Late last year, I laid out in a very clear way why universal benefits cost less than means-tested benefits and why people who think otherwise are simply getting mixed up in accounting definitions rather than dealing with economic realities.
In the first part of this argument, I iterate through three successive policies that go like this:
- Levy a $100 billion tax on the richest Americans. Pay $3k per child in child benefits.
- Levy a $100 billion tax on the richest Americans. Pay $6k per child in child benefits. Subject the $6k benefit to a 5% phaseout for family earnings above $50k.
- Levy a $100 billion tax on the richest Americans. Pay $6k per child in child benefits. Levy a 5% surcharge tax for family earnings above $50k up to $6k of tax.
In a lot of discussions about universal benefits and means-tested benefits, people insist on comparing policies (1) and (2). The reason they do this is because they think (1) and (2) cost the same amount of money, i.e. $100 billion. Put another way, they compare policy (1) and (2) because these two policies are tax-equivalent where “tax” is defined the way the national accounts define it.
But if you really want to analyze the difference between universal benefits and means-tested benefits, the proper comparison is between policies (2) and (3). The reason this is the proper comparison is because they rely upon the same effective marginal tax rates (EMTRs). The way that policy (2) is able to deliver more benefit to the poor than policy (1) is by imposing an extra 5% EMTR on families with income over $50k. But as we see in policy (3), the universalist can also impose an extra 5% EMTR on families with income over $50k. The universalist does this with a tax instead of a phaseout, but it all ultimately amounts to the same thing.
Thus we can see that, when means-tested and universal policies are EMTR-equivalent (as opposed to tax-equivalent), they do the same thing. The entire case for means-testing therefore is based on tricking people into comparing tax-equivalent policies instead of EMTR-equivalent policies.
Once this point is understood, the next move in the argument is just to observe that phaseouts are more costly to administer than taxes. Thus, since EMTR-equivalent policies are identical in all other respects, the administrative efficiency of universal policies (which use taxes instead of phaseouts) makes them less costly.
Over the last couple of days, I have been genuinely shocked to watch Scott Winship struggle to understand this really basic point about EMTRs, even as his fellow conservative think tank brethren intervene, some with spreadsheets in hand, to try to get his head around it.
Today, he wrote a short blog post explaining why he thinks I am wrong. The key paragraph is this one:
What [Matt says] both strategies really do is provide a universal benefit and then tax it away gradually for families above $50,000 in income, just in different ways. But that’s clearly wrong. The new tax is an additional tax. It makes no difference that all families see the same net change in income under both strategies, which is the point Bruenig wants to emphasize (and which is true). The fact of the matter is that the additional tax revenue could have been used to increase the targeted benefit rather than making the targeted benefit universal. That would reduce poverty and inequality even more. There is no corresponding way to increase the generosity of the targeted benefit under the phase-out strategy that is limited to $100B in revenue.
The bolded part describes a fourth policy that is as follows:
4. Levy a $100 billion tax on the richest Americans. Pay $9k per child in child benefits. Subject the $9k benefit to a 5% phaseout for family earnings above $50k. Levy a 5% surcharge tax for family earnings above $50k up to $9k of tax.
With policy (4), all Winship has done is taken the universal policy (3), added a 5% phaseout to it, and then used the “revenue” (if we can call it that) from that 5% phaseout to increase the child benefit by $3k. Put differently, in moving from policy (3) to policy (4), Winship increased the EMTR of families over $50k from 5% to 10% and then used that to increase the generosity of the child benefit.
If you look closely, you’ll see this is the exact same thing that happened when we moved from policy (1) to policy (2) in my initial example. All Winship has observed here is that, when you take the universal benefit and add a phaseout to increase the EMTR by 5 percentage points, you can get more revenue to increase the child benefit.
But we can counter his policy (4) with a new policy (5) that looks like this:
5. Levy a $100 billion tax on the richest Americans. Pay $9k per child in child benefits. Levy a 10% surcharge tax for family earnings above $50k up to $9k of tax.
In this new policy (5), all I have done is take his phaseout and turn it into a tax, which is exactly what I did with my policy (3). Policy (5) is EMTR-equivalent to policy (4) just as policy (3) was EMTR-equivalent to policy (2).
When we EMTR-equivalize the policies — such as when we compare policy (3) to policy (2) and compare policy (5) to policy (4) — there is no distributive difference between them. It’s only when we don’t EMTR-equivalize and instead tax-equivalize them — such as when we compare policy (2) to policy (1) and compare policy (4) to policy (3) — that we can generate a difference.
We can go in circles on this all day of course. To counter my policy (5), Winship could propose a policy (6) which is just my policy (5) with a 5% phaseout added, at which point I will propose policy (7), which is just his policy (6) with the phaseout replaced by a tax.
But what this round-and-round just keeps showing is what I have said already twice in this post, which is:
- When we compare EMTR-equivalized policies, there is no distributive difference between means-tested and universal benefits.
- When we compare policies that are tax-equivalized but where the means-tested benefit has a higher EMTR, the means-tested benefit comes out on top.
- Therefore, the only reason means-tested benefits appear to come out on top is due to the fact that the national accountant’s definition of “tax” counts regular taxes as tax, but does not count phaseouts as tax. The means-tester therefore is succeeding in the debate only by implementing their EMTR increases in phaseouts that, though they are economically equivalent to taxes, are nevertheless not accounted that way in the official bookkeeping.
I’ve already explained to Winship many times that the whole thing is just EMTR-equivalization v. tax-equivalization. He insists that it is not and yet in this very blog post, his response is simply to insist that we should compare two policies that are tax-equivalent but not EMTR-equivalent. I don’t know what else can be done to get through to him.