, Urban Geography, 32(4), pp. 589–605, https://doi.org/10.2747/0272-3638.32.4.589.

Magante, C. and D. Luca (2018), “Testing the link between migrants’ spatial concentration, poverty, and inequality: new micro-geographical evidence from the Netherlands”, Unpublished working paper, Gran Sasso Science Institute, L’Aquila.

Massey, D. S., and Denton, N. A. (1988). The dimensions of residential segregation. Social forces, 67(2),

281-315.

Massey, D. and N. Denton (1993), American Apartheid. Segregation and the Making of the Underclass, Harvard University Press, Cambridge, MA.

Meissner, F. and S. Vertovec (2015), “Comparing super-diversity”, Ethnic and Racial Studies, 38(4), pp. 541–55, https://doi.org/10.1080/01419870.2015.980295.

Musterd, S. (2005), “Social and ethnic segregation in Europe: Levels, causes, and effects”, Journal of Urban Affairs, 27(3), pp. 331–48, https://doi.org/10.1111/j.0735-2166.2005.00239.x.

OECD (2018), Working Together for Local Integration of Migrants and Refugees, OECD Publishing, Paris, http://dx.doi.org/10.1787/9789264085350-en.

OECD (2016a), International Migration Outlook 2016, OECD Publishing, Paris, http://dx.doi.org/10.1787/migr_outlook-2016-en.

OECD (2016b), “The economic impact of migration: Why the local level matters”, in International Migration Outlook 2016, OECD Publishing, Paris, http://dx.doi.org/10.1787/migr_outlook-2016-6-en.

OECD (2013), Definition of Functional Urban Areas (FUA) for the OECD Metropolitan Database, Paris, www.oecd.org/cfe/regional-policy/Definition-of-Functional-Urban-Areas-for-the-OECD-metropolitan-database.pdf.

OECD (2008), Jobs for Immigrants (Vol. 2): Labour Market Integration in Belgium, France, the Netherlands and Portugal, OECD Publishing, Paris, http://dx.doi.org/10.1787/9789264055605-en.

Okólski, M. (ed) (2012), European Immigrations: Trends, Structures and Policy Implications, IMISCOE Research, Amsterdam University Press, Amsterdam.

Peach, C. (1999), “London and New York: Contrasts in British and American models of segregation with a comment by Nathan Glazer”, International Journal of Population Geography, 5(5), pp. 319–47, https://doi.org/10.1002/(SICI)1099-1220(199909/10)5:5<319::AID-IJPG148>3.0.CO;2-Q.

Peach, C. (1996). “Good segregation, bad segregation”, Planning Perspectives, 11(4), pp. 379–98, https://doi.org/10.1080/026654396364817.

Peach, C. (2009), “Slippery segregation: Discovering or manufacturing ghettos?”, Journal of Ethnic and Migration Studies, 35(Aug), pp. 1381–95, https://doi.org/10.1080/13691830903125885.

Piore, M.J. (1986), “The shifting grounds for immigration”, The ANNALS of the American Academy of Political and Social Science, pp. 23–33.

Pottie-Sherman, Y. and R. Wilkes (2017), “Does size really matter? On the relationship between immigrant group size and anti-immigrant prejudice”, International Migration Review, 51 (1), pp. 218–50, https://doi.org/10.1111/imre.12191.

Reardon, S.F. and D. O’Sullivan (2004), “Measures of spatial segregation”, Sociological Methodology, 34(1), pp. 121–162.

Sager, L. (2012), “Residential segregation and socioeconomic neighbourhood sorting: Evidence at the micro-neighbourhood level for migrant groups in Germany”, Urban Studies, 49(12), pp. 2617–32, https://doi.org/10.1177/0042098011429487.

Sanderson, M.R. et al. (2015), “Are world cities also world immigrant cities? An international, cross-city analysis of global centrality and immigration”, International Journal of Comparative Sociology, 56 (3–4), pp. 173–97, https://doi.org/10.1177/0020715215604350.

Schelling, T.C. (1971), “Dynamic models of segregation”, Journal of Mathematical Sociology, 1(2), pp. 143–186.

Shannon, C.E. (1948), “A mathematical theory of communication”, Bell System Technical Journal, 27 (3), pp. 379–423, https://doi.org/10.1002/j.1538-7305.1948.tb01338.x.

Stewart, E.S. (2012), “UK dispersal policy and onward migration: Mapping the current state of knowledge”, Journal of Refugee Studies, 25(1), pp. 25–49, https://doi.org/10.1093/jrs/fer039.

Sunstein, C.R. (2017), #Republic: Divided Democracy in the Age of Social Media, Princeton University Press, Princeton.

UNHCR (2015), World at War. Global trends, Forced displacement 2014, UNHCR, http://unhcr.org/556725e69.html.

United Nations (2017), International Migration Report 2017. Highlights, Department of Economic and Social Affairs, Population Division, United Nations, New York,. www.un.org/en/development/desa/population/migration/publications/migrationreport/docs/MigrationReport2017_Highlights.pdf.

van Ham, M. and T. Tammaru (2016), “New perspectives on ethnic segregation over time and space: A domains approach”, Urban Geography, 37(7), pp. 953–62, https://doi.org/10.1080/02723638.2016.1142152.

