Neoliberalization and urban redevelopment: the impact of public policy on multiple dimensions of spatial inequality

ABSTRACT

This paper examines the impact of public policy on different dimensions of spatial inequality. We not only study residential segregation but also housing market access and inequality in terms of neighborhood status. We chart the impact of urban redevelopment policies in two Dutch cities—Amsterdam and Rotterdam—through a unique longitudinal and full-population dataset that enables us to distinguish the contributions of demolition, new construction, and tenure conversion to various dimensions of spatial inequality. We find that policy measures that reduce segregation may reduce access to housing (as happened in Amsterdam) while measures that promote upgrading may exacerbate inequalities between neighborhoods (as happened in Rotterdam). Distinguishing between different kinds of policy measures and dimensions of spatial inequality, we argue, allows for a more nuanced and comprehensive understanding of urban redevelopment and better insight into the trade-offs involved in policy decisions.

KEYWORDS:

• Spatial inequality
• segregation
• urban policy
• neoliberalism
• Amsterdam
• Rotterdam

The question how neighborhood characteristics affect their residents’ life chances has been central to urban studies (Massey, 2016; Sharkey & Faber, 2014). The relevance for policy is clear: establishing that neighborhood effects generate inequalities provides a powerful rationale for policy measures to reduce spatial inequality (Galster, 2019). But this raises a follow-up question that has received much less attention in the literature: how and to what extent can public policy reduce spatial inequality? The answer is far from trivial since research shows that patterns of segregation are remarkably persistent over time (e.g. Faber, 2021; Sampson, 2012). Only a few studies have examined whether poverty deconcentration policies indeed reduce overall levels of urban segregation (Owens, 2015; Tach & Emory, 2017). Despite these pioneering studies, we lack research that systematically examines how different kinds of policy measures affect different kinds of spatial inequality.

The current article begins to fill this gap by examining the impact of urban redevelopment policies on spatial inequality in the Dutch cities of Amsterdam and Rotterdam. The Dutch government has historically been strongly involved in urban development and the provision of social housing, with its role and that of not-for-profit housing associations steadily growing over the course of the twentieth century. The government and housing associations are still heavily involved in housing and specifically in the three policy measures we analyze in this paper: demolition, construction, and tenure conversion. While the government continues to be important, its role has changed since the 1990s as neoliberalism—understood as the policy commitment to deregulation and privatization (Peck, 2010)—gained ground. State-sponsored gentrification of deprived neighborhoods became an important policy objective, with the government promoting the demolition of social housing and the construction of owner-occupied housing in an effort to attract high-income groups to these neighborhoods (Boterman & van Gent, 2023; Uitermark et al., 2007). This article thus focuses on how the seemingly contradictory goals of promoting market forces and countering spatial inequality played out in practice. To do so, we study how both deregulatory and interventionist policy measures affected spatial inequality and distinguish between various kinds of spatial inequality.

Our approach is an answer to two challenges facing researchers in the field. First, studying how public policy impacts urban inequality requires bringing together insights from two research communities that rarely work together (Brown-Saracino, 2017). On one side, researchers in the tradition of the Chicago School of Sociology typically use positivist approaches to explain residential patterns as an outgrowth of individual preferences and resources. Although studies have done much to reveal how class and racial inequalities reproduce over time and through space (Chyn & Katz, 2021; Galster & Sharkey, 2017; Sampson, 2018), researchers in this tradition have had less to say about the role of the state in shaping patterns of spatial inequality. On the other side, scholars working in the critical tradition have offered theoretically informed empirical generalizations on state restructuring as well as qualitative case studies, but have eschewed systematic quantitative analysis (Brown-Saracino, 2017). We bring these literatures together by using quantitative methods not to study individual outcomes but how public policies have reconfigured spatial inequality.

A second challenge frustrating systematic research on public policy’s impact on spatial inequality is that the data are limited. Since we rely on micro-data, we are able to study the impact of policies at the level of individual dwellings and their inhabitants. Adapting and extending a method developed to analyze the mechanisms behind population change (Bailey, 2012), our approach uses geo-coded residence register data, allowing households and dwellings to be followed over time. It further allows us to examine the impact of different kinds of policy measures on different dimensions of spatial inequality: residential segregation as well as spatial stratification and housing market access. We need to distinguish between multiple dimensions of spatial inequality because research suggests that policies mitigating segregation often achieve their goal by displacing low-income or minority households to peripheral locations, in effect substituting one type of inequality (segregation) with another (spatial marginalization). And instead of considering a single policy measure or program, we track different policy measures simultaneously, allowing us to trace the combined impact of interventionist and deregulatory policies characteristic of neoliberal restructuring (Peck & Tickell, 2002).

In sum, we advance scholarship on spatial inequality by bridging different research traditions and addressing different kinds of policies as well as multiple dimensions of spatial inequality. In the remainder of the paper, we first elaborate our framework. We argue that segregation constitutes only one dimension of spatial inequality and suggest considering two additional dimensions: spatial stratification within cities and access to urban housing markets. We then discuss our methods. In light of the limitations of conventional measures of segregation, we develop a “spatial stratification index” that captures the distribution of income groups over neighborhoods of different status. In the empirical sections of the paper, we chart the geographies of policy intervention in Amsterdam and Rotterdam and outline the effects of policy measures on segregation, inequality, and access. It is essential, we conclude, to consider different dimensions of spatial inequality to understand policy decision trade-offs.

Spatial selectivity and the multiple dimensions of spatial inequality

How redevelopment configures the urban fabric follows broader processes of state restructuring (Brenner, 2019). The Keynesian-Fordist welfare state (cf. Jessop, 2002) attenuated spatial inequalities through investments in urban redevelopment and the construction of social housing, often remaking the city according to modernist precepts. The advent of neoliberalism marked a break with these kinds of urban policies. Investments in social housing and urban redevelopment were scaled back while the maintenance of housing estates suffered from underfunding and neglect (Hunt, 2009; Van Kempen et al., 2005). Aggravated by labor market restructuring and austerity, the inner-city neighborhoods and housing projects that had been built or renewed in the post-war period fell into decline and became concentration areas for poor and stigmatized residents (Wacquant, 2008; Wilson, 2012).

