The coalition's plan to cut child benefit for 40% tax rate payers seems to have run into confusion. Which raises the obvious question: why bother? If you want richer folk to bear a bigger share of austerity, why not simply raise the 40% rate by 2p, and raise its starting point a little, which would bring in the same ''1bn as the removal of child benefit.

Such a scheme would avoid the horizontal inequity whereby a household getting ''80,000 split evenly between two earners would keep child benefit whilst a single-earner household on ''45,000 would lose it. It would also avoid the problem of enormous marginal tax rates for people who get a small pay rise that puts them into the 40% band. It would also allow the government to boast about taking people out of the top tax bracket, and be administratively simpler.

So why not do this, rather than faff around with child benefit?

Some might say it's because the coalition wants to attack the principle of universal benefits. But I've another theory. It's about framing.

Tax is a burden - theft if you're a libertarian - whereas child benefit is, well, a benefit. And it's more comfortable - for the government and its target voters - to withdraw a benefit than to increase a burden. So we have a messy policy rather than a neat one, even though the effects are overall pretty similar.

Which brings me to my problem. Cameron has a behavioural insight team which hopes to manipulate our cognitive biases to make us behave better. But cognitive biases aren't something which ignorant citizens have and which wise governments are free of. Policy-makers are also prone to them - either because they are as irrational as everyone else or because they have to pander to an irrational electorate. One of my big complaints against Nudge is that it fails to appreciate this sufficiently, and so panders to the ideological fiction that policy-making is or can be entirely rational and evidence-based.