The analysis of long-term social and political developments in Western countries is often difficult because of a lack of sufficient survey data. Almost always official election and census statistics are available over long periods, yet the use of these data for individual-level inferences runs the risk of the ecological fallacy. In this paper we propose a method to go beyond the fallacy, the Duncan-Davis technique for area-classified data. The method is discussed and used to assess the amount of religious voting among Dutch Catholics in the 1971 general election. While the technique is only moderately helpful in this case, it is expected to be far more useful for the analysis of older elections.