I’m happy to see all the feedback on my latest means-testing post, but I wish more readers engaged my original question.  After presenting one simple means-testing formula, I asked:

In the new political equilibrium, how much do you predict
the full Social Security and Medicare benefit will fall?  Remember, this
is the benefit that everyone below the median income gets.

My co-blogger David Henderson’s not happy because I didn’t revisit earlier critiques:

My point with this post is not to take back what Bryan
persuaded me of. Rather, it’s to try to persuade Bryan and others who
share his view to admit the criticisms that many others, including Scott Sumner and I, have made. Bryan re-advocated means testing in his latest post without even hinting that these three problems I raised are indeed problems.

I think I responded to most of these complaints in an previous post.  In any case, though, the point of my latest post was not to defend means-testing, but to ask readers about the political economy of means-testing.

Kyle Thompson:

As a liberal who supports means-testing social security and Medicare, I resent the implication that I don’t exist!

I only claimed that I don’t know of any liberals I’ve converted, not that liberal supporters of means-testing don’t exist.  Glad to have you with us, Kyle!


Consider a world where SS and Medicare are means-tested.

Does Bryan Caplan (and his fellow libertarian means-testing advocates):

a) vociferously argue for the abolition of SS and Medicare


b) vociferously defend the means testing they argued for.

My bet is on a. So why bother with means testing, if we’re going to argue about abolition anyway?

Why bother?  Because means-testing has far broader appeal than abolition.  As usual, I’m happy to meet people where they stand.

Lowcountry Joe:

I think it is completely immoral to means-test old-Social Security and
take away the opportunity for those who made the most contributions to
get nothing. The ‘insurance’ was designed to provide benefits to those
who outlived the mortality table. The only fair thing to do is to
substantially raise the age at which people are entitled to full
benefits and significantly reduce the benefit for those who take it

I agree that raising the retirement age is a good idea.  But why is it uniquely “fair”?  Your reform advantages people who are longer-lived, who tend to be rich, non-black, and female.  And it’s not like Social Security or Medicare is a “promise” in any meaningful sense when a simple act of Congress has always been sufficient to change the benefits.

The noble Scott Sumner:

I think this would be a bad idea. The numbers make no
sense unless you are referring to income levels of the retired. Suppose
a couple saved a million dollars during their life and turned it into a
$50,000 annuity on retirement. Also suppose they earned $50,000
combined from Social Security. If the million dollar nest egg puts them
in the top 10% of retirees, they lose almost all their Social Security.
That’s a huge disincentive to save. And recall that the tax system
already has other saving disincentives built in, so it would make the
bias against saving much stronger.

I presented a simple means-testing formula to focus attention on the political economy.  A lifetime measure of income is obviously better.

Scott continues:

Means testing is just another high implicit marginal tax on people who work hard and save.

“Just another high implicit marginal tax”?  Not really.

1. At minimum, means-testing is also a way to sharply reduce (or avoid increases in) payroll taxes.

2. It depends on the formula; if only the bottom decile is eligible for benefits, means-testing is a high implicit marginal tax on people with low income, and zero marginal tax on everyone else.

3. If you take behavioral economics seriously – and you should – it’s easy to believe that the behavioral response to benefit reductions decades in the future will be far less than the behavioral response to taxes now.  

P.S. Thanks to Frank Howland and Peter for directly addressing the political economy of means-testing.  I’d be especially grateful for any additional feedback on this issue.