What is the case against evidence-based policy? This is one question prompted by David Cameron's refusal to legalize drugs in the face of evidence that criminalization of them doesn't reduce their use.

The question, of course, generalizes. Many of us would argue that fiscal policy and immigration policy are also less than perfectly based in the evidence - and no doubt you can think of other examples. There must, therefore, be something to be said against evidence-based policy. But what?

One answer lies in something we know from financial decision-making - that greater knowledge doesn't necessarily improve the quality of decision-making. It might instead merely increase people's overconfidence and so magnify their mistakes.

This matters because in a complex world there is inevitably a lot that cannot be known about effects of policy, particularly in the long-run and about the possible effects of policy upon social norms. For example, whilst there is good evidence that immigration doesn't reduce wages on average, and might even increase them, there is - as Diane says - rather less hard evidence about their long run effects. Might it be that, in the long-run, a more ethnically heterogenous society would lead to greater tolerance of inequality? Or take tax policy. There might (pdf) be some evidence that higher taxes on top earners would, in the short run, raise revenue. But what about the possibility, stressed by Assar Lindbeck, that redistribution might in the long-run erode social norms in favour of work?

There's a lot that cannot be known about policy effects. Stressing the need for evidence might therefore cause us to overweight partial knowledge and so lead us astray.

I'd add three other arguments:

- Evidence-based policy is, necessarily, conservative simply because there's no available evidence one way or the other about the effects of truly radical policies. For example, would a high citizens' income or market socialism work? As they've not been tried, we can't know.

- If we are to base policy upon evidence we will have to override the public's preferences in many policy areas, simply because these are founded upon egregious errors. Whilst there is a case for doing this there is also a danger. Excluding the public risks increasing elitism and undermining the democratic spirit. Ignoring people's preferences might lead to a slippery slope in which governments eventually ignore people.

- Instrumental rationality is not the only rationality. As Robert Nozick said, there is is also symbolic rationality; we do some things not because they fit a narrow cost-benefit calculus but because they symbolize who we are. We might want to criminalize drugs to express our abhorrence of wasting lives to addiction; or we might favour minimum wages even if they destroy jobs to symbolize our distaste for low pay; or we could subsidise inefficient green energy to express our concern for the climate. And so on. Whatever you think of these examples it would surely be rash to dismiss symbolic motives entirely, given that it is otherwise hard to understand why people vote or protest.

Now, I'm not sure how strong these arguments are. I mention them because both main parties attitudes to (say) immigration and fiscal policy are not based in evidence. Which must mean that there is some sort of argument against evidence-based policy.