David R. Henderson  

Landsburg on Median Income

PRINT
Evan Soltas on the Great Facto... John Taylor vs. Ryan Avent...

Steve Landsburg has an excellent post today showing that to know what has happened to incomes of various groups, you have to look behind the median. Of course, anyone who has used statistics carefully knows this, but it's amazing how many people, even economists, don't bother to do it. Steve draws on a recent book by Edward Conard, Unintended Consequences: Why Everything You've Been Told About the Economy is Wrong. It's sitting on my review pile but I haven't got to it yet. Now that I've seen this excerpt, it's higher up in my pile.

Conard points out that between 1980 and 2005, median incomes for non-white women, white women, non-white men, and white men all rose substantially, by 62%, 75%, 16%, and 15% respectively (all in real terms.) That's every adult, right? So how did the median income rise by only 3%? Answer: the composition changed. So white women and non-white women, who started out low, went up a lot, but there was a big of influx of them into the labor market during that time.

One of Steve's commenters says, in essence, "big deal." Specifically, he wrote:

As we both know 31% growth over 25 year [is] about 1% growth per annum (.10859628) whereas the real GDP has grown 2.4% per annum (.024791764), the GINI coefficient, even adjusting for government transfers, has grown accordingly.

What's interesting here is that this commenter knew enough to do compounding and knew enough to know what the Gini coefficient is, but didn't do one basic correction of the GDP growth. He/she didn't look at GDP growth per capita. Between 1980 and 2005, the U.S. population grew from 227 million to 296 million, a growth of 30.4%. That works out to an annual population growth of 1.07%. Of course, with a higher percent of the population working (which is what led to that divergence between median income growth and growth of all the "subset" income groups in the first place), I would be over-adjusting by looking at strict population growth. But you get the point, I'm sure.


Comments and Sharing





COMMENTS (19 to date)
Daniel Kuehn writes:

This is great stuff. People often forget to look at compositional changes in discussing wage cyclicality too, unfortunately.

I am still kind of left with the feeling of "so what?", I think. Certainly growth in female labor supply is a great story that I think people are widely aware of - there's an enormous literature around that. The racial picture doesn't look quite as good here because we start in 1980, but if you started in 1960 you'd see some good news there too.

But you still have the typical worker with relatively stagnant wage prospects, and families that enjoy improving family incomes do so by supplying twice the labor (both parents) that they had to previously.

I guess I'm just saying both stories are important, and we ought not to deny or obscure either.

John Thacker writes:
But you still have the typical worker with relatively stagnant wage prospects

But you still have the typical actual person with relatively good income growth. The "typical worker" is a myth here, especially because of immigration.

This is a separate issue from family incomes, where certainly some of the extra cash income goes towards outsourcing things previously provided by the inhabitants. Families would enjoy improving family cash incomes even if only one person worked, if you measure by people who were already in the country.

Daniel Kuehn, your method of counting seems to argue that the US would be better off if the borders were closed, and people weren't allowed to better themselves by coming to this country. I don't think you really believe it, but focusing on the US population (instead of world population, or change in incomes among people already here) biases the data.

David R. Henderson writes:

@Daniel Kuehn,
This is great stuff. People often forget to look at compositional changes in discussing wage cyclicality too, unfortunately.
Thanks. And, by the way, I should have noted in my post above that I was one of those careless economists who forgot to look at compositional effects.
But you still have the typical worker with relatively stagnant wage prospects, and families that enjoy improving family incomes do so by supplying twice the labor (both parents) that they had to previously.
Of course, these data, which, by their very nature, must be retrospective, can’t look at prospects. That’s the big debate about the future. And the future is hard to know.
But the data above are on individual worker income, not family income, and individual median income for the 4 subgroups increased substantially.

Jim Glass writes:

My favorite illustrative example along this line was the factoid for Mexico some years ago that while every single person alive in the country grew older (obviously) their average age became younger.

