Spatial & Geo

Theoretical foundation for all hedonic pricing models; every AVM derives from this framework.

Original Case-Shiller methodology establishing weighted repeat-sales indices used in S&P/Case-Shiller and FHFA.

Models serial correlation in house prices (momentum at 1-year, mean reversion at 5+ years); explains forecasting difficulty.

25% accuracy improvement by integrating image and text data with interpretable SHAP-based ML for AVMs.

Definitive analysis of Opendoor and Zillow Offers; quantifies adverse selection vs. liquidity tradeoffs in iBuying.

IV strategy on nationwide Airbnb data; 1% listing increase → 0.018% rent increase via supply reallocation.

Rigorous event-study finding Airbnb increased Barcelona rents 1.9% and prices 4.6%.

Comprehensive survey on household price expectation formation; documents extrapolation and social network effects.

Foundational non-parametric density estimation underpinning demand heatmaps.

Local regression allowing coefficients to vary spatially—key for heterogeneous demand.

Neural network predicting mobility flows, outperforming classic gravity models.

Uber's real-time geospatial demand prediction powering surge and dispatch.

Production GNN that reduced negative ETA outcomes by 40%+ in cities like Sydney; deployed at scale.

Foundational spatiotemporal GNN with 5,000+ citations; models traffic as diffusion on road graphs.

Definitive Hawkes process review for modeling clustered demand in surge pricing and fraud detection.

Seminal optimization formulation for facility siting under coverage constraints.

Entropy-maximization derivation of gravity and retail models.

ML-driven site selection combining foot traffic, demographics, and competition.

Authoritative review covering Hotelling spatial competition, gravity-based capture models, and location games.

Pioneering ML site selection using Foursquare data with gradient boosted trees; foundational for location analytics.

Canonical review of stochastic and robust facility location for warehouse network design.

Definitive textbook: spatial lag/error models, Moran's I, and inference.

Robust GMM estimation for spatial autoregressions—standard in applied work.

Open-source library implementing spatial weights, autocorrelation, regression.

Demonstrates spatial lag effects in short-term rental markets using hedonic models.

Foundational paper on spatial HAC standard errors ('Conley SEs'); industry-standard correction.

Causal forests for heterogeneous treatment effects with valid confidence intervals; implemented in grf and EconML.

Foundational geographic RDD methodology addressing boundary selection and spatial inference.

NLP library parsing addresses in 60+ languages—widely used in ride-sharing and logistics.

Hexagonal grid system enabling efficient geospatial joins and aggregations.

Standardized POI identifier joining disparate datasets without address matching.

Production geocoding system at Bing scale; covers parsing, candidate generation, and ranking pipeline.

HMM-based address parsing and record linkage; foundational for modern address normalization.

Large-scale mobile-phone analysis revealing power-law travel distributions.

Unifying framework explaining visitation frequency to locations across cities.

Attention-based RNN capturing periodicity and transition patterns in check-ins.

Aggregated mobility trends by place category—used globally for pandemic response.

Panel-based POI visit data enabling offline-to-online attribution and demand studies.

Uses smartphone mobility to show voluntary behavioral changes dwarfed lockdown effects on foot traffic.

Iterative Proportional Fitting (IPF) to generate representative synthetic populations for microsimulation.

CEMDAP: full-day activity-travel simulator; basis for large-scale urban demand models.