A comparative, multiscalar, and multidimensional study of residential segregation in seven European capital cities

1. Problem Statement

Despite the significance of ethnic residential segregation for migrant integration and urban functioning, comparative cross-European research remains scarce. Existing studies suffer from two primary limitations:

• Single-Dimension Focus: Most studies rely solely on the Dissimilarity Index (Evenness), ignoring other critical aspects of segregation such as clustering, concentration, and exposure.
• Single-Scale Limitation: Research typically utilizes a single administrative spatial scale (e.g., census tracts or neighborhoods). This approach fails to capture how segregation manifests differently at varying distances from an individual's home and often ignores the continuous nature of urban space, particularly the relationship between city cores and hinterlands.

The authors argue that segregation is a multidimensional and multiscale phenomenon. A city may appear integrated at a large scale but highly segregated at a micro-scale, or vice versa. Furthermore, the definition of "neighborhood" varies by urban form, making fixed administrative units inadequate for cross-city comparison.

2. Methodology

The study employs a novel multiscalar and multidimensional approach to analyze ethnic segregation in seven European capitals: Amsterdam, Berlin, Lisbon, London, Madrid, Paris, and Rome.

Data Sources and Definitions

• Data: Harmonized high-resolution spatial data (100m × 100m grid cells) from the European Commission's D4I (Data for Integration) project, sourced from national statistics offices.
• Population Groups: To ensure comparability across diverse migration histories, the study defines the minority group as "Non-Western immigrants" (originating from Africa, Asia, or Latin America) versus the "Western" population (natives + Western immigrants).
• Spatial Units: The study utilizes Functional Urban Areas (FUAs) defined by the OECD/EU, which include both the densely populated urban core and the integrated hinterland.
• Bespoke Multiscale Areas: Instead of fixed administrative units, the authors use a "distance profile" approach. Starting from a 100m grid cell, they create overlapping circular buffers with radii increasing in 100m increments up to 10km (101 scales total). This allows for a continuous analysis of segregation as the spatial context expands.

Analytical Framework

The study measures segregation using Massey and Denton's (1988) five dimensions, adapted for the multiscale approach:

1. Evenness: Measured via the Index of Dissimilarity. Indicates the percentage of a group that must move to achieve an even distribution.
2. Exposure: Measured via the Isolation Index. Quantifies the probability of a minority group member encountering members of their own group versus the majority group in their immediate vicinity.
3. Concentration: Measures the physical space occupied by a group relative to the total population density.
4. Clustering: Measured via the Spatial Proximity Index. Assesses the extent to which minority groups form contiguous enclaves (spatial proximity of similar groups).
5. Centralisation: Measures the relative concentration of groups in the urban core versus the hinterland.

3. Key Contributions

• Multidimensional Perspective: The paper moves beyond the Dissimilarity Index to show that cities can exhibit high segregation in one dimension (e.g., clustering) while appearing integrated in another (e.g., exposure), depending on the metric used.
• Multiscalar Analysis: By generating "distance profiles," the study demonstrates that segregation levels are not static; they fluctuate significantly as the spatial scale expands from immediate surroundings to the wider metropolitan region.
• Core vs. Hinterland Comparison: The study explicitly compares urban cores and hinterlands, revealing that segregation patterns often differ drastically between these zones (e.g., higher isolation in cores vs. higher concentration in hinterlands).
• Challenge to Conventional Wisdom: The study challenges the assumption that segregation always decreases as spatial scale increases. In some cases (e.g., Amsterdam), segregation increases with scale.

4. Key Results

General Patterns

• North vs. South: Southern European capitals (Lisbon, Madrid, Rome) generally show lower overall ethnic segregation compared to Western/Northern capitals (London, Amsterdam, Paris, Berlin).
• London: Exhibits the highest levels of segregation across multiple dimensions. It has a large non-Western population that is highly centralised, forms large clusters, and experiences high isolation (low exposure to Western populations).
• Berlin: Shows a unique pattern where non-Western populations are highly centralised in the core but form smaller, less dense clusters. Interestingly, Berlin shows values below 1 for clustering in hinterlands, indicating non-Western residents are sometimes closer to Western residents than to their own group.

Dimension-Specific Findings

• Evenness (Dissimilarity): In most cities, unevenness is higher in the hinterland than the core. However, the scale at which this peaks varies (e.g., <1km in Berlin vs. up to 5km in Amsterdam).
• Exposure (Isolation): London has the highest isolation (lowest interaction potential). Conversely, Rome has the lowest isolation despite high micro-scale segregation, largely because the overall share of non-Western residents is low, increasing the statistical probability of interaction.
• Concentration: Non-Western populations generally occupy smaller physical areas than the majority. Amsterdam and Paris show increasing concentration values at larger scales, contrary to the trend of decreasing segregation with scale.
• Clustering: London and Amsterdam show the strongest clustering. In Amsterdam, clusters in the hinterland form at a scale of ~3.5km, whereas core clusters are much smaller (<1km).

The "Scale Effect"

The most significant finding is that segregation does not necessarily decrease with spatial scale.

• In Amsterdam, segregation (evenness and concentration) increases as the scale expands to 3–5km in the hinterland and ~7km in the core. This suggests that while neighborhoods may look mixed, the broader metropolitan context reveals distinct, separated zones.
• In Madrid, while the city appears integrated at large scales, micro-scale analysis reveals significant local segregation in the core.

5. Significance and Policy Implications

• Methodological Rigor: The study proves that relying on a single index or a single administrative scale leads to incomplete or misleading conclusions about urban integration.
• Policy Targeting: The findings suggest that policy interventions must be tailored to specific spatial scales and dimensions.
  • Micro-scale issues: In cities like Madrid and Rome, interventions should focus on immediate neighborhood mixing (schools, local services) as segregation drops quickly at larger scales.
  • Macro-scale issues: In cities like London and Amsterdam, segregation persists or grows at larger scales, requiring regional planning interventions (housing distribution, transport links) to connect separated clusters.
• Urban Form Matters: The study highlights that the physical structure of a city (core vs. hinterland dynamics) dictates how segregation is experienced.
• Future Research: The authors call for further analysis of specific subgroups (e.g., African vs. Latin American migrants) as their residential preferences and economic drivers may create distinct spatial patterns not captured by the broad "Non-Western" category.

In conclusion, the paper provides a robust framework for understanding residential segregation as a complex, fluid phenomenon that varies by dimension, scale, and urban context, urging researchers and policymakers to adopt a more nuanced, multi-layered approach to urban integration.