David R. Henderson  

Distribution of Tax Burden by Quintile

Rand Paul on Letterman... The Hobgoblin of Simple Minds...

In comments on my post on Rand Paul and David Letterman, some commenters expressed interest in seeing the data on overall federal tax burden, not just the burden of the federal income tax. As it happens, the Congressional Budget Office reports such data. I would reprint their tables but I haven't yet figured out how to do that. So here is the link for 2006 data. Click on their data and you'll get an Excel spreadsheet that shows the following:

. The bottom quintile paid 4.3 percent of income in taxes,
. The top quintile paid 25.8 percent of income in taxes,
. The top decile paid 27.5 percent of income in taxes,
. The top 5 percent paid 29.0 percent of income in taxes, and
. The top 1 percent paid 31.2 percent of income in taxes.

In other words, it's still the case that the higher your income, the higher a percent of your income you pay in taxes.

So the existence of Social Security, while it reduces the "progressivity" of the tax system, does not come close to reversing it. I would note also, though, that most economists who examine the "progressivity" of the Social Security system also take account of the Social Security benefit formula. This gives a much higher percent of income to low-income people than to high-income people. In his article on Social Security in the Concise Encyclopedia of Economics, Thomas R. Saving wrote:

The benefit formula is set up to favor lower-income workers. For example, in 2004, someone with average monthly earnings of $624 received a benefit that replaced 90 percent of earnings. Someone whose average monthly earnings were $3,760 received a benefit that replaced 42 percent of earnings, while someone with monthly earnings at the then-taxable maximum of $7,325 received a benefit that replaced only 28 percent of earnings.
Milton Friedman speculated, in a 1972 debate with Wilbur Cohen on Social Security, that the fact that low-income people enter the work force sooner and tend to leave later and die earlier reversed that "progressivity." But that was his speculation; I've never seen anyone back it up.

Finally, I should note that to reach these conclusions, the CBO needs to make assumptions about the incidence of taxes. Here are their assumptions:

CBO's analysis of effective tax rates assumes that households bear the burden of the taxes that they pay directly, such as individual income taxes (including taxes on interest, dividends and capital gains) and employees' share of payroll taxes. The analysis assumes--as do most economists--that employers' share of payroll taxes is passed on to employees in the form of lower wages than would otherwise be paid. Therefore, the amount of those taxes is included in employees' income, and the taxes are counted as part of employees' tax burden. CBO estimates payroll taxes and individual income taxes, including refundable tax credits, with a tax "calculator" that applies the tax law for the relevant year to the tax return data from the SOI.
Excise taxes are assumed to fall on households according to their consumption of taxed goods (such as tobacco and alcohol). Excise taxes that affect intermediate goods, which are paid by businesses, are attributed to households in proportion to their overall consumption. CBO assumes that each household spends the same on taxed goods as similar household with comparable income in the Consumer Expenditure Survey.
Far less consensus exists about how to attribute corporate income taxes (and taxes on capital income generally). In this analysis, CBO assumes that corporate income taxes are borne by owners of capital in proportion to their income from interest, dividends, capital gains, and rents. Over the long term, however, some models suggest that at least part of the burden falls on labor income.

I think their assumption that is most "off" (and they seem to think so too) is this last one--on capital income. Because capital is fairly mobile across borders, much of the corporate income tax is shifted to labor.

Comments and Sharing


COMMENTS (29 to date)
Jeremy H. writes:

Even more comprehensively, we'd want to include state and local taxes. The Tax Foundation has a paper on this:


Table 16 shows average tax rates. Table 18 shows the shares. The top 20% is paying about half of all taxes, while the bottom 20% is paying about 4.3%. This is a very detailed paper, but unfortunately they don't get down to the detail of top 5%, 1%, etc.

Dale writes:

Any such analysis which considers non-pecuniary benefits as income but does not consider non-pecuniary benefits that the wealthy receive is not serious.

A poor worker has a statutory rate above their listed tax rate with no means of exemption. Unless we track where all of the profits from defense spending, road subsidies, el al. go, then we don't have a good indication of "burden" as claimed by the numbers.

In short, if we are calling medicare and medicaid "income" then we also have to call the increases in marginal productivity due to government spending income on the same scale, and exempt in our calculations any taxes paid on those "incomes".

Why must we do this? Consider the case where in-kind government benefits are taxed compared to untaxed(it is my knowledge that they are not taxed). If taxed then large benefits cycle right back into the effective rate paid by lower income people.

I.E. if someone has an "income" of $10,000 pre-tax cash and $10,000 of in-kind untaxed benefits with a statutory rate of 20% on the pre-tax cash their effective rate looks to be 10%.

But we can give them the same amount of income via higher in-kind benefits if we tax them at the same statutory rate. To get the same, final, after tax income we have in kind benefits of %12,500.

Note that the first measure of the exact same income yields an effective rate of $10 while the second measure of of the exact same income yields an effective rate of 20%.

Yet neither makes any difference to the government deficit(as revenues increase directly with proportion to expenditures).

Such any government expenditure which increases a persons marginal productivity OR is straight business that allows them to work is recorded in at a different rate than the in-kind contributions which we then use to claim that the poor have a low tax burden*.

