###
**Discussion:**
(26 comments)

Comments are closed.

Mark J. Perry is concurrently a scholar at AEI and a professor of economics and finance at the University of Michigan's Flint campus. He is best known as the creator and editor of the popular economics blog Carpe Diem. At AEI, Perry writes about economic and financial issues for American.com and the AEIdeas blog.

**More options:**
Share,
Mark as favorite

**View related content:**
Carpe Diem

Any day now, the AFL-CIO’s Executive Paywatch division will report another one of its wildly inflated CEO-to-worker pay ratios for 2017 based on a series of flawed statistical assumptions that result in a rather meaningless apples-to-oranges comparison. See my criticisms of the AFL-CIO’s report last year here, here, and here. This year, the *Wall Street Journal* just published CEO compensation online for all of the S&P 500 companies along with several related articles in today’s print and online editions here, here and here. Therefore, this year I’m able to preempt the AFL-CIO and publish a rebuttal to its anticipated massively inflated 300+ CEO-to-worker pay ratio before its analysis is released, instead of responding to their report after it has gone public and gets widely and blindly reported by the media without anyone (except John Merline of IBD) ever questioning “the statistical legerdemain used to produce it” (see below).

As I have reported many times before, here is a summary of the main flaws in the “methodology” used by the AFL-CIO to produce its wildly inflated 300+ CEO-to-worker pay ratios every year:

- The AFL-CIO uses the mean (average) CEO compensation figure each year, which is always much higher than the more realistic and more typical
*median*CEO compensation. For 2017 data from the Wall Street Journal, the mean compensation for S&P 500 CEOs was $13.5 million, which was 13.4% higher than the more realistic median pay of $11.9 million for the typical CEO (based on the actual full WSJ database, which is slightly different from the $12.1 million reported by the WSJ). - They compare total
(base salary plus all additional forms of compensation including cash bonuses, stock awards, option awards, pension benefits, etc.) for executives in their prime earning years managing the world’s largest corporations to**CEO compensation**of all ages working for companies of all sizes (including small grocery stores, independent small restaurants, etc.).**cash-only pay for part-time rank-and-file worker****s**

The analysis below summarizes my corrections for those statistical flaws above that are used by the AFL-CIO annually to get an exaggerated, inflated CEO-to-worker pay ratio:

In its Executive Paywatch release last year, the AFL-CIO reported that the average nonsupervisory rank-and-file worker earned $37,632 in 2016. To arrive at that figure, the AFL-CIO reports that its methodology is as follows:

The average annual income earned by rank-and-file workers is taken from the U.S. Bureau of Labor Statistics Current Employment Statistics survey. Specifically, it is the average hours and earnings of production and nonsupervisory employees on private nonfarm payrolls. The average weekly pay is multiplied by 52.

Here are more details of how the AFL-CIO will calculate average worker pay this year if it follows its procedure in previous years. It will use an average hourly wage of $22.05 for production and nonsupervisory workers in 2017 (BLS data here) and an * average workweek of only 33.7 hours *(BLS data here) for the average rank-and-file worker. Assuming 52 weeks of work per year: $22.05 per hour x 33.7 hours per week x 52 weeks ≈ $38,640 average annual cash-only earnings for the typical

Therefore, the AFL-CIO engages in some questionable statistical chicanery by sloppily reporting an apples-to-oranges comparison of: a) total CEO compensation for only about 500 CEOs working full-time in their prime earning years to b) the cash wages only for 1o2.5 million rank-and-file workers of all ages working for companies of all sizes, who * work an average of less than 34 hours per week*, and are therefore

For 2017, the CEO-to-worker pay using the AFL-CIO’s methodology would be about 349-to-1 ($13,472,000 ÷ $38,640) comparing average CEO compensation to cash wages for part-time workers, see top chart above.

