I don't see how they can adjust for the type of people not responding to polls unless they can force them to answer.
If you're trying to do a scientific poll that is not about an election, it can become exponentially harder unless you can find independent ways to obtain data that helps you adjust (like home ownership, which is reported to the government).