The rumor about "stagnant middle class incomes" is false

[10/10/08: See addendum 2, Real Income per Earner...]

[See the addendum at the bottom; added 10/8/08, with new information, and comments about the Obama and McCain campaigns. --Steve]

Rumor has it that middle class incomes stagnated or declined starting around the year 2000. A lot of politicians have been repeating that rumor every chance they get.

Fortunately, however, the rumor is false.

By the way, I've noticed that pundits on the left and right are getting bolder about using the "L" word (..."lie"). The Obama sycophant Keith Olberman asked his idol in a recent interview Why shouldn't it be appropriate just to start calling the opposition's campaign rhetoric "lies." Not to be outdone, McCain supporters flooded the internet with Joe Biden's "fourteen lies" shortly after the VP debate. People are getting less skittish about using the L-word — maybe as a consequence of Al Franken's book from a few years ago.

So, do the mantras about "stagnant middle class incomes" and "only the rich benefited from the economy's growth" deserve a stronger label than just "false rumor"? I don't think so; I think ignorance, belief preservation, and confirmation bias are more likely explanations than dishonesty, so I won't jump on the L-word bandwagon.

I'll begin with a chart of the official numbers that supposedly prove the rumor about the middle class's plight: The "real median household income" from the U.S. Census Bureau (adjusted to 2006 dollars using the CPI-U-RS).

Median household income since 2001 looks stagnant at best, doesn't it? Ouch! Something must have been really wrong with a growing economy that left the median household out of all that growth, don't you think? No wonder we hear so much about it from our politicians.

Hold on though; don't stop thinking just yet — even though many of them would like us to do just that. Try to think of one company — just one, large or small — that has ever written a payroll check to a "household." For example, has Microsoft Corp. ever written a payroll check to "the occupants of the house at 2345 Main Street, Redmond, Washington"?

No, of course not.

A household has a group of people in it; most of those groups contain at least one specific person who earns a paycheck. The U.S. Census Bureau (in their "Current Population Survey") calls those people "Earners." The amount of money income received by a household depends to a great degree on the number of "earners" in that household. (See their household datasets at this page.)

So the question about stagnated incomes is really a multi-part question. Here is the better question, and its multi-part answers:

Did household income stagnate or decline for households with no earners at all? YES. How about for households that had a decline in the number of earners? YES. How about those that had the same number of earners? NO. How about those that had an increase in the number of earners? NO.

By now it should be obvious that an even better question is: Did the middle class income earner participate in the overall economy's growth? It's a better question because it removes the confusion caused by differences in the number of earners per household.

So let's take a look at how "income per earner" did for each of the quintiles of household income. [For reasons I explain in the end note, I had to limit the analysis to 4 out of 5 quintiles, and 9 out of 10 deciles, of household income.]

The chart below shows the result for the period 1994-2007. Note that any possible definition of the "middle class" would show that middle class earners' incomes did not stagnate or decline. In fact, they grew in tandem with the 3.2% average growth rate of overall disposable income per capita (a derivative of GDP).

[This is a corrected chart; the original one contained a math error in the bottom quintile.]

Next is a chart that makes it easier for you to create your own definition of "middle class" if you're so inclined. It's the decile version of household incomes (9 of 10 deciles, to be precise). It won't matter which deciles you pick for "middle class": income per earner grew at pretty much the same rate for all deciles.

Last, we need to address the assertion that "the middle class is disappearing." Sounds like a bad thing, doesn't it: rich get richer, and so forth. So, I asked myself the question, Where is the middle class going? Once again using the Census Bureau's data, I think I figured it out: The middle class of yesteryear is indeed disappearing: it is getting squeezed, like toothpaste from a tube, into the income category labeled "greater than $100,000." Here's a crude movie depicting the squeezing of the middle class — into the unbounded top category of income. (Watch the number of earners in the far-right category grow over time.)

Gee, maybe the "disappearing" middle class of yesteryear isn't such a bad thing after all, huh? Do you think our politicians will be relieved to find out that the rumors about the middle class's stagnation and disappearance are false -- because the historical middle class is simply moving to higher and higher income levels? I hope so, but I won't hold my breath.