**Questions**: a) How would the AFL-CIO’s CEO-to-worker pay ratio change if we: a) calculate average worker pay for *full-time workers*, b) compare the average pay for a rank-and-file worker who works the same number of hours that a typical CEO works, e.g., 50 or 60 hours per week, c) compare** total compensation of both CEOs and rank-and-file workers working full-time including fringe benefits for the rank-and-file,** d) use median instead of average CEO compensation, and e) consider only those workers who are employed by companies with 500 or more workers to consider the workers most likely to actually work for an S&P 500 company? The charts above summarizes how the CEO-to-worker pay ratio would change, here are the details:

**1. Correcting for the AFL-CIO’s Statistical Shortcomings/Legerdemain**

a. Using median CEO compensation in 2017 of $11,900,000 would lower the CEO-to-worker pay from 349-to-1 to 308-to-1, a reduction in the ratio of nearly 12% (see bottom chart above).

b. Assuming a 40-hour workweek for full-time rank-and-file workers employed by companies with 500 or more workers, who received average hourly compensation of $36.65 last year including fringe benefits (BLS data here), and annual total compensation of $73,300, we would get **CEO-to-worker compensation ratios of 177-to-1 **u**sing average CEO compensation and 156-to-1 using median CEO compensation.** In other words, the AFL-CIO’s CEO-to-worker pay ratio is conveniently inflated by a factor of 2X by using cash wages only for part-time workers! Talk about bogus statistical chicanery! And year after year, the media never questions the AFL-CIO’s totally flawed methodology and parrots the inflated and wildly exaggerated CEO-to-worker pay ratios.

c. Since no S&P 500 CEO works only 40 hours per week, let’s assume a 50-hour workweek for full-time rank-and-file workers to make a more realistic comparison, and use hourly compensation of $36.65 like above and annual total compensation of $95,290 for workers employed by companies with 500 or more employees. That would result in **CEO-to-worker compensation ratios of 141-to-1 (average CEO pay) and 125-to-1 (median CEO pay)**, see charts above.

d. To make it an even more realistic comparison to the average workweek of an S&P 500 CEO, let’s assume a 60-hour workweek for full-time rank-and-file workers and annual total compensation of $114,348 we would get **CEO-to-worker compensation ratios of 118-to-1 and 104-to-1 for average and median CEO pay respectively.
**

**Conclusion**: By considering both total compensation (including fringe benefits) for both CEOs (and using median CEO pay) and rank-and-file workers, and by considering longer workweeks for rank-and-file workers that would be more comparable to the average hours worked by a CEO of a major multi-national corporation in the S&P 500, we can get a more accurate apples-to-apples comparison. Those more accurate comparisons result in CEO-to-worker compensation ratios of between 177-to-1 (for rank-and-file workers averaging 40-hour workweeks vs. average CEO pay, see top chart) to as low as 104-to-1 for rank-and-file workers putting in 60-hour weeks that might be the most comparable to the average workweek of a top CEO and using median CEO pay.

Let’s keep this analysis in mind in the coming weeks when the AFL-CIO’s reports something like another 350-to-1 CEO-to-worker pay ratio that will generate sensationalized media coverage even though it is hugely exaggerated by a factor of at least 2X and probably more realistically inflated by an AFL-CIO-to-reality ratio of 3.5-to-1!

**2. Confiscation and Redistribution of CEO Pay**. And what’s the whole point of the AFL-CIO’s annual reports on wildly inflated CEO-to-worker pay ratio? The sub-title of last year’s AFL-CIO’s Executive Paywatch report pretty much sums it up: “**More for Them, Less for Us.” **

The AFL-CIO’s message seems to be that if CEOs weren’t being so generously over-compensated, then the rank-and-file workers would be doing much better and making higher wages. For example, according to the AFL-CIO in 2015:

America is supposed to be the land of opportunity, a country where hard work and playing by the rules would provide working families a middle-class standard of living.

But in recent decades, corporate CEOs have been taking a greater share of the economic pie while workers’ wages have stagnated.

The AFL-CIO has fallen here hook, line and sinker for the **zero-sum, fixed pie fallacy**, one of the most common economic mistakes that falsely assumes that there’s a static fixed pie and therefore one party can gain (get a bigger slice) only at the expense of another (who’s left with a smaller slice). But let’s assume that there is a “fixed pie of wages” and do some confiscation and redistribution of CEO compensation to see how that would affect average rank-and-file worker pay.

**Q**: If the CEOs of the S&P 500 companies received $13.47 million on average last year, then as a group, those 503 CEOs in the WSJ database received about $6.77 billion in total compensation in 2017. If the AFL-CIO could wave a magic wand and confiscate that entire amount and redistribute $6.77 billion to the current 103,457,000 rank-and-file workers, what would each one get?

