Class, migrants, and the European city: spatial impacts of structural changes in early twenty-first century Amsterdam

ABSTRACT Prevailing Anglo-Saxon theories on urban segregation based on class and ‘migrant-status’ have often been rejected for continental European cities, mainly because of different economic and labour market structures and higher levels of state interventions and welfare support in the latter type of cities. As urban economies in continental Europe are growing ever more global and welfare states are in continuous restructuring we seek to investigate whether a typical European socially balanced migrant city, the city (and metropolitan region) of Amsterdam, is developing into the direction of a more outspoken ‘double dual’ condition with populations getting more spatially segregated in terms of class and migrant status. This study looks at developments in terms of the spatial dynamics of the ‘native’ and immigrant population of different classes. We find that the region is undergoing a transformation, which for now reduces spatial concentrations and inequality. As the urban core is gentrifying and some suburban neighbourhoods are declining, the typical dichotomy of a poor-migrant central city versus affluent-native suburbs is vanishing. These developments point to a different type of social-migrant city, one with a patchwork of residential milieus along social and cultural lines. However, we challenge the sustainability of that patchwork over time.


Introduction
Cities across the globe attract migrants; they move from rural to urban places, from resource-poor to resource-rich cities, and between resource-rich cities. Migrants expect to find good or improved conditions to live a decent life in the places of settlement. Especially in cities that are well-connected to global economic networks, we expect to find resource-rich and resource-poor migrants, depending to a large degree on country of origin. In both cases, they are searching for a meaningful life, in the same urban environments (e.g. Sassen 1991Sassen , 1996. However, whereas the economic structure is a very important dimension for understanding social dynamics and migration processes, the resulting social and cultural conditions in the places of settlement are to a large extent structured by the State and its policies (Bourdieu 2005). Particularly in Europe, welfare regimes have strong impacts on urban conditions (Musterd and Ostendorf 1998;Tai 2006). Policy regimes may structure the economic field to a large degree, designating 'winners' and 'losers' and thus inequality; but they may also stimulate interventions aimed at reducing social and cultural differences through redistribution and housingrelated policies aimed at bringing diverse populations in close proximity in an effort to 'integrate' them. Depending on histories and direction, the outcomes of State policy may be reflected in lower or higher levels of spatial inequalities and patterns of segregation over time (Burgers and Musterd 2002;Tammaru et al. 2016).
In the Western European context, we see a shift in direction from collectivist strategies towards a new liberal 'common sense' (see Bourdieu 2005). Most states have experienced a (further) liberalisation of state policies, among other things leading to welfare state retrenchment, but often also leading to intensified state intervention and to stimulation of economic processes, notably at the urban level. This process of liberalisation seems to have accelerated with the post-2008 financial crises (see Peck, Theodore, and Brenner 2009;Streeck 2014). But also before the crisis, primary European cities have been confronted with less generous social security programmes; stricter regulated access to health services; and housing market policies focused more on private homeownership (Bourdieu 2005;Forrest and Murie 1988). More generally 'the market' is strengthening its position in the domains mentioned. It is important to consider that the increased reliance on liberal principles is developing simultaneously-yet not necessarily causally connected -with ongoing in-migration of various population groups and with a stronger urban orientation of middle-class professionals who are drawn to globally oriented service industries (Buzar, Ogden, and Hall 2005). Cities have seen their status rise. While European urban cores have always experienced relatively strong positions as centres of production, trade, politics, culture, and (residential) consumption, compared to, for example, American cities, the attraction of European cities has further increased over the past three or four decades. This followed the changing urban economic structure of these cities, away from labour intensive standardised manufacturing production, and moving to business services, consumer services, and creative industries. The new attraction of middle-class households, together with ongoing international migration of affluent as well as poorer migrants, in combination with neo-liberalisation, created the conditions for new social and cultural divisions and accompanying social and cultural spatial patterns in the city. If liberal regimes are associated with higher levels of inequality, then neo-liberal trends suggest increasing spatial segregation along lines of class and migrant status in European cities as well.
So, the question to be answered is: Can we find support for the assumption that economically well-developed, and globally well-connected urban areas with a strong, yet waning, 'social' welfare regime are currently developing into the direction of a more outspoken divided urban condition, in which populations are getting more segregated from each other in terms of class and migrant status?
