Modelling spatial autocorrelation in spatial interaction data

Spatial interaction models of the gravity type are widely used to model origindestination
flows. They draw attention to three types of variables to explain variation in spatial
interactions across geographic space: variables that characterise an origin region of a flow,
variables that characterise a destination region of a flow, and finally variables that measure the
separation between origin and destination regions. This paper outlines and compares two
approaches, the spatial econometric and the eigenfunction-based spatial filtering approach, to
deal with the issue of spatial autocorrelation among flow residuals. An example using patent
citation data that capture knowledge flows across 112 European regions serves to illustrate the
application and the comparison of the two approaches.