TREND WATCH: What’s Happening? Experts are puzzling over the vulnerable state of the U.S. housing market, which recently turned out some of its worst monthly numbers in years. As is the case in many industries, Millennials are taking much of the heat for housing’s recent troubles. If only they would grow up and move out, some say, Millennials would be pushing the market to record heights.
Our Take: But is this account accurate? To find out, we turn to decades’ worth of Census data tracking the living arrangements of U.S. adults. As it turns out, while young adults indeed have led the way, every single age group is now generating fewer households on a per-capita basis today compared to 2007. The growing preference for co-living at every age is helping to push up the average number of adults per household—further adding to the housing market’s inescapable structural headwinds.
No one is happy about the U.S. housing market. In December, existing home sales tumbled to their lowest level in three years. At the same time, home values rose at their slowest rate in six years. And housing starts dropped to their lowest level in two years.
Indeed, industry insiders have been perplexed by the (historically) subpar performance of housing ever since the U.S. economy emerged from the Great Recession a decade ago. Many have written generally about the decline in household formation. Some have pointed accusingly at one generation in particular: young-adult Millennials. Perhaps Millennials want homes but aren’t buying them due to their high levels of student debt. Maybe they don’t want homes because they aren’t getting married and love living in cities.
All these arguments rely on a lot of hand-waving and anecdotal evidence. But what do the numbers say? Armed with comprehensive Census data on personal living arrangements by age, we have calculated the quantitative impact of behavior change on household formation and therefore on the demand for housing units.
OUR EARLIER METHOD
Recall that in our previous piece on household formation (see: “Household Formation: Why Is It Declining—and Where Is It Going?”), we revealed that the rising number of adults per household over the last decade has resulted in a significant slowdown in overall household formation. This slowdown in turn translates into an aggregate 2 million households not formed since 2007.
To identify the drivers behind this decline, we looked at the change in total households versus total adult population by age bracket. And there we found that more than all of the household formation decline can be accounted for by behavior change among Americans under age 55. Among 18- to 24-year-olds, the behavior shift was so marked that it caused the number of households to decline even though the population in this age bracket grew. Among 25- to 34-year-olds, the behavior shift reduced the percentage growth in households to less than half of their (impressive) growth in population.
But these conclusions, while revealing, give us few granular insights because we define “behavior change” as simply the residual (in each age bracket) between population growth and household growth. It’s a valid bottom line. But if we want insight into the reasons behind behavior change, we need to dig deeper.
OUR NEW METHOD
Using newly tabulated Census data, we analyzed surveys of Americans’ living arrangements—by person, not by household—to determine how exactly behavior has changed since the Great Recession.
Census sorts respondents age 18 and older into six mutually exclusive living arrangement categories. Census gets rid of people who may fall into more than one category by prioritizing the categories as follows:
- Living Alone: refers to one-adult-only households.
- Living with Spouse: two married adults, who may or may not be living with other adults.
- Living with Unmarried Partner: two adults who call themselves “unmarried partners,” who may or may not be living with other adults.
- [Child of Householder]: any adult who is living with a parent or living at college; but who is not included in any category above (e.g., if this child is also living with a spouse).
- [Living with Nonrelatives]: any adult who is living with another non-relative adult; but who is not included in any category above.
- [Living with Relatives]: any adult who is living with another adult relative; but who is not included in any category above.
The first three unbracketed categories above take priority (i.e., these categories are treated as the primary living arrangement for individuals who would otherwise overlap), while the next three bracketed categories are subordinate.
We sought to find the percentage change in the number of householders created or lost due to behavior, both among the entire population and within each age bracket. We were also interested in seeing the relative impact of each living arrangement on the total household formation. This required us to assume that each household type is associated with a countable number of like-aged adults per household. So, for example, a “living alone” household contains (by definition) one adult. A married or “partner” household contains two adults. A person living with nonrelatives contains 1.5 to 2.5 adults depending on the age bracket. And so on.
To be sure, this isn’t an exact science. No one (not even Census) actually counts the number of adults living with other adults in the same age bracket. But on average, across age brackets containing millions of people, plausible assumptions are possible. What’s more, the exact number for each household type has less of an impact on household formation than the change over time in the share of adults in each household type. Overall, our approach is good enough to explain the broad trends that most interest us.
INSIDE THE NUMBERS
Now we can turn to the data. While we are interested mainly in the post-2007 period by age bracket, it’s helpful to begin by looking at population-wide trends over a longer time period.
