Estimating location premia using fixed effects—A semi‐local approach
提出一种半局部回归方法，通过重叠回归和对称性构建都市区内相对区位价值曲面，解决传统享乐回归中固定效应的遗漏变量偏误，并以凤凰城数据证明区位溢价是非单中心、不对称且高度动态的。
Abstract In this article, we develop a novel method for measuring the value of location within a metropolitan area. By estimating overlapping “semi‐local” regressions and exploiting the symmetry among them, we are able to construct a surface of the relative value of location within a metropolitan area. This methodology maintains the simplicity of using location fixed effects while resolving the omitted variable bias that arises from the traditional application in hedonic regressions. We demonstrate our semi‐local approach using a flexible, hex‐based geography. We also show that it can be generalized and applied to other geographies such as census tracts. Using data from Maricopa County (Phoenix), we find that every geography we employ supports the same conclusion: the premium for location is decidedly nonmonocentric, asymmetric, and highly dynamic. This dynamism conflicts with the typical assumptions associated with the use of fixed effects in the construction of aggregate price indexes, and potentially complicates the interpretation of difference‐in‐difference methods.