Update: How changing household composition, household work hours, and retirement explain median household income

From Mark Perry.

"One of the most frequently reported economic trends is the gradual decline in US real median household income from its 1999 peak of about $57,843 (in 2014 dollars) to below $54,000 in each of the last five years (see dark blue line in top chart above). We hear a number of reasons from politicians and pundits for the decline in median household income over the last 15 years, mostly reasons that involve a narrative about economic stagnation and growing inequality caused by the progressives’ usual suspects: most of the gains in worker productivity, income, and wealth going to corporations and “the rich” instead of being shared by average workers; failure to increase the minimum wage or pass “living wage” laws; the combined effects of globalization, free trade and outsourcing putting downward pressure on middle-class incomes in America; excessive CEO pay; Wall Street greed and fraud, and other narratives of economic malaise and pessimism.



About a year ago, I featured several posts on CD (here and here) about how demographic changes over time in the composition of US households might help explain the stagnation and decline in US median household income starting around 1999. Those posts were inspired by Alex Pollock’s excellent essay “If income is going up, can median household income go down? It’s possible” where Alex explained how the changing composition of US households could result in declining median real household income even if all Americans’ real incomes are rising. That is, the decline in median household income might not be caused by decline in real income for the average American worker, but rather by changes in the composition of the average US household.

More specifically, the demographic changes that Alex and I analyzed include: a) the declining share of households with two or more earners starting around 1999 (see top chart), b) the increasing share of no earner households starting around 1999 (see top chart), c) the increasing number of retirees in the US as a share of the US adult population starting around 2007 (see middle chart above), and d) the decline starting around 2000 in average weekly hours of work per US household (see bottom chart above). Actually, those demographic trends could all be related since the increasing number and share of US retirees would obviously result in: a) an increase in the share of households with no earners, and b) a decrease in the average number of work hours per US household. Now that Census and Social Security data are available for another year — 2014 — (data in my previous posts were through 2013), I thought it would be a good time to update the charts and update the analysis on how the changing US household demographics and retirement trends can help explain the 7.2% decline in median household income from the peak of $57,843 in 1999 to $53,657 in 2014. Here’s the update:

1. Households with No Earners and Two or More Earners. One example of a major dynamic change in household composition is the significant increase in the share of US households with no earners, from fewer than 20% of all US households in 1980 to 24% of households in 2014 (see light blue line in top chart above, Census data here from Table H-12). At the same, there’s been a significant decrease in the share of US households with 2 or more earners from above 45% of all households in 1999 (when median household income peaked) to fewer than 40% of US households in each of the last five most recent years starting in 2010 (see brown line in top chart above).
Technical Note: In a linear regression model, those two variables (no earner and two earner household shares) explain 90% of the decline in median household since 1999.

In summary, over the last several decades, there’s been an increasing share of no-earner households and a decreasing share of married and two-or-more-earner households. That major demographic shift in household composition would naturally depress median household income over the last 15 years, even though it’s possible, as Pollock showed in his essay, that the income of individual working Americans could have been the same or rising since 1999.

2. Increasing Number and Share of US Retirees. Another key demographic shift is the increasing number of retired Americans as a share of the adult population based on Social Security data. As the light blue line in the middle chart above shows, US retirees represented a pretty stable 15% share of the adult US population from 1990 to 2007. Then, starting around 2008 when the early “baby-boomers” – those born in 1946 — reached early retirement age of 62, the share of retirees started increasing from less than 15% of the adult population in 2007 to nearly 17% in 2014.

In the six-year period between 2008 and 2014, the number of retired Americans increased by 7.4 million, which was the largest six-year increase in US history, and more than triple the 2.4 million increase in the previous six-year period. Given that wave of recent retirements, there have been millions of older, experienced, highly paid workers going from their peak earning levels to a much lower retirement income that would typically include Social Security payments, pensions, and distributions from retirement accounts. As those millions of retirees are replaced in the workforce by younger, less experienced, lower paid workers, median household income would naturally be falling even though the average and median incomes of working Americans could be rising.

It’s probably no coincidence that the recent increase in retirees, both in absolute numbers and as a share of the adult population, along with the other demographic changes in households described above, has naturally coincided with a decline in median household income. It would be hard to imagine that an aging population with a significant increase in the number and share of retirees, wouldn’t depress median household income, for purely demographic reasons.

