Saturday, March 17, 2007

More Tax Analysis

Using the same IRS data for my recent postings on taxes, I have produced the two charts above showing the adjusted gross income (AGI) floors for the top 50% of taxpayers (top chart), and for the top 1% (the IRS also has data on the top 5%, top 10% and top 25%, those are not displayed here).

Note that these are in constant dollars, adjusted for inflation. Further, because the IRS calculates annually the threshold level of income for the top 50% of taxpayers, those threshold values are actually the median levels of income (AGI), since 50% of taxpayers have higher income than the threshold, and 50% of taxpayers have lower incomes. Therefore, although it is not the direct intention of the IRS, its tax data produce a measure of real median income using actual tax return data.


1. Real median income, measured by AGI from tax returns, has remained relatively constant over the last 19 years at just under $16,000.

2. The real income of the richest 1% has increased significantly between 1986-2004, and the threshold level to be in the top 1% has increased from $108,000 to $174,000, a 60% increase.

3. The richest 1% have gotten richer over the last year 19 years, but the income of the average, or median taxpayer has remained relatively constant and stable, in real dollars. The middle class has not disappeared, and the median AGI over time has not changed much. There are more high-income taxpayers in the extreme right-hand tail of the AGI distribution, but that has not come at the expense of the median taxpayer. That is, the median income taxpayer has NOT been affected by the top 1% of taxpayers getting richer.

4. I wouldn't pay a lot of attention to the likes of Paul Krugman and his constant opining about this topic, like this recent talk "A History of America's Disappearing Middle Class." If there really was a disappearing middle class, the real median AGI should be declining, when it has actually been increasing slightly since the mid-1990s.

5. Note that the taxpayers in income groups like the top 1% or bottom 50% are not the same taxpayers from year to year, there is actually significant income mobility over a lifetime. "The rich" is not a club closed to new members. Many of the taxpayers in the top 1% in 1998 might now be retired and in the bottom 50%, and many taxpayers in the bottom 50% in 1998 (medical students, law students, business students, etc.), might now be in the top 1%.


At 3/19/2007 4:09 AM, Blogger Brian said...

Making conclusions on the 50% mark and 99% mark of a dataset won't give any meaningful results. If we were dealing with a Gaussian distribution (bell curve) then we could conclude all sorts of things, but there isn't any information presented which would lead me to believe so.

That means all we've established is the 50% mark and 99% mark for a data set which gives us very little except for bounds on certain points in the data.
It only says that 50% of people make less than ~16k, 49% of people make between ~16k and 174k, and that last 1%, well the sky's the limit! It doesn't tell you anything about that bottom 50% that might have some meaning. For all we know, everyone up to the 50% mark could be making absolutely nothing or the 50% mark and we would still come up with this exact data set. The same goes for the top 1%. Everyone in it could be making 174k, or it could be just that one guy at the 99% mark and everyone else is an order of magnitude higher.

My conclusion is that you really didn't get enough information here to conclude what you did. If you don't believe me, go ask someone in your statistics department and see what they think.

On another note, your conclusions appear to have an amount of confirmation bias in them since it appears as though your interpretations of limited data have a common theme to them.

At 3/19/2007 8:08 AM, Anonymous Walt G. said...


Professor Perry included a link for you to conduct your own analysis. It’s an in-depth data set that could probably be disaggregated for the information you seek. I’ve always considered someone else’s conclusions as my starting point into research.

I was intrigued by Conclusion 5. One would probably need longitudinal data to study that income distribution angle, though. I thought it was an excellent and rational insight. After all, don’t we all aspire to be in the top 70%, 80%, or maybe even the 90% group at least once in our lives?

At 3/19/2007 2:56 PM, Blogger Mark J. Perry said...

It is well known that economic variables like income and housing prices are not distributed normally (non-Gaussian), and are skewed positively to the right. In those cases with extreme positive outliers (the "Oprah effect"), the median value is a much better measure of central tendency than the mean/average. Since the IRS calculates and reports the threshold income to be included in the top 50% of taxpayers, they are actually calculating the median income by AGI. It is exactly because income is a NON-Gaussian distribution that the median income is a better measure of "typical income," versus using the average/mean income.

Although median housing prices are usually reported, median income values are not usually reported - mean income is more common. That is why I thought the IRS data was interesting, because they inadvertently calculate and report median income by AGI, which is much harder to obtain than average/mean income.


Post a Comment

<< Home