STUMP » Articles » Taxing Tuesday: What's the Real Tax Rate? » 22 October 2019, 12:35

Where Stu & MP spout off about everything.

Taxing Tuesday: What's the Real Tax Rate?  


22 October 2019, 12:35

There was a piece out in the NYT about the supposed historical trend of all taxes paid by people at various income levels.

It was an excerpt from a new book, but as with Piketty’s big book, I have no intention of getting the book or reading it. I just want their “data”.

I’m not linking to the NYT piece, but I will link to the Actuarial Outpost thread on it.

Because I just want to grab this:

You see that 1950 spike? That the highest income folks were supposedly paying 70% of their income out in taxes?


This is the claim:

They have constructed a historical database that tracks the tax payments of households at different points along the income spectrum going back to 1913, when the federal income tax began.

I have a story, from my college days, with regards to constructed data.


When I was in college [back in 1992 or 1993], I joined a math modeling seminar, where we thought we’d be looking at different models, yadda yadda. What we really did was help the professor put together a rebuttal against a paper that had been presented to CITES.

The paper was by a team of Cuban and Japanese scientists/mathematicians, who were promoting a sea turtle population model, to make the argument that they should be allowed to start a sea turtle fishery off the the coast of Cuba.

[The prof told us not to worry too much about the importance of the paper we were putting together — he said that the decision had already been made, it’s totally political, but it would be nice if our paper did support the decision already made. I learned a lot from this project, and institutional politics like this was one of the lessons.]

One of the biggest sticking points of the model was to have any reliable data of sea turtle population, in terms of age, sex, and count.

If I remember correctly, the data they actually had was of turtle count & size of turtles. I didn’t get to see that data (that was one warning sign).

They said they had fitted their model, so that one could infer the age of turtle by its size. I thought that meant they had tagged some turtles, measured their size over a number of years, and then figured out their growth curves.

Nope. That’s not what they did.

There was a table in an appendix of supposed “data” of the turtle populations, which was just a table of numbers. I decided to graph the “data”, and realized that it was a bit too smooth in pattern to be raw data. I took the formulaic model they had… and it fit perfectly. Their “data” was just the output of their model. Oh, and the model hadn’t been fit to much of anything. I dug into their sources, found they had used a fish growth model (which, in case you weren’t aware, is not appropriate for reptile growth, aquatic or not.)

Once I realized all their “data” were simply things that were produced by models they created, I found that sort of behavior all over the paper.

One thing I’ve always been on the lookout for is lack of raw data. Even when it’s not so blatant as to present data that is simply model output, usually you get some form of “massaged” data, if you get any data at all. When people transform data to make it “usable”, too often they transform it using assumptions that are actually the results they want to show.

Then there’s the case of changing definitions of various portions of the data, as well as stitching together data from different sources.

The bizarre (to me) taxation levels of the 1950s are a case in point: I do not believe they are using a constant definition of “income”. I doubt that the top people by income were really paying 70% of their gross income to taxes. 70% after all the exemptions, deductions, etc. in the old tax code? Sure. It was written that way.

But if the definition of income changes over time, there’s not much comparability here.


Tax Foundation makes a graph based off federal income tax stats:

Here is some detail of why this graph does not match the animation above:

There are obviously several large differences between this IRS data and the new data being debated, which comes from a new book by economists Emmanuel Saez and Gabriel Zucman.

First, the IRS only considers the federal individual income tax. The data currently being discussed includes not just the federal individual income tax, but all federal, state, and local taxes. Some of these taxes, such as the corporate tax, are more progressive than the income tax. Other taxes, such as sales and excise taxes, are less progressive.

Second, the IRS data uses adjusted gross income as the income measurement. AGI is a rather narrow income measure that leaves out important sources of income for taxpayers at the bottom of the distribution, such as transfer payments, and for taxpayers at the top of the distribution, such as unrealized capital income.

And finally, the 2017 data does not account for changes made by the new tax law passed at the end of 2017, the Tax Cuts and Jobs Act. Data for 2018 will likely show a drop in average taxes rates across all income levels, but a somewhat larger drop for the highest-income earners.

Except those being bitten by the SALT cap, of course.

There are many things to question in the data created for the book, including their definition of income and whether it includes transfer payments (Social Security payments included? Welfare benefits? EITC?)

But the biggest question I have is the source of the “total tax” number. One can have a well-defined data set, such as federal individual income tax. But if you look at, say, corporate income taxes, which specific people do you allocate those taxes to? All stockholders pro-rata? And how do you appropriately distribute that among income levels?

My big question relates to finding all the non-federal income taxes. Not all state and local taxes are in federal tax returns, for instance. I doubt itemization has remained level throughout income tax history… and SALT itemization definitely isn’t uniform across the states.

I do not have the time… and I don’t even know if their raw, non-transformed data are available. I am not paid to dig through somebody else’s research.

I will just say that if you’re trying to tell me that the 1% paid 70% of their “income” out in taxes back in ye Golden Era, you need to do a lot of work to convince me of that.


And now for something completely different:

You’d better not be taking food pumpkins and decorating them!


Some people did not agree.

Yeah, I do some work thinking for free, but to explain something I really dislike?

Pay me.

Now I need to know if the pumpkin tax was incorporated in the book data.




(He’s referring to the SALT cap)

Well, it definitely leads to rich folks paying for the services of people like me. Thanks!

Pretty sure that’s not enough money.

Picky picky.

You first.

Related Posts
Taxing Tuesday: California Wants All the Money, Chicago Needs All the Money
The Amazon Escape: On Negotiation and Leverage
Taxing Tuesday: DOOOOOOOM and more DOOOOOOOOM