They probably only called phone numbers in reliably democrat area codes and still could only show Illary with a 6-point lead. I don’t put any faith in polls on this side of the debates. A lot will change between now and election day that will render them irrelevant.
And since it’s an NBC poll, you know the only point of doing the poll is to produce their desired result. If they run 6 polls and 5 show Trump with a significant lead they’ll just toss those out and run a story on the outlier result. Don’t fret, Trump will win and it’ll be somewhere between a comfortable win and a landslide come November.
I am thinking about how they reach their respondents. Polls are conducted on land lines in an age when more people are exclusively on cell phones. I believe cell phone users are politically different than those who use land lines. Also, those more likely to answer on land lines are those who are home and do not screen their calls from caller I.D. If they had caller I.D. they would less likely to answer the call. This makes the possibility of live contacts much reduced. As one who has conducted numerous phone calls for my candidate, live contacts are a rarity. Those who do answer the call will rarely participate in a telephone poll. So, out of how many attempts does a pollster get a respondent for their questions? And, when a respondent is reached, what is their political bias and do they represent the political population at large? I strongly think not.