what is the problem with it?
Landlines: what are the demographics? Not good and not representative, as even the pollsters admit.
Cell phones: pollsters aren’t supposed to be calling cell phones. In fact, I think the pollsters must obtain consent before calling cell phones for a poll just as telemarketers are banned from cold-calling (even if the telemarketers often ignore the law).
In both cases there is a strong likelihood that robocalling was used. What self-respecting sentient being sits through any robocall? What self-respecting sentient being sits through a robocall poll, especially one that may ask his political affiliation/registration and even more intrusive questions such as occupation, income, zip code, race, religion, etc.?
Yes, the FCC have posted rules for pollsters who robocall but I doubt compliance is voluntary or complete. As a dodge, the robocall/autodialer is still used but calls are handed off to an operator if someone is bored or foolish enough to consent to the poll.
And, of course, there is the veracity of the responses. How do we know the Carson effect isn’t the Wilder effect? How do we know the respondents know a damned thing about either candidate or their positions on the issues?
This is a nation of 350 million persons. Polling 575 individuals especially via antiquated means and declaring it a valid sample is too ludicrous for words, scientific statistical horsestuff reassurances notiwthstanding.
The poll also has a margin of error of 6 points i.e. 23% of Carson’s total and 27% of Trump’s. Ridiculously broad especially when the choice is not a binary one. It also, by definition, means that Trump could well be in front of Carson. Carson’s lead could be even larger as well but the point is that margin cannot be dismissed.