Anyone that's taken any higher level statistics could tell you that with the right premise and choices you could make any bit of data support your claim. That is why you're supposed to gather your data and make a conclusion about it (which is a bit backwards from the scientific way of make your claim then test). Doing it in reverse order will allow you to model your results to support your claim. Even then, it's so easy to lie about the same thing ten different ways for it to say what you want it to say, even after the data is collected. Multiple scale axes, truncation, defining correlations where there really is none, extrapolating from a specific to a general, etc. Of course you can high-brow a lot of this so that your average Joe would not be able to understand it.
I almost never trust a "questionnaire poll" because if the poll wasn't developed without a bias, then who knows how the questions were phrased? The way a question is phrased could make all the difference in the answers.