Vertovec, S. (2007), “Super-diversity and its implications”, Ethnic and Racial Studies, 30(6), pp. 1024‑54, https://doi.org/10.1080/01419870701599465.

Weber, H. (2015), “National and regional proportion of immigrants and perceived threat of immigration: A three-level analysis in Western Europe”, International Journal of Comparative Sociology, 56(2), pp. 116–40, https://doi.org/10.1177/0020715215571950.

Wong, D.W.S. and S.-L. Shaw (2011), “Measuring segregation: An activity space approach” Journal of Geographical Systems, 13(2), pp. 127–45, https://doi.org/10.1007/s10109-010-0112-x.

Wright, R.A., M. Ellis and M. Reibel (2008), “The Linkage between immigration and internal migration in large metropolitan areas in the United States”, Economic Geography, 73(2), pp. 234–54, https://doi.org/10.1111/j.1944-8287.1997.tb00069.x.

Zelinsky, W. (1971), “The hypothesis of the mobility transition”, Geographical Review, 61(2), pp. 219‑49, https://doi.org/10.2307/213996.

Zelinsky, W. and B.A. Lee (1998), “Heterolocalism: An alternative model of the sociospatial behaviour of immigrant ethnic communities”, International Journal of Population Geography, 4(4), p. 281.

Zhou, M. (2006), “Revisiting Ethnic entrepreneurship: convergencies, controversies, and conceptual advancements”, International Migration Review, 38(3), pp. 1040–74, https://doi.org/10.1111/j.1747-7379.2004.tb00228.x.

Annex 4.A. Calculation of the segregation and diversity indexes

The calculation of segregation along the two dimensions of isolation and clustering was done using the Spatial Isolation index $I_{A}$ and the Spatial Information Theory Index H defined by Reardon and O’Sullivan (2014) and using the R package seg.

The spatial Isolation of a group A is defined as:

$I_{A} = \int_{p \in R}^{} \frac{D_{A p}}{T_{A}} P_{A p l o c}$

Where:

$D_{A p}$ is the population density of the group A at the point p (cell centroid)
$T_{A}$ is the total population in the region R (LAU)
$P_{A p l o c} i$ s the proportion of a group A in the local neighbourhood for the point p

$P_{A p l o c} = \frac{P o p u l a t i o n o f a o r i g i n g r o u p A i n t h e l o c . n e i g h .}{T o t a l p o p u l a t i o n i n t h e l o c . n e i g h .}$

The Spatial Information Theory Index H is defined as:

$H = 1 - \frac{1}{T E} \int_{p \in R}^{} D_{p} E_{p}$

Where:

T is the total population in the region (LAU)
$D_{p}$ is the population density at point p
$E_{p}$ is the Entropy of the local neighbourhood for the point p for G mutually exclusive groups:
${E_{p} = - \sum_{G = 1}^{G} (P_{A p l o c}) \log_{G} (P_{A p l o c})}_{}$
$E$ is the Entropy of the total population for the entire region R:
${E = - \sum_{G = 1}^{G} (P_{A}) \log_{G} (P_{A})}_{}$

$P_{A}$ is the proportion of the group A:
$P_{A} = \frac{P o p u l a t i o n o f t h e A g r o u p}{T o t a l p o p u l a t i o n}$

The input dataset is the residential population by origin at 100 m x100 m resolution. The grid data is considered as a population density surface. It is assumed that the population data is located at the centroid of the cell and that the population is uniformly distributed across the entire cell.

The local neighbourhood population of each grid cell corresponds to the total population included in a search radius (max distance) from the cell centroid weighted using a distance function:

$w (d) = e^{- 2 d}$

where $d$ is the Euclidean distance between two cell centroids.

The Spatial Isolation Index ranges between 0 and 1. The highest value of isolation of 1 is obtained when all local environments of a LAU are composed only by a migrant group and no natives. The Information Theory Segregation Index again ranges between 0 and 1. The maximum value of 1 is obtained when each local environment has only one group of migrants, and the minimum value of 0 when each local environment has the same composition of the LAU.

For the calculation of the diversity indexes, the Shannon entropy index which is calculated according to the following formula:

${H = - \sum_{i = 1}^{R} p_{i} \log p_{i}}_{}$

where p_i is the proportion of population in the region (LAU or FUA) of a given country of origin i.

Annex 4.B. Regression estimates for clustering and isolation of migrants

Sample: 24 495 combinations between 9 297 LAU in France, Ireland, Italy, the Netherlands, Portugal, Spain, the United Kingdom and 115 countries of origin.

Method: OLS regression

Explanatory variables:

Relative size of community: migrant population by country of origin/total population of the LAU;
Size of city: total population of the LAU;
Diversity of the city: Shannon entropy index of the LAU;
Contiguity country of origin: dummy variable, equals 1 if countries of origin and residence have a shared border;
Distance country of origin: weighted geographical distance between the country of origin of the migrants and their country of residence;
Refugees/Migrants: share of refugees in respect of the total migrant population in 2010.

