Bernie Grofman and I have submitted a paper on what I feel is a fundamental methodological issue in the study of electoral systems, namely how to infer district-level vote changes given national-level vote changes. This has direct relevance to election forecasting and study of partisan gerrymandering, for example. We show that uniform swing and proportional swing, by far the most commonly used methods, fail various natural axioms, and we give an alternative method that does satisfy the axioms. Interestingly, we show that on real data there is very little difference between the predictions made by these methods.