Boundary-Swapping Optimisation and Region Design

The project described in this chapter, which was conducted in the mid 1990s, had some unusual characteristics. From a regional science perspective, it lay outside the core area of spatial economics. Analytically, it had as much to do with algorithms as models. Procedurally, it was more concerned with education than model implementation. And for the client, its value was as much institutional as technical. The project was conducted for a national Electoral Agency. It was part of a redistricting exercise on the boundaries for national elections. The country, which for reasons of client confidentiality must remain nameless, was regarded as having suffered in the past from widespread gerrymandering and the Electoral Agency was responsible for preparing a new, defensible, electoral map. The larger part of this exercise, in terms of labour and computing time, involved the production of a spatial database containing digitised boundaries and statistical information for census tracts. The particular element with which we1 were concerned was the assembly of census tracts into constituencies. We were not asked to set up and run a procedure for generating constituencies but to advise on the scope for, and difficulties inherent in, the use of so-called active redistricting methods. In particular, we focussed on the use of boundary-swapping optimisation methods in a geographical information system (GIS) environment (for more details on GIS see chapters 15–17).