Quantifying urban socio-economic segregation through co-residence network reconstruction

TL;DR

The study addresses urban socio-economic segregation in Vienna by constructing a district-level co-residence network from administrative migrant data and quantifying cross-national co-residence with weights $w_{ij}^d = \kappa_i^d \kappa_j^d$, standardized as $z_{ij}^d$ and aggregated to $z_{ij}$. Using Infomap on the positive links, it identifies two distinct country clusters that are differentially associated with neighbourhood wealth and district diversity, with cultural homophily and proximity playing key roles. The findings reveal that one cluster concentrates in wealthier, less diverse districts while the other aligns with more diverse, less affluent areas, illustrating a multifactorial, dynamic pattern of segregation and integration in a major European city. The approach, grounded in a null multinomial baseline and robust clustering, provides policy-relevant insights into how wealth, diversity, and national ties shape residential sorting, with implications for designing integrative urban interventions.

Abstract

Urban segregation poses a critical challenge in cities, exacerbating inequalities, social tensions, fears, and polarization. It emerges from a complex interplay of socio-economic disparities and residential preferences, disproportionately impacting migrant communities. In this paper, using a comprehensive administrative data from Vienna, where nearly 40% of the population consists of international migrants, we analyse co-residence preferences between migrants and locals at the neighbourhood level. Our findings reveal two major clusters in Vienna shaped by wealth disparities, district diversity, and nationality-based homophily. These insights shed light on the underlying mechanisms of urban segregation and designing policies for better integration.

Figures (4)

• Figure 1: Melting Pot in Vienna.a Population of migrants in the districts of Vienna on 22nd September 2023 (in %). Districts 15 and 13 have the highest and lowest percentage of migrants, respectively (see Supplementary Table \ref{['tab:tabs1']} for a detailed list of values). b Population distributions of the top 19 most populous migrant nationalities in Vienna represented by at least ten individuals per district, the label "Others" is used for the rest. The colour bar indicates the fraction of people living in Vienna on a log scale (see Supplementary Table \ref{['tab:tabs2']} for a detailed list of values). Countries are sorted from highest to lowest population fraction in Vienna.
• Figure 2: Clustering countries sharing similar residence interests.a Diagram of the method used to build the network of co-residence. b$\textit{z}$-score network. Edges with $\lvert z_{ij} \rvert \le 20$ are not shown for better readability. The width and colour transparency of the links is proportional to the $\textit{z}$-score values, while the size of the nodes reflects the total population of each country, except for Austria. The network has been produced using Noverlap layout of Gephi software gephi. Two prominent clusters are marked with a dashed line. c World map of residence preferences. Countries that belong to the same cluster have the same colour. Countries coloured with grey lines are either subsumed within Others or for which we lack data on immigrants residing in Vienna. The larger cluster, comprising the majority of the population residing in Vienna, is shown in purple and referred to as the "majority cluster". The smaller cluster, with fewer residents, is depicted in yellow and termed the "minority cluster". d World network of residence preferences. The size of the nodes represents the total $\textit{z}$-score of the clusters and countries. The links represent the connections between nodes obtained from the cluster analysis with Infomap mapequation2023software; the thicker the line, the stronger the connection.
• Figure 3: Socio-economic and diversity factors underlying nationality clusters.a Correlation between district wealth and population fraction for each nationality. b Correlation between diversity and population fraction for each nationality. Nationalities are ranked by their correlation values, from highest to lowest for wealth-population fraction correlation in a, and from lowest to highest for diversity-population fraction correlation in b. The nationality clusters are colour-coded consistently with Fig. \ref{['fig:fig2']}c.
• Figure 4: District diversity and national homophily in Vienna.a District diversity. Districts are ranked from highest to lowest based on the estimated diversity of Vienna districts. b Vienna diversity map. c National homophily. Nationalities are ranked from highest to lowest based on the estimated homophily of countries. d World homophily map.