Saturday, July 30, 2011

Talking about Tax Progressivity (part 2)

Whether you soldiered through the technical stuff in part 1 or not, welcome to the fun part.[1]

In our first graph, we use the data to look at just the payroll and income taxes, that is, the taxes that can be directly attributable to households.  That the data is presented below[2]

You’ll quickly see in comparing the two graphs that something quite interesting has happened since 1970 in terms of how taxes are distributed.  The line for 2004 is actually above the line for 1970 at most points across the line.  The only places it is below is at the very top of the distribution and for most of the bottom half.  This is not surprising as effective income and payroll tax rates in total are actually higher in 2004 than they were in 1970.  The second thing you will notice is that the absolute magnitude of change is largest at the bottom of the income distribution.  For the P20-40 group, the effective income and payroll tax rate today is only half as high as it was in 1970.  Yes, the rates are lower for the top 0.1 percent but the magnitude of the change is by far the largest at the bottom of the distribution.

So to the degree that one would want to argue that income and payroll taxes have become less progressive since 1970, one could only be focusing on the fact that tax rates for the top 0.1% have gone down.  Visually, it is quite difficult to argue that the red line is less progressive than the blue line despite this fact.  One way to look at this is the ratio of the P20-40 group’s tax rates to the P99.99-100 group.  In 1970, this ratio was 2.1.  By 2004, it had increased to 3.8, hardly a signed of declining progressivity.

It is important to remember that most discussions of the Federal tax code focus on income and payroll tax rates and therefore an understanding of this graph is quite important as one considers proposals in those areas.

Adding imputed corporate taxes changes the picture a bit.  Now the answer to whether the tax code has become more or less progressive depends a bit on where you want to look.  Were we to focus on the bottom 99 percent of the population, we’d conclude that the tax code has become more progressive.  Rates have declined for the bottom part of the distribution and increased from the 80th to the 99th percentile.  Rates have also decreased from the 99th to the 100th percentile, and dramatically so at the top.  And yet, the ratio we constructed in the first chart between the second quintile (P20-40) and the very top of the distribution is 2.8 in 1970 and 3.7 in 2004, indicating that progressivity, at least by that measure, is increasing even with the sharp reduction in rates on the top income earners.

All of which brings us back to the final chart, the most suspect chart because of the issues with the allocation of the estate tax.  I would note however, that even in that case, the progressivity index goes from 4.03 in 1970 to 3.96 in 2004, not exactly a massive change in progressivity.  And we still see that the only people who are paying higher taxes today than they were in 1970 are the people between the 80th and the 99th percentiles.

It is also worth looking back at the original chart from the first post again.  On that chart, you'll see lines drawn by year as opposed to using the end point but you won't see the horizontal data series so that chart masks the fact that tax rates have declined dramatically for the 20-40 percentile group and that rates have increased from the 80-99 percentile group.  Another reason why it is good to go back to the base data.  You get to see all of the data and not just some of it.

What are we to make of all this.  Personally, I take away three conclusions:
  1. The Federal tax code has become more progressive over the last 40 years.  I draw this from an assessment of the any of the graphs above but particularly the ones excluding the estate tax that, in my view, cannot be fairly allocated given the issues with incidence.  Any decrease in progressivity can only be argued at the absolute top (e.g., 99.99+ percentile) of the income distribution.
  2. There is a case to be made for increasing taxes at the top of the income distribution (somewhere between the top 1 percent and the top .5 percent) given the decline in tax rates on this group.  This probably should come through higher levels of taxation on capital income, particularly capital gains.  A movement in the capital gains tax rate from 15 to 20 percent would likely address much of the change in tax rates up to the top 0.1% of the income distribution.  While this case is real, it relies on the notion that the 1970 effective rates are an appropriate baseline.
  3. There is an equally strong (perhaps stronger) case to be made that taxes at the bottom of the income distribution should be increased as the decline in tax rates in the bottom half of the distribution has been larger than any group outside of the top 0.1%.
Tax incidence is a complex subject and is yet another area where you need to see the data behind the chart in order to understand it.  Without going back to the source, you'd never know.

[1] The source data for all of this is Piketty and Saez, “How Progressive is the U.S. Federal Tax System? A Historical and International Perspective”, Journal of Economic Perspectives—Volume 21, Number 1—Winter 2007—Pages 3–24.  Throughout this post, I use the 1970 and 2004 end points.  There are three reasons for this.  The conceptual reason is that both 1970 and 2004 are after the Medicare and Medicaid programs were instituted whereas 1960 is before and therefore less directly comparable.  The practical reason is that Piketty and Saez do not present in tabular form the data for any of the other years.  In addition, the total Federal tax rate in the 1970 and 2004 series are nearly identical, increasing the focus on changes in the rates driven by distribution as opposed to average rate increases or decreases.

[2] Two analytical methods are applied to the original Piketty and Saez tabular data.  Starting from the data set in table 2 which reports total tax rates by quintile and subsets of the top quintile, estate tax rates and corporate tax rates from tables 3a and 3b were subtracted to get data excluding the estate tax and both the estate tax and corporate tax.  Since the estate and corporate tax data are reported only for the 0-90 percentile band, each group within this band was assumed to pay at the same rate.  While this introduces error into the analysis, the error is small because estate and corporate tax rates for these groups are small.  Any error would tend to overstate the tax rates of those at the lower end of the spectrum relative to those at the upper end of the 0-90 spectrum.

No comments:

Post a Comment