Multidimensional housing inequality across eight European countries: a cross-national cluster and risk analysis based on an online survey

Technical Summary: Multidimensional Housing Inequality Across Eight European Countries

Problem Statement
Housing inequality in Europe is a pervasive issue with significant implications for health, wellbeing, and life chances. While existing literature has increasingly moved beyond income-based measures to assess deprivation through observed living conditions, a critical gap remains in understanding how housing vulnerabilities intersect with environmental and infrastructural dimensions at the household level. Previous research has often examined dwelling quality, green-space access, and transport access in isolation or adopted spatial perspectives that overlook how these disadvantages co-occur within specific households. Furthermore, there is limited comparative evidence on whether dwelling vulnerability consistently correlates with inequalities in external domains (green space and transport) or if these relationships are shaped by context-specific processes. This study addresses these limitations by examining multidimensional housing-related vulnerability across eight European countries.

Methodology
The study utilizes data from a 2025 multinational online survey conducted across Belgium (Flanders), Czech Republic, Germany, Estonia, Finland, Great Britain, Italy, and Romania. The final analytical sample comprises 7,169 households, stratified to approximate Eurostat distributions for income, age, and gender.

The analysis employs a two-stage strategy:

1. K-Prototypes Cluster Analysis:

   • Objective: To identify recurring household vulnerability profiles based on combinations of dwelling, green-space, and transport indicators.
   • Variables: The clustering model included variables representing dwelling conditions (e.g., sanitation, space shortage, leaks/mould, energy efficiency, heating/cooling affordability) and immediate environmental access (green space and transport). Socio-demographic and health variables were excluded from the clustering algorithm to ensure profiles were driven by housing-related factors, though they were used subsequently to characterize the clusters.
   • Algorithm: The K-Prototypes algorithm was selected to handle mixed data types (numerical and categorical), utilizing a distance function combining squared Euclidean distance and simple matching dissimilarity. The optimal number of clusters was determined using the elbow heuristic and silhouette coefficients, resulting in a four-cluster solution.
   • Validation: Principal Component Analysis (PCA) and Uniform Manifold Approximation and Projection (UMAP) were used to visualize cluster separation and internal structure.

2. Category-Based Relative Risk (RR) Analysis:

   • Objective: To assess the cross-domain co-occurrence of inequalities, specifically testing whether dwelling vulnerability is associated with green-space and transport inequalities.
   • Procedure: Ordered vulnerability categories were created for three domains: dwelling (cumulative score), green space (access levels), and transport (mobility constraints). These were recoded into binary exposure (dwelling vulnerability) and outcome (green-space or transport inequality) variables.
   • Metrics: Relative risks (RR) and 95% confidence intervals (using the Katz log method) were calculated for the pooled sample and stratified by country to examine cross-national variations.

Key Contributions
The paper makes three primary contributions to the literature:

1. Extension of Scope: It moves beyond dwelling-centred approaches by simultaneously examining dwelling conditions, green-space access, and transport disadvantages at the household level across a multinational sample.
2. Typology Development: It identifies and characterizes four recurring configurations of multidimensional vulnerability across diverse European contexts, providing a comparative typology of housing-related disadvantage.
3. Cross-Domain Association Testing: It empirically tests whether dwelling vulnerability is more strongly associated with specific external domains of inequality (green space vs. transport), offering evidence on which dimensions of disadvantage tend to co-occur versus those shaped by distinct contextual processes.

Results

• Cluster Profiles: The analysis identified four distinct household profiles:

  1. Accumulated Housing and Environmental Vulnerability: Characterized by the broadest concentration of dwelling problems (damp, mould, space shortage), limited green-space access, high financial strain, and poor health outcomes. This group is predominantly urban and includes many single-person households.
  2. Thermal and Energy Vulnerability: Defined primarily by heating difficulties and cooling discomfort rather than structural defects. This cluster consists largely of older populations in urban settings with moderate financial strain.
  3. Resource-Advantaged with Thermal Strain: A socio-economically advantaged group with favorable dwelling quality and high green-space access, yet still experiencing cooling discomfort. This group is predominantly suburban/rural with high car ownership.
  4. Dwelling Problems and Transport Vulnerability: Characterized by persistent structural dwelling problems and weak transport access, despite an intermediate socio-economic position. This group shows high rates of chronic illness and is often located in suburban or rural areas.

• Relative Risk Findings:

  • Green-Space Inequality: There is a consistent positive association between dwelling vulnerability and green-space inequality. In the pooled sample, households with dwelling vulnerability were 57% more likely to experience green-space inequality than those without (RR = 1.57; 95% CI: 1.39–1.77). This pattern held across all national contexts, with point estimates above 1 in every country.
  • Transport Inequality: The association between dwelling vulnerability and transport inequality was weaker and less consistent. In the pooled sample, dwelling vulnerability was associated with a lower likelihood of transport inequality (RR = 0.86; 95% CI: 0.81–0.92). Country-specific estimates varied, with some showing no clear association and others showing negative associations, suggesting that transport disadvantage is shaped by context-specific processes (e.g., urbanicity) rather than a uniform correlation with housing quality.

Significance and Claims
The authors claim that these findings support a multidimensional approach to housing inequality, demonstrating that housing-related vulnerability extends beyond the physical dwelling to encompass environmental and infrastructural access. However, the paper modestly asserts that different domains of disadvantage do not necessarily overlap in the same way. Specifically, the consistent link between poor housing and lack of green space suggests these issues are structurally intertwined, whereas the weak link to transport inequality indicates that housing disadvantage does not automatically imply transport disadvantage.

The study concludes that policy responses should be integrated but differentiated: housing assessments must consider the wider living environment (particularly green space), while transport policies require context-specific strategies that account for the spatial distribution of services and car dependence, rather than assuming a direct correlation with housing quality.

Limitations
The authors acknowledge several limitations: the cross-sectional nature of the data prevents analysis of vulnerability trajectories over time; the clustering solution relies on methodological decisions and variable weighting (with dwelling indicators potentially dominating due to their quantity); and the non-probability sampling strategy limits the generalizability of prevalence estimates, though the sample was adjusted for demographic balance. The clusters are presented as heuristic summaries of recurring configurations rather than discrete population categories.