The lesson is that economic growth has not benefited only the rich; all classes, however one chooses to define them, have participated.  Now all we have to do is get the economy back on a growth track — after we dispose of today's more pressing problem of how to minimize the depth and duration of the recession.

=======================
End note about quintiles and deciles...

The bottom 4 quintiles of household income are easy to analyze, but the 5th (top) quintile is not. The Census Bureau posts forty-one categories of household income; forty of them are $2500 wide, and the last one is of unlimited width. For example, the bottom category is for households with annual money income in the range "0-$2499"; the next one is for the range "$2500-4999", and so on up to the top category, which is labeled "$100000 or more." Each quintile has the same number of households in it (as does each decile), but only four out of five quintiles (or nine out of ten deciles) have easy-to-calculate, weighted-average income per household, and income per earner. That's why the charts don't show the highest one.

Also of interest to some: These income numbers are "money income"; they do not include the value of in-kind transfers, such as food stamps, school lunches, and other programs.

Lastly, my inspiration for diving into the Census Bureau's numbers came from Alan Reynolds's excellent book, Income and Wealth.

*-*-*-*-*-*-*
Addendum:
Additional information that helps clarify income distribution among household quintiles. Plus, my editorial comments, concluding with thoughts on the Obama and McCain campaigns.



Note the inferences on the two additional charts above.  Is it any wonder that making the "household" the basis for evaluating income distribution automatically increases the difference between rich v. middle class v. poor -- as opposed to controlling for earners per household or persons per household?

Said another way: Why should it be a surprise to anyone that a nearly-empty household is likely to have a lower "household income"? Answer: It shouldn't be a surprise. Not only that, it shouldn't be buried as an explanation for the distribution of household income. But that's what is happening.

Oversimplified campaign rhetoric implies that rich households got the benefit of the economy's growth, and non-rich households didn't get any of it; therefore, the market failed us, and we need the government (which we can count on not to fail) to jump in and make the income distribution fairer.

But wait a minute: Did rich workers get all the income growth, while middle class and poor workers stagnated? No, the growth was distributed evenly among those who worked, regardless of their household quintile.

Did full-time workers have higher incomes than part-time workers? Duh. Did full-year workers have higher incomes than part-year workers? Duh. Did two-income households have higher incomes than one-income households? Duh.

Did "failed economic policy" cause all of the difference between household incomes? Absolutely not. Did it cause middle and lower quintiles to "miss out" on the benefits of growth? Absolutely not. Did it cause any difference whatsoever in the way growth was distributed to workers in the five quintiles? Apparently not.

But do we ever hear anything besides "failed economic policy" as the explanation for the income differences among household quintiles? Absolutely not.

Conclusion:
I think we definitely have a fairness issue: we need "fairer" political discourse, instead of oversimplified rhetoric designed to fuel the politically-lucrative class-warfare propaganda machine.

The message should be this:

Our economic record clearly demonstrates that you will get more than your so-called fair share of the nation's economic growth if you get an education, keep educating yourself, work full time, work all year, and stay married -- regardless whether you're rich, middle class, or poor. 

But that's not the message we hear, is it? Instead, we hear, "Most of you got screwed by 'trickle down' economics; you should elect me, because I'll see to it that you get your fair share."

Do I expect the Obama campaign to soften the economic class-warfare rhetoric and start talking more sensibly about the successful track record of growth economics? No, not this close to the election; the class warfare message is working too well — because it's simple, and very effective at stirring up emotions. Do I expect the McCain campaign to challenge it? No, I don't think he gets it. Besides, he apparently would rather spew the Concord Coalition balderdash about the national debt, as his campaign descends into the abyss.

My hope is that rational analysis can have a small, positive effect on the Obama administration's approach to growth economics. Let's face it: some of Obama's proposals are supply-side economics, believe it or not. He can't admit it, because that label has already been bastardized for political purposes, but supply side is nonetheless the proper label. I'll explain that in a future article.

There, I've said my piece. My editorial commentary is now finished.
*-*-*-*-*-*-*

*-*-*-*-*-*
ADDENDUM 2: *REAL* INCOME PER EARNER

The readers have spoken: We need to see real income per earner by quintile and by decile, for an easier comparison with real median household income.

