Please note. A slightly altered and edited version of this paper was published as:
Musterd, S. & W.P.C. van Gent (2016) "Changing welfare context and income segregation in Amsterdam and its metropolitan area, 2004–2011." In: T. Tammaru, S. Marcińczak, M. van Ham & S. Musterd (Eds.), Socio-Economic Segregation in European Capital Cities; East meets West. (London: Routledge).
Please refer to book for correct citation and page numbers.
Segregation based on income differences is central to this chapter. We aim to understand the most recent (2004-2011) socio-economic segregation process in (metropolitan) Amsterdam, while connecting it to the changing Dutch welfare regime, which from around 1990 is moving into a liberal direction. The social rented sector and the ‘regulated’ parts of the private sector are under pressure, while owner-occupancy is growing. Social housing is increasingly catering for just those who have a very low income. Since housing is not spatially distributed in an even way, this affects the level of socio-economic segregation.
When it comes to spatial inequality in Western European cities the need to address high concentrations of poverty have been recurring political memes over time. While this is understandable from a socialist or social democratic point of view, it is also of concern for social and conservative liberals. The aversion against segregation among centre-right politicians, policy makers and commentators might appear surprising as their advocacy of free market reign contradicts the notion of intervention and regulation to limit spatial inequality. A liberalised housing market inherently means more segregation based on wealth and income. It is important to consider this because over the past two decades Western Europe (and beyond) has clearly been confronted with a neo-liberal turn (Peck & Tickell 2002; Harvey 2005). Neo-liberalisation has placed the interest of capital over that of labour, resulting in a political project that seeks to restructure welfare state arrangements. To be clear, neo-liberalisation does not imply less state intervention but a re-arrangement of state institutions, policies and public resources to serve market interests (Harvey 2005). As an important factor in current financial capitalism, housing and housing policy are key to this restructuring and may serve as a lever (Malpass 2008; Ronald 2008; Aalbers & Christophers 2014). Expanding owner-occupancy allows for cuts in health care and pensions, as growing housing equity and lower housing costs in later life are presumed to cover more of individual welfare expenses. Furthermore, in financially liberal countries, the expansion of owner occupancy also serves interests of financial actors and pension funds dealing in mortgage loans, derivatives and securities (Schwartz & Seabrooke 2008). So, the expansion of owner occupancy is not only a symptom but also a cause of welfare state change.
Still, why would those who advocate free market capitalism also rally against segregation? One explanation is that politicians insufficiently define key concepts to leave room for manoeuvre in political discourse. They might have mixed up segregation with social inclusion and exclusion. Also, it may be that many do not fully grasp how socio-economic segregation comes about. In addition to political manoeuvring and ignorance, the practice of desegregation policies is also in line with the wish to expand owner occupancy. In the last two decades, policies which seek to tackle segregation and poverty in European cities have come in the form of area-based interventions (Van Gent, Musterd & Ostendorf 2009). These interventions have sought to alter designated deprived areas’ social compositions through renewal and housing market restructuring. This typically entails lowering the share of affordable and social rental dwellings in favour of more expensive and owner occupancy housing to accommodate more middle class. For this reason, gentrification scholars have cast social mixing policies as a form of state-led gentrification (Lees 2008; Bridge, Butler & Lees 2012). Indeed, together with privatisation of social housing citywide and the lowering quotes for new-built social housing, social mixing policies are part of a policy which seeks to change the city’s tenure and social composition, effectively reducing the number of low income households (Van Gent 2013).
Yet, explanations which rely on the political economy of capitalism as just explained are insufficient to fully understand the call for ‘undivided cities’ in the Netherlands and elsewhere in Western Europe. Here, anti-segregation policies are related to the issue of integration of immigrant communities (Van Kempen & Murie 2009; Van Eijk 2010; Uitermark 2014). While politicians and policy-makers express the need to manage ethnic concentrations, anti-discrimination laws prohibit targeting ethnic groups directly. Instead, urban policies are formally targeting social-economic differences. Regardless, both political rhetoric and policy ‘wisdom’ assume a strong causal relation between ‘social’ and ‘ethnic’ segregation (Andersson 2006; Van Gent, Musterd & Ostendorf 2009). It seems irrelevant to them that this assumption can be challenged, first, because an increasing share of ‘migrants’ earns a middle or high income; and second, because the socio-economic position only explains ‘ethnic’ segregation to a limited extent (Peach 1999).
