It can’t be said too often that statistics are indispensable for better understanding the economy. Nor can it be said too often that when gathering, reporting, and encountering statistics, extreme care is required. Few intellectual feats are as easy as creating false impressions with true statistics. Equally easy is drawing false conclusions from true statistics.
An example of one such false impression is the often-heard lament that the American middle class is disappearing. If true, it is taken to be very bad news, for this claim conjures up the impression that larger and larger percentages of Americans are falling into the lower income-earning classes. And in fact, clear evidence does indeed exist to support the claim that the American middle-class is disappearing. Yet middle-class Americans are disappearing, not into the lower classes, but into the upper classes.
The United States Census Bureau reports the percentage of U.S. households earning different levels of annual pre-tax incomes. These incomes include some, but not all, of the transfers that households receive from government welfare programs. These incomes also are adjusted for inflation. Data on these incomes can be found in Table H-17 (“Households by Total Money Income, Race, and Hispanic Origin of Householder”) of “Historical Income Tables: Households.” According to the Census Bureau’s own income-bracket classifications — for example, “Under $15,000,” “$15,000 to $24,999,” and “$25,000 to $34,999” — the percentage of American households earning (in 2019 dollars) annual incomes that are unambiguously middle class (“$50,000-$74,999” and “$75,000-$99,999”) was lower in 2001 than in 1973, and lower still in 2019 than in 2001. (I use 2019 as the most-recent date to avoid the economic, fiscal, and monetary distortions caused by covid lockdowns and covid-inspired monetary and fiscal extravagance.)
This trend seems ominous. But a look at all the lower household-income brackets reveals that the percentage of American households that are today in lower-middle-income and low-income brackets is also smaller than in the past. The American households who ‘left’ middle-class brackets didn’t fall into lower-income brackets; they rose into higher-income brackets. In 1973 — the year that many people identify as the one when ordinary Americans reached their economic zenith — the percentage of American households annually earning more than $100,000 (again, in 2019 dollars) was 16.7; in 2019 this figure was 34.1 percent. Over this same time period, the percentage of households earning more than $200,000 grew from 1.9 to 10.3.
The only sense in which the American middle-class is disappearing economically is that an ever-increasing percentage of American households earn annual incomes that are in the upper brackets.
One must be very careful when interpreting statistics.
Beware of Averages
Each semester I ask my undergraduate students what would happen to the average height of people in the classroom if a newborn baby were brought in. They immediately recognize that the average height would fall. I then ask them if, upon getting a report of this decline in the average height, they would call their physicians in a panic to see how to stop themselves from shrinking. They laugh and say “of course not.”
As this simple example shows, a calculated average can change in ways that give quite misleading impressions of what’s happening to each of the individuals in the group for which the average is calculated. For example, as more immigrants enter the ranks of the employed in America, the average wage rate will be lower than it would have been if fewer immigrants worked. It’s possible that the average will fall because the immigrant, by increasing the supply of labor, pulls down everyone’s wage rate. But it’s also possible – and economically more plausible – that the average will fall without any decline in any worker’s wage rate. The reason is that the typical immigrant is not (yet) as productive on the job as is the typical long-time participant in the American labor market. The typical immigrant’s wage, therefore, is below average. The immigrant’s employment thus causes the calculated average wage to be pulled down below the level it would have attained without the immigrant. But of course this statistical reality lowers or suppresses your or any other worker’s wage no more than does the entry of an infant into a college classroom lower or suppress the height of a college student in that room.
If the workforce becomes more open to workers with lower skills without closing off opportunities to workers with higher skills — as arguably seems to have occurred in America over the past several decades — this development is unquestionably good. Nevertheless, because one result is downward pressure on measured average wages (and even on measured median wages), the economic statistics might tell a misleadingly pessimistic tale.
Location Is a Normal Good
Economists define a “normal good” as a good that people demand more of as their expected purchasing power rises. Examples of normal goods are fine wine (rather than box wine), hotel (rather than motel) accommodations, and new (rather than used) automobiles. When we observe people consuming greater amounts of a normal good when that good’s price hasn’t fallen, one possible explanation for this increase in demand is that people’s expected purchasing power has risen — and, in turn, that people are economically better off than they were earlier.
Suppose that workers today in locales hit with significant negative employment ‘shocks’ suffer longer periods of unemployment than did similarly situated workers in the past. One result is an increase in the average duration of unemployment. What might explain this increase in the duration of unemployment? One possibility is that the economy’s rate of creating new jobs has slowed relative to the rate at which it loses jobs. If so, this reality would be evidence of worsening economic performance.
But this reality in America today is likely better explained by another, very different possibility: working-class people are wealthier than in the past; their lifetime purchasing power is higher. Powerful evidence that ordinary Americans are today much wealthier than in the past is found (among very many other places) in Michael Strain’s 2020 book, The American Dream Is Not Dead (But Populism Could Kill It), and in Phil Gramm’s, Robert Ekelund’s, and John Early’s 2022 study, The Myth of American Inequality: How Government Biases Policy Debate.
Compared to worker Jones today, worker Smith and his family in, say, 1974 would have suffered deeper economic distress if he remained unemployed for many months. So worker Smith didn’t wait long before moving himself and his family from the hometown that they love to a different town where his employment prospects were brighter. In contrast, worker Jones and his family today have higher purchasing power than did worker Smith and his family in the 1970s. If worker Jones loses his job, he can better afford than could worker Smith to ‘consume’ for a longer time his love of his hometown.
Yet worker Jones’s decision not to move in search of employment causes economic statistics to look worse than they would have looked had worker Jones been less wealthy. Specifically, because worker Jones’s greater prosperity allows him to remain longer in his hometown waiting for new employment to come to him — rather than him finding new employment sooner by moving to another town — the measured average duration of unemployment is higher than it would have been had Jones been poorer and moved away in search of a new job.
In short, the rising material prosperity of even ordinary Americans makes the consumption of locational preferences more attractive. Although this outcome is unambiguously desirable, it can result in statistics seeming to indicate a worsening of ordinary Americans’ economic lot.
The above three examples of how true statistics can fuel false conclusions are only the tip of the empirical iceberg. Insist, when appropriate, on data. But beware always that these can be highly misleading if handled carelessly.