3. Decline in Average Number of Hours Worked per Household. In a December 2014 Real Clear Markets op-ed (“The Obvious Reason for the Decline In Median Income”), economist Jeffrey Dorfman points out another very important demographic change that has significantly contributed to the decline in real median household income in the US over the last decade: the average number of hours worked per US household has been declining. So we would naturally and logically expect median household incomes to decrease when average household work hours are falling. Here’s the opening of Jeffrey’s article:

Much has been made recently of the fact that real median household income has been stagnant over the past twenty years and falling for the past seven. While there are many problems with using median household income as a measure of the economic health of the middle class, it is still important to examine what is causing this middle-income stall. A major, and overlooked, part of the answer appears to be quite simple: Americans are working less. When people and families work fewer hours, they earn less money.

Following a procedure outlined by Dorfman in his RCM article but using a slightly different dataset, the bottom chart above shows the relationship over time between annual US median real household income and the average weekly work hours per US household in each year from 1980 to 2014. To calculate the average work hours per household, I used: a) the BLS series “Average Weekly Hours at Work in All Industries” (from Table 22 in this BLS report), b) the BLS series “Civilian Employment” for the number of employed Americans, and c) the number of US households in each year from the US Census. Using the average weekly work hours and the number of Americans employed, the total number of hours worked annually were calculated and divided by the number of US households in each year to determine the average number of weekly work hours per household in each year from 1980 to 2014, and those values are represented by the light blue line in the bottom chart above.
As can be seen in the top chart, the 7.2% decline in real US median household income (in 2014 dollars) between the 1999 peak ($57,843) and 2014 ($53,657) was accompanied by an even greater 8.8% decline in average hours worked per household, suggesting a very close statistical relationship between those two variables. In fact, a linear regression model reveals that more than 90% of the decline in median household income between 1999 and 2014 can be explained by the decline in average household work hours, so there is a very strong statistical relationship between median household income and average household hours worked, as expected.

Bottom Line: It’s an important point that most of the discussions and hand-wringing about declining US median household incomes completely ignore the demographic realities that the composition of American households is not static. US households in 2014 are significantly different from US households in 1999 in important ways (size, age, number of earners, hours worked, marital status, etc.) that affect household income, so we can’t accurately compare the $53,657 in median household income in 2014 to the much higher peak of $57,843 in median household income in 1999 (both median incomes are expressed in 2014 dollars). Specifically, the composition of US households is changing over time due to the natural consequences of an aging population and an increasing share of households with retirees, along with more (fewer) single-earner (multiple earner) households that reflects an ongoing trend of smaller US households dating back to at least WWII – and those factors can’t be ignored when discussing trends in US household income.

Most explanations of the recent decline in US median household income are based on some variation of a narrative of economic stagnation, rising inequality and pessimism. But what is almost always overlooked are the very significant demographic changes that have taken place in the composition of US households over time that would significantly impact the income of the median US household. Taken together, a) the increase in the share of no-earner, single-earner, single-parent households, b) the increase in the number and share of retirees, along with c) the decline in the share of two-earner-or-more and married households, would all logically and necessarily depress the income level of the median US household over the last 10 years.

In conclusion, the composition of US households is not static, fixed and permanent; rather it’s dynamic, evolving and ever-changing. Discussions on changes in median household income over time that ignore the changes in household composition over time will always be incomplete, distorted and misleading. Perhaps the decline in median household income this century is not a narrative of economic pessimism and stagnation after all, but a more upbeat story of a greater number of Americans living longer lives, and enjoying periods of time in retirement that were never possible until this century.

Update: The chart below is provided in response to comments below the post from Sprewell and Marque2, and shows that during the 1999 to 2014 period when real US median household income decreased by -7.2% real hourly earnings increased by +7.5% and real compensation increased by +13.5%. Those data support Alex Pollock’s original position that real incomes can be going up at the same time that real household incomes are declining, if the composition of US household is changing. To summarize what we know for sure:

1) The size and composition of US households has been changing over time, but especially starting in 1999 when median household income started to decline, and in ways that would naturally lower median household income even if average incomes were the same or rising: a) a decrease in the number and share of US households with two or more earners and b) an increase in the number and share of US households with no earners.

2) During the period when real median household income fell by 7.2% between 1999 and 2014, real wages increased by 7.5% and real compensation increased by 13.5%.

That doesn’t necessarily mean that the 1999-2014 wasn’t a period of economic stagnation for some Americans, since the increases in real wages and real compensation only apply to those Americans who were employed during that period.  It’s certainly a complicated analysis to determine exactly how much of declining real median household income between 1999 and 2014 is because of economic stagnation and how much is explained by the changing composition of US households and the increasing number and share of retired Americans. My main point is that the economic stagnation story usually takes center stage in the hand-wringing about declining median household income, while the equally, or maybe even more, important changes in the household size and composition usually get overlooked.

"