The assumption referred to should be tested, since empirical evidence for the transformations described and for their impacts is rather thin. Consequently, local authorities may continue to regard high levels of social spatial segregation and migrant concentrations as problematic, leading to calls for 'social integration' and for the 'integration of migrants' (see Dukes and Musterd 2012), yet it is unsure what patterns of social and ethnic segregation are emergent and whether migrant concentrations may be affluent. Hence, the objective of this paper is to contribute to the understanding of how the social and cultural geography of the changing European 'migrant city' in the early twenty-first century is developing. More specifically, we are interested in what kind of changes have occurred in terms of the spatial and social (income) dynamics of the working-age population in the Amsterdam urban region (the Netherlands). The focus will be on 'natives' and on (first and second generation) 'migrants', as well as on lower, middle and higher social classes; and on the combination of migrant status and class. Micro-level data will be used to test assumptions derived from the theoretical insights which will be elaborated upon in the following section. In this paper we will evaluate the formulated question with data gathered for the Amsterdam metropolitan region. The Dutch capital is a typical 'migrant city', embedded in a still fairly universal welfare state, yet also experiencing transformations in terms of the population and housing market (Van Gent 2013). While using the concept 'city' we actually will analyse what happens in the city and urban region of Amsterdam. We have chosen Amsterdam because it a very international and connected city in an economically strong environment. The outcomes of the analysis may provide insights that could also be relevant for European cities such as Hamburg, Cologne, Lyon, Vienna, Zürich, Copenhagen, Oslo, and Stockholm; all experiencing growth of the number of migrants and changes in their (formerly strong) welfare regimes.
Before evaluating the assumptions, we will first, in the next section, elaborate on the theory about how structural conditions may impact spatial segregation of social classes, and of migrants versus 'natives'. We will also present the connection between the theory and the design of the analyses. This will result in the formulation of the research questions to be answered. Since we are looking at population processes as a way to test the theoretical assumptions, we have to take into account migration dynamics and social mobility, but also demographic changes that affect the population (Bailey 2012;Finney and Simpson 2009). After a brief discussion of the data and methodology we applied, we turn to the empirical investigations of the spatial dynamics of social and migrant categories in the city and region, provide information whether stronger spatial concentrations are developing or not, and to what extent there are differences between different class-migrant status groups.

Theoretical framework
Here, we will review how structural conditions and welfare regimes connect to the segregation of social classes and to the segregation of migrants and natives, also in combination with class positions. We will also discuss emerging demographic trends in cities and regions as well as the importance of socio-spatial dynamics.

Economic restructuring and urban reinvestment
We already briefly referred to the consequences of economic restructuring for cities. Typically, in European countries, economic restructuring that took place from around the 1950s, had major effects on the economic activities and social compositions. Labour-intensive and standardised production, supplying many 'blue-collar' jobs in manufacturing industries, were steadily replaced by capital-intensive, unstandardised, and specialised production. High-tech and 'white-collar' jobs started to dominate manufacturing industries. Moreover, business and consumer services activities grew in numbers, and also creative industries, and more generally, knowledge intensive industries experienced growth. The new types of activities generally fit better with urban environments. The nuisance from the older-often polluting-manufacturing production, which pusheds production to the edges of metropolitan areas, mostly disappeared, while creative-knowledge and services industries could function within the existing urban fabric. For some activities, accessibility (a location near highways) remained important; but for others, mainly those in the cognitive-cultural industries and services sectors, densely populated areas offered more opportunities for business interaction, symbolic value, connectedness, and proximity to workers. Importantly, the changing economic structure also implied new labour structures, with more middle-class professionals who preferred specific areas, built forms and public spaces for interaction and leisure (Burgers and Musterd 2002). The rising share of one or two person households, which developed parallel to the economic restructuring, contributed to the urban orientation of large parts of the population. This has, eventually, resulted in growing demands by various distinctive categories of urban middle-class households (Butler and Robson 2003;Hamnett 2003), and fuelled new gentrification processes in centrally located districts with ample access to a wide array of consumer functions (Lees, Slater, and Wyly 2007;Van Gent 2013). With their stronger income positions and rising demand, gentrifiers have increasingly come in competition with lower class urbanites, which, it is argued, resulted in replacement and displacement of population segments and new patterns of segregation (Lees, Slater, and Wyly 2007). In short, the economic transformations resulted in a wider attraction of urban life for various categories of middle-class households in an increasingly competitive urban space.