In this chapter we provide a recent description of social inequality (in particular income inequality) in the metropolitan area of Amsterdam and try to reach a better understanding of socio-economic segregation changes. We will also refer to spatial segregation of immigrants based on country of origin broken down by income category to investigate to what extent ethnic and income spatial inequality relate to each other in the Amsterdam context. We will analyse the spatial patterns and changes of levels of segregation for (grouped) income deciles and quintiles by applying standard indicators of evenness and isolation; this will be followed by a context-sensitive section in which we dig deeper in terms of analysing social spatial patterns of segregation of low-income households. Before turning to these empirical sections, we will first elaborate on a selection of the literature in which theories of socio-economic segregation and changes of patterns have been dealt with that may help to frame the empirical findings.
Framing the dynamics of socio-economic segregation
Studies on the explanation of segregation are rooted in different approaches to science. They include viewpoints developed in behavioural science, including those derived from human ecology and choice-based research (Robson 1975; Clark 2009); in structural theory, with reference to economic restructuring, globalisation, and mismatch theory (Wilson 1987; Hamnett 1994); and in institutional theory, much inspired by Esping Andersen’s (1990) work, which also was the basis for a new understanding of urban social segregation (Musterd & Ostendorf 1998). Theories on segregation also refer to the impact of historically grown and path dependent developments (Burgers & Musterd 2002).
Here we argue that individual households will try to realise some preferences based on life style and, more importantly, income and wealth. However, they will be constrained by the opportunities that are available, by economic and other structures that have developed over time, and by rules, regulations and institutions that affect the development of specific social and physical spatial structures. These dimensions are not independent from each other. Welfare state restructuring and a reliance on owner occupancy for welfare provision will influence the distribution of economic capital within and between generations, and will therefore influence the level of choice people have (McKee 2012). We argue that especially the type of welfare state is crucial, because this may change the attitudes, values, rules, and opportunities; the type of welfare state also impacts on how economic restructuring and globally organised forces ‘play out’ in certain places. Effects may be cushioned through redistributions (through social and unemployment benefits, high-level minimum wages, collective pension systems, and housing subsidies or individual rent subsidies) – as is common in universal types of welfare states. However, effects may also get more pronounced, because the welfare state model at stake may amplify inequality. This seems to be more common in residual types of welfare states (Forrest & Murie 1988; Musterd & Ostendorf 1998).
Because we take the position that the type of welfare state takes on a special position in the social organisation of cities, the change of type is obviously regarded to be a crucial dimension. As mentioned before, several Western European states, including the Dutch, currently undergo a neo-liberal turn (Peck & Tickell 2002). Welfare state restructuring will likely create more choice for the affluent and less opportunity for those with less economic capital (Kadi 2014). This is expressed by the Gini coefficient. In a recent study on income inequality of the Netherlands, Salverda et al. (2013) found that over a thirty year period the Gini coefficient rose from 0.242 in 1977 to 0.284 in 2010 (p. 10). Much of the rise occurred just before 1990, followed by a new increase around 2001; in between these years and after 2001 the coefficient turned out to be fairly stable, between approximately 0.280 and 0.290. However, the authors showed that the tails of the distribution went further apart. Moreover, the upper income decile owns a fairly stable 70 per cent of total financial wealth (p. xi). The literature suggests that those who have a stronger economic position use that position to match their own characteristics as much as they can to the characteristics of the environment, which will result in higher levels of segregation, especially in liberalised housing markets (Reardon & Bischoff 2011; Clark, Van Ham & Coulter 2013; Boterman & Van Gent 2014). This is helping the development and sustenance of socio-economically homogeneous residential environments (Sampson 2012). In a recent study, Musterd et al. (2014) found additional support for such types of dynamics. Based on large-scale data at individual level for the entire population of the four largest urban regions of the Netherlands, they provided new evidence for the homogenisation hat is currently evolving. They found that the larger the ‘income distance’ (positive or negative) between an individual and the median income of their residential neighbourhood, the higher the odds that the individual will move from that neighbourhood. Individuals who move tend to select destination neighbourhoods that reduce the social distance between themselves and the destination neighbourhood.
The liberalisation of urban housing markets has particular strong impact on the behaviour of middle class and upper class households. Compared to other households, they are more than others able to realise their wish for living in socioeconomically homogeneous environments. Qualitative research has shown that middle class households do not always directly refer to a wish to better match their own individual income with that of their residential environment, but use indirect arguments for such behaviour. Bridge (2006) and Boterman (2013) have shown that many middle-class households with children mention the quality of schools as among the prime reasons for moving into a homogeneous middle-class neighbourhood. The gentrification literature places the core of its attention to this middle class behaviour and politics, with a particular interest in the issue of displacement (Smith 1996; Lees 2008, Bridge, Butler & Lees 2012). Whether the gentrification process would entail forced displacement has been disputed in some contexts, especially where households with relatively moderate socioeconomic positions are more firmly protected by institutions and regulations (e.g. Van Weesep & Musterd 1991). This was typically the case for the Netherlands and other states known for their regulated housing market and practices in support of households with relatively limited economic capital. This is not to say, however, that low-income households who are living in gentrification neighbourhoods will always stay at that level. Recent research has shown that many households with relatively low incomes rapidly realise social upward mobility after settling in such neighbourhoods (McKinnish, Walsh & White 2010; Teernstra 2014). Eventually, this may result in higher levels of homogeneity and ultimately in higher levels of segregation.
The demise of strong welfare states in which redistribution of affluence and social solidarity were key elements explains why ever firmer concepts are being used to describe the processes of segregation. When inequalities in society are increasing, this may ignite circular processes that further contribute to rising inequality. Social inequality is said to be related to negative developments in a range of spheres. Wilkinson & Pickett (2010) showed evidence that in more unequal contexts crime levels are higher, health conditions of part of the population are worse, housing opportunities for the poor more limited, etcetera. This may create more fear for ‘the other’, and further disaffiliation of the middle class from the rest of the population. As a result we may see the development of so-called encapsulated homogeneous environments, insulated gated or semi-gated communities (Atkinson 2006). Boterman & Musterd (2014) suggest that this stretches beyond the residential sphere. They studied ‘cocooning’ behaviour and looked at segregation from the perspective of individuals who are exposed to diversity. They investigated levels of disaffiliation (withdrawing in one’s own socially homogenous environment) of the middle class in residential, workplace and mobility spaces and concluded that higher income natives are the most frequently cocooning population category, in all of the domains. Obviously, when exit strategies in the form of living in gated communities, cocooning, disconnection, or disaffiliation are taking extreme forms, such as in some North and South American metropolises, very negative processes of estrangement may occur, resulting in the rise of the number and size of no-go areas. This will eventually impact on the urban condition and result in a devaluation of life in urban settings (Atkinson & Blandy 2013). This will be harmful to the poor, especially those with poor and very local – primary ties – networks, as Marquez (2012) has shown for São Paulo, but fear for certain urban spaces will also reduce the use of urban territory for the middle-class itself.
So far, the experience in (continental) West-European cities is different. Here, state intervention has been more extensive over a longer period of time, assuaging segregation levels. Even though these welfare states are changing, they still experience the legacies of a period of strong intervention. They continue to have institutions and practices that help to reduce difference and segregation and continue to create difference from contexts such as those of the US (Van Kempen & Murie 2009). It seems that irrespective of the wider societal changes and regime shifts, not all middle-class households develop firm disaffiliating behaviour. Based on a series of interviews, Andreotti, Le Galès & Moreno Fuentes (2013) argue that even the upper middle class is applying only ‘partial exit’ strategies. They seem to be willing to share some dimensions of life with other social groups, while simultaneously searching for stronger separation in other domains of life. They conclude that “The responses of our interviewees were consistently more nuanced and complex than any simplistic theory about a drive to free-ride or withdraw from society” (p. 594). Even under ‘neo-liberalism’ there may be limits to the extent people want to segregate themselves from others.
