The impact of spatial aggregation on urban development analyses

Abstract

This paper illustrates the impacts of spatial data aggregation on the analysis of urban development. Spatial econometric methods are used to control for spatial autocorrelation in the data and existing weighting methods are used to overcome aggregation dependencies that are due to differences in sizes of areal units. The analyses show that shape dependencies can be partially removed by the used weighting methods, and that even regularly latticed areal units need such weighting in practice. Aggregating to coarser resolutions does not affect the order of magnitude of coefficients estimated for variables that are aggregated by averaging, if the aggregation process maintains sufficient variance within variables. We argue that small-sized areal units approximating the true characteristics of the studied process are to be preferred in urban development analyses, because such micro-data allows the exploration of highly local factors alongside higher scale linkages. We demonstrate that spatial autocorrelation and scale dependencies interact and that spatial econometric methods can help explain variance in analyses of small-grained land-use data.