Welfare state and segregation
The welfare regime also played a role in structuring residential space and-thus-social and migrant types of segregation. In liberal welfare regime contexts, such as that of the US, where most of the urban theory on social and cultural spatial segregation originates, social inequalities and place of origin of their migrants seem to translate directly in urban space through spatial patterns in which social categories, and also cultural groups are rather separated from each other (Musterd 2005). Social space seems to be shaped by economic resources, cultural space by the language people speak, the religion they have, and the cultural attitudes they expose. In such a regime the state often facilitates entrepreneurialism. This will result in higher levels of social and cultural segregation (Phillips 2006;Rex and Moore 1967). The most affluent migrants, accompanied by a growing category of 'native' resource-rich inhabitants, will claim specific parts of the city and urban region. They have the resources to realise their residential objectives-and contribute to the upgrading of such areas. Poorer migrants and 'natives' will experience less choice and have to do with the remaining areas of the urban region. In short, differences between these migrant-and non-migrant-categories with regard to economic, cultural and religious capital may create dualities or inequalities, which reflect in the urban regional space where they settle.
In more or less social-democratic institutional contexts, many of which can be found in continental Europe, (state) efforts to reduce inequality are common (Esping-Andersen 1990; Musterd and Ostendorf 1998;Van Kempen and Murie 2009). The inequality debate is often translated into an integration debate; urban social-economic problems are defined as 'lack of integration' of the less well-off. Integration in its operational form frequently refers to participation in the labour market, in education, and in cultural realms. Regarding the cultural dimension we see 'multi-cultural' models of integration, whereby different cultures may-to a certain extent-continue celebrating their own original cultural norms, attitudes and behaviours; and 'assimilationist' models of integration, where newly arrived migrants are pushed to quickly adopt cultural norms, attitudes and behaviours of the country of destination (e.g. Alexander 2007). Through anti-segregation policies-both a-spatial and spatial-and pro-social mix policies, European states have made efforts to avoid the development of highly segregated or divided cities with fragmented social and cultural spaces. The result of these interventions is that social and ethnic segregation tends to be less pronounced in 'social democratic' forms of welfare states compared to 'liberal' welfare states (Musterd 2005;Arbaci 2007).
In short, theoretical insights from various contexts suggest that more socially and culturally segregated cities are developing in liberal contexts, such as the US and UK; and less socially and culturally segregated cities in social-democratic contexts, such as-so far-the Netherlands and Sweden. A key assumption is that such welfare regimes have their own specific associations with processes such as international migration, residential mobility, social mobility, social redistribution, and cultural dynamics, which will be reflected in social and cultural segregation outcomes.

New urban demographics
Residential patterns in the urban region are also related to a range of other (structural) processes. In the urban literature of the past half century we find a structural increase of the share of small one or two person households. These households have 'always' had an urban orientation. Therefore, it comes as no surprise that they gained territory in the most urbanised parts of European city regions. Conversely, family households lost territory in these areas. The outdoor behaviour of small households requires proximity to facilities and services they need. These can be found in dense and highly urban residential quarters. Households raising children moved to more spacious dwellings and neighbourhoods on the edges of the city or in suburban settings (e.g. Robson 1975). Recent research has shown, however, that patterns have been changing once again (see Buzar, Ogden, and Hall 2005). In Switzerland, Rérat (2012) identified three categories (on top of the small old and young (student) households) which are driving the (new) urban growth: international migrants who respond to economic opportunities, global connectedness, and crisis; new middle-class gentrifiers; and non-traditional households such as single-parent families and same-sex couples. Evidence from the Netherlands and Sweden suggests that not just middle-class small households, but also middle-class families are becoming more likely to settle in urban settings (Magnusson Turner 2008; Boterman and Karsten 2014).
Self-reinforcing dynamics: concentrations, self-selection, and non-migration change Before moving to the analyses, we are stressing that social and cultural spatial outcomes also function as structural forces in themselves. In a recent American study Sampson and Sharkey (2008) identified a structural pattern of mobility flows between residential environments, which appeared to reproduce ethnic segregation patterns. A Dutch study finds similar trends; immigrants from non-western origin were more likely to move to concentrations of non-western migrants and less likely to move out of them than natives (Bolt, Van Kempen, andVan Ham 2008, 1376). A cross-European analysis revealed that preferences to reside apart from ethnic minorities among native Europeans are fairly similar to preferences of whites in the US (Semyonov, Glikman, and Krysan 2007). Such discriminatory practices are supported by other European research outcomes, which signals 'white flight' and 'white avoidance', continuously feeding ethnic spatial segregation (Bråmå 2006;Hedman, Van Ham, and Manley 2011).
