Green Gentrification in European and North American Cities

The study investigates whether urban greening contributes to or accelerates neighborhood gentrification, thereby exacerbating social and racial inequities in access to greenspace benefits and climate justice. While greenspaces provide ecosystem services for climate adaptation and mitigation and foster social cohesion, their distribution is often unequal by race and income. Recent scholarship posits a green space paradox wherein high-profile greening and climate-adaptive infrastructure can raise property values and displace working-class and racialized residents, leading to green/environmental/climate gentrification. The central hypothesis tested is that greening in a neighborhood during a given period drives gentrification in the immediately subsequent period. Prior quantitative studies largely examined single indicators (e.g., income, housing prices, racial composition) rather than multivariate constructs of gentrification, and few assessed the timing and citywide extent of green gentrification across international contexts. This study addresses these gaps using a composite, multivariate gentrification measure and a comparative, multi-city, multi-decade design.

The paper reviews extensive environmental justice literature documenting racial and class disparities in access to park area, quality, and maintenance in Global North cities. Quantitative green gentrification research often uses single outcomes: increases in housing prices, median income, college-educated or professional residents, and changes in racial/ethnic composition (e.g., loss of African-American or Hispanic residents). Some studies reframe as neighborhood change, linking new parks to changes in composition, flows, housing costs, vacancies, and reputation. A key unresolved issue is greening’s role relative to other drivers (e.g., proximity to already gentrifying areas, density of greenspaces, centrality, housing stock quality, design). Findings include: proximity to already gentrified neighborhoods heightens susceptibility (Philadelphia); critical density/area of greenspaces correlates with property values and gentrification (New York City); park location near city centers can trigger gentrification regardless of size/function (US-wide); and park effects vary with housing stock quality and amenity design (Barcelona). The literature suggests context matters and greening may be primary in some cities but subsidiary or integrated with transit and development in others.

Design: International comparative analysis of 28 mid-sized (500,000–1.5 million residents) cities across 9 countries in Western Europe and North America, selected to maximize diversity in geography, growth trajectories, urban form, and greening agendas. Three periods approximating decades from 1990 to 2016 were analyzed (1990s, 2000s, 2010s). Data: For each city, the team compiled spatial boundaries and inauguration years (≥1990) for all public greenspaces: parks, greenways, preserves, gardens, and recreation areas, validated via municipal records, reports, media, archival sources, and imagery. Greenspace inauguration was defined as acquisition for public use or opening year. Greenspace exposure by tract used the tract polygon plus a 400 m buffer to represent a standard walkable catchment. Gentrification measurement: A composite gentrification score per small-area unit (e.g., US census tracts; IRIS in France; Secció Censal in Barcelona) was calculated for each period as a diversity-weighted sum of standardized social change indicators plus standardized rent change. Social indicators included: increase in non-vulnerable ethnic group share (nvul), university-educated share (uni), professional occupation share (prof), high-income share (hises), and non-poverty share (npov). Vulnerable groups were defined contextually (e.g., immigration/nationality markers where race data absent). Each variable’s change was normalized by citywide change, then converted to Z-scores. A Shannon’s Equitability Index (HE) weight captured the evenness/diversity of concurrent changes across the five social indicators, emphasizing areas with multifaceted change. The final score: Gtract = HE(Zvul + Zuni + Zprof + Zhises + Znpov) + Zrent, where Zrent is the standardized change in the share of households paying above-median rent. Temporal alignment: Greenspace area attributed to a period included inaugurations during that period plus two subsequent years (to capture announcement effects) and was matched to subsequent-period gentrification outcomes: H1 tests Period 1 greenspace → Period 2 gentrification; H2 tests Period 1 greenspace → Periods 2–3 gentrification; H3 tests Periods 1–2 greenspace (compounded) → Period 3 gentrification. Covariates: Small-area and city-level controls included new residential buildings (by decade), new transit stops since 1990, distance to city center, residential density, prior green coverage (1990), city-level population change, and GDP change (from 2001 due to availability). Models: Global analysis used a Bayesian linear mixed model (city and country as random effects; country ultimately non-relevant) with covariate selection via BayesVarSel for fixed effects, followed by inclusion of random effects. City-level analyses applied a three-step procedure: (1) BayesVarSel variable selection; (2) test residual spatial autocorrelation with Moran’s I; (3) fit a Bayesian model with a spatial conditional autoregressive (CAR) random effect orthogonal to covariates when needed. Inference used vague priors and Integrated Nested Laplace Approximation (INLA). The modeling emphasizes temporal precedence of greening relative to gentrification and accounts for unobserved spatial processes.

