Well, it depends on where the 1000 people are drawn from as well. Your assumptions only hold if for both the original survey and the repeat, the pool of people being surveyed are an accurate sample of all of the United States.
Good point. On the other hand, the confidence interval I provided is correct as long as the original survey and the repeat are sampled in the same way (regardless whether it is un-biased sampling or not)
