This paper explores methods that might be used to construct new geographies for data reporting that are in some sense optimised. The objective of the paper is to evaluate an automatic method for generating robust statistical output zones that fulfil pre-specified optimal characteristics such as compactness of shape, minimum population size, standard mean population size, and constrained nesting within larger areas.
We apply a spatially aware method from the computational geography literature to design new reporting geographies, using 2006 Census data as the main input. We compare the new geographies with existing, official geographies, addressing two criteria: firstly, the stability and performance of the optimisation algorithm, the impact of different control parameters, and the robustness of the solutions generated; and secondly, the statistical quality of automatically generated geographies in terms of population distributions and minima, shape compactness, and internal homogeneity.
We find that our new geographies substantially out-perform the current geographies across almost all of our optimisation criteria. The algorithm that we use is stable, and is able to repeatedly generate high-quality solutions in a timely manner. We conclude that the ArcGIS / AZTool toolkit developed here would form the basis of a viable workflow for the automatic production of optimal geographical areas.