The Swing Ratio as a Measure of Partisan Gerrymandering
Richard G. Niemi
In a previous paper ( Niemi, 1985) I argued that if the Supreme Court is to consider squarely the question of political gerrymandering, "sooner or later it will have to take a position on the significance of the relationship between votes and seats won by each political party" (p. 191). If in doing so the Court holds to its frequently stated argument against proportional representation (reiterated in Davis v. Bandemer, 106 S.Ct 2797 ( 1986), p. 2809), then facile comparisons--that a party won 58% of the vote but only 52% of the seats--are inadequate. So the question arises, how can one look at the seats-votes relationship without reducing it to a question of proportional representation?
Political scientists have developed such a way--now generally referred to as the swing ratio. Though the idea originated as the "cube law" some 40 years ago, the swing ratio is a relatively new concept as it is applied to political districting, and there are a number of problems about its use in this context. But because the seats-votes relationship is at the heart of the gerrymandering issue, I continue to pursue its possible applicability. In this paper I describe the swing ratio, drawing heavily on a previous description ( Niemi, 1985), but showing for the first time the swing ratio for the New Jersey legislature at the time of Karcher v. Daggett. I then discuss the applicability of the concept to tests of political gerrymandering.
The swing ratio is the rate at which seats change as votes change. More formally, it is the change in the proportion of seats won by a