Jim Glass writes:

But you still have the typical worker with relatively stagnant wage prospects,

I'm not sure that looking at the changes in average hourly compensation (even in the aggregate) tells us that.

http://research.stlouisfed.org/fred2/series/COMPRNFB

If there was a time when hourly compensation was stagnant, it looks like it was during the Reagan/Bush I/early Clinton years. Since then it's picked up a lot.

and families that enjoy improving family incomes do so by supplying twice the labor (both parents) that they had to previously.

Oh, I remember when the complaint was that women were being "denied their right" to work. Now it's that they are being "forced" to.

Ryan V writes:

"Of course, with a higher percent of the population working (. . .), I would be over-adjusting by looking at strict population growth. But you get the point, I'm sure."

There's actually a smaller percent of the population working now (58.6%) than 1980 (60.0%). But that's probably the recession, since workers/capita was higher in 2005 (62.8%). (Source)

But anyway, using the 1980 to 2005 numbers, wouldn't you be under-adjusting by using just population growth? Presumably we want to compare median wages to GDP/worker. And GDP/worker will increase more slowly than GDP/capita if the number of workers grows faster than the number of people.

So if I'm doing the math right (1.845 GDP change / 1.365 change in # of workers), that's a 35% increase in GDP/worker. Almost bang on the 31% increase in adjusted median wage.

Daniel Kuehn writes:

John -
I have no idea what you're talking about. Thinking about both national and world economies in no way implies the borders should be closed, and in thinking about immigration policy the content of one blog comment not addressing immigration policy is obviously not going to be exhaustive of what should be considered. Why are you trying to (inaccurately) reconstruct an immigration policy from what I wrote???

David -
That's a good point, and I really don't want to overstate mine. Women are improving because of secular trends - increased education, moving out of the home and into occupations that were not open to them before.

But white men grew 15% over this period, and non-white men a little more than that. Now if I'm doing my compound annual growth rates right, that's half a percent annual growth in real wages. Does anyone think white men's productivity grew that slowly over this period?

I'm just saying even if we chuck those stark "no real wage growth" stories as being misleading, something is wrong here.

Joe Cushing writes:

The people who don't look are those who's favorite story is supported by the median income. Many people love to play victim and to say that things are not as good as they used to be. This is where the phrase "good ole days" came from

Tony N writes:

I’m probably the most economically illiterate reader of this blog, so I ask the following with a genuine suspicion that it is a stupid question:

Should one necessarily expect a commensurate rise in wages with increased productivity? Can’t the increase in a employee’s productivity be the result of efforts beyond the his own, such as investments whose costs were incurred by the employer?

It seems likely true that the assembly-line worker at Ford Motors 2012 is much more productive than his Ford Motors 1912 counterpart. And it seems likely true that the assembly-line worker of 2012 had a much more difficult and hazardous job. Doesn’t it stand to reason that there is a larger supply of people capable of doing the job today than there were in 1912?

The ditch digger equipped with a mattock is far less productive than the ditch digger equipped with a jackhammer. Must the latter nevertheless be paid better than the former?

Again, I’m probably missing something here.

I'll second the praise for Conard's book. There are some clunky sections to it, but it's a good antidote to the sophistries of Joe Stiglitz. Especially his argument that investors don't enjoy the bulk of the benefits from their capital; consumers and workers who have more tools with which to improve their wages do.

He also has a very interesting take on AAA rated MBS and how the ratings agencies made a rational decision in so rating them.

David R. Henderson writes:

@Tony N,
I’m probably the most economically illiterate reader of this blog,
Probably not.
I ask the following with a genuine suspicion that it is a stupid question
Your suspicion, while genuine, is not justified. That’s a good question. For an example of a stupid question, consider this one, that a woman once asked at a Weight Watchers meeting I attended in the late 1980s, after we were told that a can of some food had 120 calories: “How many calories are there in half a can?”
Should one necessarily expect a commensurate rise in wages with increased productivity?
No, one shouldn’t. It depends, as your instincts tell you, on how productivity increases.
Can’t the increase in a employee’s productivity be the result of efforts beyond the his own, such as investments whose costs were incurred by the employer?
Yes, it can. Interestingly, though, increases in capital often increase the productivity of labor and so we would expect in those cases, which are common, for wage rates to rise also.

