Statisical Bias

Northern Ireland has proposed a total smoking ban that includes restaurants and pubs, citing overwhelming public support. Read this:

Of the 70,000 responses, 91 percent said they favored a ban on smoking in all enclosed public places and workplaces.

An absolutely inviolable rule in public opinion statistics: if your approval/disapproval ratio ever exceeds 90%, your numbers are useless. You have biased the poll--deliberately or otherwise--beyond any semblance of reality. On an issue as controversial and politically charged as this one, I would expect a maximum 65 percentile margin.

There's an old saw: "Tell me what result you want, and I'll write a poll to produce the numbers." It's still very true; this article proves it.