Proof-of-Concept Land Value Models: A Spatial Econometric Approach

2023–2024 · Robert Schalkenbach Foundation · Co-PI (PI: Chester Harvey) · Completed

Land value is rarely observed directly, since vacant land seldom sells, so most land values are modeled as a residual: a property’s total value minus the value of its structure. As a regression, that residual requires a perfectly specified model, because any misspecification is indistinguishable from land or housing value. Land prices are also spatially dependent, directly through spillovers from neighboring parcels and latently through price speculation.

The project’s approach, developed by Knaap, fits a spatial Durbin hedonic model to property sales and specifies it as fully as possible, then sums the land components (the intercept and the land terms, with their direct and indirect spatial effects) and leaves the value of the structure aside. The proof of concept built the hedonic specification on property sales in Baltimore City, where every test for spatial dependence rejected the non-spatial model. The method is set out in the chapter The Hedonic Housing Price Model of Knaap’s Urban Analysis & Spatial Science, and its prediction step is implemented in neighbayes.