No, we are simply recording the tax burden of the rich in a more favorable light than we are recording the tax burden of the poor.

*The easiest example to see is government employees. They are taxed on income and payroll. If they were not taxed on that income and payroll the government could reduce their wages accordingly and be at a net zero position for the budget. If we record their income as we do for poor people then the people who have no other sources of revenue will have an effective rate of 0%. If we record their income as we do for currently they have a rate depending on their income and exemptions.

Such, counting one form of government aid as tax free, and another as not tax free, skews the numbers and makes them worthless.

Scott Sumner writes:

I'm surprised that there is a consensus among economists that the burden of direct taxes like income taxes actually falls on those who pay the tax. Much of labor income is a return on human capital investment.

David R. Henderson writes:

@Scott Sumner,
Good point, Scott.

Dale writes:

True Scott. But drawing the line somewhere makes sense. Especially if we accept norms with regards to the inputs on that human capital investment as the result of some sort of market mechanism. That is to say that the laborers have purchased in some way that investment. If a single person "owns" the investment in the human capital then we don't have to distribute between them.

More difficult is tax burdens on other forms of capital and the assumptions are likely largely for expedience. It would be quite a task to back out the relative burdens from elasticities over the entire nation.

Practically I am not sure its even possible. If we could track the relative burden we wouldn't know to whom it transferred. And if we knew to who it transferred we wouldn't have an easy way to assign it into a statistical model without information on everyone in the United States and all of their cash flows and burdens.

frankcross writes:

The progressivity of SS benefits as reported is far too simplistic. The rich tend to live much longer than the poor, some of whom may die before receiving any benefits. I recall an NBER paper that showed that when life expectancy was considered, the progressivity pretty much went away.

Yancey Ward writes:

Shouldn't SSA and Medicare taxes be netted against expected benefits properly discounted? In other words, how much is the Social Security tax if one pays $100,000 over a life time and receives $120,000 in benefits before dying?

R Richard Schweitzer writes:

In The Ethics of Redistribution (Liberty Fund) de Jouvenel touches briefly on the point that the issue never seems to have been raised that those Corporations or enterprises with greater incomes should be taxed at "progressively" higher marginal rates; or perhaps the more "profitable" should pay at progressive rates.

Wonder why that is?

John V writes:


You are trying too hard to make the numbers say less than they really do.

Just on the face of it, you are trying to compare widely received benefits to the poor with income from a proportionally small amount of companies that get government contracts as part of their income. How can you bring up a point like that and expect to carry as much as weight as you want it to?

So many businesses, like my own, fall into the overwhelming majority of business that doesn't get any such income. Moreover, a lot of that work is going to get done one way or another. If roads are private, we can say that the burden is even more regressive since it would user-based with tolls. If defense-spending were private (which it can't be...but it's just an example) then many would get no benefit at all. And since defense spending, in theory, is supposed to be paid with income taxes and corporate taxes (and NOT payroll taxes...which are supposed to pay for their intended programs) then the "bang for the buck" on an individual level favors those who pay the least.

David Friedman writes:

You can find an estimate of tax burden by income for federal, state and local at:


The organization producing the data has, like most, its own axes to grind, but if you follow the links you can find a fairly detailed explanation of how the calculations were done.

If their numbers are correct, the combined tax burden still increases with income, but not as sharply as the CBO estimate for combined federal taxes.

Incidentally, one of the things I find dubious about the CBO calculation is that they treat the entire payroll tax as a tax on employees. It's true that the division between employer and employee portion is bogus, but the whole thing is a tax on the transaction of hiring labor; how the burden is distributed depends on the relevant elasticities, as with other taxes.

Dale writes:

John V wrote

Just on the face of it, you are trying to compare widely received benefits to the poor with income from a proportionally small amount of companies that get government contracts as part of their income. How can you bring up a point like that and expect to carry as much as weight as you want it to?

Run the numbers, i think you would be surprised as to the raw value of those benefits and who they go to compared to the rest of the budget.

At the end of the day it doesn't much matter just how spread out those benefits are, as the recipients are in increasingly small and the amount significant.

Imagine for a second that we spread those benefits out evenly across 50% of the population. At 1 trillion dollars that comes to about 6500 dollars per person. If we assume that the other 2.8 trillion in benefit is to poor people then we are still looking at significant percentage point modifications to the listed rates which is simply not recorded

Also, private roads don't work on a national scale, and whether or not tolls are used is irrelevant to the size of the subsidy and benefit(which is significant for those who use roads more). If you are of the mindset that the government should do less, knowing just who benefits from those aspects should be high on your priority list.

David R. Henderson writes:

Thanks. You're making the point Milton Friedman made, a point I referred to. Do you have the NBER cite?

Bill writes:

The issue of the progressivity vs regressiveness of Social Security taxes AND benefit payments brings up another point, and that is which groups see the largest return in benefits from government expenditures for the tax dollars they pay. A 2007 Tax Foundation study (http://www.taxfoundation.org/publications/show/2282.html) reported on that very issue. Using data from 2004 (the latest available at the time of the study) researchers found that, considering local, state and federal tax and expenditures, those in the bottom quintile of income distribution received back $8.21 in benefits from government expenditures for each tax dollar they paid. Those in the middle quintile received $1.30 in expenditure benefits per tax dollar paid, while those in the top income quintile received $0.41 in expenditure benefits per tax dollar paid.