Although governments in the 1980s scaled back investments in welfare and social housing, in the 1990s, the “neoliberal project … gradually metamorphosed into more socially interventionist and ameliorative forms” (Peck & Tickell, 2002, pp. 388–389). While governments pursued deregulation—including the scrapping of rent protections, the reduction of housing subsidies, and the transfer of public housing to the private sector—they also developed interventionist policies to upgrade deprived areas and mitigate spatial inequalities. In the United States, the HOPE VI program sought to deconcentrate poverty by demolishing public housing projects and replacing them with racially and economically integrated neighborhoods containing a mix of rental and owner-occupied housing (Goetz, 2013). Australia (Cheshire, 2017), Canada (August, 2008), France (Lelévrier, 2013), Israel (Shmaryahu-Yeshurun, 2022), and the United Kingdom (Arbaci & Rae, 2013), among other countries, adopted variants of this approach. As we explain below, the Dutch government followed a similar strategy, combining privatization and deregulation with targeted investments in an effort to create “balanced neighborhoods” and “undivided cities.” This points to the need to examine how the contradictory combination of deregulatory and interventionist policy measures affects spatial inequality.

Spatial inequality: segregation, status, access

Before explaining how we gauge the impact of policy, we need to unpack the concept of spatial inequality. Segregation has been central to the study of spatial inequality, and researchers have advanced a host of measures to determine to what degree groups cohabit. The concept’s intuitive appeal is that disadvantaged groups are indeed segregated as they are relegated to stigmatized areas with substandard provisions (Jargowsky, 1996; Wilson, 2012). However, at least two additional dimensions of spatial inequality warrant consideration.

First, we need to consider inequality in neighborhood status. Do groups secure a place in the city’s most coveted areas or are they relegated to its unpopular fringes? The dissimilarity index, the isolation index, the rank-order information theory index, and other commonly used measures of segregation (see, e.g. Reardon & O’Sullivan, 2004) do not explicitly take into neighborhood status. Given the limitations of the available measures, we need a metric that directly measures inequalities between groups with respect to neighborhood status.

Second, we need to consider access to housing markets. Especially in areas with high market demand, housing systems can create equality among established residents while erecting barriers to outsiders (Kadi & Musterd, 2015). Newman and Wyly (2006) suggest that areas with low churn will not show rapid population changes but may experience drastic changes in the groups that can access them. Tach and Emory (2017: p. 723) find that the redevelopment of neighborhoods with public housing succeeded in altering hierarchies among neighborhoods but did so largely by displacing poor and non-White residents. In ignoring access to housing, we may fail to apprehend that reduced segregation and spatial stratification can come at the cost of exclusion. To grasp the inequalities produced by housing systems, we must examine access to the city’s housing market alongside segregation and inequality within the city.

Reconfiguring spatial inequality: three scenarios

Having outlined the kinds of policy measures and effects we wish to study, we now consider the patterns we expect to find. Although the literature does not offer a comprehensive theory of how the contradiction between interventionist and deregulatory policy measures plays out, it does suggest three scenarios. First, a number of authors have suggested that governments only promote “social mixing” when this means that stigmatized and poor households are displaced from central and valuable land (Bridge et al., 2012; Lees, 2008). Policies promoting social mixing are then used opportunistically to facilitate gentrification in previously marginalized areas, with the government orchestrating private investment and changing rules and regulations to allow upscale development (Bernt, 2022). While measures of segregation may temporarily decline as gentrifiers move into low-income areas, spatial stratification is exacerbated as poor and minority households are displaced from coveted central areas and relegated to the periphery (Elliott-Cooper et al., 2020; Walks & Maaranen, 2008). In this first scenario, policies serve the market and fortify the inequalities it generates.

In a second scenario, public policies aim to counter segregation but lack leverage. While policy interventions might deconcentrate affordable housing, efforts are generally insufficient to override the many other structural forces engendering segregation. Research in the United States (Owens, 2015; Sampson, 2008) and Europe (Tammaru et al., 2016) suggests that territorial interventions have limited effects and typically reshuffle rather than attenuate spatial inequality. While urban policies might mitigate spatial inequality, they are insufficient to counterbalance the spatial polarization resulting from deepening housing financialization and the uneven geographies of capital investments (Hochstenbach & Arundel, 2020). Writing about the Netherlands, Bolt et al. (2009, p. 502) suggest that even comprehensive policies do not affect overall levels of ethnic and class segregation: “while the social make-up of neighborhoods is altered, and low-income households shift in space, the displacement does not contribute to desegregation.” In this second scenario, public policies have in situ effects but do not significantly alter spatial inequality in cities at large.

The literature on European cities suggests a third scenario based on the idea that forces of integration continue to be strong in centralized welfare states such as the Netherlands (cf. Cox, 2016; Le Galès, 2002). This strand of literature argues that policies promoting social mixing in urban areas mitigate spatial inequalities by promoting “social housing in bourgeois areas” and making “social housing available to the lower-middle as well as the working class” (Bridge et al., 2014, p. 1134). Social mixing policies do not necessarily involve privatization or displacement but might also rely on selective allocation or the construction of affordable units in new neighborhoods (Alves, 2022; Söderhäll & Fjellborg, 2022). Alves (2022) argues that social mixing measures in Copenhagen are not meant to promote gentrification but instead form an integral part of broader efforts to ensure social and spatial justice. Even though urban redevelopment policies in the Netherlands often involve displacement, relatively strong tenant protection and resident participation ensure that affected residents can either return to redeveloped areas or move to areas of higher status (Kleinhans, 2003). In this third scenario, the government’s policies reduce spatial inequalities.

Data, methods and analytic strategy

Case selection: The Netherlands

Neoliberalization in the context of Dutch housing and urban policy represents a rupture in a long history of regulation and decommodification. While social housing remained the norm in the early 1980s, the belief that the market should have a central role in the production and allocation of housing gained ground in the 1990s. By the end of the 1990s, the policy consensus was that urban redevelopment centered on social housing had successfully tackled issues of physical dilapidation but had failed to address social downgrading: the built environment was in good shape, but the neighborhoods were not. The dominance of social housing in deprived urban neighborhoods thus came to be seen as a key factor contributing to the concentration of poverty, best addressed by altering the composition of the housing stock. Although urban and housing policy formally only considers residents’ socio-economic status, ethnicity featured prominently in policy debates, with politicians and policymakers fearing that the spatial concentration of ethnic minority groups would destabilize neighborhoods and integration (Musterd & Ostendorf, 1998; Uitermark, 2003; Uitermark et al., 2017).

As the literature on neoliberalization suggests, this process is more complex than the simple scaling back of the state; what we see is a remaking of the institutions governing housing and urban development (cf. Brenner, 2019). The government and housing associations pursued an ambivalent strategy: promoting market forces in housing and urban development through deregulatory measures while using interventionist measures—in the form of subsidized demolition of social housing and construction of owner-occupied housing—to boost poor neighborhoods and deconcentrate poverty. The Dutch urban restructuring policy, introduced in 1997, did not only redevelop social housing estates but sought to restructure entire neighborhoods and cities, targeting 170 neighborhoods with a combined population of 750,000 households or roughly 12% of the national population (Uitermark, 2003, p. 532). In 2007, at the beginning of our study period, the national government and housing associations intensified the restructuring policy, committing an additional 750 million euros per year over a four-year period as part of an “Action plan for strong neighborhoods” (Actieplan Krachtwijken). The end result would be “balanced neighborhoods” with a greater mixture of income groups and “undivided cities” with less spatial inequality. The government decided to discontinue the policy in 2012 but since redevelopment plans take several years to materialize, it was in effect until around 2014, which is the end of our study period.

