It's a growing issue. I saw a tweet from a pollster trying to poll MN8 last fall, and they were literally having to make 200 calls to get one response. They openly said that they had little confidence in the data. They are trying to find ways around it. All that said, the 2018 polls were fairly accurate. In MN8, that firm was the only one to poll. They had one D+1 and one R+14 4 weeks later. Ironically the average of those ended up pretty close to the result.