Monday, January 11, 2016

Defining the “Disappearing Middle Class”

Michael J. Perry
(Image source:
On December 30, 2015, economist Mark J. Perry published in his American Enterprise Institute blog Carpe Diem a couple of charts purporting to show that the American middle class, so far as it can be said to be disappearing, is doing so into higher-income households. Said Perry:

Over the last nearly 50 years the biggest gain for US households has been the 16.6 percentage point increase in the share of high-income households earning $100,000 or more per year, which accounts for the declining share of low-income and middle-income households (by two different measures). Yes, the middle-class has been disappearing over the last generation or more, but they have moved into higher-income categories of household income, not moving down into lower-income categories of household income.

“Cooking the Books”

Of course, Perry is a recognized economist, and I’m just a smart-aleck with a computer and three credit-hours in Econ 201. But I’m also a son of a bookkeeper, and have seen many interesting tricks people can play with numbers. Science is heavily dependent for its effectiveness on the honesty by which it applies numbers to phenomena, and is therefore vulnerable to anyone who knows how to “cook the books”. And the “dismal science”, like the others, tends to suffer when the numbers collide with policy preferences.

The picture Perry paints is of a middle class that was better off in 2014 than it was in 1967 — at the very least, that said middle class is making more money even after inflation is taken into account. However, to get an apples-to-apples comparison, he has to account for inflation. And here’s where the problem begins: there are a number of tools an analyst can use for inflating and deflating number … but none of them are 100% accurate. (For a comparison of four common price indexes used in policy analysis, see this post in The FRED Blog.)

Deflating Poverty

The graphs Perry and his research associate Tom Sullivan prepared concern the percentage of US households that had incomes within certain ranges. The incomes from 1967 to 2013 have been adjusted, most likely using the Bureau of Labor Statistics’ CPI research series (about which more here). Both graphs have a “high income” threshold of $100,000/year; one has a “low income” threshold of $35,000/year, while the other has a “low income” benchmark of $50,000/year.

Lookit that! Sure looks like poverty is declining, right? And so is the middle class! Why? Because everyone’s becoming richer! See, once we take inflation out of the equation, we get what the financial world calls “real dollars”, as opposed to “current” or “nominal” dollars”. In theory, once we account for inflation, we can tell whether consumers can buy more or less stuff. And it looks like they can buy more stuff!

As reasonable as Perry’s benchmarks for high and low income may seem, the fact is there’s no accepted income definition of “middle class”; these benchmarks are his arbitrary choice. They might be good choices, too, if we could show that they really reflect “constantized” values (pardon the neologism).

What Things Cost vs. What We Spend

To illustrate the problem, let me bring in a different index: the federal poverty threshold.

Consumer price indices are derived by putting together a theoretical “basket” of goods and services, measuring their price changes over time, and averaging out their changes. But what we actually spend on these things is a different matter from what they cost; there are necessities we must spend on regularly and luxuries we might (or might not) buy if we have the money. So while the CPI functions as a measure of price change, the poverty threshold will be our measure of cost of living change.[*]

The Census Bureau breaks the population down into fifths, or quintiles. The charts above show the top income values for each of the bottom four quintiles, plus the bottom value for the top five percent of incomes. Although the top income of the fourth quintile breaches the $100k benchmark, seemingly validating Perry, two things we should notice right away: 1) In both charts, the bottom value of the top 5% increases dramatically and disproportionately to the movements of the bottom 80%. 2) Despite the fact that the bottom chart has been controlled for inflation, the poverty threshold still rises, though not by much — roughly 13.1% over forty-seven years.

Why is there an increase at all? First, the CPI doesn’t ask how much money we spend on hairspray or housing in a given month; it merely asks what the difference is between their prices in 2014 and their prices in 2013. Needless to say, a 3% increase in the rent of an apartment is a larger money amount than is a 3% increase in a bottle of Paul Masson. Second, prices don’t all inflate at the same rate; healthcare and college tuition, for example, have been outpacing the CPI for a couple of decades or more. Because the poverty threshold calculation accounts for necessary expenditures, any necessary good or service which inflates at a faster rate will push the poverty threshold up faster, while luxury items that inflate at a slower rate won’t affect it at all.

