I have long argued that voters' preferences are systematically warped by cognitive biases or ideology. It turns out that things are even worse than I thought. A new survey by the Royal Statistical Society shows that people vastly over-estimate spending on unemployment benefits, welfare fraud and the number of migrants, among other things. This corroborates the TUC's finding that hostility to welfare spending is based upon ignorance.

This, though, brings into question the nature of our democracy. If voters are massively wrong about basic facts, what hope can there be for good policy-making?

Now, we shouldn't be too gloomy here. Alex points out that, for all their faults, British voters don't much support demagogues or the worst extremists. And Sue notes that, in at least one respect, public attitudes to welfare spending are quite humane.

What's more, we should remember that the virtue of democracy is not that it leads to better government. Instead, it is an intrinsic value; democracy recognizes the inherent equality of all people, by acknowledging that none is entitled to govern without others' consent. Bad government is the price we must pay for this principle. If moral values didn't have a cost, they wouldn't be values at all, but mere expedients.

Nevertheless, we should ask: is it possible to improve out democracy? The problem here is that bad ideas based on big errors do not get filtered out of the public debate. In fact, the opposite happens. Politics is a marketing exercise in which the media and politicians pander to public prejudice, with the result that bad ideas are actually selected for. Could things change?

In theory, yes. More associative or deliberative democratic institutions would give us a hope of filtering out bad ideas. Another possibility - which is consistent with Sam's call for more markets and less politics - is to use demand-revealing referenda. If people have to pay their own money to enforce a policy preference (say, immigration controls) they have an incentive to gather information. Making people pay is a way of making them think.

There is, of course, zero chance of such an institutional change.

There is, though, a paradox here. To see it, just ask: in which areas of public life are people most likely to be well-informed? The answer is not political affairs, where rational ignorance rules. Instead, people are more likely to have useful knowledge about their workplaces: they know, from the ground up, where there is inefficiency and waste. If we wanted to harness individuals' dispersed knowledge and expertise, wouldn't we therefore think that workplace democracy is (at worst) a reasonable idea, whilst political democracy is rather odd?

John Stuart Mill famously said that "liberty is often granted where it should be withheld, as well as withheld where it should be granted." Viewed from the perspective of efficiency - which I concede is just a partial view - you could say the same about democracy.