Thinking about tax share

Kevin Drum posts this showing tax share by income in 1979 vs 2007.

This purports to show that:

1) Our tax system is not progressive (i.e. the rich don't pay outsized tax rates)

2) Tax rates on the rich have declined.

The biggest differences between this graph and graphs favored by the right:

1) "Average" is probably a very misleading way to present a comparison.

2) This graph includes state/local taxes.

3) This graph includes payroll taxes (Medicare, Social Security)

Thinking about "Average" tax share

It's not clear what "average" mean here. I'm presuming it's the simple mathematical average: tax rate = sum(all taxes) / sum(all income)

If that's the case, then it's a highly misleading way to try and show the progressiveness of a tax system. The average almost by definition cannot show the progressiveness of a distribution.

Consider the most progressive tax system imaginable:

Top 1%: 50% of income, 100% of all taxes, with 30% effective tax rate.

Other 99%: 50% of income, 0% of taxes, (0% effective tax rate)

This is the most progressive tax system imaginable. So what would EPI's numbers look like?

Top 1%: 30% tax rate

Average: 15% tax rate

That is, in the most progressive tax system possible, comparing the average to the top results in numbers that don't look that progressive! So hat's off to EPI for loading the statistical dice.

State and Local taxes

There's not much to say here. The EPI is right- when talking about tax share, we should almost certainly include local data. To be fair, the policy debate is about Federal tax policy, but the larger ideological debate is really what matters.

Payroll Taxes - what is a tax?

This is the most conceptually challenging area, and there's no clear answer. The problem is, taxes aren't as simple as adding up what you pay. Why is that? Consider the following thought experiments:

Imagine a world with a flat tax, everyone pays a fixed percent of income (10%). This is neither a progressive, nor regressive system. Now imagine that the government imposes a $5000 poll tax, the most regressive of taxes. But, in a sop to those on the left, it also creates a new "entitlement" which grants every taxpayer exactly $5000.

Under this scenario, every taxpayer is exactly as well off. This scenario may sound crazy, but a) governments do this sort of thing all the time and b) it has big implications for measuring tax share. Under the most naive accounting (EPI's methodology by the way), the system has become much more regressive, since the $5000 poll tax counts as a tax, but the $5000 entitlement is govt spending and not accounted for. But of course, the net effect of this change is nothing! Every taxpayer is exactly as well off as before, but yet the tax system (as measured by EPI) has become more regressive.

This is of course a contrived example. But it happens all the time. For instance, the gov't has raised taxes on everyone to pay for Medicaid, but then turned around and given some (low-income) taxpayers a huge medical benefit worth thousands of dollars. This is largely the same as the example above, except instead of $5000 in cash, the gov't has given (some) taxpayers an in-kind benefit worth $5000. There must be some adjustment for this in the tax share, otherwise tax share gets totally distorted by in-kind and non-tax benefits.

Or consider what happens if instead of giving each Taxpayer $5000, it gives that $5000 to the taxpayer the next year (plus interest). Now taxpayers are essentially giving the government a loan and getting paid back with interest the next year. How does that impact tax share? What if instead of a 1-year loan, you loan the money to the government until you turn 65? That's essentially what Social Security and Medicare are. What's the proper way to account for that?

SS and Medicare are of course more complicated than that, but those complications make the system more progressive.  As one final thought experiment to see just how meaningless "tax share" can be, imagine that the Bernie Sanders administration decides to make SS & Medicare benefits completely independent of contributions.  That is, everyone gets the same benefit regardless of what they've contributed.  President Sanders has made the US system more progressive, but now SS & Medicare payroll taxes are really pure taxes, not tied to any sort of future benefit (since those benefits are guaranteed).  My previous analysis breaks down- there's no reason to exclude these taxes when computing tax share.  After this change which makes the overall system much more progressive, the measured "tax share" seems more regressive!  Clearly "tax share" is not a meaningful metric for measuring the progressiveness of a tax code.

A true measure of tax share would be taxes paid net of direct benefits received. This  is fiendishly difficult to do. It probably makes sense to try and ignore those taxes most directly tied to benefits. Fortunately in the US system that's made easier since they are broken out as explicitly.

SS and Medicare benefits are highly progressive (they transfer wealth from the rich to the poor).  Ending SS & Medicare would make the US system much more regressive, but counting them as a pure tax liability makes the system seem less progressive, which is exactly EPI's goal.

The simpler approach is to just remove SS and Medicare taxes from the mix. This still doesn't account for those programs progressiveness or for things like Medicaid, but it's at least a lower-bound on the progressiveness of the tax system.