Intra-urban house prices in Madrid following the financial crisis: an exploration of spatial inequality

The study investigates whether and how spatial inequality in house prices within Madrid evolved across the post-2008 crisis period, focusing on the bust (2010–2015) and subsequent boom (2016–2019). Motivated by the challenge of spatialising ‘wealth’ in financialised housing systems, the paper uses house prices as a locationally fixed proxy for wealth inequality and as a mediator of socio-spatial inequality. Prior evidence indicates that uneven house price appreciation amplifies wealth disparities and shapes access to amenities, with European cities showing strengthening spatial polarisation. Spain’s boom-bust cycle was acute, with exceptionally large pre-2008 price inflation followed by sharp declines and widespread foreclosures. Against this backdrop, the paper aims to measure changes in intra-urban housing price inequality over time, assess their spatial structure, and identify emerging sub-markets, thereby shedding light on urban segregation and accessibility implications in a major European capital.

The paper situates its contribution within research on housing financialisation and rising wealth inequality across Western Europe, highlighting housing’s central role in modern wealth disparities and intergenerational mobility. It notes evidence of intensified spatial polarisation in European cities since 2000, with high-value neighbourhoods driving divergence. Spain’s pre-crisis state-led promotion of homeownership, rapid mortgage expansion, and low interest rates fuelled a dramatic boom followed by a severe bust, precipitating mass evictions and social unrest. The limited provision of social housing (about 2.8% of stock) and post-crisis privatisations exacerbated affordability pressures and segregation. Studies show the suburbanisation of poverty, environmental risks associated with low-quality housing, and strong links between accessibility to amenities and house prices. Methodologically, prior work has applied inequality metrics (Gini, Theil), spatial autocorrelation (Moran’s I, LISA), and explored spillovers and sub-market dynamics; however, there is a relative paucity of intra-urban, high-resolution, longitudinal analyses using listing data for large cities like Madrid. This study extends the literature by combining standard and spatial decompositions of inequality with local clustering to trace the spatio-temporal evolution of house price polarisation.

Study area: The City of Madrid (604.3 km²; ~3.2 million population; ~384,364 residential buildings) with a monocentric structure featuring high central prices. Administrative spatial units at three hierarchical scales are used for aggregation: 21 regions, 131 districts, and 2442 neighbourhoods (census sections with ~2,500 residents). These official boundaries facilitate policy relevance and alignment with segregation statistics while acknowledging MAUP concerns.

Data: Housing listings from Idealista (Spain’s leading real estate portal) covering 872,016 listings in the Madrid metropolitan area from 2010–2019. The key variable is listing price (Euros), timestamped by year/quarter. Prices are inflation-adjusted using Spain’s Housing Price Index (HIPI). To improve data quality, listings that remained for over three months without price changes were removed (n=10,455). Listings that persisted with price changes were retained (n=278,991), as indicative of market dynamics. Idealista’s market dominance suggests broad supply coverage, though the dataset is not the full housing stock. Listing prices are used as a proxy for market value given final sale prices are not openly available; prior work indicates a typical asking-to-closing discount of roughly 5–13.8% in Madrid over 2010–2018. Annual variation in listing counts reflects both market conditions and portal popularity.

Inequality metrics: The Gini coefficient is computed annually at multiple intra-urban scales: dwelling-level, neighbourhoods (n=2442), districts (n=131), and regions (n=21). To incorporate spatial structure, a spatial decomposition of the Gini (Rey & Smith, 2013) is applied at the neighbourhood level using rook-contiguity spatial weights, separating ‘near differences’ (between adjacent units) and ‘far differences’ (between distant units). Statistical significance of the decomposition is assessed (reported p<0.01 for neighbourhood-level over time; not significant at larger geographies). Kernel Density Estimation (KDE) is used to visualise annual shifts in the neighbourhood house price distribution.

Spatial autocorrelation and clustering: Global Moran’s I measures overall spatial autocorrelation in neighbourhood house prices over time. Local Indicators of Spatial Association (LISA) identify local clusters: High-High (hot spots), Low-Low (cold spots), and instances of negative autocorrelation (High-Low, Low-High), with a non-significant category capturing areas without significant spatial autocorrelation. Rook contiguity defines neighbours for spatial weights in Moran’s I and LISA. Annual LISA classifications quantify the changing shares of hot/cold/non-significant clusters.

Aggregation and MAUP: Micro-level listings are aggregated to neighbourhood, district, and region scales to evaluate scale effects on inequality measures. The analysis acknowledges that aggregation can mask micro-level variation and that zonation/scale choices can influence results.

Computation and reproducibility: Analyses draw on methods and code adapted from the open textbook Geographic Data Science with Python (Rey, Arribas-Bel & Wolf, 2023), particularly sections on Local Spatial Autocorrelation and Spatial Inequality Dynamics.