Seth writes:

It might be good to point out that the bottom quintile paid 4.3% in combined SS and income taxes, which is less than the SS tax rate by itself, because of the earned income tax credit, I think.

I have yet to see those who bring up the non-progressiveness of the SS tax system also bring up the hyper progressive earned income credit, which somewhat offsets the SS tax.

Ironman writes:

David Henderson wrote:

I would reprint their tables but I haven't yet figured out how to do that.

If you have data in a spreadsheet format, try Tableizer.

Bill writes:

I should note that the Tax Foundation study to which I referred is the same as the one referenced by Jeremy H. We just happened to focus on different elements of the study in our respective posts.

HispanicPundit writes:

What about capital gains taxes? Most income earned today, especially by the rich, is earned via stocks and investment. Do you have data that includes capital gains taxes?

David R. Henderson writes:

Yes. All of the income tax data include whatever capital gains taxes are paid.

Richard Spencer writes:

It wouldseem that a present unacceptable and dirty little secret of Social Security, since it has salary caps for payments into the system, is that the end result is poorer workers subsidize the richer, at least in percentage of income.

I am not suggesting there should be an unlimited cap for workers as I favor the elimination of SS in favor of a 401 type savings program.

Joe Cushing writes:

The proportion of entitlement taxes born by employers and employees is not what anybody I've ever heard talk about it thinks it is. It's not evenly split as average people think and it's not all borne by employees as many libertarians and economists think. The proportion is unknowable. It depends on the slope of the supply and demand curves for labor at a specific company. Which ever is more inelastic bears more of the burden. One thing is certain: The cost is shared. The cost of taxing a transaction is always shared by the two parties involved. The only way for this not to be true would be for the two slopes to meet at a right angle where one is vertical and the other is horizontal. This just doesn't happen.

Bill writes:

Joe C. ~ David Friedman already made your point about the incidence of the SS tax being dependent on the comparative elasticities of supply and demand for labor services. Beyond that, you have your economics wrong in asserting that full shifting requires one curve to be vertical when the other is horizontal. Such is not the case. If, for example, labor demand is horizontal, the entire burden of the tax will be borne by employees for any positively sloping supply curve. Likewise, if the demand for labor is vertical, the entire burden of the tax will be borne by employers for any positively sloping supply curve.

mark writes:

I think if one were to add in estimated adjustments for
a) the underground economy going untaxed,
b) the tax subsidy of untaxed employer-provided health benefits;
c) the value of free k-12 education

d) the non or limited deductibility of capital losses, which by definition don't flow through taxes.
e) the losses of businesses that, unlike business income, are not passed through to individual taxpayers in this model,

you would find even greater progressivity as between the very top and the bottom quintile although it would probably make for a wavier line in betweent the two as the second item and arguably the third are of greatest benefit to upper middleincmoe taxpayers.

Dale writes:

Mark, the numbers account for all of those things(except a, not sure how they're going to measure that). Except that i am not sure what e means, capital losses can be counted against capital gains for tax purposes and if someone makes net capital losses, but doesn't entirely offset their capital gains for that period then that will show up in this data.

Richard Stewart writes:

I've heard comments from some accountants that they are amazed at how much income rich people don't pay taxes on. I explained that much of that income is likely from tax-free bonds which are not truly "free" since they earn a lower interest rate.

The CBO mentions they include tax-free income but I imagine it would be rather complicated to add back in (and then take out again in taxes) the income that would have been earned if there were no tax preferences.

I'm curious if this effect is taken into account.

Aaron Durst writes:

I am glad that you included all taxes in your analysis. The payroll tax is the most inefficient tax possible, and it deceives the American people into not realizing how much they are actually being taxed by their government. It is a pet peeve of mine when other conservatives ignore it and act as if only the income tax matters when showing the progressivity of the federal tax burden. More than half of all federal revenue comes from taxes other than the income tax, and I wish conservatives would stop focusing exclusively on the income tax. As you have shown it hardly makes a difference in calculating the progressive nature of the federal tax burden, so stop ignoring one of the federal government's worst and most inefficent forms of taxation (i.e., the payroll tax).

Steve Roth writes:

Because state and local taxes are so regressive, it's worth looking at the total tax burden: fed, state, local combined.

It's essentially nonprogressive above about $60K/year.


IMO, that's a fault in our tax system. It sacrifices both equity and efficiency.

Steve Roth writes:

And really: no amount of hand-waving is going to change the basic facts on the ground:


Jim writes:

I support your underlying message but I don't think the numbers make sense. If the top 5% paid 29% of the taxes, then the top 1% (a subset of those in the top 5%) should pay something less than 29%, but you have them paying 31% of the taxes. This is a mathematical non sequitor.

chris writes:

Jim - it's not a percentage of total taxes, it's the percentage of their income that they pay in tax.

Comments for this entry have been closed
Return to top