• Global model: Across all 28 cities, greenspace inaugurated during the 1990s–2000s is positively and relevantly associated with gentrification in the 2010s; greenspace from the 1990s is positively associated with gentrification spanning 2000–2016, while the relationship to 2000s-only gentrification is weakly negative. New residential development and city-level GDP growth also show positive relevant associations; greater distance from city center relates negatively to gentrification (i.e., closer-in areas gentrify more). Country effects are non-relevant, indicating variation is best explained at the city level.
• City-level results: 17 of 28 cities show that greening from an earlier period is a relevant positive predictor of gentrification in the immediately following period for at least one decade between 1990 and 2016. Of these, 11 are in the US, 2 in Canada, and 4 in Europe, indicating greater prevalence in North America but not exclusively so.
• Temporal patterns: Two primary patterns among green gentrification cities: (1) Long-term sustained citywide green gentrification across at least two decades (e.g., Atlanta, Louisville, Seattle; Vancouver, Boston, San Francisco; Washington, DC; Denver, Philadelphia; Milwaukee; Barcelona). (2) Short-term patterns limited to one decade, often in the later period (e.g., Austin, Montreal, Nantes, Copenhagen), or earlier only (Detroit, Edinburgh).
• Typology of roles: Three types of green gentrification based on greening’s role relative to other drivers: • Lead Green Gentrification (8 cities): Greening is the standout, sustained explanatory factor citywide; other built environment changes like new development/transit are not relevant (e.g., Atlanta, Austin, Copenhagen, Louisville, Milwaukee, Montreal, Nantes, Vancouver). • Integrated Green Gentrification (6 cities): Greening’s role is on par with new development and/or transit (e.g., Barcelona, Boston, Denver, Edinburgh, San Francisco, Seattle). • Subsidiary Green Gentrification (3 cities): Greening is relevant but secondary to other interventions (e.g., Detroit; also Philadelphia and Washington, DC classified as subsidiary in figure).
• Cities with no clear green gentrification (11 cities): Greening is negative or not relevant; gentrification is linked instead to new development and/or transit (e.g., Baltimore, Bristol, Cleveland, Portland, Sheffield, Valencia, Vienna). Many of these are European (Amsterdam, Bristol, Lyon, Sheffield, Valencia, Vienna, Dublin). In some cases, data limitations constrained temporal coverage.
• Spatial pattern: When relevant, distance from center typically negatively explains gentrification (closer-in areas more affected), suggesting redevelopment in post-industrial or underused areas adjacent to historic centers. Maps of posterior predictive means illustrate localized areas with strongest greening–gentrification relationships (e.g., Île de Nantes and adjacent neighborhoods in Nantes).

Findings provide a conservative, citywide affirmation of the green gentrification hypothesis in a majority of the 28 cities studied, highlighting that greening often contributes to gentrification with a temporal lag and increasing influence into the 2010s. The city-specific analyses reveal heterogeneity in greening’s role: in some cities greening is the principal driver, in others it is integrated with or secondary to new development or transit investments. Overall prevalence is higher in North America, potentially reflecting weaker housing rights, greater reliance on property tax-driven growth, and the coupling of greening with investment-oriented redevelopment strategies. In several European cases (e.g., Copenhagen, Nantes, Barcelona), more recent investment-led greening and climate-livability initiatives coincide with emerging green gentrification patterns, particularly where social protection and affordability measures have been weakened or were historically limited. Conversely, cities with robust social housing and anti-displacement policies (e.g., Amsterdam, Lyon) may be buffered against green gentrification. The spatial tendency for effects near but not within historic centers underscores redevelopment of post-industrial areas. Importantly, even integrated and subsidiary types indicate greening is part of the causal mix, suggesting that mitigating displacement requires addressing greening alongside other growth-oriented interventions. The study’s composite gentrification index and multi-decade, multi-city modeling provide a high-level, comparative test that complements site-specific studies of pathways.

The study demonstrates that across many mid-sized cities in the Global North, urban greening—while beneficial for climate, health, and livability—often contributes to gentrification, reinforcing social, racial, and health inequities and potentially undermining climate equity and justice. Contributions include: (1) an internationally comparative, multidecadal analysis across 28 cities; (2) a composite, diversity-weighted gentrification index capturing multi-dimensional social and real estate changes; (3) spatially explicit Bayesian modeling separating greening’s role from other drivers; and (4) a typology of greening’s relative role (lead, integrated, subsidiary). The authors call for anti-displacement and inclusive greening policies to ensure climate-responsive cities center equity and long-term health. Future research should broaden greening operationalizations (types and sizes of green spaces), extend to Global South cities and smaller towns, and refine spatial and temporal data to capture localized effects within and beyond city boundaries.

• Temporal truncation: The most recent period (2010–2016) is shortened, possibly underestimating recent gentrification extent.
• Data heterogeneity: Variable availability and definitions differ between Europe and the US; some cities lack complete time-series or variables (e.g., transit data not in global model due to coverage).
• Endogeneity/causality: Some covariates (e.g., GDP growth) may also be consequences of gentrification; while temporal ordering is enforced, causality cannot be fully disentangled.
• Spatial units and boundaries: Analyses within formal city boundaries may suffer from modifiable areal unit problems and miss cross-boundary dynamics.
• Greening operationalization: Focus on five greenspace types may omit relevant forms of greening; density/design/quality nuances are not fully captured.
• Data access limits: Some city datasets (e.g., Copenhagen, Vienna) are restricted; US demographic data used a pre-processed standardized product; gaps exist in later periods for some European cities.