	Clustering	Isolation
Relative size of community	-0.0778*** (0.0010)	0.0500*** (0.0003)
Size of city	0.0511*** (0.0009)	0.0001 (0.0003)
Diversity of the city	-0.0282*** (0.0010)	-0.0058*** (0.0003)
Contiguity country of origin	-0.0543*** (0.0014)	-0.0108*** (0.0005)
Distance country of origin	0.0231*** (0.0011)	0.0017*** (0.0004)
Refugees/Migrants	0.0193*** (0.0010)	0.0010** (0.0003)
R2	0.6817	0.6521
Observations	23402	23402

Note: Stars indicate statistical significance of the variable (* P ≤ 0.05, ** P ≤ 0.01, *** P ≤ 0.001).

Both models include fixed effects for countries of destination to control amongst others for systematic differences in the segregation indexes due to the characteristics of the original data sets.

Explanatory variables are rescaled to have a mean of 0 and a standard deviation of 0.5.

Source: Elaborations based on data sources detailed in section “Data processing and measurement”.

Notes

← 1. The statistical data for Israel are supplied by and under the responsibility of the relevant Israeli authorities. The use of such data by the OECD is without prejudice to the status of the Golan Heights, East Jerusalem and Israeli settlements in the West Bank under the terms of international law.

← 2. Throughout the text the term city is often used in alternative to LAU. The LAU represent the main unit of analysis of our study. In some cases, LAU may correspond to portions of a large city (see London or Paris) and this determines the absence from charts and tables of specific measures for such cities. LAU stands for Local Administrative Unit, while FUA is a shorthand for Functional Urban Area. For a more extensive explanation, see (OECD, 2013). Eurostat defines LAUs as “a low level administrative division of a country, ranked below a province, region, or state. Not all countries describe their locally governed areas this way, but it can be descriptively applied anywhere to refer to counties, municipalities, etc.” (Eurostat, 2015).

← 3. Particularly in the context of a comparatively high incidence of single-parent families.

← 4. Musterd reaches similar conclusions for the Surinamese in Dutch cities (Musterd, 2005, 335).

← 5. “The ‘melting pot’ model […] envisages a progressive assimilation and convergence of the autochthonous white and minority populations over time. Its spatial concomitant is a progressive reduction in the level of segregation of the minority from the rest of the population over time” (Peach, 1999, 320).

← 6. “The structural pluralist model, on the other hand […], envisages economic integration, but also social distinctiveness or closure, which would be manifested in continuing high levels of spatial segregation. Upward movement in class terms would not produce regular spatial diffusion throughout the class, but the maintenance of distinct ethnic enclaves within the class” (Peach, 1999, 320).

← 7. Musterd (2005b, 332) defines ethnic segregation as “the spatial separation of population categories that are characterized by different countries of origin”

← 8. The measures of segregation - along the two dimensions of isolation/exposure and clustering/evenness - presented in this report are respectively based on the Spatial Isolation Index and the Information Theory Segregation Index (Reardon and O’Sullivan 2004). The Spatial Isolation Index ranges between 0 and 1. The highest value of isolation of 1 is obtained when all local environments of a LAU are composed only by a migrant group and no natives. The Information Theory Segregation Index again ranges between 0 and 1. The maximum value of 1 is obtained when each local environment has only one group of migrants, and the minimum value of 0 when each local environment has the same composition of the LAU.

← 9. This makes this measure directly dependent on the size of the group (this is not the case for the clustering measures because they are relative), as explained in Reardon and O’Sullivan (2004).

← 10. Portugal was excluded from the calculation of the diversity indexes since the original data from the census statistics was only providing figures aggregated by continent and not by specific countries of origin.

← 11. Portugal was excluded from the calculation of the segregation indexes since the original data from the census statistics was only providing figures aggregated by continent and not by specific countries of origin.

Featured topics

Agriculture and fisheries

Climate change

Development

Digital

Economy

Education and skills

Employment

Environment

Finance and investment

Governance

Health

Industry, business and entrepreneurship

Regional, rural and urban development

Science, technology and innovation

Society

Taxation

Trade

Energy

Nuclear energy

Transport

Featured topics

Agriculture and fisheries

Climate change

Development

Digital

Economy

Education and skills

Employment

Environment

Finance and investment

Governance

Health

Industry, business and entrepreneurship

Regional, rural and urban development

Science, technology and innovation

Society

Taxation

Trade

Energy

Nuclear energy

Transport

Countries A - C

Countries D - I

Countries J - M

Countries N - R

Countries S - T

Countries U - Z

Regional and global engagement

Countries

Countries A - C

Countries D - I

Countries J - M

Countries N - R

Countries S - T

Countries U - Z

Regional and global engagement

Publications

Publications

Featured publications

Data

Data

Featured data

News & Events

News & Events

Featured Events

About OECD

About

Engage with us

Work with us

Featured topics

Agriculture and fisheries

Climate change

Development

Digital

Economy

Education and skills

Employment

Environment

Finance and investment