I’ve always found it intuitively a little odd that after a positive event for one side in a political race the poll numbers will bounce up and then, usually, come back down pretty close to where they were before.
Since most polls are cold-calls, and most people who are cold-called for polls hang up the phone, could that positive event temporarily pump up the enthusiasm of one side (and perhaps temporarily reduce the enthusiasm for the other side) so that they are more likely to stay on the phone and talk to the pollster?
So my hypothesis is that, at least for most events leading to a poll bounce, very few voters actually changed their minds- they just temporarily became more or less likely to speak to a pollster depending on who they support.
If this is the case, could it be corrected for in some way?