**A**: An annual increase in pay of about $66 for each rank-and-file worker before taxes, or about $1.27 more per week and only 3.2 cents per hour.** In other words, complete confiscation and redistribution of S&P 500 CEO compensation would make almost no difference for the average rank-and-file worker.**

**Bottom Line**: The AFL-CIO can only get a distorted and wildly inflated CEO-to-worker pay ratio of something like 350-to-1 every year with a flawed apples-to-oranges analysis that compares the **total annual compensation** **of a small, select group of CEOs in their prime earning years heading America’s largest multi-national corporations**, who probably typically work 50-60 hours per week or more, to the average **annual cash wages of part-time rank-and-file employees** who work less than 34 hours per week on average. Once we make a more statistically valid apples-to-apples comparison, the CEO-to-worker compensation ratio falls by 50% from the AFL-CIO’s expected 350-to-1 ratio to 177-to-1 once we consider total compensation for both CEOs and full-time (40 hours per week) rank-and-file workers employed by companies with 500 or more workers. If we assume a 60-hour work week for the average worker (to be comparable to the workweek of an average CEO), the CEO-to-worker compensation ratio falls by two-thirds to only 118-to-1 for average CEO pay and 104-to-1 for median CEO pay. Further, even if we could confiscate 100% of the compensation of all S&P 500 CEOs, the typical rank-and-file worker would probably get about $1 per week in additional after-tax earnings. Big deal.

Just like last year, the CEO-to-worker pay ratio reported by the AFL-CIO gets my annual “**Biggest Blindly Accepted Statistical Fairy Tale of the Year Award**.” Well no, it’s actually a tie with the **gender wage gap myth** and the incessantly repeated “**77 cents on the dollar**” statistical falsehood. What’s disappointing is that much of the mainstream media seem to blindly accept both of these statistical falsehoods without ever challenging the “statistical legerdemain” that are used to produce and perpetuate these statistical myths. One exception was this excellent article in 2015 by *Investor’s Business Daily*’s John Merline, (“Do CEOs Make 300 Times What Workers Get? Not Even Close“) who concluded:

What’s not understandable is why the mainstream press keeps repeating the massively inflated 300-to-1 number without noting the statistical legerdemain that produced it.

Comments are closed.

AEI

1789 Massachusetts Avenue, NW, Washington, DC 20036

© 2018 American Enterprise Institute

- Samuel J. Abrams (11)
- Joseph Antos (334)
- Leon Aron (280)
- John P. Bailey (23)
- Claude Barfield (371)
- Michael Barone (758)
- Robert J. Barro (8)
- Roger Bate (547)
- Eric J. Belasco (5)
- Andrew G. Biggs (457)
- Edward Blum (63)
- Dan Blumenthal (275)
- Babette Boliek (39)
- John R. Bolton (630)
- Andrew Bowen (16)
- Karlyn Bowman (1910)
- Alex Brill (226)
- Arthur C. Brooks (593)
- James C. Capretta (251)
- Timothy P. Carney (367)
- Walton Chaney (4)
- Lynne V. Cheney (39)
- Zack Cooper (4)
- Jay Cost (3)
- Sarah Crain (1)
- Jason D. Delisle (40)
- Sadanand Dhume (312)
- Robert Doar (158)
- Thomas Donnelly (467)
- Colin Dueck (11)
- Mackenzie Eaglen (203)
- Nicholas Eberstadt (400)
- Jeffrey Eisenach (115)
- Kate Farrar (2)
- Lynn M. Fisher (2)
- Clay R. Fuller (34)
- Joseph B. Fuller (15)
- R. Richard Geddes (51)
- Joseph W. Glauber (7)
- Jonah Goldberg (630)
- Barry K. Goodwin (14)
- Phil Gramm (21)
- Eric A. Hanushek (10)
- Robert B. Helms (109)
- Frederick M. Hess (834)
- R. Glenn Hubbard (111)
- Gus Hurwitz (61)
- Benedic N. Ippolito (23)
- Mark Jamison (85)
- Matt Jensen (48)
- Frederick W. Kagan (298)
- Leon R. Kass (55)
- Paul H. Kupiec (97)
- Jon Kyl (21)
- Desmond Lachman (904)
- Roslyn Layton (121)
- Adam Lerrick (47)
- Phillip Lohaus (50)
- Daniel Lyons (77)
- Kirsten D. Madison (7)
- Nat Malkus (97)
- Oriana Skylar Mastro (37)
- Aparna Mathur (287)
- Michael Mazza (156)
- Rachel M. McCleary (2)
- Michael Q. McShane (1)
- Bruce D. Meyer (7)
- John W. Miller (3)
- Thomas P. Miller (341)
- Charles Murray (232)
- Roger F. Noriega (398)
- Stephen D. Oliner (107)
- Ed Olsen (3)
- Norman J. Ornstein (665)
- Brent Orrell (2)
- Anthony Paranzino (1)
- Mark J. Perry (8145)
- James Pethokoukis (2756)
- Edward J. Pinto (295)
- Danielle Pletka (815)
- AEI Events Podcast (15)
- Kenneth Pollack (26)
- Ramesh Ponnuru (332)
- Angela Rachidi (115)
- Vincent R. Reinhart (93)
- Gerard Robinson (99)
- Dalibor Rohac (165)
- Michael Rubin (1125)
- Sally Satel (320)
- Naomi Schaefer Riley (36)
- Gary J. Schmitt (437)
- Mark Schneider (69)
- Derek Scissors (236)
- Neena Shenai (1)
- Sita Nataraj Slavov (34)
- Andy Smarick (123)
- Vincent H. Smith (121)
- Christina Hoff Sommers (189)
- Katharine B. Stevens (59)
- Michael R. Strain (278)
- Ryan Streeter (54)
- Phillip Swagel (68)
- Bret Swanson (76)
- Jim Talent (42)
- Shane Tews (63)
- Marc A. Thiessen (981)
- Sean Trende (4)
- Jessica Trisko Darden (20)
- Stan Veuger (244)
- Alan D. Viard (242)
- Peter J. Wallison (686)
- W. Bradford Wilcox (100)
- Paul Wolfowitz (140)
- Di Xu (1)
- John Yoo (226)
- Paul Zajac (9)
- Roger I. Zakheim (19)
- Weifeng Zhong (30)
- Katherine Zimmerman (150)
- Benjamin Zycher (118)

