Ben Goldacre (of Bad Science and Bad Pharma fame) was on BBC's Newsnight last night (Wednesday 4th) advocating severe punishments for politicians who misuse or abuse statistics. I am not sure I would go that far, but I think the point about misuse of statistics is a good one.
One possible response to statistical misbehaviour is people being more statistically literate, so they can spot the misbehaviour. There is a discussion of the problem in Wikipedia.
How much knowledge would we need to be statistically literate? If we are a healthcare practitioner trying to apply the results of a piece of research to our own practice, then I suspect quite a lot, and the sort of thing that is covered in health education courses. But someone looking at government statistics about, say, unemployment, or waiting times might just need to be able to identify things like graphs whose vertical axis doesn't start at zero, using different time periods or starting points, and the different sorts of "average".
So, what would help? There are books, like How to lie with statistics, and The tiger that isn't. There are revision guides on the web, for GCSE, for example, but they rather assume you know what statistical technique you need. There is the BBC programme "More or Less", which has pages that discuss various things in the news, as well as clips and episodes. The Open University, a partner in the programme, have a site with resources.
Making sense of statistics, from Sense about Science and Straight Statistics , looks rather useful. It covers percentages and certainty, among other things.
The work of Straight Statistics is now carried on by Full Facts, a fact checking organisation, whose site looks at statistics in all sorts of areas including health.