However, in France Pan Ké Shon (2010) found that spatial outcomes of residential processes were less connected to cultural and more to social positions. He argued that it was not 'white flight' or 'white avoidance', but a 'flight of all colours' that shaped ethnic residential space. In line with Rex and Moore (1967), when a certain social and socio-economic position was achieved, all ethnic groups moved out to higher status neighbourhoods. Pan Ké Shon argues that first generation Africans stayed behind because they find it harder to achieve socio-economic success. This suggests that in the French context social and spatial assimilation of migrants and economic success are highly related. Musterd and De Vos (2007) found a similar association between residential mobility and socio-economic positions of migrants in migrant clusters in Amsterdam. No migrant group is persistently staying in the same concentration. Those who moved out of clusters moved to areas with better quality housing and higher rates of ownership, which arguably reflects an improved socio-economic position. This interpretation is supported by other research, on residential behaviour of Surinamese and Antillean migrants in the Amsterdam region. Between 1999 and 2006 they moved in large number to the suburbs (especially to the New Town of Almere). Moroccan and Turkish migrants did not follow that pattern (Zorlu 2009;Musterd and Van Gent 2012). This is partly explained by the finding that family ties have a comparatively higher impact on the out-migration for second-generation Moroccans and Turks (Zorlu 2009).
These findings signify the importance of social-economic position in processes of migration and self-selection of neighbourhoods. Recent research in the four largest urban regions of the Netherlands shows that a focus on the social dimension may be relevant for the understanding of social spatial segregation as well (Musterd et al. 2014). Without looking at migrant status, the study investigated the relation between household income and the median income of the neighbourhood it moved from and moved to. It showed that a household's probability to move is higher when the distance between their own social position relative to the neighbourhood median social position is larger. Moreover, the social position better fitted the social level of the neighbourhood the household moved to than that of the neighbourhood left behind. The outcomes show that segregation tends to increase, as households are searching for socially homogeneous neighbourhoods on the basis of their own social position. Of course, spatial processes are also influenced by forms of capital other than economic or financial (social, cultural, etc.), which, in conjunction with characteristics of the residential environment, may other produce spatial outcomes. Consequently, at aggregate levels, social diversity, at least along some lines, may still ensue.
Lastly, much of the research cited here looks at residential mobility as the material cause of segregation. While moving is very important in both sustaining and changing spatial patterns (Sampson 2012;Musterd et al. 2014), recent studies have looked into the aggregate effects of in situ social mobility and ageing. Teernstra (2014) found that social upgrading in a neighbourhood, without moving, likely has a significant impact on the potential gentrification of that neighbourhood. This implies that the changing social spatial structure may also be triggered by social mobility 'on site' (also Hochstenbach and Van Gent 2015). For Scotland, Bailey (2012) demonstrated the importance of the ageing in and out of population groups to explain levels of social segregation, much like a 'demographic conveyor'. Likewise, in addition to migration, new residential patterns at the regional urban level may also result from a substantial second-generation immigrant population growing into adulthood (Finney and Simpson 2009).

Research questions
In sum, structural and institutional factors and changes therein, may produce and reproduce certain social and cultural spatial outcomes. Structural changes in the economic structure, but also shifting welfare regimes and demographic transitions will have impacts on a range of processes that shape and reshape class and migrant segregation in cities. How the different forces that operate at various levels will work out and how the residential structures are developing can be theorised, as we did, and may or may not be supported with empirical outcomes of segregation analysis. In the remainder of this paper we will show such outcomes for a primary European urban region which has been successfully adapting to economic change, a shifting welfare state, and increased attraction for resource-rich and -poor newcomers from domestic and foreign origins. The analyses that follow will reveal whether stronger concentrations of social or cultural ('ethnic') categories occur and how they are changing over time due to residential mobility in, out and within the region; due to demographic change, and due to social mobility of those staying in the city or region. The following empirical research questions will be answered in this paper.

Data and methodology
To assess social and socio-spatial inequality we will exclusively look at the behaviour of the working-age population (25-64-year-olds) in the Amsterdam region. This study draws on register data from the Dutch System of Social-statistical Datasets (SSD) for the years 2004, 2008, and 2011. Data were obtained from Statistics Netherlands (CBS). The SSD contains individual level register data on the entire population of the Netherlands, on income from work, benefits, student subsidies and pensions, as well as several individual characteristics such as neighbourhood of residence, migrant status, age, gender, and household characteristics. This study follows the neighbourhood classification of Statistics Netherlands. Our data set comprises of 302 neighbourhoods, which have at least 100 working-age individuals in 2004, 2008, or 2011 (the average neighbourhood population was 2770 in 2004). In this period, 16 neighbourhoods were added due to new residential development.
In the region of study, statistical boundaries are relatively stable over time, yet seven neighbourhoods changed boundaries. Residents in these areas where excluded from our concentration analyses. Furthermore, in Amstelveen one new neighbourhood was carved out of two older ones. These were combined to one in our concentration analysis, but excluded in the spatial analyses.