Case selection: Amsterdam and Rotterdam

We compare the Netherlands’ two largest cities to examine how similar policies played out in different urban contexts (cf. Burgers & Musterd, 2002). A second-tier global city, Amsterdam has strong financial, service and leisure sectors; its rapidly growing population and demand for housing have led to growing price differences between Amsterdam and the rest of the country (Hochstenbach, 2017). In contrast, Rotterdam’s economy has historically been organized around heavy industry, and the city is struggling to adapt (Van den Berg, 2017). Its population has been growing only modestly while house prices have increased roughly in line with national trends. While both cities have similar proportions of low-income residents (around 24% belonged to the lowest quintile in 2014), Amsterdam has many more affluent residents (23.5% in the top income quintile compared to 17.5% in Rotterdam). To simplify, Amsterdam is a thriving post-industrial city, and Rotterdam is a struggling post-industrial city. Despite their different socio-economic profiles and geographies, Amsterdam and Rotterdam have comparable housing systems. In both cities, rental dwellings owned by not-for-profit housing associations constituted around 46% of the housing stock in 2013, while homeownership rates in Amsterdam and Rotterdam stood at respectively 28% and 35% (Hochstenbach, 2017).

The city governments of Amsterdam and Rotterdam both implemented the national restructuring policy introduced above: they upgraded the housing stock and encouraged owner-occupied dwellings at the expense of social housing. While their policy programs were broadly similar, there were differences in emphasis that reflected the political orientation of their governments. In the study period, Leefbaar Rotterdam, a populist party founded by Pim Fortuyn, was the largest party in the city of Rotterdam. The party and its coalition partners, fearing the over-representation of deprived ethnic minorities would destabilize the city, wanted to attract and retain middle-class, native Dutch households (Uitermark & Duyvendak, 2008). Local authorities promoted controversial policies such as the exclusionary “Rotterdam Act”, essentially forbidding low-income residents from settling in selected disadvantaged neighborhoods, in an effort to reduce poverty and immigrant concentrations (Uitermark et al., 2017).

Amsterdam’s left-leaning government did not adopt such a revanchist discourse, but it did want to increase owner-occupation on the grounds that the city’s housing stock had to be restructured to better accommodate upwardly mobile households (Boterman & van Gent, 2023; Hochstenbach & Ronald, 2020). While Rotterdam presented its housing policy measures as a means to change the composition of the city, in Amsterdam the government presented social mixing as a way to better meet the needs of its residents and pursue an undivided city. In short, our two cases allow us to study how comparable policy ambitions worked out in cities with different socio-economic and political profiles. Based on their political profiles and stated policy objectives, we should expect that spatial inequality increased in Rotterdam while it decreased in Amsterdam.

Data

To trace population changes and measure the impact of housing policy interventions, we use micro-data from the System of Social-statistical Databases (SSD) managed by the Dutch national statistical agency (Centraal Bureau voor de Statistiek or CBS). We constructed a dataset of individual dwellings and assessed whether these dwellings were subjected to different kinds of policy measures (see below). We subsequently coupled the dwellings to the households inhabiting them, allowing us to track residential changes over time. We focus on the 2007–2014 period due to the availability of data and because the urban redevelopment policy was at its most ambitious.

Neighborhoods. We examine policy interventions at the level of neighborhoods, following official CBS demarcations. As policy-makers refer to CBS neighborhoods, this is the relevant scale to investigate the state’s spatial selectivity. After excluding sparsely populated neighborhoods, we are left with 84 neighborhoods in Amsterdam and 66 in Rotterdam. In both cities, neighborhoods on average contain around 4,000 dwellings.

Income groups. We measure the socio-economic status of households living in the dwellings and track changes over time, using equivalized household income to correct for differences in household size and composition. Using percentile groups relative to the national population in both 2007 and 2014, we construct five income groups. Households in the bottom quintile (q1) belong to the 20% of households with the lowest incomes relative to the entire Dutch population; households in the top quintile (q5) have the highest incomes. When constructing composite indicators, we use percentile rather than quintile groups for greater accuracy.

Neighborhood status. A common way to establish neighborhood status is to examine the status of its residents, but this is inappropriate for our purposes as we want to know to what extent the status of neighborhoods matches that of its residents. We thus use the average square meter sale price of housing as an indicator of status. To measure the distribution of inhabitants over areas of different status, we group neighborhoods in city-level quintile groups based on the average square meter sales price of houses in 2013.

Deregulatory and interventionist policy measures

Our micro-data allow us to identify three types of housing interventions: demolition, new construction, and tenure conversion. While there are parts of the housing stock over which the government holds little control (including the owner-occupied stock and the high-end private rental market), municipal governments are strongly involved in all the three policy measures we study here. Through detailed stipulations on tenure, size, and price, municipalities have a decisive influence on the kind of housing that gets built. The government is also heavily involved in converting social rental housing units to owner-occupied units. For social housing units owned by housing associations, the volume and location of sell-offs are negotiated by the local government, housing associations, and tenant representatives. When private landlords wish to convert rent-controlled housing into private housing, they must typically apply for government approval. In the period under study, such approval was conditional on the landlord upgrading the dwelling prior to sale. In effect, the policy incentivizes landlords to upgrade and sell off their properties (Boterman & van Gent, 2014). Although the technical details of housing policy are complex, tenure conversions (broadly speaking) exemplify deregulatory policy measures while demolitions and constructions exemplify interventionist policies, enabling us to track different aspects of neoliberalization.

Outcome measures: segregation, spatial stratification, access

We assess the impact of policy interventions on spatial inequality in several ways. First, we map the geography of policy measures and population changes at the neighborhood level using Geographical Information Systems (GIS). Dynamic and interactive versions of all maps presented in this paper, along with supplementary maps, can be accessed online at www.uva.nl/urbaninequality. Second, we compare different parts of the housing stock (i.e. transformed and untransformed) in neighborhood quintiles of different status. Third, we establish the impact of policy measures on residential segregation, neighborhood status, and access to the housing market in both cities at large.

