Currently Online 
| Cerebralbore101 said: but Pennsylvania polls are always wrong by 2% to 3% in Trump's favor. |
Systematic polling errors have no correlation between election cycles because the methodology changes to accommodate for the systematic sampling errors of the previous cycle. The fact that R-favorable polls like Rasmussen and Trafalgar are much closer to the standard polls this cycle, hints that the sampling bias resulting from undersampling Trump supporters has likely been resolved this cycle.
What is more likely going to be the systematic error is that pollsters seem to be throwing out what they deem as "outlier" polls in favor of ones that show a close race.
The issue with doing that is that no single poll is very accurate in itself, but the aggregation of polls tend to be because biased sub-samples tend to balance each-other out. We're likely going to see a situation where the conservativeness (not in a political sense, in a statistical sense) of the pollsters this cycle (because no individual pollster wants their ratings reduced) will make any single pollster more accurate, but the aggregation of polls less accurate because they're likely underestimating the winning candidate, systematically. This could be Trump or Harris. Past cycles tell us nothing about where this systematic error resides, because of the bolded point at the start of this post.
Last edited by sc94597 - 5 days ago