All Ages. Across all age brackets, the share of adults living with a spouse has declined over the past half-century by 19.1%—from 70.3% in 1967 to 51.2% in 2018. Even including the unmarried as “couples,” the share has declined by 11.8%. Filling most of this decline have been persons living alone (+6.4%). There has also been a significant rise in adults living in larger groups, that is, with “relatives” and “nonrelatives” (+5.4%). But the net effect of substituting so many single homes for couple homes has been an overall reduction of the number of adults per household.
It’s worth noting that most of this reduction occurred early in this period, between 1967 and 1987. The “seventies” was when the divorce revolution accelerated before reaching its inflection point. It was also when Americans of all ages began to prefer living alone—first the G.I. Generation as “senior citizens” now able to live on their own thanks to newly raised federal entitlements; then the Silent Generation as divorcees; and finally the young-adult Boomers who couldn’t wait to live apart from their families (and from each other).
It’s also worth noting that, over the last decade, things have begun to swing the other way. Total couples are down by -1.9%. But now filling in for nearly all of that decline (+1.6%) is a rise in larger households (“relatives” and “nonrelatives”). The net effect, since 2007, is that the number of adults per household has begun to decline.
When we subject Americans of all ages between 2007 and 2018 to our decomposition of change by household type, we confirm exactly this.
Here, the slight rise in adults per household translates into a slight decline in expected households per 100 adults. Yes, the rate of couple households is declining slightly and the rate of single households is rising slightly. But what’s tilting everything the other way is the rise in “children of householders” (who by definition generate no households of their own) and the rise in “relatives” and “nonrelatives.” As we’ll see shortly, most of these togetherness trends are being accelerated by younger households.
Age 18 to 24. Now let’s look at each age bracket, starting with the youngest: adults under age 25. The absolute number of householders in this group is relatively small, both due to the small seven-year age span and to the fact that a large and growing share of these adults (54.9% in 2018, according to Census) live with parents or in college dorms. A very small share (0.9% in 2018) even lives with their parents and a significant other.
What else is going up? More of these youths are living together with relatives and nonrelatives, which tends to concentrate more adults into one household. What’s going down, on the other hand, are less concentrated households like singles (living alone) and couples (either married or not). Back in 1967, 39.3% of this age bracket was married. Today, that has dropped to only 7.0%. Even unmarried couples, which are now more numerous in this age bracket than married couples, have been declining per-capita over the last decade.
Understandably, this shifting mix is generating a lot fewer households per 100 persons over time. More persons living with parents reduces the ratio by simply taking them out of consideration. And more persons living in more concentrated households—and fewer in less concentrated households—also reduces the ratio. All told, we find that expected householders per 100 persons dropped by 9.4% over the last decade.
Age 25 to 34. Here we find a lot more young adults living on their own either as couples (54.0%) or as living-alone singles (10.3%). A smaller share (17.8%) are living with their parents. (This includes, again, some 1.1% who live with parents while cohabitating.) Still, the trends over time are very similar. Both the married share and the total “couple” share (including unmarrieds) are declining over time. The living-with-parent share is rising over time. And the share living in more concentrated households (relatives and nonrelatives) is rising.
For all these reasons, overall household concentration is similarly growing over time. Much of the growth is due simply to the large jump (+5.7 percentage points) in the living-with-parent share. The increase (+1.4 points) in high-density nonrelative living also helped, although this was partially neutralized by a slight rise (+0.4 points) in living alone. All told, we find that expected householders per capita (x100) dropped by 5.9% over the last decade. That’s a bit less percentage-wise than the younger age bracket, but it applies to a vastly larger number of households.
Apparently, much of what people say about Millennials delaying the onset of adulthood is rooted in truth. Their low marriage rates are leading to fewer new householders. In part, this trend can be attributed to a lack of financial well-being, as Millennials are content to wait until they’ve established their careers and their finances before taking the plunge. (See: “Are Finances Keeping Millennials from Marrying?”) In the meantime, this generation is content living together with parents or with friends, whether in conventional apartments or new co-living spaces (see: “Does Co-Living Threaten the U.S. Housing Market?”).
To be sure, these trends may reverse if the economy continues to improve. But we haven’t seen any reversal in the yearly trend through 2018—and we’re surely nearer to the next recession than we are to the last. Another possibility is that young adults are adapting happily to a more communal lifestyle with same-age peers and that both young adults and older generations are becoming increasingly comfortable pooling resources like housing. Nuclear families with independent youth are, after all, a very recent phenomenon. Multigenerational families with dependent youth are the time-tested historical norm. Perhaps America is moving back there.