Whether the ratio is 300x or 100x doesn’t really matter. The reality is that these CEOs, many of them useless, are getting paid a truckload more than the average fella.

Is it right? Should it matter? Would the cure be worse than the disease?

All pertinent questions.

Share

And the average NBA player makes many, many truckloads of cash more than the average ticket-taker or peanut sales person at the arena, just like Oprah makes a lot more than her accountant and limo driver, and Michael Moore makes a lot more than the servers at the restaurants he eats at, etc.

Share

I don’t disagree. It’s even more starc for Michael Moore who portends to be a battler and then makes millions and lecturers everyone else on our moral failings.

Only other observation is that the working man – rightly or wrongly – is quite comfortable with sport stars etc making good dollars. Not so true for businessmen who – and again, rightly or wrongly – are viewed suspiciously as getting rich by engorging themselves at the expense of the common man.

I just think, does it matter if someone drives a fancier car than me or live in a bigger house? Not really – it doesn’t affect my wellbeing. And I know the cure of equality makes us all worse off – so not fussed.

Share

You know, if my boss made less money, I’d be better off. I think.

Share

If your boss made less money then he’d likely not be as good a boss as he is, making the company you work for less competitive, making everyone working there less well off … including you.

If you want to be paid more than you are, you need to make yourself worth more than you are.

Reality is not optional.

Share

Diane

I agree with everything you wrote, but I think Sam was being sarcastic.

Share

Why not compare CEO compensation (any method you want) over time compared to the CPI or other notable items as we often see here for things such as college textbooks or medical procedures?

Share

What would you expect such a comparison to show, Walt?

Share

Ron, I have no preconceived notion or bias about what it would show. Let the data and charting speak for themselves — that’s research.

Maybe it’s more telling that the charting and comparison methods commonly used for textbooks and other times such as medical procedures used here isn’t shown likewise here for CEO compensation percentage increase over time than what it would show?

Try this comparison percentage increase for example: all U.S. CEO median compensation, CPI, all median U.S. non-salary employees full-time compensation for years 1978-2018.

Share

Walt

Your first and second paragraphs are in conflict. Without any preconceived notion of what a comparison of CEO salaries to CPI would show, why on earth would anyone choose to make that particular comparison?

Why is it telling that Dr. Perry hasn’t made such a comparison? There are literally millions, if not billions of things for which the good professor hasn’t made a comparison to CPI over time.

“

Try this comparison percentage increase for example: all U.S. CEO median compensation, CPI, all median U.S. non-salary employees full-time compensation for years 1978-2018.”Have you done such a comparison? If so, what was your result, and what is the significance (if any) of that result?

Share

“Without any preconceived notion of what a comparison of CEO salaries to CPI would show, why on earth would anyone choose to make that particular comparison?”Ron, do you normally research to confirm what you already think? I don’t need answers before the work is done, do you?

