That depends on the purpose, and (as you mentioned) who they polled. In fact, who they polled (and how they worded it) is far more important than the sample size, as long as the sample size is 'big enough' for the purpose.CrazyBlaze said:The big issue here is the claim that a 2000 sample, well still large does not represent the entire U.S. of A.

.

In fact, I've noticed that some people mistakenly believe that a proper sample size has to be a certain percentage of the population being sampled. That is NOT true. The increase in reliability in statistics is subject to harsh diminishing returns once you go beyond a certain sample size. In fact, the amount of possible answers is more important to determining how important a sample size needs to be (a simple yes/no question needs a lower sample size than a question with 8 different answers, unless those answers are point-based and you're just trying to compute the average amount of points given). If you sample 5000 people out of a population of 30 million, you don't need to sample 50000 people if the population is 300 million to get as reliable results.