As I said before, the only thing that changes is the Y-axis scale. Adjusting all numbers for the same amount of inflation changes neither the shape of the trend lines nor their relative positions.

Here are the two new charts (quintiles and deciles). All groups shown:
• beat inflation by virtually the same amount, and
• grew at virtually the same rate.

The rich got richer, the middle class got richer, and the poor got richer. 





*-*-*-*-*-*

Comments

[Optimist123]

Higgs:
Thanks for your comment on the previous version of this article. I found a math error in the calculations for quintile 1, and corrected it. When I deleted the original and reposted the corrected version, I forgot to save your comment. Feel free to repost it.

[goy]

Thanks for another excellent linkable, Steve. Always shining a light in the corners the MSM and politicians would rather remain dark, you are. I might quibble with the notion of using Heinlein's Razor in this particular case though. It's been apparent for a long time that contemporary journalistic accountability doesn't include investigations even as seemingly straightforward as the ones you provide here on your site. As such, it's hard not to see active malice in the noise they spew. At some point unwillingness to seek the truth on behalf of one's audience crosses the line from spreading false rumors to actively lying, IMHO. We've certainly seen plenty of that recently.

On a related note, how does the $1.x trillion loss last week affect the debt-to-GDP ratio clock? I noticed a recent jump in the ratio, is that what caused it?

Thanks again!

[Steve C]

Interesting article. Why is it better to pay attention to earners rather than households?

[Optimist123]

Steve C:

In short, to eliminate what statisticians call "spurious" relationships. It might look as if X caused Y, but in reality X might cause Z, which may or may not cause Y. The early-morning crowing rooster might look like the cause of the frost on the grass melting, but there's really an intervening variable: the sunrise.

For example, if a two-earner married couple gets divorced, thereby splitting one household into two, "household" income stats appear to have deteriorated (same income divided by 2 instead of 1), even though earner stats may have improved.

Even the "income per earner" stats can be improved in accuracy -- by controlling for full-time vs part time, and full-year vs part-year work. I didn't have time for that (yet).

[mark]

Steve,
Could you clarify a few points for me. The median income chart is in constant 2006 dollars. The income group charts are nominal so they are not corrected for inflation. True? And the last dynamic chart showing income per earner does not specify real or nominal. Do you know which it is? And lastly, the real income only corrects for core inflation, unfortunately most of us have to put gas in our cars and heat our homes. Do you think that should be factored into the equation?

[Optimist123]

Mark:
Yes, the median chart is "real" HH income adjusted to 2006 using CPI-U-RS. The others are not adjusted, because it doesn't change the outcome: no matter which inflation indicator one uses to adjust the numbers, the lines are the same shape, are in the same relative position, and are virtually identical to inflation-adjusted disposable personal income per capita.

Another way to say it: Overall average growth in real disposable personal income per capita (i.e., the economy's average growth) was distributed essentially equally in the growth of real money income per earner across the household quintiles and deciles shown.

I wish I could think of a shorter way of saying it.

In any case, the following two statements are true (...and the second one deserves a big "duh"):

1: "Income earners within the middle class and lower quintiles (and deciles) shared the overall economy's growth rate."

2: "People who earned no income experienced stagnant incomes."

[Bob]

Steve,

Great facts. Only thing is, right now folks are ignoring facts and rational thought.

A number of years ago I was in "Change Management" training (ugh). The best thing I learned was management speaking facts but the rank and file heard feelings. That's right. Emotions play more.

Right now and, for some time actually, emotions have been in play. Never underestimate the power of emotions.

BTW, it's the last bar that has folks riled up. Make it smaller by giving the increases to the middle guys. Wrong headed, yes. Reality, absolutely.

[Steve C]

This gets into a more complicated question of what numbers are meaningful in what context.

To take an extreme example, if there were a total of three earners in town, and their incomes were rising, but there were 10,000 households full of other people, you wouldn't describe that town as doing well.

I don't think it's evident without further information that the household "unit" is inappropriate to use. Maybe people who get paid to think about this have looked at it from all the angles and decided that this household incomes are the thing to track - the household unit best captures some aggregate information that best indicates what's happening with income equality. Maybe someone is choosing households because it tells the story they want to see told. I don't know. I think you're asserting the latter, but I can't tell why.

