If your point is that American political polls, on the whole, are inaccurate as well, you might have a point. However you are certainly overcounting the U.S. Population, AND assuming every person counted in the Population either chooses to vote, or is able to vote (whether blocked by age or legal disenfranchisement). Barely above 54% of the American Voting populaces vote in reality. Also, there are NUMEROUS political polls (AP, Pew, Zogby, Rasmussen, Gallup, as well as as each major news station having its own on occasion), each with their own methodology. And these only measure opinion, an intangible quality.
In contrast, there is only one Game Sales group in the US, and it doesn't even count the largest contributor to said game market. Its inaccuracy is distressing, especially since they don't say "well, this is only half the market that we have surveyed, take it or leave it." They simply crudely estimated based on the data given and then say "this is how everything sold last month."