Evaluating the use of entropy-maximising procedures in the study of voting patterns: 1. Sampling and measurement error in the flow-of-the-vote matrix and the robustness of estimates

Entropy-maximising procedures were introduced to urban and regional studies as a means of providing maximum likelihood estimates of traffic flows and other interaction patterns. They have recently been adapted to the study of spatial variations in voting patterns. In this paper, the first of a series, the impact of certain aspects of sampling and measurement error on the pattern of estimates is evaluated. The results suggest that relative variation over space is insensitive to the error terms, but absolute variations are not.