Numbers Divisible by 5 and 10

My Accounting 201 professor gave us an old rule of thumb: “Never trust numbers divisible by 5 or 10.” While they do occur naturally every now and again, they usually indicate either carelessness or malfeasance. In fact, prices and sales taxes are set precisely so such numbers rarely occur.

Perry’s benchmarks, which seem so reasonable, actually reflect 144.4%, 206.4%, and 412.7% of the 2014 poverty threshold (in ascending order). If we deflate the 2014 benchmarks to their 1967 nominal equivalent, they would be $5,645, $8,065, and $16,130 (rounded to the nearest dollar). These in turn reflect 163.4%, 233.4%, and 466.9% of poverty level. Graph out the “real” values as multiples of the poverty threshold — and lookit this! They decline, much like the middle and lower class groups in Perry’s graphs! How ’bout them apples!?

It’s arguable that the differences are marginal … but they’re there. The difference between $14,259 (144.4% of the 1967 poverty threshold) and $16,130 (163.4%) works out to approximately $11,621 in 2014 dollars, almost an extra thousand a month. And, in fact, when we hold the 1967 proportions constant, the 2014 benchmarks become $39,592, $56,560, and $113,119 respectively; the fourth quintile’s top income only breaches the “high income” benchmark in 2006 and 2007.

Ultimately, the benchmarks Perry and Sullivan chose have no real analytical rationale other than that, in 2016, $100,000 is a lot of money. However, when we look at the difference in the cost of living, it’s not as much money now as $16,000 was almost fifty years ago; it won’t go quite as far.

Looks Can Be Deceiving

There’s another way in which the graphs are deceptive. Remember that sliver of the fourth quintile that breaches the $100k benchmark? That’s the “4.7%” in Perry’s 24.7% of American households that are “high-income”. But because it’s still part of the fourth quintile, it’s part of the 80% of Americans that share less than half of America’s aggregate income.

The way Perry presents the information makes it appear that, by controlling for inflation, he has flattened the top incomes. And indeed, the fourth quartile’s mean real income rose 43.1%; compared with the poverty threshold, the nominal mean income rose about 26.5%. However, its share of the income pie went down about 4.1%, from 24.2% to 23.2%; overall, the lower 80% lost about 13.5% of their share to the highest quintile, with just over three-fifths of it going to the top 5%.

The truth is, no deflator can hide the massive increase in income among the top 20%. Yes, there are more people making over $100k in “real” dollars; but there are also more millionaires, and even some billionaires, which was unheard-of in 1967. Keep in mind as well that we’re talking household income; two people making $50k each combine to make a $100k household income. And there are many metropolitan areas where $100k doesn’t get you much; it may get you a McMansion in Ogallala, Nebraska, but it won’t get you a condo in Manhattan.

Change in mean income for each quintile. The bottom numbers show that the apparent increase in “real” income is offset by the increase in the cost of living. (Data source: Census Bureau.)

Wrapping It Up

Trying to define an income bracket may be misunderstanding what’s meant by the “disappearing middle class”. Try thinking of what the French originally meant by the bourgeoisie — the small business owners, the shopkeepers, tailors, artisans and such that formed a class apart from laborers and the investor class … people whose money and lives are invested in only one business. In an age where everything from big-box stores to lingerie shops and lunch counters are chained and franchised to spread like the “pod people” in Invasion of the Body Snatchers, the bourgeoisie aren’t so much disappearing as they’re being squeezed out of the market. Yet they’re still there; and, unless I miss my guess, will still be there for many years to come.

In any event, until we have a clearer, more objectively defined concept of the middle class, it won’t do to simply pull benchmarks out of a hat and call everyone that falls between them the “middle”. Certainly the majority of us are still better off now than we were during LBJ’s presidency. But there are still some issues that won’t be “cooked” away by flattening the aggregate US income. And that’s the biggest disservice Perry does us.

[*] The number I use is the most common one — weighted average of a household of four. Prior to 1980, this number was broken down by the sex of the head of household and by farm/non-farm; for 1967 to 1979, I used non-farm, male head of household.