The study of short-term electoral change at a fine spatial scale is frequently hampered by the unavailability of suitable ecological data, let alone data on individuals. In Britain, most studies of change have used the measure of swing. An alternative is presented here, using national survey data to predict short-term electoral change in each English constituency. The deviations between the observed and predicted values are the dependent variables in an analysis of the ecological determinants of short-term change. Three independent variables are suggested: the neighbourhood effect; campaign spending; and constituency type (reflecting the spatial division of labour). All three contribute substantially in regression analyses, although some of the coefficients relating to campaign spending are difficult to interpret.