Segregation. We measure segregation through the rank-order information theory index, the H^R index for short (Reardon & Bischoff, 2011). The H^R index has some important advantages viz-a-viz other common measures of segregation. In contrast to e.g. the dissimilarity index (e.g. Duncan & Duncan, 1955), the H^R index does not presume two categories, which is helpful when examining segregation between income groups. For both cities and both time points, we calculate the H^R index overall, as well as for the 20% and 80% of lowest-income households versus all other households.¹

Spatial stratification. We want to know whether income groups live in the same neighborhoods but also what kinds of neighborhoods they inhabit: do low-income groups live in low-status neighborhoods or are they evenly distributed across neighborhoods of different status? To answer this question, we developed a spatial stratification index (SSI). To calculate our index, we first sort the population of households in percentiles according to income. We then measure the cumulative distribution of neighborhood-level house prices, based on households’ place of residence, over the percentiles. For our index, we adapt the formula for Gini (see Witlox, 2017) as follows: $\mathrm{SSI}=1-\sum _{k=0}^n(X_{k-1}+X_k)(Y_{k-1}+Y_k)$In this formula, X is the cumulative household population sorted by income percentiles, and Y the cumulative neighborhood-level house prices taken up by this population. With n we denote the total number of population groups (percentiles, so n = 100), while k represents individual percentile groups. While the spatial stratification index closely resembles the well-known Gini index, there is an important difference. Whereas the Gini coefficient sorts the population according to the same quantity for which inequality is measured (e.g. it sorts according to income to measure income inequality), the spatial stratification index sorts the population according to income and then measures inequality in terms of neighborhood-level house prices. The Gini index thus runs from 0 to 1, the SSI index from −1 to 1. A score of 0 implies a perfectly equal distribution of neighborhood prices, that is, the income percentiles are distributed equally over neighborhoods with different prices. A score higher than 0 means that the richer percentiles live in neighborhoods with higher values, while a score lower than 0 means that the lower percentiles live in neighborhoods with higher values.²

Housing market access. The H^R index and status inequality index measure different dimensions of spatial inequality within the city; they do not capture exclusion from the city. This means that both indexes can go down when particular groups are excluded. While measuring access with a single metric is virtually impossible as we only know who resides in the city—and not who wants to move into it—we approximate access in two ways. First, we examine the direct effects of policy measures by looking at the differences in population dynamics between parts of the housing stock subject to policy measures (the transformed stock) and parts of the stock not subject to these measures (the untransformed stock). While this difference is a proxy for how far policy measures have closed or opened opportunities for different income groups, we note it only captures the direct effects of policy measures within the transformed stock. For instance, if a household moves out of a dwelling due to demolition, we register this as an effect of policy. But if the same household then moves to a dwelling in the untransformed stock, this is not registered as an effect of policy. This is why we also look at, second, the presence of income groups in the cities as a whole. Although this is a rough indicator, we surmise that, under conditions of high demand, decreasing presence reflects reduced access for low-income groups.

Analytic strategy: the impact of policy interventions

To account for changing patterns of spatial inequality in Amsterdam and Rotterdam, we first chart overall changes in population composition in neighborhoods, classes of neighborhoods, and cities as a whole between 2007 and 2014. We then disaggregate broader changes by distinguishing changes within parts of the housing stock affected by policy interventions (the “transformed stock”) from changes in parts of the housing stock unaffected by policy interventions (the “untransformed stock”). Here we adapt a method developed to disentangle the influence of residential moves, in situ social mobility, and demographic change on neighborhood trajectories (Bailey, 2012; Hochstenbach & van Gent, 2015) in order to unravel the influence of tenure conversions, demolitions, and new constructions.

The basic idea of our method is to chart population change within neighborhoods (or classes of neighborhoods) with and without taking into account the dwellings subjected to different kinds of policy interventions. In other words, we compare population changes in the different parts of the housing stock: the “stable” (untransformed) stock as well as those parts of the stock that have been subject to transformation (i.e. newly built, demolished, or changed tenure). We similarly calculate our indices for segregation and spatial stratification with and without taking into account parts of the stock subjected to different kinds of policy measures. Since we track policy interventions at the level of individual dwellings, we can break down developments within neighborhoods and do not have to rely on comparing intervention and control neighborhoods.

Limitations

Three caveats are necessary when interpreting our findings. First, our approach provides an approximation of the effects of policy, not a direct measurement. Since the different parts of the housing stock are interdependent and residential mobility is affected by a range of factors beyond our view, our research design is not experimental but rather provides a best-possible approximation of the direction and extent of policy impacts on spatial inequality.

Second, our approach estimates the direct effects of policy. Systematically measuring spill-over effects in our case studies is challenging for at least two reasons. A first reason is that policies generally aim to transform entire classes of neighborhoods, which means it is by definition impossible to find suitable control neighborhoods (cf. Tach & Emory, 2017). A second reason is that spill-over effects work in different directions. One would reduce the proportion of low-income groups by boosting prices in adjacent neighborhoods; another would entail the relocation of low-income groups within the neighborhood. Since these issues cannot entirely be resolved, we focus on direct policy effects.

Third, this paper focuses on spatial inequality between income groups. In both policy debates and practice, income interacts with ethnicity and race. As noted above, Dutch urban redevelopment policy can in part be traced to fears that the spatial concentration of ethnic minority groups will destabilize neighborhoods, and ultimately society at large. This begs the question of how policies differentially affect various ethnic and racial groups. Although exploring these questions would require more space and a different theoretical angle, our approach could be amended to pursue this line of research.

Results

Our presentation of the results first considers the patterns and effects of policy measures for different areas and groups. These analyses of different areas and groups provide stepping-stones for studying the impact of policy on residential segregation, spatial stratification, and access to housing in the cities as a whole. The following sections show a selection of the maps that inform our analysis. A full presentation, with interactive maps, can be found at www.uva.nl/urbaninequality.

The geography of policy measures

The policy measures we studied have distinct geographies. The maps in Figure 1 show that in Amsterdam, the government has facilitated the sale of social housing in centrally located areas: the historical center and especially the neighborhoods surrounding it, the so-called nineteenth-century ring—the city’s premier location of gentrification. This suggests that the government and property owners (housing associations and private landlords) have capitalized on increasing demand by converting dwellings in these areas. The pattern is similar but less pronounced in Rotterdam, where historical districts in and around the center have seen conversions but at lower rates than in Amsterdam’s historical districts.

Figure 1. The geography of policy measures in Amsterdam (1a) and Rotterdam (1b) (2007-2014).

