Is nobody suspicious in the least that the polls all differ so radically?
It used to be that polls used to all come up with similar numbers, but now polls represent the bias of those who commission them enough to make them all almost worthless.
And if can't tell what the numbers are supposed to be coming in, how can you trust the numbers that come out?
I don't believe that's true. I think that at least all the major pollsters make a real effort to be unbiased. There are differences in their methodologies that explain some of the gaps. Some of them use primarily telephone polling that excludes mobile phone users, and those polls contain a natural bias in favour of the Republicans because they miss younger voters. But that is not an intention to be biased in favour of the Republicans, that is just a feature of the methods used by that pollster.
I also think people expect more of polls than is realistic. They AREN'T that accurate.
Polls typically come with a warning that they are within 5 percent 19 times out of 20. A lot of pollsters have tried to improve on that with larger sample sizes and better methods, and now claim to be within 3 or 4 percent. But still, two polls that are within 3 or 4 percent, using identical methods, can differ from each other by as much as 6 or 8 percent. Compare polls that used different methods, and on different days, and even if there were a gap of 10% between two polls that wouldn't necessarily indicate that anybody is biased.
If you are trying to collect public attitudes on the latest Star Wars movie, or figure out whether the latest AXE commercial is working out--which is what most pollsters do most of the time, that's how they pay the bills--then that sort of margin of error is probably fine. Learning that 70% liked your new brand of toothpaste is not really all that different from hearing that 76% liked it.
But when you are trying to figure out who is going to win a very narrowly split election, where a difference of 1 or 2 percent really matters, then a poll with a margin of error of 4% has a problem.