There's an old engineering saying, faster, better, cheaper, pick any two. Meaning that if you want your product to be better and developed faster, it's not going to be cheaper, if you want it to be better and cheaper you're not going to get it faster and so on through the possible combinations. As Arnold Kling points out, this also applies to welfare systems:

1. Use “means testing” in order to provide a significant benefit that is aimed at the poor.

2. Keep the marginal tax rate low.

3. Keep the budget cost low.

If you've got a significant benefit to the poor and you want a low marginal tax rate as people earn their way out of the means test then the taper must be shallow, making the cost of the program high. If you've got that low taper and you also want a low budget cost with a means test then you cannot have a significant benefit and so on. Andrew Biggs tries to plot a way out of this:

a limited but universal government-provided retirement benefit, set at the poverty threshold and paid to every retiree regardless of income or work history. Since the plan is universal, there’s no phase-out and thus no implicit marginal tax. Since the benefit is relatively modest, the explicit taxes need to finance it are lower than what we currently pay for social security. And unlike the current program, which leaves around 9% of seniors in poverty, the universal benefit would essentially eliminate poverty in old age.

OK, he's talking about the US pension system only and he also agrees that it doesn't make Kling's problems go away. It's a balancing of the various options, that 's all.

Moving this over to here and then to a consideration of the basic idea of welfare at all I think I'd make the same sorts of trade offs. I think that it's the marginal tax rates (combined with benefit withdrawal of course) that make the problem so pernicious. We absolutely do have people stuck in situations where they're very definitely over the peak of the Laffer Curve. Heck, the budget documents themselves point out that there's millions who face tax and benefit rates of over 60%, some even of over 100%.

Which is why I find myself moving over to the idea of the citizens' basic income, or the universal basic income to taste. We're going to have a benefit system whatever else happens so we should have one which distorts the rest of the economy the least. And the only way to get rid of those high marginal rates, at a budget cost that is anything like possible, is to make it an unconditional income. Everyone gets that (low) amount no questions asked. Given a tax allowance on top of it we've thus got people facing marginal tax rates of zero: which would rather encourage people into work I would have thought.

For this to be palatable you do indeed need to be assuming that there's going to be redistribution and a welfare system run by government, come what may. But once you have, I put it to you that a cbi or ubi is going to be the least bad of possible options.