Regarding Quinnipac:
Here is a long answer. So if you look at the spread of each pollster's results from survey to survey, in the same geography, Quinnipiac tends to be very high. Eg take the 9 point swing in 3 weeks compared to their last MI poll. It's fine if you get percentages moving by 4.5 points (margins by 9) between surveys on occasion, but it was 8 points before that in MI and 7 points poll-to-poll in their most recent national survey. This is high compared to other pollsters. Generally this indicates that something about the methodology is producing more noise than you'd expect based only on the sample size of the poll. And the answer is that they are using random digit dialing and doing a lot of weighting, so the MOE is about 30% higher than it would normally be per their own disclosure, and the low response rates and representativeness are also probably causing some sort of non-response that isn't captured by these statistics. None of this is to say that Q polls are inaccurate, but they are seemingly less reliable than other methods that do more work to equalize response rates on the sampling end or use other design approaches to decrease effects of weighting and other non-sampling biases.