David’s not convinced by my case for means-testing.  I’d like to reply to his objections, point-by-point:

1. The phase-out issue.
Bryan recognizes that you wouldn’t want to give benefits below income
threshold x and then zero benefits to people above that threshold… So what’s the alternative?
Phase out the benefit over some income range… Take Bryan’s
proposal of giving a benefit to the bottom decile and then phasing it
out over the next decile. Let’s say the benefit is a $5,000 government
check. Just eye-balling the data,
I estimate that the bottom decile (about 8 million families) in 2008
had income between $0 and $17,500 and the second from the bottom decile
had income from $17,500 to about $28,500. So that’s $11,000 of income
over which to phase out a $5,000 benefit. The implicit marginal tax
rate from that phase-out alone is, therefore, $5,000/$11,000 or 45%.
That’s on top of other tax rates.

This isn’t nearly as bad as it seems. 

First, marginal rates on the poor are relatively low, so even a 45-percentage-point boost leaves tolerable work incentives. 

Second, as free-market people rightly emphasize, most people in the bottom quintile don’t stay there for long.  One of the main payoffs people in low-paid jobs get, in fact, is upward mobility; by putting their foot on the first rung of the economic ladder, they have a much better chance of graduating to a much better life.  Since this massive hidden benefit would remain untaxed under means-testing, there would still be decent work incentives for people in the second-to-last decile (and even the bottom decile).

Third, to be brutally blunt, the lower deciles don’t contribute that much to the economy, anyway.  Suppose means-testing led half of the second decile to stop working.  That destroys a lot less value than higher overall taxes that lead to 5% less work effort across the entire income distribution. 

2. What are means?
Without exception, every time I’ve seen someone advocate means testing,
he uses income as his measure of means. This completely ignores wealth… Therefore means testing would
discriminate in favor of wealthy people with low income. Because the
most expensive programs for which means testing is advocated tend to be
for the elderly, this is an even bigger problem. Among the retired
population, the correlation between income and wealth is even weaker, I
believe, than among the population in general.

Actually, Medicaid already means-tests funding for nursing homes on the basis of assets as well as income.  There’s no reason this practice couldn’t be broadly extended.

3. The fairness issue.
Again, because the programs at issue tend to be for the elderly, there
can be huge differences in income because one family saved a large
percent over the years, and is earning interest and dividends on that
income, and the other family saved 0 and relies on Social Security.
This could be so even though the two families had a similar
age-earnings profile. It’s unfair. I know that the program per se is
unfair, with or without means testing, but this one seems particularly

Think about it this way: Universal programs are basically just (means-tested programs + accounting fiction + waste).  There’s almost always some genuine redistribution.   But to preserve the illusion of universality, universal programs primarily take money from the non-poor, waste some resources in administration, then return the remainder.  I don’t see how the accounting fiction and waste make means-testing “fairer” for the non-poor.  Indeed, don’t they just add insult to injury?

Of course, I’d rather just abolish the welfare state.  I suspect David agrees.  But as long as the welfare state is around, means-testing is a common-sense way to keep it under control.