Lastly, to incorporate both social and migrant status dimensions we focus on six categories of the working population, combining the two dimensions. First, we distinguish between native-Dutch and migrants, following the definition of Statistics Netherlands. Migrants are individuals who were born abroad or have at least one foreign born parent. The second dimension relates empirically to the social-economic status (SES) and is based on household income from work, benefits, and (early) pensions as well as income from self-employment. This income measure is 'equivalised' to account for household size. We compare three income categories based on quintiles of the national income distribution. Low-income categories comprise of the lowest two quintiles. Middle-income categories fall in the next two quintiles, and high income is the top 20%. A fourth category was defined to account for missing income data, comprising 5.1% of the working-age population in the region in 2004 and 3.0% in 2011. As our income data are based on Dutch tax records, missing data are higher for migrants because a share of international employees pay income tax in their country of origin. The SES unknown category has been omitted from our presentation below, but have been included in all analyses.
The urban region of Amsterdam has been defined as Amsterdam municipality, including the adjacent urban municipalities of Amstelveen, Diemen, Haarlemmermeer, Ouderkerk aan de Amstel, Zaanstad, and Almere.
In the subsequent analyses we will describe the development of different social and migrant categories over the period 2004-2011. We included a 2008 measurement to be able to see pre and post global financial crisis start effects. This will be followed by an analysis of the development of spatial concentrations of the main categories we investigate. Through a series of maps we will then present the dynamic spatial orientations of the most important categories and finally will investigate the relative impact of underlying processes shaping the changes in the residential structure, focusing on demography, migration and in situ social mobility.

Social and migrant category development: shifts in city and region
Tables 1 and 2 display the distribution and relative change of social and migrant categories for the municipality and region of Amsterdam, for 2004Amsterdam, for , 2008Amsterdam, for , and 2011. Over this period, there has been sustained population growth in the Amsterdam region. Indeed, all income groups are showing growth (Table 2). Interestingly, the high-income group is growing most rapidly in the region as a whole. The growth in income groups is not equally strong within the region. If we split the region between Amsterdam and the rest of the region, it becomes clear that the highest increase rates are for high-income individuals in the city (+18.1%) and low-income individuals in rest of the region (+21.1%). We should bear in mind that the city of Amsterdam still has more low-income individuals within its borders than the other municipalities (Table 1). This is due to the high share of smaller and affordable housing (often social rent) in Amsterdam. When comparing growth in 2008-2011 with 2004-2011, we see the effect of the global financial crisis. The initial decline of low-income households in Amsterdam (between 2004 and 2008) and the moderate growth of middleincome households turned into stronger growth after 2008. This may indicate a reluctance to move due to economic uncertainty. High incomes continued to grow steadily before and after 2008. When compared with the rest of the region we see a strong increase of low incomes, stabilising development of middle incomes, and more moderate but steady growth of higher income households. These figures suggest that when there would not have been an economic crisis, the share of low incomes in Amsterdam would have further declined to the benefit of the rest of the region. Middle-income growth in Amsterdam would have been more moderate, while growth in the rest of the region would have continued. In contrast, the strong increase of higher incomes in Amsterdam and the more moderate growth in the rest of the region seem to be more structural processes.
In terms of migrant status, there is a similar trend. While the native-Dutch population stayed at roughly equal numbers in the region over time, population growth is mostly among the migrant groups. Consequently, the share of natives in the Amsterdam region population has dropped from 62% to 58% between 2004 and 2011. So far, the Amsterdam population has the highest share of migrants (48% in 2008). Yet, the increase of migrant categories is twice as high in the rest of the region. Between 2004 and 2011 the municipality of Amsterdam saw a 14.6% increase in migrants; in the surrounding region the increase was 29.6%. When combining income groups and migrant/native categories into six population categories, we see that growth trends for the city and surrounding region are differing as well. The data reveal some interesting differences for within migrant and native categories between city and region over 7 years. In Amsterdam we see declining numbers of lowincome and middle-income native-Dutch (4.5% and 3.8% decline), and sustained growth of native high-income individuals (+18.1%). The surrounding region also shows, surprisingly, a decline of native middle-income groups (-3.0%) and a steeper growth for low-income natives (+9.4%). Here again these processes might have been stronger without the economic and financial crisis. The trends regarding low and middle-income natives and low-income migrants in Amsterdam were negative before the crisis started but became positive after 2008. Likely, more low and middle-income people will leave the city when the economy will pick up again.