Tony N writes:

@David R Henderson,

Thanks for responding. I thoroughly enjoyed the post.

joshua writes:

Every group performed better than the average! This has also happened with average Texas school scores getting worse. I believe it is known as Simpson's paradox

Lauren writes:

Thanks, joshua, for your observation. Though the way you put it, it was reminiscent of Garrison Keillor's famously charming satire about Lake Wobegon, where all the children are above average.

Simpson's paradox--that's Edward Simpson, not Homer Simpson for you other humorists out there--is very common when studying international economic data. It is common for data collected within a single country consistently to show one direction of correlation no matter which country is examined individually; and yet when the same regression is run across the identical group of countries' joint data--aggregated across multiple countries--the results show a completely opposite correlation. The differences can go from positive to negative, or vice versa. It typically means that the data are not commensurately collected.

It's easy to think that more data means better results. But that's not the case when the aggregated data are inconsistently collected, ranked, or weighted across different groups. Not to mention different possible cultural matters. It sometimes helps to put in a statistical 0-1 variable by nationality/data-collection-group. That doesn't always solve the whole of the statistical problem.

I am not sure that Texas scores' getting worse is an illustration of Edward Simpson's paradox. I do not think that scores that were collected and grouped in one way within Texas were then regrouped into a larger pool without taking differing collection matters into account. Of course, it could be. Maybe within Texas, polls were done in different districts or schools that showed one kind of correlation, and then aggregated together in a way that reversed or contradicted the individual districts's experiences. Let me know if you think that happened.

joshua writes:

I may have slightly misremembered the issue with Texas (I didn't try to look it up on my phone yesterday). Here is one article about it. It says that Texas overall test scores were equal to the national average even though every demographic beat the national average for that demographic... Texas just had a disproportionate number of lower-scoring Hispanics.

The statistical details may be different, but I generally understand Simpson's paradox to be when every group of a set performs better than some other set's groups, but the average doesn't look as impressive because it has larger lower-performing groups - whether we're comparing Texas test scores to the nation, or modern incomes compared to past incomes.

Michael Rulle writes:

@David Henderson

I also read Tony N's questions with great interest. I was able to answer each of them easily except the question on why workers "must" be paid more when increased productivity is caused by better technology. I was initially not able to answer that question (so to Tony N, far from stupid---not that my inability to answer it demonstrates that!). I also am not sure you, David, fully answered it either. When I read your answer, it was a statement not an explanation. Upon reflection, I believe the explanation is that the price of labor is bid up in a competitive marketplace as productivity increases. It is bid up because the market will "not permit" excess profits to exist just because productivity rose. Competition comes in to lower profit, by raising the prices of inputs (including Labor) and lowering the price of output.

At least I think that is correct.

steve writes:

I have doubts about Conard's data. I think that the data actually show that wages for men have deteriorated over this time period.

http://www.hamiltonproject.org/files/downloads_and_links/07_milken_greenstone_looney.pdf

Tony N writes:

@ Michael Rulle

Thank you too for addressing my typo-riddled question.

gic writes:

I was the poster on Steve's site and you are right that I didn't add the increase in working age population which data is readily available at FRED.

I just checked: working age population increased 26.86% in this period, this means that even assuming it is necessary to throw this in and the adjustment made is the obvious one (and I am not sure this is the right one to make, I will think on this) , the average person has done at least .4% a year worse than GDP growth, which when compounded for 25 years also helps show why the GINI coefficient continues to worsen.

I think Conard's numbers and similar attempts to obfuscate do everybody a disservice. The fact is that the average person has done worse in the last 25 years and when you look at the top 1%, ..1% and .01% each have done *much* better than the previous group has done and all have done much better than the 99%.

Some may argue that this is well deserved but at least lets not obfuscate that it happened.

Comments for this entry have been closed
Return to top