Nobody From Nowhere (@i8dc)

Occasional Common Sense

Why Perry’s Chart Was Dishonest, Lazy, or Thoughtless

with 6 comments

or “I’m Glad Duels are No Longer Commonplace”

Mark Perry took issue with my characterization of his work as dishonest, lazy, or thoughtless.  Fair enough – here’s my reasoning.

The source Perry cites is a spreadsheet put out by the IRS Statistics of Income (SOI) Division, which provides various statistics broken down into the following categories:

  • Top 1 percent
  • Top 5 percent
  • Top 10 percent
  • Top 25 percent
  • Top 50 percent
  • All (or Top 100 percent)

Perry’s chart depicts two different statistics that are included in the IRS spreadsheet – average tax rate and adjusted gross income (AGI) floor for each group.  For all of the groups except the bottom 50%, Perry uses data provided by the IRS.  But for the bottom 50% average tax rate, Perry makes a choice: rather than use the11.06% figure provided for all returns (or the top 100%, which would be consistent with his other columns), he chooses to do a little math to come up with the average tax rate paid by the bottom 50%.  The result is Perry’s chart:

This chart would be far less dramatic if the left column were 11.06%.  But is this a fair way to present the data?  Not at all.  If you don’t believe it, here’s the same trick going the other way:

Compare this chart with the one above – they show the exact same data.  The difference is that for the middle categories, this chart makes them appear smaller by counting all of the returns at the low end, while Perry’s makes them appear larger by including all of the returns at the top end.

It’s quite easy to avoid these kinds of biases in charts.  Perry could have derived meaningful splits just as easily as he derived the bottom 50%.  That’s what I did in my critique, shown below:

So back to my characterization of Perry’s work here as “dishonest, lazy, or thoughtless.”  Why did he derive the single bottom 50% figure and not the others?  If he did it intentionally, then it was a dishonest treatment of the statistics.  If he didn’t realize that his treatment was unfair, then it was thoughtless (perhaps not exactly the right word – maybe negligent or careless would have been better).  If he figured the bottom 50% number and then thought “gosh, that was hard, I don’t want to do any more of that,” then it was lazy.

I couldn’t think of any other options late last night. But I’m open to the possibility than I was being dishonest, lazy, or thoughtless.

Advertisements

Written by David Clayton

October 13, 2011 at 11:44 pm

Posted in Debunkery

6 Responses

Subscribe to comments with RSS.

  1. “There are three kinds of lies: lies, damned lies, and statistics.”

    I can tell that you have some animosity towards Mark Perry. Far be it from me to judge. My Italian roots mean that holding a grudge is part of my DNA, but I know that you know exactly where he got the data for chart 1.

    The IRS published data in both descending and ascending income percentiles in the two tables under this heading: “Individual Income Tax Returns with Positive Adjusted Gross Income (AGI) Returns Classified by Tax Percentile”

    Your chart (chart 2) makes it look like the top 1% are getting screwed. Your should probably take that down, or else you’ll be accused of supporting the rich.

    Your presentation in chart 3 is the only one that involved any calculation with the IRS data. Fun with Excel I assume. It more accurately shows the breakdown of average tax rates between various income levels. Point for you, but accusing someone of being dishonest or thoughtless is a stretch. I’ll concede that it could be classified as lazy to just take the raw data from one sheet, add a column from another and graph it.

    You had previously scolded me for my characterization of Warren Buffett’s motives. I think a little restraint is in order here too. Perry clearly plotted data provided by the IRS in a way that supported his point. That’s pretty much what everyone does.

    The real problem presented in this data is that my effective tax rate (IRS definition: Income Tax / AGI) is way higher than the average value for my income level in chart 3.

    Michael

    October 15, 2011 at 3:54 pm

    • I hold no more animosity towards Perry than he holds towards Warren Buffett.

      “The IRS published data in both descending and ascending income percentiles…”

      Perry provided his source, to which I linked above. It’s called “Table 1.–Returns with Positive Adjusted Gross Income (AGI):
      Number of Returns, Shares of AGI and Total Income Tax,
      AGI Floor on Percentiles in Current and Constant Dollars, and
      Average Tax Rates, by Selected DESCENDING CUMULATIVE PERCENTILES…” [emphasis added]

      in this file you will find every figure in Perry’s chart, except one: the average income tax rate of the bottom 50% of returns with positive AGI. Caculating it was easy. But no easier than calculating all of the splits in the third chart.

      [At least, that’s how it looks, but I’m using a 3.5″ screen to write this.]

      “accusing someone of being dishonest or thoughtless is a stretch. I’ll concede that it could be classified as lazy to just take the raw data from one sheet, add a column from another and graph it.”

      I didn’t find these other files you’re referring to when I wrote any of this. Perry said this preliminary data was his source, provided a link, and I didn’t find this data elsewhere in the usual sources,but I could have missed it. Please provide links.