Overall, the discussion of the literature, combined with the perhaps somewhat special situation we are coping with in the Netherlands, more precisely Amsterdam, triggers the first question to be answered in this chapter:
To what extent do we (as expected) find strong and increasing segregation between social classes in the metropolitan area of Amsterdam?
The mechanisms that are producing segregation are not all operating at the same level. Individual preferences of those who are able to exert some choice may result in search processes in which a certain level of social homogeneity is aimed for; this will likely occur at small area scale and become visible in segregation levels at that scale. However, when institutions at a state level play a significant role, this may not just have consequences for the local level of segregation; it may also imply that larger areas such as the entire metropolitan area are affected. In contrast, when due to liberalisation and decentralisation the implementation of uniform regulation principles is not obligatory anymore regional variation in levels of segregation within metropolitan areas may develop. Due to historically grown structures this may lead to sharper distinctions between the core city experiences (with a long and varied urban history) and the rest of the metropolitan region (with often shorter and less varied urban histories). Moreover, the core areas in larger Dutch cities have experienced long periods of social democratic governments. Although the local political arena has changed over the past decade, the welfare state legacy remains. Its lasting impact is noticeable in social housing and part of the private rental sector which are incorporated in the housing allocation system; in Amsterdam, in 2015, this still was way over fifty per cent. The city of Amsterdam also still owns almost ninety per cent of the land. This has effectively enabled the realisation of mixed housing programmes in small scale neighbourhoods. Nevertheless, over recent years we see new discussion about land ownership with pleas for selling land to individual home owners. Social housing too is under considerable pressure (Boelhouwer & Priemus 2014). These trends reflect and accelerate a wider transformation process towards a more neo-liberal urban regime (Van Gent 2013), which may already be manifest in new levels of segregation at certain scales. This results in the second question:
Is there a difference between levels of segregation at different levels of scale?
Apart from focusing on changing levels of income segregation, on individual level interactions between income and ethnicity, and on the geographical scales at which processes occur, segregation issues can also be approached in a different way if detailed spatial level data is available. In the second empirical section we will analyse micro-level data on the spatial distribution of households classified in terms of income. In that section we will particularly focus on the lowest income category: households with a so-called minimum income (limits set by law), typically the very poorest households, and on the municipality of Amsterdam only. We will also be able to investigate to what extent the changing patterns of concentrations of minimum-income households associate with the changing housing market structure. Two questions will be guiding that analysis:
To what extent are the lowest income households increasingly marginalised spatially?
What are the housing related factors that impact on the changing spatial pattern of low income households in Amsterdam?
We will provide answers to these questions in two empirical sections (below), after a short introduction of the data and methods used.
Data and methods
For answering the first two questions we used data from the SSD, the Social Statistical Database of Statistics Netherlands. This is an individual level longitudinal database (total population, no sample), which enabled us to calculate precise distributions of the disposable household income, to select only those who could, in principle, be employed (those between 25 and 65 years old), to combine income and country of origin information before aggregating the information to the neighbourhood level, to differentiate between calculation for the Amsterdam metropolitan region1 (in short: region) and for the municipality of Amsterdam separately, and to select only those neighbourhoods that counted at least 150 households (Figure 1). The average neighbourhood size was (in 2010) approximately 3,700 inhabitants, but in the city of Amsterdam almost double that size.
< Figure 1 about here >
We used income deciles and income quintiles for mapping location quotients and for calculating the index of segregation (IS), dissimilarity indices (D), the index of isolation P*, and, taking into account changing sizes of income groups over time, also the (modified) index of isolation at the regional and municipal level on the basis of the income information, for 2004, 2008 and 2011. This year-division enabled us to include the potential impact of the financial and economic crisis, which also hit the city and region of Amsterdam (Engelen & Musterd 2010).
