Nate Silver has an odd way of prioritizing his pollster, making it hard to see momentum, but sometime it just shows up non-the-less
There is something to be said for both RCP (whom only uses high quality pollsters) and 538 and Nate Silver (who uses everyone and their brother, but provides an influence number). With RCP, you are not bogged down with 1001 pollsters you have never heard of. But with the others, you at least are given access to "all" pollsters, regardless of whether or not you have ever heard of them. I guess this is why I have always welcomed the concepts of both.
The problem I have with 538 and Nate Silver is "how" they prioritize their pollsters. It doesn't appear to have much to do the historical accuracy as much as it has to do with the method being used. I would think that after some period of time you would accept that certain pollsters are more accurate using less favorable methods (short and sweet) than those who are supposedly using the gold-standard of polling (long drawn out phone interviews). Accuracy is measurable from an objective standpoint. Methodology is simply a subjective manner to determine what pollsters you like and don't like.
As far as RCP, I am of the feeling that just dropping off older polls prior to those pollsters providing new polls seems a bit lazy. But including the same pollsters several times because the poll more often, doesn't seem like a fair "average" either. This also causes problem when many of your more favorable polls to Democrats are traditionally polls that come out weekly or even more often than that. This is why I have held on to my own method of keeping the latest poll from as many pollsters as I can find that will put out some semblance of an ongoing poll. It may not be perfect as you do end up with polls that can get out of date, but I think it is better than keeping almost nothing or keeping duplicates.
Anyways, back to Pennsylvania
Right now, RCP shows Trump back in the lead in Pennsylvania, based on the averages of 12 different pollsters who have polled within a certain period of time. Meanwhile Silver shows Trump down over a point, based on the results of polls 24 polls, in many cases including the same pollster more than once. In the RCP average, only two out of the most recent 12 polls show Harris ahead. Meanwhile Nate Silver is including 15 different polls showing Harris ahead (including the same pollster multiple times with Harris lead). I am not sure why Morning Consult deserves three different cracks at influencing the polling number, while most others are only shown once.
Comments