Letters to the Editor
-
@Ondelette (on the poll discussion from yesterday)
No, it wouldn't average out, but its effects would be minimized by it being short w/resp to the window width.
Yes that minimization must occur if the rapid changes are impulsive in nature, or similar to a delta function. That is, some event causes support for a candidate to change quickly , and then that event tends to lose control, and support returns to its previous level.
But is that the right model? A very different model would use a step function. This model says that some event changes the support for a candidate, and that change tends to "stick". But then other events happen, and they can make changes of different magnitude and direction. And then the effects of these multiple events tend to cancel out over the length of the poll window, or perhaps over a much longer time.
The advantage of the step function model is that it can explain long term trends. The step functions of one sign tend to be bigger or more frequent on average than those with the other sign, and so support tends to move in one direction on average.
The disadvantage of the impulsive or delta function model is that it does not explain trends. You need to add something else to the model, a different kind of event.