I really think it has to do with mechanics. Sampling procedures are just not accurate. In fairness to Rasmussen, that seems to be a problem for everyone. On the other hand, I do think Rasmussen is intentionally skewing its sampling process to promote the narrative that the GOP/Trump is doing better than it is. Sort of use the poll results to pull the train.
One of the great untruths of the 2016 campaign is that "the polls were wrong." People's
interpretations of some polls and statistical data were wrong, but the polls themselves gave a fairly good indication of the odds of each candidate of winning.
On the eve of the election, Hillary Clinton had a 2-4% lead in the national poll. She won the popular vote by 2.1%. That poll was basically dead on, but also meaningless. It isn't how the election is decided.
On the eve of the election, Donald trailed within the MOE or had a lead in virtually all the states that would decide the election (Florida, North Carolina, Virginia, Ohio, Pennsylvania, Michigan, Wisconsin). Now, it would have been a stretch to
expect him to win all, or almost all, of those states. The polling was very close. That's why he was rightly considered an underdog, but he should have never been as big of an underdog as he was. The way I saw it, there was no statistical reason for the NYT to have Hillary at or above a 90% chance of winning all summer long and right up until election day. I never one thing in the polling that indicated that. If you looked at an average of polling in the swing states (RCP has a great one) and tracked it over time, about the highest chance of winning you could give Hillary was 75%. Personally, on the eve of the election, I figured she had about a 60-65% chance of winning. Based on the fact that Donald had to thread the needle and win all/almost all of the swing states, while Hillary only needed a few. That meant Donald had a 30-35% chance of winning, so while he beat the odds, it shouldn't have come as this incredible shock. He was not a 5 TD underdog that won the game - he was more like a 10 or 14 point underdog. Still an upset, but not as incredible as it seemed.
For Hillary to have had a 90%+ chance of winning (sometimes, the NYT Upshot poll was giving her a 98-99% chance of victory), she would have to have had consistent, large, outside-the-MOE leads in all the key states. She never had those. This "the polling was so wrong" narrative I think is an attempt by some in the media to make sense of the fact that a person of Donald Trump's demeanor and temperament won the presidency. A lot of inside the Beltway types still haven't come to grips with it, so they conclude things like the polling must have been wrong.
Overall, the key is to look at an average of polling over time, not just one in particular, and in a Presidential election you have to look at an average of polls in states that are going to decide the election. People got way too hung up and read way too much into the results of the national poll, or the results of one particular poll in one particular state. For example, Hillary did much better than previous Republicans did in Georgia, and pre-election polling was indicative of that. However, I saw a bunch of thinkpieces saying that this indicated support for Donald was low in a traditional red state, and this portended doom for him in the general election, because if he's a little weaker in a place like Georgia he'll get killed in Pennsylvania or Michigan. It was a dead wrong conclusion drawn from a poll that was actually accurate.
I think the same thing happened with Brexit polling in the UK too. On the eve of that vote, the polling was within the MOE. When "leave" won, the media seemed absolutely floored, which I never understood because polling was indicating the entire time that it could go either way.
