For Christmas, I got Lucius some cat treats. Which has given me a problem. In an effort to keep his weight below a level that would distort space-time, I'm trying to ration the treats. But this leads to a irate cat meowing to demand them.

Which illustrates a problem with welfare spending. Benefits quickly become regarded as entitlements, and so reductions in them cause far more anger and resentment than we'd have if the benefits were never granted at all.

What I'm seeing with Lucius is just what we see with attempts to reduce the pensions of public sector workers or housing benefits in London, or with opposition to fiscal restraint in Greece and Italy. Payments which would be regarded as quite reasonable if they were promised instead of nothing are regarded as mean if they replace more generous payments.

It's easy to see the psychology here. There's a framing and anchoring effect.; ''340 a week housing benefit is mean, when the anchor is ''438. And there's the status quo bias and loss aversion; people hate losses more than they like equivalent gains.

All of this means that there is an asymmetry in public spending. It is easy to expand it, but very difficult to significantly cut it, thanks to what Vito Tanzi and Ludger Schuknecht call the 'tyranny of past commitments'.

As Matthew Taylor has written, there is a problem with entitlements.

All of this leads to sympathise with Daniel Sage's call for a degree of conditionality in welfare payments. My instincts say this should be a bad idea. The state should aim at setting people free from drudge work and increasing their bargaining power. Universal benefits - a basic income - would do this whereas conditional benefits mean the welfare state operates as a human resources department for capitalism.

And yet, despite this, mightn't there be a case for conditionality, slightly different from what Daniel suggests? The case is this. If governments anticipate that benefits will become entitlements, they will be loath to increase them, for fear of them becoming hard to cut should the perceived need arise. If, however, the benefits are conditional, this reluctance will be smaller; governments could figure that the lack of an entitlement culture will make it easier to reduce future spending. They might, therefore, pay larger benefits. This effect would be augmented by voters' greater willingness to pay benefits if they believe (rightly or wrongly) that they are not giving handouts to scroungers.

In other words, conditional benefits - flawed as they are in principle - might have the virtue, from a leftist point of view, of leading to more generous provision than unconditional ones.

I say this tentatively, partly because I'm arguing against my own instincts, and partly because so many of those who want more conditionality also seem to want a smaller welfare state too. My point is simply that we should compare the merits of conditionality and entitlements not just in static terms, but in dynamic ones: which would lead to greater redistribution over the long-run?