Reliability vs. validity
Conducting a poll and reporting its results in accordance with the requirements of random sampling should make that poll reliable, but will not necessarily make it valid. A poll is reliable if it can be replicated again and again, producing essentially the same result. A valid poll is an accurate description of what actually exists in the universe being examined. Any given survey could be reliable, or valid, or both.
A highly biased question may consistently produce the same response, but that does not make it an accurate description of the views of the respondents. On the other hand, the results reported by a poll may be accurate even though the methodology is faulty. The 1928 and 1932 Literary Digest polls are examples.
As experience gradually improves polling techniques there is less likelihood that a poll, like the Literary Digest, will be right for the wrong reasons. The improvement in methodology also means that polls will be more reliable. Still, the problem of validity remains. Errors, where they exist, are more likely to be caused by the way questions are asked or interpreted than by the way a sample is drawn.