Evaluating an entropy-maximizing solution to the ecological inference problem: Split-ticket voting in New Zealand, 1999

For geographers, one important aspect of the ecological inference problem relates to the analysis of spatial variations in individual behavior. Obtaining estimates of this behavior, in the absence of direct data, is often difficult, and the "solution" to the ecological inference problem propounded by G. King is not relevant in all circumstances. An alternative, using a different approach, has been used for some time in electoral studies but has lacked "real" data against which to assess the accuracy of its estimates. The availability of such data for New Zealand's 1999 general election allows such an assessment to be made - with very favourable results. The estimates are then used to test hypotheses regarding the spatial variation in split-ticket voting, again with considerable success.