Source: Authors’ elaboration. Note: Brown indicates that demolition was the predominant measure in the neighborhood (between 7.5 and 29.9% of the stock in 2007); gray that new construction was the predominant measure (between 6.3 and 87.8% of the stock in 2007); and yellow that conversion from rent to ownership was predominant (between 4.9 and 13.2% of the stock in 2007). The color gradients represent the scale of the interventions relative to the housing stock. Interactive versions of the maps can be found at www.uva.nl/urbaninequality. SSD data, own calculations.

Figure 1 shows maps of neighborhoods in Amsterdam and Rotterdam colored according to the policy intervention most prevalant in the area. In Amsterdam, construction and demolition are more common on the periphery whereas in Rotterdam construction and demolition occur in the central areas and along the Waterfront. Tenure conversion from rent to owner-occupied occurs especially in the central and southern areas of Amsterdam whereas Rotterdam does not have a pronounced pattern.

While there are differences of degree with respect to conversions, we find qualitative differences between Amsterdam and Rotterdam for demolitions and constructions. While both cities added large numbers of owner-occupied dwellings (a net increase of 6,465 units in Rotterdam and 8,966 units in Amsterdam), Rotterdam’s additions were heavily concentrated in areas with the highest status (+2,568 units in the top neighborhood quintile) (Table 1). In contrast, Amsterdam added stock mostly in lower-status areas (+3,966 units in the bottom quintile and +3,665 units in the second quintile). We see a similar pattern with rental units: whereas Amsterdam added rental units mostly in low-status areas, Rotterdam demolished (mostly social) housing in low-status areas while adding (mostly market) rental units in high-status areas. These differences align with our expectation that the cities’ political philosophies translate into different into different policy measures, with the government of Amsterdam creating new socially mixed neighborhoods representative of the city’s population and Rotterdam aiming to create high-class neighborhoods in a deliberate effort to promote new-build gentrification (cf. Davidson & Lees, 2005). These include new constructions on the waterfront featuring luxury high-rise complexes (Doucet et al., 2011) as well as more peripheral newly built areas catering to well-off households (see Figure 1).

Table 1. The spatial selectivity of policy measures (2007-2014) in Amsterdam and Rotterdam.

		Net demolitions and new constructions				Net conversions		Total stock 2007
		Rent		Owned		Rent → Owned
	Neighborhood quintiles	N	%	N	%	N	%	N
Amsterdam	Lowest status	213	0.2	3,965	4.6	2,848	3.3	85,342
	Low	3,456	4.5	3,665	4.8	2,861	3.7	76,578
	Medium	681	1.0	699	1.0	3,361	4.8	69,384
	High	−462	−0.6	157	0.2	2,985	4.0	73,884
	Highest status	−159	−0.3	268	0.5	630	1.1	55,860
	Total	4,279	1.2	8,966	2.4	12,795	3.5	366,344
Rotterdam	Lowest status	−893	−1.6	875	1.6	670	1.2	55,742
	Low	−822	−1.6	851	1.6	1,070	2.1	51,729
	Medium	−247	−0.3	948	1.2	2,448	3.1	77,988
	High	−102	−0.2	860	1.8	903	1.9	48,798
	Highest status	1,395	4.0	2,568	7.3	425	1.2	34,978
	Total	−593	−0.2	6,465	2.3	5,478	2.0	275,618

Note: net changes are additions minus subtractions. For conversions, positive net figures indicate more rental dwellings have been converted into ownership than vice versa. Percentages refer to the proportion of the total stock in a segment that has been subjected to policy intervention. SSD data, own calculations.

Policy interventions and population changes per neighborhood class

While the findings above already suggest that the governments of both cities deployed deregulatory and interventionist measures to facilitate high-income groups, we now look more with more precision at the effects of policy measures on different income groups. As explained above, we gauge the impact of policy measures by comparing population changes with and without taking into account the transformed stock (Table 2). An example aids interpretation: the upper-left cell (top row, first column) shows that the share of the lowest income group in Amsterdam’s lowest-status neighborhoods decreased by 0.3 percentage points. This is the overall change between 2007 and 2014. To pinpoint the impact of policy, we disaggregate this figure by examining changes in both the transformed and untransformed stock. We see that the share of the lowest income households (q1) in Amsterdam’s lowest-status neighborhoods (q1) decreased by 2.5 percentage points due to policy measures. This figure can be interpreted as the isolated and direct impact of all interventions (conversions, demolitions, constructions) on the share of low-income households residing in these neighborhoods. We can further disaggregate this figure by differentiating between the impact of demolition and construction (−2.1 percentage points) and conversions (−0.6 percentage points).³ While policy measures reduced the proportion of low-income groups in these neighborhoods, their overall presence decreased only slightly as they took up a higher proportion of the untransformed stock (+2.2 percentage points).

Table 2. Change in income composition by neighborhood type, and the impact of different policy measures.

	Amsterdam						Rotterdam
						Change overall
Income quintiles	Neighborhood quintiles (from low to high)						Neighborhood quintiles (from low to high)
	q1	q2	q3	q4	q5	total	q1	q2	q3	q4	q5	total
Lowest	−0.3	−0.4	−2.0	−1.6	0.7	−0.8	1.6	1.8	0.7	1.1	0.0	0.8
Low	0.1	−0.4	−1.0	0.3	0.2	−0.1	−0.2	−1.5	−0.3	−0.1	−0.5	−0.6
Medium	−0.3	−0.9	−1.0	−0.9	−0.6	−0.7	−0.8	−0.8	−0.5	−0.5	−0.7	−0.6
High	0.1	0.0	0.2	−0.2	−0.1	0.0	−0.5	−0.3	−0.1	−0.8	0.1	0.1
Highest	0.6	1.7	3.7	2.3	−0.2	1.5	−0.1	0.8	0.1	0.3	1.1	0.6
Change due to all measures
Lowest	−2.5	−1.9	−1.9	−1.2	−0.5	−1.7	−1.1	−1.7	−1.4	−0.9	−1.3	−1.4
Low	−1.1	−1.2	−1.0	−0.7	−0.2	−0.8	−0.1	−0.5	−0.5	−0.5	−1.2	−0.6
Medium	0.5	0.0	0.0	−0.1	−0.2	0.1	0.5	0.5	0.4	0.2	0.2	0.3
High	1.4	0.6	0.8	0.4	0.2	0.7	0.4	0.9	1.0	0.3	1.0	0.8
Highest	1.7	2.4	2.1	1.7	0.7	1.7	0.2	0.8	0.6	0.9	1.3	1.0
Change due to demolitions and constructions
Lowest	−2.1	−1.1	−0.6	−0.3	−0.2	−0.9	−1.1	−1.3	−1.0	−0.6	−1.2	−1.2
Low	−0.9	−0.8	−0.2	−0.1	0.0	−0.4	0.0	−0.5	−0.3	−0.4	−0.9	−0.5
Medium	0.1	−0.4	−0.2	−0.1	−0.1	−0.1	0.3	0.2	0.1	−0.2	−0.1	0.1
High	1.2	0.1	0.1	0.0	0.0	0.3	0.4	0.6	0.6	0.2	0.7	0.5
Highest	1.7	2.2	1.0	0.4	0.3	1.1	0.4	0.9	0.6	1.1	1.4	1.1
Change due to conversions
Lowest	−0.6	−0.9	−1.4	−1.2	−0.6	−0.9	−0.5	−0.5	−0.6	−0.7	−0.2	−0.5
Low	−0.2	−0.4	−0.8	−0.8	−0.3	−0.5	−0.1	−0.1	−0.3	−0.2	−0.3	−0.2
Medium	0.3	0.3	0.2	−0.1	−0.2	0.2	0.3	0.3	0.3	0.4	0.1	0.3
High	0.3	0.5	0.7	0.5	0.1	0.4	0.2	0.3	0.4	0.4	0.2	0.3
Highest	0.2	0.4	1.3	1.6	0.9	0.8	0.0	0.0	0.1	0.2	0.2	0.1
Change in the untransformed stock
Lowest	2.2	1.5	−0.1	−0.4	1.2	0.9	2.7	3.5	2.1	1.9	1.2	2.2
Low	1.2	0.8	0.0	1.1	0.4	0.8	−0.1	−1.0	0.2	0.4	0.6	0.0
Medium	−0.9	−0.9	−1.0	−0.7	−0.4	−0.8	−1.3	−1.2	−0.8	−0.6	−0.9	−0.9
High	−1.4	−0.7	−0.6	−0.5	−0.3	−0.7	−0.9	−1.3	−1.0	−1.1	−0.9	−0.6
Highest	−1.1	−0.7	1.6	0.6	−0.9	−0.2	−0.3	0.0	−0.4	−0.6	−0.1	−0.3