In short, while the image of a poorer and more diverse city still holds true in 2011, growth trends indicate a demographic and social shift between core city and periphery. The region is attracting more low-income households and is becoming ethnically more diverse. Amsterdam municipality is also still becoming more diverse but is attracting more high-income individuals. This process may have been tempered by the economic/financial crisis.

Development of spatial concentrations
Taking the six category distinction between population groups, we will investigate how the population shifts and changing patterns have affected concentrations of these groups. We Table 2. Relative change in % of income groups and social-migrant categories for working-age population in the municipality of Amsterdam, its surroundings municipalities, and the Amsterdam region, for periods 2004-2008, 2008-2011, and 2004-2011 will look at the development of living in one's own-group concentrations in the core city and in the Amsterdam region. With a growing population of migrants and an almost steady number of native-Dutch in Amsterdam proper and in the Amsterdam region, the question is to what degree these changes are leading to new residential patterns. The increase of the migrant population across the region may lead to the consolidation of existing concentrations and the formation of new concentrations of different income-ethnic categories in Amsterdam and surrounding municipalities. To compare population categories and developments over time, we have defined, for each population category, a neighbourhood as a concentration neighbourhood when the share of a population category of a neighbourhood population exceeds 1.5 standard deviations above the mean. Standard deviations were calculated based on all neighbourhoods with more than 100 working-age inhabitants for the entire region and for Amsterdam municipality. Shares refer to the category's percentage of the neighbourhoods working-age population. On a methodological note, shifting percentages may be the result of using standard deviations as a threshold. Changing population sizes imply shifting thresholds each year. To ensure robustness, we checked these with fixed threshold percentages as well (results not shown). These confirmed our findings, except for low-income native-Dutch in Amsterdam municipality. Table 3 shows the share of the total population living in own-group concentration neighbourhoods in Amsterdam and the region in 2004, 2008, and 2011. If concentration definitions are calculated for just Amsterdam municipality, we see that nearly all categories are de-concentrating. The only exceptions are middle-income migrants and low-income natives. The former are most concentrated within the city and seem to increasingly live in neighbourhoods where other middle-income migrants are concentrated. The latter group shows versatility related to changing size and shifting thresholds. When these are kept constant, the share of this group living in concentrations remains constant between 2004 and 2008 and increases between 2008 and 2011.
At regional level we see that, in 2004, low-and middle-income migrants and lowincome native-Dutch are the most concentrated groups. However, the trend lines show significant de-concentration for these groups, particularly for low-income native-Dutch. We see fairly stable concentrations for middle and high-income native-Dutch and highincome migrant groups.
To conclude, at both the municipal and regional scale nearly all categories show a deconcentration or stable concentration trend between 2004 and 2011. Nevertheless, increases are most likely not evenly distributed across the region's neighbourhoods. Therefore, in the next section we will gauge the reforming residential structure in more detailed geography.

Dynamics of spatial orientations of six population categories
In Figure 1(a)-(f), the structure and direction of geographical change are shown of the six population categories referred to before. The average refers to the distribution of neighbourhood proportions in 2004. Growth and decline are based on an increase or decrease of numbers. An above-average number of low-income natives (Figure 1(a)) (working population) mostly live in Amsterdam neighbourhoods, Zaanstad and in a few high density areas in Haarlemmermeer and Almere. Decline is mainly shown in the most central parts of the urban region. Those which show an increase are mostly student housing areas where some students are over 24 years old. The increases in the periphery seem ubiquitous. The suburban areas which show a decline are mostly newly built suburban neighbourhoods.
Like low-income native-Dutch, middle-income native-Dutch continue to show decreasing numbers in the city of Amsterdam. The only neighbourhoods with an increase in numbers are gentrifying areas, mostly in the nineteenth-century belt (Figure 1(b)). In the periphery, it is interesting to see that most neighbourhoods in Almere and Zaanstad show a decline in numbers, while quite a few neighbourhoods in Haarlemmermeer (South-West of the region) show an increase. This may be connected to a stronger local economy in the latter municipality due to proximity to the international airport.
Neighbourhoods with above-average shares of high-income native-Dutch are mostly located in the periphery (Figure 1(c)). The city of Amsterdam has a few high-end residential areas with pre-war town houses, and with recently redeveloped waterfront areas, which show as above-average. In addition, above-average shares in the periphery are mostly areas with a rural character (villages), and in newly developed suburban neighbourhoods. In terms of growth, the city centre of Amsterdam (within the ring road area) shows increasing numbers of high-income natives. This may be attributed to processes of gentrification and shows the new appreciation of urban living. High density post-war areas show a decrease, and so do the ageing kernels of post-war New Towns Hoofddorp, Nieuw-Vennep, and Almere.