[OBloodyHell]

Another couple elements to the quantitative analysis --

1) The cost of just about everything has been steadily dropping for the last 75-odd years, in terms of percentage of income required (or number of hours of work required... both are similar). Over at Carpe Diem there have been several graphs of this sort of thing in the last six-odd months, for Clothing, Food, and Gasoline.

2) Another element is that family/household size has steadily decreased, which means that a "stagnant" income is being distributed among a smaller group of people, meaning that the individuals involved each has more disposable income, regardless of the household aggregate.

[mark]

I am still a bit confused on some of the numbers Steve. I found nominal GDP for the 1st qtr 1994 to be 6.9 trillion and 1st qtr 2008 to be 14.0 trillion. That's an increase of 103%. So if nominal incomes kept pace with nominal gdp growth all the income groups should have slightly more than doubled in income. But it looks like most are up around 60%. If nominal gdp growth was 60% over the time period, it would be 9.6 trillion. So 4.4 trillion is missing. Did all of it go to inflation and population (work force) growth? Anywhere else it can go?

[Optimist123]

All:
See the new addendum to the article.

[Optimist123]

Mark:

There's more to GDP than personal income. Disposable personal income (per capita) is about as close an aggregate indicator of growth as we can get to money income per earner. That's why I used disposable personal income per capita, which grew at an average annual rate of 3.2% for the thirteen-year period.

I will say again that the shape of the lines and the spread among quintiles will look the same, regardless of which deflator is chosen to turn nominal numbers into "real" numbers. I don't have time to do it. Volunteers are welcome to take up the challenge; all it will do is change the numbers on the Y-axis, but be my guest.

Here are a few growth rates, for reference, regarding the 13-year period in question:

(HHI = household income; DPI = disposable personal income; GDP = gross domestic product)

nominal median HHI: 3.5%
real median HHI: 0.9%
nominal GDP per capita: 4.1%
nominal DPI per capita: 3.2%
real DPI per capita: 2.1%

Also, "per capita" is important. The BEA publishes the population numbers in its tables for personal income.

[STS]

Interesting post. If earner incomes at all levels were growing fairly steadily and at similar rates, why does the household income graph go through that dip during the past 8 years?

I understand that you are attributing the apparent stagnation to changing distribution of earners per household. But what has been happening to reduce earners per household?

[spencer]

Why is it that all the data you use to disprove the point that incomes stagnated after 2000 run from 1994 on.

Aren't you able to use the same time periods for the two types of series?

I haven't bothered to look at your data because I always just jump to the conclusion that someone pulling data from one time period to prove something about another time period is playing fast and loose with the data.

Can you show me why this assumption is not true in this case?

[Optimist123]

STS:
I'll list some possibilities, but it's a good question for further research by someone. First of all, the household income series is "real" money income, so it doesn't slope up as steeply as nominal income, due to inflation of around 2% or so. But there are still several demography variables not isolated that could explain a lot of the dip besides changes in earners per household, such as: addition of lower-earning immigrants to the working population; the proportion of earners working at part-time vs full-time jobs; the proportion who worked "50+ weeks" vs "26-49 weeks" vs "less than 26 weeks"; and the proportion of households in which nobody worked. I didn't have time to control for those, but eventually plan to do just that in a multiple regression that will find the correlation of quintile-by-quintile earnings growth to overall per-capita GDP and DPI growth, controlling for all that stuff.

[Optimist123]

spencer:
It's a lot simpler than the conspiracy theory you seem to have formulated. First, I used 1994-2007 for every time series. Second, I started with 1994 because that's as far back as the Census Bureau data goes (...try clicking on the link I posted to their data).

I respectfully suggest you reconsider just jumping to your standard conclusion all the time.

[Baggi]

I think the answer is right in front of our faces. The housing boom.

The housing boom created more households but didn't really create that much more income. In San Diego many immigrants share a home until they are able to move out on their own. During the housing boom, some married friends of mine lived with their parents until they were able to get their own home.

So during the housing boom of the last 8 years, many workers were able to get their own homes creating more "households".

Now that the housing boom is over and all these homes are going into foreclosure, one can reasonably expect the "Household" numbers to go back up, as earners will consolidate back into the homes they originally came from.

