Spatial Econometrics

Quantitative geographers, as well as other social scientists, use the methodological techniques of spatial econometrics to analyze data in ways that help understand how humans are interrelated with one another across different geographic areas. More specifically, researchers use spatial econometric techniques to build regression models that attempt to mathematically represent the effects of human interdependence when explaining the factors leading to the geographic distribution of a variable. Spatial econometric techniques have been applied by researchers from a wide variety of fields, such as criminology, archaeology, regional economics, election analysis, international political conflict, and real estate analysis. Common to all these fields is the use of observational data that can be mapped (or geographically referenced) to infer interesting spatial (or geographic) associations.

Strongly associated with the fields of regional science and quantitative geography, spatial econometrics emerged in the 1970s to deal with the problems social scientists were having in estimating regression models using cross-sectional data in the presence of spatial autocorrelation. The techniques developed out of a concern that it was unrealistic to accept the spatial independence assumption of standard econometric inference, which posits that the values at any one location are unrelated to those at other nearby locations. Like time-series methods, which account for how a data-generating process may be characterized by temporal dependence across lagged time periods, spatial econometric methods were developed to account for spatial dependence across geographically lagged locations, those locations defined as being syn-chronically within close proximity of one another. A feature distinguishing spatial analysis from time-series analysis is that the causal associations are not unidirectional but are multidirectional.

From a methodological perspective, spatial econometrics comprises a set of techniques to incorporate spatial dependence and heterogeneity into a regression model. Luc Anselin has described the techniques of spatial econometrics as falling into four broad areas: (1) specification of how to represent different forms of spatial dependence and spatial heterogeneity in a regression model, (2) estimation of model parameters, (3) model specification tests and diagnostics of the potential presence of dependence within the data-generating process, and (4) model prediction. A key requirement of the estimation techniques of spatial econometrics is that the data be based on observations that have been geographically referenced at the time they were recorded, often as the centroids of a lattice of regions.