Replies

Even an honest pollster can be wrong. Say, for example, there are a hundred million voters. For comparison, put 50 million red marbles and 50 million blue marbles in a drum, and mix them up thoroughly. Now draw out a sample of marbles from the drum. The sample, very likely will not have exactly half red and half blue marbles. The "Margin of Error," MOE, estimates how far off you might be. If you take this sample 20 times, your measurement will likely be within the MOE 19 times, and outside the MOE once. That's right, one time out of 20, your sample will lie outside the MOE.

Now, let's put some reality into your stirring of the drum. Some regions of the drum will have more blue than it has red marbles or vise versa. Cities, states, zip codes, land lines, area codes, time of day, internet providers, caller ID, etc., and more, can all alter the mix within the drum. It is very easy to either intentionally or inadvertently focus your sample on one or the other of these "hot spots". You can even do this without the people actually taking the samples knowing. You can do it without even knowing it yourself. Some pollsters might do this intentionally.