Semiparametric Estimation of Land Price Gradients Using Large Data Sets

Abstract

The exact nature of land price gradients, the surface describing how land prices change with location, can be difficult to uncover. This is particularly true for cities with few vacant lots or in more rural regions where the number of land sales in a given area is limited. This article outlines a semiparametric method to construct the land price surface given a large set of residential property sales, and investigates properties of this surface in Richmond, Virginia, and three surrounding counties. Despite recent concentrations of housing in suburban areas, we find that Richmond remains largely a monocentric city. Nevertheless, the price surface that we estimate features a complex topography, and high prices near suburban interstates and lakes are clearly evident.