• Market trajectory: Average listing price declined from €443,970 (2010) to €280,500 (2015), then partially recovered to €324,209 (2018) and €306,594 (2019).
• Inequality growth across scales: Gini increased at all intra-urban scales over 2010–2019, with the largest rise at the regional level (+6.93 percentage points: 18.82% to 25.75%). District Gini rose by 6.65 points; neighbourhood by 2.67; dwelling-level peaked at 50.49 in 2016 (from 40.87 in 2010), then declined modestly by 2019.
• Temporal pattern: Inequality rose year-on-year through the bust, peaking around 2016/2017, then decreased slightly during the boom years, indicating uneven impacts of the bust across neighbourhoods.
• Distributional shifts: The neighbourhood price distribution changed from approximately bell-shaped (2010) to positively skewed during 2010–2015. From 2016–2019 it became bimodal, indicating polarisation with concentrations of low- and high-priced neighbourhoods.
• Spatial Gini (neighbourhoods): Far differences increased until ~2014, plateaued to ~2017, then slightly declined; near differences decreased continuously, implying growing similarity among adjacent neighbourhoods but increased disparities across distant areas. Spatial decomposition was significant over time at p<0.01 at the neighbourhood level, not at larger geographies.
• Spatial autocorrelation: Global Moran’s I rose from 0.50 (2010) to 0.80 (2019), with a sharp increase early in the decade (to 0.78 by 2013), evidencing intensifying clustering of similar price levels.
• LISA cluster dynamics: Share of cold spots rose from 21.7% (2010) to 31.0% (2019); hot spots increased from 8.6% to 11.3%; non-significant areas declined from 67.6% to 57.3%. In counts, cold-spot neighbourhoods grew from 532 to 758; hot spots from 211 to 277.
• Spatial patterning: Hot spots are concentrated in central and northern areas (e.g., Retiro, Salamanca, Chamartín; also Moncloa-Aravaca, Chamberí; and parts of Barajas, Hortaleza, northern Ciudad Lineal) with average property prices ~€700,000–€950,000. Cold spots are largely peripheral in the southern semicircle (Usera, Villaverde, Puente de Vallecas, Carabanchel) with average prices ~€90,000–€170,000, plus smaller cold spots in parts of Ciudad Lineal and Tetuán.
• Sub-market wealth inequality: Using listings in consistently hot/cold neighbourhoods, hot spots (26% of listings) hold 53% of total housing wealth; cold spots (38% of listings) hold only 16%.
• Differential crisis impact and recovery: 2010–2014 per m² prices declined by 24% in hot spots vs 47% in cold spots. By 2019, hot-spot prices were ~€5,000 per m² (about €1,000 higher than 2010), whereas cold-spot prices were ~€2,300 per m² (about €400 below 2010), indicating slower and incomplete recovery in low-priced areas.

Findings show that spatial inequality and polarisation in Madrid’s housing market intensified following the financial crisis, with the sharpest inequality increases during the bust. Rising Moran’s I and the growth of hot and cold LISA clusters indicate stronger spatial segmentation into high- and low-priced sub-markets. The spatial Gini decomposition suggests that widening disparities are driven primarily by differences between distant neighbourhoods, while adjacent areas became more similar, consistent with clustering and possible spillover processes. The distribution becoming bimodal during the boom years further evidences polarisation.

These dynamics have significant socio-economic implications. The hot sub-market, concentrated in central/northern areas with better access to employment and amenities, proved more resilient during the bust and achieved stronger recovery, capturing a disproportionate share of housing wealth (53%). In contrast, cold sub-markets, largely peripheral and near industrial zones, experienced larger value losses and slower recovery, exacerbating wealth gaps and likely reinforcing socio-spatial segregation. Such uneven trajectories mirror patterns documented in other financialised urban markets.

While inequality began to ease modestly after 2017, continued monitoring is warranted, including evaluation of subsequent shocks (e.g., COVID-19) that may have altered spatial demand. The observed patterns align with hypothesised spillovers, but drivers require further investigation, including the roles of accessibility, amenities, and neighbourhood socio-demographics.

The study demonstrates that intra-urban housing price inequality in Madrid increased notably during the post-2008 bust and remained elevated, with clear spatial polarisation into persistent hot and cold sub-markets. By leveraging high-resolution listings data and combining standard and spatial inequality metrics with local clustering, the research provides fine-grained evidence of how market shocks reshape urban wealth geographies. The unequal resilience and recovery between sub-markets intensify wealth disparities and segregation, with central, amenity-rich areas consolidating advantages.

Future research should: (1) identify causal drivers of increasing spatial inequality and test for spillover mechanisms; (2) profile the socio-economic and demographic characteristics of hot/cold sub-markets; (3) examine the rental market where financialisation has shifted tenure patterns; and (4) assess impacts of subsequent shocks (e.g., COVID-19). Policy should target equitable access to homeownership, expand affordable and social housing—particularly in central areas—and more evenly distribute public services and accessibility, aligning with initiatives such as the 15-minute city to mitigate the suburbanisation of poverty and rising segregation.

• Data represent listing prices rather than final sale prices; while correlated, asking-to-closing discounts (approx. 5–13.8% in 2010–2018) may bias levels though likely preserve relative patterns.
• Coverage is broad but not the full housing stock; Idealista’s dominance suggests strong market representation, yet unsold or unlisted properties are not captured.
• Deduplication is partial: listings reposted after removal cannot be fully identified; listings without price changes >3 months were removed, which may affect counts but is unlikely to alter main trends.
• Aggregation to administrative units introduces MAUP; aggregation can mask micro-variation, and zonation choices may influence spatial statistics.
• Spatial Gini significance held at neighbourhoods but not at larger geographies; some spatial structure is thus scale-dependent.
• The study is descriptive and exploratory; drivers (e.g., amenity accessibility, socio-demographics) and spillover effects are not causally identified or formally tested.
• Focus is on ownership/listing markets; rental market dynamics, which may differ, are not analysed.