[Jarrod Myrick]

Great work, Skeptical Optimist. Explosive growth of upper middle class the most under-reported, under-recognized event of our times.

Is it simply a matter of being on the right side of compounded interest?

[bill]

mark, thanks for the charts. i'm puzzing over two issues.

you said: "the median chart is "real" HH income adjusted to 2006 using CPI-U-RS. The others are not adjusted, because it doesn't change the outcome: no matter which inflation indicator one uses to adjust the numbers, the lines are the same shape, are in the same relative position, and are virtually identical to inflation-adjusted disposable personal income per capita."

you don't mention "same slope". if inflation was greater than zero for those years, then the lines on the chart would be flatter, indicating less growth. if inflation was more than about 3 1/2% (i don't know if it was or not) then the lines would be flat, indicating stagnation.

[sorry, i'm always cautious when monetary items are not inflation adjusted.]

the other aspect i'm trying to get my mind around is your leaving off the uppermost quintile/decile. since you are only talking about middle class incomes, that does make some sense. but if you wanted to know if the middle class incomes are stagnant "compared to the top incomes" then you need the figures for that top quintile/decile.

[STS]

Bill,

I think Steve is simply missing those numbers for top decile/quintile because they aren't collected by the Census Bureau.

Jarrod,

You're making the same mistake as the Census Bureau -- adopting the artificial convention that income greater than $100k makes you "upper middle class". It depends where you live and inflation has probably been higher than reported. Cheaper/faster computers don't really make your cost of living lower, but official statistics pretend they do.

If incomes were bucketed by $2500 increments from 100k up to maybe 500k, Steve could create a much more detailed version of his nice animated graph which would show the "toothpaste" gradually shifting to the right while retaining the same humped shape. That's a picture of growth and inflation, not people winning the lottery (so to speak). The spike at the right end is an artifact of limited data collection.

[Jarrod Myrick]

STS,

Town i live in median income 83k--where i'm from 38k. point is: if a man can't define and pursue his own happiness making 100k/yr in this economic-political union he ain't tryin hard enough. 100k is arbitrary sure, of course it is--this is the comments section of a man's blog neither of us's ever met. but if we wanted to meet him we could trade five hours of labor for a ticket to take an aeroplane trip to meet anywhere in our land in probably five hours or less. i don't know how to drive a horse carriage or operate a gatling gun or fear the dark. in relative terms to the history of humanity, we're gods.

cost of living? how many cars do you need? how many tvs do you want?

[piannaf]

I keep reading this blog because I love the way you work with data.

I recently came across "The Coming Collapse of the Middle Class" http://economistsview.typepad.com/economistsview/2008/04/the-coming-coll.html and it worried me how the speaker was working with data.

Especially worrisome for me was her idea of a debt treadmill which she assumes goes in the opposite direction of your animated graph.

She also compared a housewife in ~1970 who doesn't need an extra car or outside childcare to a working mother in ~2000 who does need both those things. Now, they may both be "the middle class" for each time period but it's a personal choice which changed their situations so I wonder what consequence that choice has in ~1970 vs ~2000.

[Optimist123]

piannaf:

Thanks for the link. Her analysis included both the income and expenditure sides of the household budget. She had some interesting data, but didn't control for the two basic causes of family insolvency: (1) bad luck, versus (2) financial incompetence. Health problems, job problems, spending problems, and education problems can fall into either category. For example, it is easy to feel badly for the person who was talked into a mortgage he couldn't afford -- until finding out he had a $250,000 salary, and was talked into a $2.5 million mortgage he couldn't handle. I place that one under the heading of incompetence, and have no desire whatsoever to pitch in to ease the pain of that particular insolvency.

I'm still wondering whether she thinks the solution would be more government-provided safety nets for families that stumble into insolvency for whatever reason, or for better education as to how to avoid insolvency due to incompetence.

Also, regarding the middle class: The census data leaves several questions unanswered, as you can see from my chart. For example, it is entirely possible that the distribution has become bimodal (twin peaks, with the higher one masked by the $100,000 cutoff). I'd be surprised at that, but cannot completely rule it out without seeing how the "above $100K" earners are distributed.