Whenever an economist or libertarian opposes means-testing of Social Security and Medicare, I immediately ask: “So should we extend all currently means-tested programs to the entire population?”  Listeners often admit that it’s a persuasive challenge.  At our last lunch, however, Tim Kane, one of my favorite libertarian economists, bit the bullet.  In his view, any program that we make available to the poor should be available to everyone.

If I understand Tim correctly, his main argument is that means-testing is a high implicit tax rate, with all the standard effects.  As a behavioral economist, I suspect this a serious overstatement.  Framing and transparency matter. 

Even if Tim is right, though, we face a choice of evils.  With means-testing, we’ll have a high implicit marginal tax rate on everyone who might plausibly collect benefits.  Without means-testing, however, we’ll have a high implicit marginal tax rate on everyone.  Why the difference?  Because means-testing saves a ton of money, allowing lower average tax rates.

Or so it seems to me.  My challenge for Tim and anyone who agrees with him: Offer a plausible estimate for the total cost of expanding all means-tested programs to the entire population – year in, year out.  Start with Medicaid, then add unemployment insurance, TANF, and everything else.  Even ignoring disincentive effects, can you really come up with a figure under 10% of GDP?

Next step: Ignoring disincentive effects, calculate the average tax increase you’d need to pay to end means-testing.

Penultimate step: Adjust the final equilibrium to account for (a) the better incentives at the bottom of the income distribution, and (b) the worse incentives for the remainder of the income distribution.

Final step: Tell me if you’ve changed your mind.

Please show your work.