High densities of low-income migrants are mostly within Amsterdam municipality, and, to a lesser degree, in Almere and Zaanstad (Figure 1(d)). As we have seen, this category is growing in numbers. Interestingly, within Amsterdam we see a decline in the nineteenth-century belt. These areas see an increase in middle and high-income groups (both native-Dutch and migrant, as shown below). Low-income groups seem to be displaced from these neighbourhoods.
Like their low-income counterpart, neighbourhoods with an above-average number of middle-income migrants can be found in Amsterdam, Zaanstad and the New Town of Almere (Figure 1(e)). Particularly, Amsterdam and Almere neighbourhoods are showing an above-average presence and growth. In addition, Diemen and Amstelveen also boast quite a few middle-income migrant neighbourhoods. The growth is nearly ubiquitous.
Lastly, high-income migrants show a different pattern compared with affluent native-Dutch as well as other migrant groups (Figure 1(f)). They are mostly located in the canal belt and affluent southern neighbourhoods of Amsterdam and in Amstelveen. Most of these affluent migrants originate from other relatively affluent countries. A few newly built areas in Hoofddorp and Almere also show an above-average presence of highincome migrants.

Relative impact of three underlying processes
The analyses above show differing spatial distributions of migrant population in the region of Amsterdam as well as varying growth trends. For a better understanding of these structures and processes we would like to know what is driving these spatial changes. Social patterning is driven by several processes, as we have seen in the theoretical section of this paper. In this section we will, in particular, be looking at three main processes that on their turn derive from structural and institutional changes as described in the theoretical section. These processes are shaping changes in population at the Amsterdam regional and municipal scale: . Demographic change refers to ageing in and ageing out the working-age population (25-64 years old) within the region or municipality, as well as deaths. . Migration refers to net migration in and out of the Amsterdam region or municipality.
For the municipality we also look at intraregional migration, meaning the net balance of moves between Amsterdam municipality and the regional periphery. . Social mobility refers to shifting of income categories of those who stay in the region or municipality. This is a zero sum for each population category.
In addition, there is an 'other' category which represents missing data with regard to neighbourhood of residence and, most importantly, income for either 2004 or 2011. As the quality of data is better in 2011, this leads to additional observations. The population balance for the Amsterdam region shows that demographic change, that is the changing age structure of the working population staying within the region for the entire period of seven years, is leading to a decline of population for low and middle-income native-Dutch (Table 4). High-income native-Dutch and all migrant  categories show growth in terms of demographic change. Migrant growth is likely due to larger families in some migrant communities and the ageing offspring of the low-skilled international migrants who settled in Amsterdam and Zaanstad in the 1960s, 1970s and 1980s. In terms of migration in and out of the region, we see a positive net balance for all categories. The influx is higher on the lower end of the income scale. Particularly low and middle-income migrants are growing because of migration into the region. As for social mobility, for those who stay, the net effect is upward social mobility. This is to be expected as the research population concerns a working-age population. At the regional level, we see internal upgrading for both natives and migrants, but the shift from low and middle income to high income seems larger for natives. However, in relative terms, native-Dutch high-income groups gain 3.3% from internal social mobility between 2004 and 2011 while migrant high-income groups gain 6.1%. When we just look at the municipality of Amsterdam (Table 5), the picture is somewhat different. For all native-Dutch income categories we see a negative impact of demographic change, whereas for migrants a positive impact can be shown. This implies that an ageing population also affects the city's socio-economic population structure.
As mentioned, we distinguish between two types of migration flows. First, the exchange between city and surrounding region, the intraregional migration, is leading to a loss of population for all categories. This shows that processes of suburbanisation are still ongoing. The net balance from outside the region to the city is roughly similar to that in the region as a whole. Particularly, low-income groups, often relatively young, migrate to the Amsterdam region from the Netherlands and beyond. Indeed, when comparing the inflow into the municipality to that of the region, which is shown in Table 6, we see that immigration from outside the region to the region is mostly focused on Amsterdam municipality; this is particularly the case for low-income groups.
As expected, the process of upward social mobility is also present in the municipality. For native-Dutch the internal shift from low and middle income to high income is more pronounced than for migrants, which shift from low income to middle and high income in lower numbers.
