Designing a Better Electoral System for Kenya?

Examples From Kenya

To design a better electoral system for Kenya, we use the SDSS to model alternative configurations of electoral districts according to certain specified criteria. These consist of:

The number of electoral districts or constituencies that will elect representatives to the new National Assembly.
The criteria to be maximized. In these illustrations we first delimit a new set of constituencies that maximizes for the criterion of geographic compactness (our surrogate for community of interest) while minimizing variations of populations across districts. We then delimit a second set of constituencies that maximizes for the criterion of equal numbers of people per district while minimizing deviations from compactness.
Constraints that exclude the construction of electoral districts that cross provincial boundaries and physical boundaries such as major lakes and mountains.

The SDSS constructs alternative sets of districts according to these criteria, by evaluating data from a large number of small geographical units for which there is appropriate population data and electoral data from a recent election. The SDSS thus configures and reconfigures alternative sets of constituencies by combining and recombining contiguous units. For this exercise, we use data obtained for 3,715 sub-locations, the smallest administrative unit in Kenya.

The next three maps show Kenya's current constituencies (black borders) composed of groups of contiguous sub-locations (yellow borders). A close-up is presented of the constituencies around Mt. Kenya to demonstrate the implementation of a constraint for geographical features and administrative boundaries-districts do not straddle the peaks of mountains or provincial borders. A close-up is also presented of the constituencies in and around Nairobi.

The population data analyzed is from the 1988 census, because results from the 1999 census were not yet available. The electoral data analyzed is for the 1997 election. Details of the data preparation required for the SDSS are discussed below.