      “Perry clearly plotted data provided by the IRS in a way that supported his point. That’s pretty much what everyone does.”

      There are honest ways to present data. This was not one. But as I said, I’m open to other interpretations: Perry derived one figure, which showed something fundamentally different from all the rest, and charted them together, and presented it as meaningful. I think leaving lazy and careless/unthinking as options is charitable.

      “The real problem presented in this data is that my effective tax rate (IRS definition: Income Tax / AGI) is way higher than the average value for my income level in chart 3.”

      I can certainly understand how that would be a problem! But these are broad averages, with lots of variation, especially in any split that includes the top end — which is the point made quite definitively in the CRS report I wrote about (but Perry didn’t – big surprise).

      David Clayton

      October 17, 2011 at 7:50 pm

      • David,

        Here is the source I was using: http://www.irs.gov/taxstats/indtaxstats/article/0,,id=133521,00.html. Scroll down a little. “Individual Income Tax Returns with Positive Adjusted Gross Income (AGI) Returns Classified by Tax Percentile—Early Release”

        Perry used the same source, but linked to the Excel file. It looks like he used all of the data from Table 1: “Selected Descending Cumulative Percentiles of Returns Based on Income Size Using the Definition of AGI for Each Year, Table 1”

        He then used the bottom 50% from:
        “Selected Ascending Cumulative Percentiles of Returns Based on Income Size Using the Definition of AGI for Each Year, Table 2.”

        I don’t think he did any calculation at all for the bottom 50%. It’s the 1st column in Table 2.

        “But these are broad averages, with lots of variation, especially in any split that includes the top end”

        It would be nice if the IRS provided some more information about the distributions. That way we would really know what effective tax rates are “typical” and which are outliers. I’m my opinion, which doesn’t actually count for much, the fact that the the average rates do represent such a broad range of effective rates for similar incomes is a fundamental “unfairness” in the tax laws, which is lost in the argument over “who pays their fair share”

        Michael

        October 17, 2011 at 8:28 pm

      • Thanks – I should have found that. Though I don’t think it matters here; I think it would probably be worse to pull the one column from the second sheet rather than do a quick calculation from the first. I mean, doing that would be almost certainly intentionally deceptive. At least if it was calculated from the first, then it might have been unintentionally misleading.

        You’ll notice that my chart 2, which I derived from the table 1 data, has the same figures as table 2 (except the top 1% data, of course). The math here is quite simple.

        The great variation in tax rates is due to the enormous range of possible deductions and exemptions/credits available, and because of the difference between regular and investment income.

        There are some better (more granular, smaller slices) data on effective tax rates in a table here: https://voxrationalis.wordpress.com/2011/09/22/warren-buffetts-secretary-fact-checking-the-fact-checkers/, which is mostly from one source and then deriving the top two slices using the top 400 taxpayers data. Sources are linked in there. This is the data I keep referring to when I say that the tax code turns regressive at the very top end of the income scale; this effect is primarily due to the prevalence of income from capital at the top end, though there are other factors, including higher charitable giving and tax-free income.

        David Clayton

        October 17, 2011 at 10:29 pm

  2. David,

    The Tax Policy Center retracted the information in the chart you included in that previous post:

    http://taxvox.taxpolicycenter.org/2011/09/23/retracting-some-recent-estimates/?CFID=112878868&CFTOKEN=27271023&jsessionid=b230f30b8fc336465e3a

    “On September 21, TPC published several tables (T11-0359 through 0362) that examine how effective tax rates vary within income groups. Those tables were intended to shed light on recent claims that some high-income taxpayers face low tax rates.

    Unfortunately, we made an error in our calculations. That error, which involved rollover distributions from 401(k)s and similar retirement plans, caused us to significantly overstate the income of some high-income taxpayers and thus understate the tax rates they paid. We have therefore retracted the tables while we work to fix our estimates.

    We regret this error and apologize to our many users for any inconvenience. TPC is committed to providing the highest-quality information about America’s tax system. We fell short this time. But we have sharpened our pencils and look forward to releasing new estimates.”

    I hope they correct it soon because is a very concise way to present the distribution of tax rates at various income levels.

    FYI, I commented as Itchy on some previous posts. I don’t have a WordPress account, but decided to at least use my first name when commenting as a guest.

    Michael

    October 18, 2011 at 7:18 am

    • That’s interesting. If I remember correctly, they’ve got a model that estimates the entire country’s distribution of income and tax characteristics. Sounds like they found an error in the model and are correcting it. I suspect that the overall impact will not disqualify that analysis, though it probably will say that the distributions are not as pronounced at the top end. Which will mean that there aren’t as many “Buffett rule” millionaires.

      I found the CRS report to be more important as soon as it was released, as I think it’s based less on a model and more on actual data. I should go back to see if this assumption was correct.

      David Clayton

      October 18, 2011 at 10:42 am


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: