The Luxury Effect Is Weakening (U.S., 87 Cities, 1990–2023)
Summary
Question. Do wealthier neighborhoods still have more greenness and cooler surfaces—and if that pattern is changing, how is equity shifting?
What we did. Built a tract-level, multi-decade panel (1990–2023) for 87 U.S. cities using Landsat NDVI & LST + Census income; estimated within-city luxury-effect slopes by year and tested trend/weakening with pairwise city models and pooled mixed-effects.
Outcome. By 2023, the luxury effect had weakened sharply nationwide (median city decline: −71% for NDVI, −78% for LST); rich–poor gaps narrowed—NDVI via both bottom-up greening and top-end loss, LST via faster warming in affluent tracts.
Research Challenge
Planners often assume a stable “rich = greener & cooler” gradient. We needed to test whether that relationship is moving over time across climates and cities, and determine whether apparent equity gains reflect improvement in underserved neighborhoods (“leveling up”) or loss of advantage at the top (“leveling down”).
Approach
Data
Landsat 4/5/7/8/9 Level-2 → NDVI & LST (May–Sep windows).
Census tract-level socio-demographics (1990, 2000, 2010, 2020/2023).
Context layers: GHSL building volume, NLCD impervious/tree canopy, TerraClimate/SPEI water-balance indicators; SRTM elevation.
Methods
For each city–year, regress tract NDVI or LST on standardized income to get annual luxury-effect slopes.
Test 1990→2023 change per city using a tract-level interaction model (y ~ income_z * year).
Estimate continental trend with a linear mixed-effects model pooling all cities/years.
Quality controls: Landsat QA, ≥50% tract pixel coverage, light winsorization (1st–99th pct).
Tools
Google Earth Engine (ingest/masking), ArcGIS Pro (zonal stats), MATLAB/Python (modeling & LME).
Key Findings
Nationwide weakening. In 1990, 87% of cities had a significant income→NDVI slope and 82% income→LST; by 2023 that fell to 37% and 30%. Median city weakening: −71% (NDVI), −78% (LST). No city significantly strengthened.
Mechanisms of equity change.
NDVI gap shrank by 0.028 NDVI/decade, split between bottom-tail greening (+0.0144/dec) and top-tail loss (−0.0137/dec).
LST gap shrank by 0.69 °C/decade because high-income tracts warmed faster (+1.67 vs +0.98 °C/decade).
What drives the weakening. Built intensity and urban form were the strongest correlates of slope attenuation (e.g., higher building volume → much higher odds of weakening).
Pathway for heat. The mean thermal luxury effect is largely mediated by income→greenness; background climate/urban form add smaller, context-dependent contributions.
Outputs
Deliverables: city dashboards & tract-level maps (NDVI, LST, luxury-effect slopes, tail diagnostics), city summary briefs (4–6 pp), CSV/GeoPackage with all metrics, and reproducible code (GEE + MATLAB/Python).
Status: Manuscript in review (Landscape & Urban Planning). Public summary/preprint upon request.
Impact / How it’s used
Targeted equity planning. Distinguishes where gaps narrowed by gains vs losses, informing whether to invest in bottom-up greening, prevent top-end canopy decline, or both.
Program evaluation. Provides a 33-year baseline to track whether tree-canopy and heat-mitigation programs are producing true leveling up rather than parity via decline.
Design guidance. Where built intensity predicts weakening, pair greening with form/reflectivity/shade strategies; where aridity limits vegetation, align with water management.