Note: SSD data, own calculations

The overall pattern is one of policy measures reducing the proportion of the lowest income groups (−1.7 percentage points in Amsterdam and −1.4 percentage points in Rotterdam) while boosting the presence of the highest income groups (+1.7 percentage points in Amsterdam and +1.0 percentage points in Rotterdam). In Amsterdam, construction and demolition particularly altered the population composition of lower-status neighborhoods, increasing the presence of top-income households by 1.7 and 2.2 percentage points in the bottom two neighborhood types, while decreasing the share of low-income households (−2.1 and −1.1 percentage points). Conversions mainly facilitated changes in middle (q3) and upper-middle (q4) neighborhoods, increasing the share of high-income households (+1.3 and +1.6 percentage points) and reducing the share of low-income households (−1.4 and −1.2 percentage points). In Rotterdam, demolitions and constructions especially facilitated the growth of high-income households in already expensive neighborhoods (+1.4 percentage points in top quintile neighborhoods). Conversions in Rotterdam did not have the same impact as in Amsterdam. While Rotterdam was more zealous in promoting gentrification, the relatively low demand for housing limited the possibilities of promoting gentrification through deregulatory measures.

Figure 2 maps the impact of all policy measures on the share of low- and high-income households at the neighborhood level. Overall, we find that Amsterdam and Rotterdam have created more space for higher-income groups in the vast majority of neighborhoods. Figure 3 disentangles the impact of conversions from those of demolition and construction. The maps clearly show how policy impacts are geographically differentiated. In Amsterdam, the impact of conversions is especially palpable in more central and older parts of the city while construction and demolition create the biggest changes in more peripheral and recently built neighborhoods. The peripheral areas contain a large proportion of the city’s poor and ethnic minority populations and are often portrayed as places where drastic interventions are necessary to create more “balanced neighborhoods.” The neighborhoods where most conversions take place are where gentrification is most intense (Figure 3). The government’s deregulatory policy measures thus feed on and reinforce processes of gentrification already underway. The pattern for conversions in Rotterdam is similar, although less pronounced; conversions disproportionately take place in the gentrifying areas adjacent to the city center, albeit at a lower rate than in Amsterdam. In Rotterdam, demolition occurs in low-status areas, especially on the city’s southern periphery. Construction occurs mostly in high-status areas, both in the central areas along the Nieuwe Maas and on the northern periphery.

Figure 2. Policy effects on the presence of the lowest income group (maps on top) and the highest income group (maps at bottom) in Amsterdam and Rotterdam.

Source: Authors’ elaboration. Note: Hot colors indicate that policy interventions led to a decrease, cold colors indicate an increase. Interactive versions of the maps can be found at www.uva.nl/urbaninequality SSD data, own calculations.

Figure 2 shows maps of neighborhoods in Amsterdam and Rotterdam colored according to the impact of policy on the presence of different income groups. The maps show that in most areas that high-income groups increase their presence relative to low-income groups.

Figure 3. The impact of rental housing sales (maps at the top) and demolition/construction (maps at the bottom) on the presence of high-income groups (top quintile).

Source: Authors’ elaboration. Note: Hot colors indicate that policy interventions led to a decrease, cold colors indicate an increase. Interactive versions of these maps as well as maps for low-income groups can be found at www.uva.nl/urbaninequality. SSD data, own calculations.

Figure 3 shows maps of neighborhoods in Amsterdam and Rotterdam colored according to the impact of different kinds of policy measures on the presence of different income groups. Whereas tenure conversion reduces the presence in central and expensive areas, demolitions and new construction have the same effect in peripheral and lesser wanted areas.

These findings suggest that both deregulatory and interventionist measures increase the proportion of high-income households at the expense of low-income groups. High-income groups consolidate their presence in the most prestigious areas while expanding their presence in other neighborhoods. Gentrification spreads like Hamnett’s fountain metaphor suggests: “The water falls into the top bowl but, as prices rise, this is soon filled and the water spills over into the next bowl which in turn spills over down to the lowest and broadest bowl of the fountain.” (Hamnett, 2003, p. 2416). The water, however, does not simply flow; it is actively channeled by the government as it issues permits to convert and commodify social housing. Nor is the government’s role limited to accommodating demand wherever it arises; it directly intervenes by demolishing social housing in low-status neighborhoods and adding housing in higher-status areas.

