Can we use the ‘swingometer’ to predict US Elections?

Some surprising results

The conventional wisdom is that using national voting figures in US Elections, especially in terms of voting swings, is a futile device. This is because individual US states are so different in terms of ethnicity, demographics and individual issues that you cannot simply assume that all 50 states will move in the same way. Instead, the conventional wisdom is that you must treat each state as an individual election. However, while this is practical for someone with the time and energy to study US statewider polling trends in depth, it is impactical both for the average punter and for someone with a PhD to finish and a job to find. It is also a dirty little secret that the British method of applying uniform national swing (simply assuming that a national swing will be repeated at every level) works pretty well, if there isn’t a factor which massively distorts the picture (such a third party with regional strength).

Indeed, although the average overall state swing to Bush in 2004 was lower than the national swing, applying either of those two swings would have called 46 out of 50 states correctly and predicted that Bush would remain in the White House. Indeed, the use of a swingometer would have slightly overestimated Bush’s lead as Wisconsin, Oregon and Minnesota would all have become ‘red’ states, while New Hampshire would have stayed in the Republican column. This crude device performs as well as the worst of the projections made by this sites competitors. For instance, Scott Thomas of ‘The Blogging Caesar’ got only 44 states correct (his formula only miscalled 2 state – showing the limitations of ‘gut feel’) while the main algorithm of Andrew “Electoral-vote.com” Tanenbaum got 46 right (although other algorithms actually worked better). Of course they didn’t have the actual national vote tallies but it does show that the Swingometer isn’t as crude as many people make it out to be.