The income groups are based on national income distributions. This means that the groups may vary in size at the lower regional and municipal level between categories and over time. Changing sizes may be the result of uneven population growth as a result of gentrification in Amsterdam’s city centre, urban renewal at the city’s periphery and new housing developments in city and region. Also, there may be an uneven spatial impact of the economic crisis at regional and lower levels after 2008. Indeed, Table 1 shows an increase of higher income groups between 2004 and 2008, followed by stabilisation, and an increase of lower income groups after 2008.
< Table 1 about here >
For answering the third and fourth questions we applied data provided by the Department of Research and Statistics of the municipality of Amsterdam. This regards six-digit postcode level data in which, among other data, the number of people with a minimum income per postcode was available, for 2004 and 2012. Amsterdam has approximately 18,000 six-digit postcodes – with on average some 40 inhabitants. Detailed spatial concentrations of households with a minimum-income have been constructed with a GIS application that has been developed by Urban Geography of the University of Amsterdam, together with the Department of Research and Statistics of the municipality of Amsterdam.
Empirical section 1: questions 1 and 2
The income dimension only
Starting with the (modified) index of isolation (Tables 2a and 2b) we see that in each of the years of measurement, households who find themselves in the lowest and in the highest income brackets (quintiles and deciles) are spatially relatively most isolated from the rest, whereas those who find themselves in the middle income categories are least isolated. The most isolated category is the category with incomes in the highest decile. This general picture is rather similar for the Index of Isolation and for the Modified Index of Isolation (MI).
< Table 2a and b about here >
The difference between isolation levels in the region and in the municipality is limited, except for the lowest decile and lowest quintile when calculating the Modified Index. The general picture is that those who find themselves in the extreme categories of the income distribution are relatively most isolated; this holds for the regional and municipal levels. The one exception is the lowest decile at municipal level when calculated as a Modified Index. That decile shows a low level of segregation. This will likely be due to (older) students who are living in relatively mixed neighbourhoods in the core areas of the metropolis. Also, smaller clusters of social housing in relatively affluent neighbourhoods may play a role. We will return to these areas below. At municipal level those who are in the middle and high income categories (except for the very highest decile) are a bit less isolated than at regional level. The highest income decile is most isolated at municipal level, both when calculated unmodified and modified.
Before the crisis started, levels of segregation defined as isolation, were stable or slowly increasing for all income categories, while after the start of the crisis for most categories the level was stable or slightly decreasing, except for those with the lowest income level (unmodified scores) who saw an increasing level. The pre-crisis experience seems to reflect the shift towards more neo-liberalisation, resulting in slightly higher levels of isolation. The limited drop in isolation of the poor may be a combined effect of the increasing share of the population in general that belonged to the poor and the policy of social mixing. The rising share of the poor after 2008 (unmodified) logically increases isolation, because the probability of encountering someone from one’s own (low income) category increases; when modified for the share in the region/municipality as a whole, the isolation in fact decreases (Table 2b). The policy of social mixing, which was particularly directed at neighbourhoods with the highest shares of low income households, is expected to have a dampening effect. The post-crisis experience is characterised by postponing the social mix policy in poor neighbourhoods. Urban restructuring in these neighbourhoods used to rely on the demolition of a substantial volume of old social housing, while rebuilding on the spot new and more expensive ownership housing, thus forcing more social spatial mix. However, with the crisis this policy came to a halt. In the Amsterdam region, demolitions dropped with 75 per cent between 2007 and 2011; in the municipality of Amsterdam with even 78 per cent, according to Statistics Netherlands. These processes implied increasing segregation levels for those with the lowest incomes (unmodified), but decreasing segregation for that category when modified. For the middle and higher income categories the crisis implied a strong reduction of residential mobility. Difficulties in selling housing for pre-crisis prices or in obtaining new mortgage loans meant that residential adaptation strategies were put on hold. Also, job loss likely led to a drop in income levels of the entire pre-crisis working population, regardless of place of residence. These two developments would lead to moderately increasing ‘spontaneous’ income mix in the neighbourhoods they lived in. This is in support of findings of Bailey (2012, p. 718), who states that “over time, the fit between individuals and their area weakens through social mobility. If people were not able to move, segregation would tend to fall as a result”.