Segregation, inequality, access

Segregation

How do policies impact different dimensions of spatial inequality? We measure overall developments in residential segregation using the H^R index and then gauge the contribution of policy interventions to changing segregation levels by re-calculating the H^R index without considering the transformed stock (Table 3). This gives us an approximation of the isolated, direct impact of policy measures on the H^R index. Income segregation in Amsterdam (already lower than in Rotterdam) decreased between 2007 and 2014, with income groups at the very bottom and top mingling more with the rest of the population. All three kinds of policy measures reduced income segregation. First, newly built areas were highly mixed in housing tenure and thus income groups; new constructions drove down income segregation in the city as a whole. Second, tenure conversions resulted in the influx of high-income groups into low-income areas. Conversions mainly took place in high-status areas that were undergoing rapid gentrification. Since these areas were previously mostly inhabited by low-income groups, the result is a greater mixing of income groups. Third, demolition and construction in more peripheral areas, stimulated by nationally subsidized urban redevelopment programs, added comparatively expensive owner-occupied housing to the stock and facilitated the move of high-income groups into areas where low-income groups were prominent. In Rotterdam, segregation levels remained relatively stable and the influence of policy interventions on segregation was comparatively weak. Whereas measures in Amsterdam opened space for high-income groups in areas predominantly inhabited by low-income groups, measures in Rotterdam tracked already high levels of segregation.

Table 3. Rank-order information theory index in 2007 and 2014, and the direct impact of different policy interventions in Amsterdam and Rotterdam.

	Amsterdam			Rotterdam
	Overall	20th percentile	80th percentile	Overall	20th percentile	80th percentile
2007	5.54	3.16	8.43	5.87	4.88	8.3
2014	4.65	2.36	7.1	6.17	5.41	8.23
	−0.9	−0.8	−1.33	0.31	0.53	−0.07
Influence of:
All interventions	−0.5	−0.43	−0.93	0.01	0.01	−0.14
Tenure conversions	−0.1	−0.17	−0.09	−0.06	−0.1	0.05
Demolitions and constructions	−0.39	−0.25	−0.77	0.04	0.09	−0.21
Change in stable stock:	−0.4	−0.37	−0.4	0.29	0.53	0.07

Note: “Overall” provides an overall measure of income segregation; “20th percentile” provides a measure of segregation for households with incomes in the bottom 20% versus all other households; “80th percentile” does the same for households in the bottom 80% versus all other households. SSD data, own calculations.

Especially the findings for Amsterdam show why a narrow focus on segregation limits our understanding of the dynamics of spatial inequality: gentrification, stimulated through tenure conversions as well as new construction, brings high-income groups into low-income areas and thus reduces segregation, at least in the short run. Although Amsterdam’s neighborhoods have indeed become more mixed in terms of income, two qualifications are in order. First, reduced segregation will prove temporary if gentrification in the long run displaces low-income groups (Walks & Maaranen, 2008). Second, segregation is not the only or even the most important dimension of spatial inequality. We now turn to two other dimensions of spatial inequality.

Neighborhood status

While segregation in Amsterdam declined, this does not mean that low-income groups are now more often residing in high-status neighborhoods. Whereas segregation (as measured using H^R) dropped, the spatial stratification index remained roughly stable. Table 4 shows that this apparent stability results from two kinds of policy measures pushing in different directions. Demolition and construction brought high-income groups into low-status neighborhoods, thereby pushing down the inequality index. In contrast, tenure conversions drove up inequality as they expanded housing opportunities for high-income groups in more expensive and already gentrifying areas. In addition to these direct effects of policy, our data also suggest an indirect effect: changes in the untransformed housing stock drove up spatial inequality, meaning that low-income groups were pushed to less-wanted housing units.

Table 4. Spatial stratification between income groups in 2007 and 2014.

	Amsterdam	Rotterdam
2007	2.08	2.84
2014	2.08	2.97
change	0.00	0.12
Influence of:
All interventions	−0.23	0.11
Tenure conversions	0.13	0.00
Demolitions and constructions	−0.31	0.10
Change in stable stock:	0.23	0.02

Note: SSD data, own calculations

Rotterdam was already a more divided city than Amsterdam, as seen in its higher 2007 score on the spatial stratification index: 2.66 versus Amsterdam’s 2.16. Furthermore, inequality increased more in Rotterdam (+0.12) than it did in Amsterdam (+0.05), largely due to its interventionist policy measures. The construction of new higher-status neighborhoods led to the concentration of affluence, with households in the highest income bracket concentrating (even) more in neighborhoods with the highest status and households in the lowest income bracket further concentrating in neighborhoods with the lowest status. The differences in how policy measures affected spatial stratification are striking: in Amsterdam’s booming housing market, deregulatory measures led to greater stratification; in Rotterdam, interventionist measures had that same effect.

Access

We now turn to the third dimension of spatial inequality: access to the housing market. We approximate the impact of policy measures in two ways: by looking at population changes in different parts of the housing stock and by looking at population changes in the cities as a whole. Table 4 suggests that policy interventions have played a key role in restricting the access of low-income groups to Amsterdam’s housing stock. While the impact of conversions on the presence of the lowest income group (−1.7) is stronger than the impact of demolition and construction (−0.9), both open up space for higher income groups. Here we get a sense of how policy measures that reduce segregation and spatial stratification may have reduced access to the city. We should note, however, that these numbers only document the immediate effects of policy interventions; it is plausible that households displaced by demolition or conversion moved into the untransformed stock. While the proportion of the lowest income group in the untransformed stock indeed shot up (+2.2), it does not compensate for their decreased presence in the transformed stock: in the city as a whole, the lowest income group sees a drop (−0.8).

In Rotterdam, conversions, as well as demolition and construction, reduced the presence of low-income groups. Although the combined effect of these interventions is comparable to Amsterdam, their weight is different, with demolition and construction contributing more to the reduction of low-income groups (−1.2) than conversions (−0.5). The modest impact of conversions in Rotterdam is not due to the city government’s lack of commitment to gentrification but to market conditions: compared to Amsterdam, there was less demand for housing from high-income groups. Although policy measures in Rotterdam indeed limited options for low-income groups, their presence in the untransformed stock increased to such an extent that their share in the city at large increased.

While low-income groups continued to have a strong presence in both cities, policies did reduce their housing options (Hochstenbach & Musterd, 2018). As an immediate effect of policies, the two lowest income groups saw their share decrease by 1.7 and 0.8 percentage points in Amsterdam and by 1.4 and 0.6 percentage points in Rotterdam (Table 2). We also see the indirect effects of policy with the untransformed parts of the housing stock increasingly accommodating the low-income groups displaced by urban redevelopment. Regardless of the trends for segregation or inequality within the city, policies have contributed to reduced access to the city for low-income groups.

Discussion

What do our findings reveal about the impact of policy on spatial inequality? To summarize, the literature suggests three scenarios: (1) the government attempts and succeeds in promoting socially mixed neighborhoods; (2) the government pursues gentrification and its policies effectively exacerbate spatial inequalities; and (3) the government attempts to promote socially mixed neighborhoods but fails to effectively counter spatial inequality. Our findings align with the first two scenarios, although we see important differences between Amsterdam and Rotterdam that, we suggest, reflect their different socio-economic profiles.

