1. This is not a "fact check" via @APFactCheck, it's a hack job that omitted some pretty significant details. @Rasmussen_Poll may have different methods but they also performed better on Election Day. They were more accurate than the "Gold Standards." apnews.com/59ef870461a841…
2. Secondly, we have to start a conversation about that term "Gold Standard" used to describe pollsters who use live-interviewers and call cellphones. You cannot legitimately claim that mantle if you continue to SUCK election after election. Methods matter and results matter.
3. Third, it may come as a surprise to many, but calling cellphones with live interviewers aren't a proven, reliable method for survey data collection. Some of us are making real attempts to develop accurate and reliable methods. And some in the media like AP are hindering us.
4. This thread isn't even so much a defense of Rasmussen itself as it is a defense of what they did post-Scott. 20% of their sample is online to reach younger voters. Well, guess what? Calling cells w/ live interviewers doesn't reach older, con rural voters. Do media polls care?
5. No, they don't. No one has asked how media polls missed those Trump voters in MI, PA & WI. They were older, white voters who FOR YEARS voted Dem on the presidential level. Trump flipped them, but media polls didn't know that b/c they weren't reaching them, particularly in MI.
6. From journalists, yes. But from pollsters, we haven't received a single call asking, "How the hell did you do that?" Anonymity was a big part of it. I'll give you that clue. But our panels consistently outperformed cell responses in battlegrounds & generally more rural areas.
7. Even if media polls fix their issues w/ weighting for education, income, region, etc., which they botched, it still won't matter if their sample isn't representative of the population b/c they're missing a big chunk of it. That's also how they all blew NC, OH and FL.
8. The polling industry is going to have to evolve. The term "Gold Standard" doesn't mean anything anymore. Come Election Day, you'll still look like a fool if you poll a population reflective of what you want it to look like, rather than what it really does look like.
9. To sum up what we've found in our own post-election analysis of what we did right & what we did wrong, we found cellphones were not at all accurate in rural America, and were less accurate (as was IVR) than the panel in burbs.
10. Dems were much more likely to identify as Trump voters if they were surveyed via OSP (online survey panel), and far less likely if interviewed live, which we tried if they failed to respond to the panel after 7 attempts. Even IVR performed better. Why? We concluded anonymity.
11. We can't expect larges percentages of the electorate to be truthful if there's 24/7 media claims that their views are racist, backward, xenophobic, bigoted, etc. Many will not be comfortable answering a live-caller truthfully. They'll say "undecided" or just lie. End thread.
12. Thread reopened to briefly say... Good for @greggutfeld. Cannot dismiss a polling firm that outperformed nationally. They were right, almost all others were wrong. No one else made significant changes, so why would we expect their accuracy to improve? mediaite.com/tv/fox-news-gr…