Analyzing how almost anything compares in percentage increase to the CPI or other like items over time is a fairly common topic of discussion in economics (not just here). Is failing to present information in a format that it is very often presented a Legerdemain, too?

I would add another line to the chart I proposed earlier — yearly revenues (to answer the question: Is CEO pay possibly linked to the amount of revenue they control?)

Share

You have avoided answering the question as is usual with you, Walt, and you are conflating the terms research, confirmation, and falsification.

Share

Why not compare the S&P 500 CEOs with the top 500 athletes or entertainers over time?

Then compare the growth of responsibilities over time between CEOs and athletes over time.

Share

Cit

I suspect that the amounts paid to both CEOs and athletes is based on their contribution to the bottom line of the company they work for. Supporting and improving that bottom line may be their one and only primary responsibility.

Share

Rkn, I am going to give you a partially correct for your reply.

Athletes are not planning for future organizational success with strategic allocation of assets. But yes, Dodger legend will be enhanced with Clayton Kershaw’s amazing pitching.

Share

Cit

Only partially? 🙁

Share

“Why not compare CEO compensation (any method you want) over time compared to the CPI or other notable items…I’d suppose it’s because he’s making an argument based on what the AFL-CIO is comparing CEO compensation to and the AFL-CIO is not comparing CEO compensation to the CPI or other notable items.

Share

Re: “The AFL-CIO uses the mean (average) CEO compensation figure each year, which is always much higher than the more realistic and more typical median CEO compensation.”

Since pay is a positive number and we are interested in ratios, it seems that the geometric mean would be the natural way to arrive at an “average”, better than either the arithmetic mean or the median.

Share

Anytime you have outliers on the high end, e.g., home prices or household income, you always use the median home price or median household income as a better measure of a “typical” home price or household income.

If there are four average income households (or home prices) in a sample, and you add Oprah’s income (or home price), you get a much better idea of typical household income or typical home price using median vs. mean.

Likewise, median CEO pay is a much better measure of a typical SP 500 CEO’s pay vs. mean. For 2017, CEO Hock Tan of Broadcom’s pay of $103.2 million is an outlier and artificially raises the mean, since the next highest paid CEO is $69.3M. It’s just an accepted statistical truth that median is more representative than mean when there are outliers on the high end.

Share

Yes, but the geometric mean might do an even better job than the median, because it changes a little whenever one of the salaries changes a little. And the way it changes is understandable: if we increase one of the salaries by a small percentage r, then the geometric mean increases approximately by a percentage r/n, where n is the number of earners.

For example:

Salaries of 1, 10, 100, 1000, 5000: median 100, geometric mean 87.

Salaries of 1, 10, 100, 1000, 10000: median 100, geometric mean 100.

Share

Geometric mean would be fine, and in fact, would lower the AFL-CIO’s CEO-to-worker pay ratio from 349-to-1 (arithmetic mean) to 284-to-1 (geometric mean). The problem is that geometric mean is somewhat obscure and isn’t used as much as arithmetric mean and median, which everybody is familiar with and understands. If you depart from using the standard mean/median measures and introduce what is a new measure of central tendency for most people, it might be looked at with suspicion. The average person can understand why we use median home prices and median household income, but would likely glaze over if a new, unfamiliar measure (geometric mean) was introduced.

Share

I imagine your readers are above average. One could explain it by saying that it averages the number of digits in the salaries.

Share

The problem is that I can’t find a single example of anybody else using geometric mean instead of median (or arithmetic mean), except in some cases for annual returns over time to account for compounding. In fact, if you do a Google search for “geometric mean” the first result is your comment on this blog post! So I don’t mind being a trail blazer/innovator, but I’m sure I would confuse half the readers, and arouse suspicion among the other half that I was doing something irregular to advance a position. That’s why I usually don’t even show graphs using a log scale, because I’m certain that logarithm scales would confuse many/most readers.

Share

You may be right. I wonder how we can move it into the main stream.

Share

Here’s the first hit google gives me: https://en.wikipedia.org/wiki/Geometric_mean

Share

This is the BLS page for the Chief Executive category May 2017

https://www.bls.gov/oes/current/oes111011.htm

There are over 210,000 positions in this group with mean annual wage of $196,000 and median annual wage of $183,000 which means with your maximum adjustments included the the real apples to apples ratio gets down to under 2:1

Share