The most important lesson of the 2012 presidential campaign, in my view, was not that polling-based models are foolproof ways to assess the political environment, but instead that undisciplined ways of evaluating polls and political events can lead to flawed conclusions. On several occasions during the race, news media commentators either overrated the amount of information contained in outlier polls and jumped the gun on declaring a change in momentum — or insisted that a candidate had the “momentum” in the race when there was little evidence of it.
The past year-and-a-half hasn’t made me optimistic that things are getting better. …
The general flaw is in overestimating the importance of recent events and assuming that short-term trends will continue indefinitely: that a candidate rising in the polls will continue to do so, for example. In fact, especially in general elections, candidates gaining in the polls see their position revert to the mean as often as they continue to gain ground. CONT.
Nate Silver, FiveThirtyEight