In Table 3 we present the Index of Segregation for income quintiles and deciles. At regional and at municipal level, segregation levels for the lowest and highest income quintiles and deciles were almost stable before the crisis, and then dropping. The post-crisis postponement of residential migration clearly seems to have had a desegregation effect.
< Table 3 and Table 4 about here >
Such an effect of the crisis is also clearly noticeable when we focus in some greater detail at the dissimilarity levels between income quintiles for each of the three years (Table 4). Segregation of the first quintile category, relative to each of the other quintiles, is generally stable or in a downward direction before the crisis and then further down during the crisis. For the second quintile the pre-post crisis rhythm is stable, then down. For the fifth quintile we find that the rhythm is generally up, then down. These dynamics are shown for the regional level, but are similar for the municipal level.
Including the ‘native-migrant’ dimension
When the ‘native-migrant’ dimension is included in the analysis, a more complicated picture emerges, showing that there is not a one-to-one relation between income and the ‘native-migrant’ status. We confine the analysis to comparing the first income quintile and the fifth income quintile within and between ‘native-migrant’ categories, to allow for measuring effects of each dimension separately and jointly. Figure 2 provides the graphs for the various combinations of dissimilarities between sub-categories, for the region and the municipality.
< Figure 2 about here >
In general again a dampening effect of the crisis can be shown on levels of segregation. Highest levels of dissimilarity can be found for the most affluent natives relative to the migrant poor. The indexes are much higher than for the affluent natives relative to the native poor; and for the most affluent migrants relative to the native poor. Lowest levels of segregation can be found for affluent natives relative to affluent migrants. We also see an effect of the scale, between regional and municipal. Whereas in the region the dissimilarity index for the first income quintile for natives and the first quintile for migrants is stable, we see a slight increase within the municipality of Amsterdam. In addition, at regional level de D for the fifth quintile natives and first quintile migrants was stable and then downwards before and with the crisis, while at municipal level the segregation between these two groups was first increasing and then downward before and with the crisis. The fact that the segregation between the first and fifth income quintile for migrant categories is much higher than between the first and fifth income quintile for natives, seems to reflect the process which has been described by Wilson (1987) for the Chicago case: middle class migrants were fleeing the poor migrant neighbourhoods, with – in the Chicago case – very negative effects on those who were left behind. Simultaneously, the middle class migrants have been able to realise their housing ambitions.
The overall picture is that segregation in the municipality is generally a bit lower than at regional level, suggesting a stronger ‘melting pot’ at municipal level; however, just before the crisis started there was more often a tendency of increasing segregation at that level. This reflects the political changes whose social impacts seem to be more outspoken at municipal level than at the regional level.
Local patterns of three income categories, low-income migrants and high-income natives
The location quotients of low, middle and high-income categories are presented in Figures 3, 4 and 5. Relative to the share of low income households in the metropolitan region, low-income households are still clearly overrepresented in much of the older parts of the core municipality of Amsterdam (the core of the metropolitan area) and in large sections of the post-war social housing dominated neighbourhoods; with some exceptions they are vastly underrepresented in suburban areas (Figure 3). Middle-income households are overrepresented in suburban areas, and in large sections of the urban post-war neighbourhoods (Figure 4). The highest income categories seem to have found their own ‘niches’ both in some of the older neighbourhoods in the inner city (including the Canal Belt) and some adjacent areas, and in specific suburban areas (Figure 5).
< Figures 3, 4, 5 about here >
When income is (at individual level) combined with the ‘native-migrant’ dimension the patterns show a stronger contrast; these are almost ‘complementary’. Low-income migrants are overrepresented in specific post-war neighbourhoods and in some of the older 19th century neighbourhoods of the core city, and in a few neighbourhoods in more peripheral locations (Figure 6). High-income natives are overrepresented in four areas of the core city: the Canal Belt (A in Figure 1); the relatively new Eastern Docklands (B) neighbourhoods; the older inner-urban eastern neighbourhood Watergraafsmeer (C); and the south-western wedge from the inner city Canal Belt to the most important economic area (South Axis)(D). In addition some specific peripheral and suburban neighbourhoods show strong overrepresentation (Figure 7).