In short, there are four mechanisms which help to understand the changes and changing population geography in the Amsterdam region. First and foremost, there is the impact of the positive net migration of migrants to the region and (especially) the city. There are large net gains for migrants, in relative and absolute terms, compared to native-Dutch groups. There appear to be different demographic trends among the native-Dutch and migrant populations. Native-Dutch are ageing out of the working-age population (also a cohort effect in Almere) without sufficient young people ageing within the region. Conversely, migrants are relatively young and are ageing in greater numbers. There is a higher internal upward social mobility of migrants in the municipality and region when measured as relative growth in high income. Also, migrants are gaining position in middle and higher income groups while natives are only gaining in the highincome category. Finally, we also found that there is sustained suburbanisation going on: not only middle and high-income natives continue to move to suburban municipalities, but also migrants, and to the same relative degree (Table 6).

Discussion and conclusions
In this paper we investigated processes of change in Amsterdam to enhance the understanding of the changing (spatial) positions of households according to their social-economic position and immigrant status in view of economic change, welfare state change and demographic transitions. We posed the question whether in light of these changes, we would see a more segregated city and region. Indeed, some trends are in line with our expectations. In socio-economic terms, central urban areas are growing more affluent, particularly those in Amsterdam's pre-war centre, which is arguably related to economic success, housing deregulation and an influx of middle-class households. Also, the migrant working-age population is growing, but this is not only the result of in-migration but also uneven ageing processes. Yet, rather than merely accentuating existing segregation patterns and concentrations, we also see new socio-spatial patterns emerging, indicating a transformation of the social relationship between various household categories and between the core city and the surrounding region of Amsterdam. The newly forming social geography can no longer be understood by looking at migrants and social-economic groups separately, or by conflating migrant groups and low-income groups. Simply put, the city is no longer a typical residential domain of low-income groups and migrants, nor is the rest of the region a typical residential domain for middle and high-income groups and 'natives'. It is becoming more affluent and in some gentrification areas the share of native-Dutch is increasing. Conversely, the surrounding region has seen an influx of migrants, particularly low-income migrants, and is slowly losing its homogeneous middle-class native-Dutch character, a group long associated with suburban living.
We showed that the transformation process goes along with a spatial de-concentration of most of the categories we distinguished (combinations of socio-economic and migrant status groups). The question is, however, whether this 'social mix' is sustainable or whether we are witnessing a tipping point phase of a process which is heading to new segregation and new homogeneous milieus. The continuing trend of liberalisation in general and housing market liberalisation in particular, low rates of new development, and sustained demand for housing in Amsterdam, suggest that population structures and the relationship between city and suburbs will continue to shift in the direction as described above. If this will be combined with ongoing social spatial processes leading to social homogeneity, as shown in recent research, the next phase of the diversifying European migrant city may be another type of migrant city. To be clear, these qualifications are to be understood grosso modo. Suburban affluence will not disappear and the city will likely continue to house a large share of low-income and migrant households due to social and physical structures related to housing market characteristics, reputations and enduring social fabric (see Sampson 2012), as well as due to the rich opportunity structure a city has to offer for first-time entrants from abroad (Musterd et al. 2008). Moreover, cities and urban regions are varied in themselves, in many respects. There are big differences between residential environments as a result of planning and social histories. At the population side, a big role may be played by the national and local welfare regimes. These have, in principle, the capacity to avoid large inequalities, in terms of resources, and in terms of access to good quality housing for many, and not just a few. The organisation of the welfare regimes, therefore, must be taken into account, next to globalisation and economic restructuring forces.
Our analysis raises a number of questions and venues for further research. First, while we place a theoretical emphasis on structural factors, our analyses have mostly focused on documenting and understanding the population and socio-spatial changes. Future analyses should be complemented with in-depth research in which the various mechanisms that operate between the structural and institutional level and the 'making of social and cultural segregation on-the-ground', is required to create a more comprehensive theory of segregation. In this sense, comparative urban studies may also help in revealing salient mechanisms and outcomes (cf. Tammaru et al. 2016).
Second, our study focused on class and migrant status and found evidence for more diversity, at least for the moment. We should point out that these are not the only markers of social diversity. Although methodologically challenging, a full analysis of the urban socio-spatial fabric would have to include life course, class traits other than income (e.g. cultural capital), and even sexuality. These three are arguably affecting on individual mobilities distinctively and, consequently, on segregation patterns.
Third, considering that housing market structures are rigid and demographic and mobility processes take time, the shifts presented above are relatively persuasive. The question is whether the population shifts in the region, particularly the growth of low-income migrants in the urban periphery, is without political effects. Recent evidence suggests that support for radical right-wing parties is contextually higher in suburban areas close to older cities like Amsterdam (Van Gent, Jansen, and Smits 2014). Further research is needed to investigate whether the regional shift feeds suburban support for populist parties which use vitriolic discourse in relation to migrants groups.