The reason why surveying a different number of people or different population of people can possibly change the percentages that you calculated is because these surveys are simply "samples" of the general population as a whole, and are not exact in terms of their results given any random sample.