The Amsterdam government pursued an ambivalent strategy. In line with its left-leaning political profile, it used interventionist measures to create areas with different tenures and a mix of income groups: it upgraded areas with low market demand, added affordable rental housing to expensive areas (Zuidas), and constructed new neighborhoods mixed by design (IJburg). While the city government did thus not simply reinforce market forces, some of its efforts to diversify the housing stock did have the effect of inducing displacement. The government’s deregulatory measures—specifically the conversion of social to owner-occupied housing—accelerated gentrification and accentuated spatial inequality. Overall, considering all interventions, we identify a clear policy trade-off: replacing social-rental housing in deprived neighborhoods reduced income segregation but did so at the cost of limiting affordable housing options. While the intense demand for housing in the environs of Amsterdam placed the city government in a strong position to channel gentrification and promote social mixing, attempts to reduce spatial inequality within the city often came at the expense of access for low-income groups. We suspect that this trade-off is not unique to Amsterdam but can be found in other cities with overheated housing markets.

Rotterdam presents a fascinating counter-case. The city government’s strategy was fully in line with what we would expect from the literature on neoliberal urbanism: investments find their way to areas with the highest market value while social housing is demolished. Rotterdam had more pronounced spatial divisions from the outset, and these divisions were further deepened by policy interventions that reflected the right-leaning political orientation of the Rotterdam government (also see Custers & Engbersen, 2022). Although we might expect such policy interventions to be detrimental to low-income groups, we observed neither increased income segregation nor a decline in the proportion of lower-income groups in the city’s population. We suspect that the trade-off we observed in Rotterdam—that attempts to attract and retain affluent residents come at the expense of deepening spatial inequality—applies to struggling post-industrial cities more broadly.

We should note that we generally found rather small percentage changes—the populations of Amsterdam and Rotterdam changed, but not dramatically. One reason why the numbers are generally not spectacular is that different developments do not always push in the same direction. We saw this clearly when comparing the transformed and the untransformed stock in Amsterdam and especially Rotterdam: while low-income groups are less present in the former, their proportion increases in the latter. Another explanation for the seemingly modest changes is that policy in the field of housing takes a long period to materialize. Even when policy efforts are intense, as in the period under study, they affect only a small proportion of the total stock. While this could be taken as a reason to relativize the impact of policies, it also means that it will take a long time and concerted effort to reverse the trends of displacement that we found in Amsterdam and of increasing spatial stratification in Rotterdam.

Conclusion

Although the impact of public policy on spatial inequality is of key concern to urban studies, it is notoriously difficult to conceptualize and measure. In this article, we have presented an approach to unravel the impact of different kinds of policy measures on different dimensions of spatial inequality. Beyond our empirical findings on Amsterdam and Rotterdam, our research supports three more theoretical conclusions.

First, we need to distinguish between different dimensions of spatial inequality. Had we relied exclusively on metrics of segregation, we would have falsely concluded that Amsterdam has grown more egalitarian over our study period. When we consider spatial inequality’s other dimensions, a more complex picture emerges: segregation may have decreased, but low-income groups did not move up in the housing market hierarchy and suffered from decreased access to housing in the city. In Rotterdam, low-income groups did not suffer as much from reduced access to housing, nor did we observe growing segregation. But there was a trade-off, with low-income households increasingly confined to low-status areas. Although residential segregation certainly matters, our findings suggest that spatial stratification and access to the housing market deserve greater attention than they have hitherto received. Both essentially pertain to exclusion: from neighborhoods of privilege and associated resources, and from the city as a whole, respectively. Assessments of social and spatial justice should thus look beyond segregation to take into account different dimensions of spatial inequality.

Second, studying the impact of public policy on spatial inequality requires simultaneously examining different kinds of policy interventions. Neoliberalization—which is more than the simple retreat of government—combines deregulation and intervention (cf. Peck & Tickell, 2002). We have shown how such contradictory tendencies play out in the housing market. Deregulation and intervention (in the form of government-directed and subsidized demolition/construction) have markedly different geographies. Since deregulatory and interventionist policy measures are two sides of the same process of state restructuring, analyzing them separately provides, at best, a partial picture. The spatial selectivity of the state is complex and contradictory (Jones, 1997), with different policies pushing in different directions. Middle-class politics promoting housing commodification and neighborhood gentrification may co-exist with efforts to preserve a social mix and dampen spatial inequality (Boterman & van Gent, 2023). This complexity also has implications for how we assess policy. Instead of asking whether individual policy measures are successful or questioning whether ambitions to promote social mixing are sincere, the task becomes to understand policy measures as part of a broader process of state restructuring and investigate their differential effects.

Third, the approach we have developed in this paper goes some way towards meeting the challenge of mapping and quantifying displacement. While recent studies have advanced different methods to study how gentrification affects displacement (Easton et al., 2019; Preis et al. 2021), much work remains to be done to investigate the impact of policy and differentiate between various dimensions of spatial inequality. Although methodological and data challenges can partially explain the dearth of research on how policy impacts spatial inequality, another reason is that quantitatively minded scholars have prioritized different research agendas, notably those revolving around neighborhood effects and residential mobility. Bringing different approaches into the conversation is necessary to examine the causes and patterns of spatial inequality and, more broadly, to develop an urban studies approach that is both theoretically reflexive and statistically rigorous (Derudder & van Meeteren, 2019; Kwan & Schwanen, 2009).

Acknowledgments

We thank Thijs Bol, Petter Törnberg, and Wouter van Gent for their comments and Takeo David Hymans for his textual edits. We also gratefully acknowledge feedback from colleagues at University of Amsterdam’s Center for Urban Studies.

Disclosure statement

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

Additional information

Funding

Cody Hochstenbach acknowledges the financial support of a VENI grant (VI.Veni.191S.014, “Investing in inequality: how the increase in private housing investors shapes social divides”) from NWO, the Dutch Research Council. This work was supported by NWO (Dutch Research Council).

Notes

1 We use the rankseg Stata module to calculate the H^R index (Reardon & Townsend, 2018).

2 We should expect the spatial stratification index to return lower scores than the Gini index since the former runs from -1 to 1 and house prices are more evenly distributed than wealth or income. Furthermore, we calculate the SSI over data aggregated to the neighborhood level. We are less interested in the absolute value than in the direction of change (more or less inequality) and the impact of policy interventions.

3 The sum of individual interventions does not necessarily equal the impact of interventions taken together, e.g. due to conversions from owner-occupation to rental (buy-to-let).