Spatial influence on voter transitions in Swedish elections: An application of Johnston's maximum entropy method

Abstract

A major problem facing students of electoral behaviour is the scarcity of data on individual behaviour. Election results are reported only for aggregates; access to individual voting returns is not possible under secret ballot arrangements. Survey data, because of relatively small sample sizes, do not allow one to study local influences. To obtain detailed survey estimates of local flows of voters between parties would require resources far beyond the means of any academic or survey organization. The contents of analysis is therefore closely prescribed by available data, and there are many areas of relative ignorance about the geography of voting, most of them stemming from the absence of relevant data (Johnston, 1985). Having to rely almost entirely on aggregate data means that most areal election studies involve what are known as ecological methods. In this context, the ecological fallacy of reading into the data inferences about individual behaviour that may be unwarranted is an obvious danger.