Re: 3% margin of error
Although the pollsters claim a 3% margin, that is clearly shown to be false. You just have to look at the results of the various polls, where the spread from one poll to another was much greater than 3%. When Sky News was broadcasting before the election they had a rolling graphic on the screen that summarised the most recent polls. Looking at the numbers as they went by, one could see that few of the polls were within 3% of each other.
It would be easy to dismiss the spread by saying that the polls were taken on different days. However, if there really was so much variation: one day to the next, then the pollsters were measuring a quickly changing variable to too high a degree of accuracy. The poll might have been accurate to the stated degree, but its shelf-life at that accuracy was so short that it was probably obsolete before it was even published.