Why the shifting make up of the sample space then? That number is well known at federal, state and local levels.
Two answers. (1) That number isn't as well known as you might think, if you're referring to the dem/rep/ind breakdown of the population; the number changes based on economic conditions and other news and cannot be assumed static. (2) The samples are random, and fluctuations in those numbers should be expected with random samples - part of the reason for the +/- 3% uncertainty bounds normally reported for a poll of 1000 people. Some pollsters adjust for the difference between their sample demographics and the population, but that runs the risk of missing the fact that party identification is actually changing. Others intentionally slant their samples, but the sleaze in some polling is not the only reason for these results.