Age 35 to 64. In 2018 we need to get to the core midlife age bracket before we see couples as the dominant household lifestyle. And if we group the unmarried together with the married couples, the “coupled” share of this age bracket, after dropping during the earlier postwar era, has stabilized—losing only 1.3 percentage points over the last decade. Even here, though, we see a slight growth in living with parents (from 3.2 to 3.7%) and in living with relatives and nonrelatives (from 13.1 to 13.9%). Single living remains unchanged.
Once again, the net impact on adults per household is positive, though just barely. Midlife 35- to 64-year-olds (x100) created 46.7 expected householders in 2007 versus 46.6 in 2018. This generates a very slight increase (+0.3%) in adults per household.
Keep in mind that these figures track only adults per household within each age bracket. But we can be certain that the vast majority of young adults age 18-24 and age 25-34 who are living with their parents are in fact living with householders in this age bracket. All told, counting all adults living in parent households, the actual rise in total adults per household is probably about 3% less in the younger age brackets and about 4% more in this 35- to 64-year-old age bracket. This means that the actual rise in total adults per household is roughly equal across all nonelderly households (that is, households headed by a person under age 65).
This age bracket today is mostly occupied by Generation Xers and late-wave Boomers, many of whom were ravaged during the Great Recession and have been slow to recover. (See: “Xers ‘Still Carry Scars’ from GFC.”) A rising share of this age group view group living as an attractive option that enables them to save what they can before retirement. But, as we have seen, the biggest reason these households are getting more crowded is the aging in place of their own adult children.
Age 65-74. As Boomers have aged past age 65 over the last decade, the total number of adults in the young-seniors age bracket has surged. The percentage breakdown of this group by household lifestyle, however, has remained remarkably stable. As expected, the share living as unmarried couples has grown dramatically (by nearly half), but it remains small (at 2.9%). For the first time, this age bracket boasts the highest share of married couples, though this is due entirely to the decline in the midlife age bracket.
Overall, like midlifers, this age group saw a slight decline in expected householders, which translates into an equivalent slight increase (0.3%) in adults per household. Also like midlifers, the total number of adults per household in this age group was undoubtedly boosted further (though not as dramatically) by an increase in adult children living with their parents.
Age 75+. Among the old-seniors, the biggest long-term shift has been the sizable growth—especially in the first two decades between 1967 and 1987—in the share who live alone. This shift happened when the G.I. Generation was entering this age bracket and was fueled by the G.I.s’ greater wealth and their access to newly boosted federal entitlement programs. Unlike the impoverished Lost Generation before them, the G.I.s were no longer compelled to live with “other relatives.”
Since 1987, this shift has gradually backtracked. Living alone has declined, and living with a spouse has grown. Why? It’s mostly due to the contemporaneous rise and fall in the share of widowed women over age 65—and that in turn is due to rise and fall in the gap in life expectancies between men and women at age 65. Over the early postwar era, until the early 1980s, women’s life expectancy rose much faster than men’s. (Widows used their larger Social Security checks to live alone and thereby unburden their grown children.) Since then, men’s life expectancy has been catching up fast. To explain this rise and fall, most demographers point to the parallel rise and fall—with roughly a 40-year delay—in the differential smoking rate of males over females.
This is the only age bracket in which living alone accounts for more than half of the expected householders per 100 population (35.8 in 2018). But, due to the ongoing decline in this share over the last decade, we find that expected householders per capita (x100) dropped by 1.7%, with an equivalent rise in adults per household.
THE RESULT: RISING ADULTS PER HOUSEHOLD
Let’s now summarize our conclusions. Over the last decade, every age bracket has shown some decline in expected householders per capita (x100). These declines are largest for the youngest two age brackets (-9.4% for age 18-24 and -5.9% for age 25-34) and smallest for midlifers and young-seniors (-0.3% for both age 35-64 and age 65-74).
These declines translate into equivalent increases in the number of adults per household. Moreover, as we have explained, they only track adults per household in each age group. When we adjust for changes in the share of adults who live with householders in other age brackets—primarily adult children living with parents—this undoubtedly makes the rise in all adults per household a lot more equal across all nonelderly age brackets.
Our new method of decomposing household formation by age bracket aligns fairly well with the findings of our earlier report. Compared to a simple comparison of population and households by age bracket, our new method slightly understates the decline in householders by a slight margin across age brackets. Without precise figures for the average number of adults per household for every living arrangement within each age bracket, we will never get an exact match. (The comparison suggests that overall household “crowding” may be a bit more severe than indicated by the per-person survey.) But the match is good enough to illuminate most of the salient household lifestyle trends within each age bracket.
It bears repeating that, since the Great Recession, behavior change has pulled down the number of expected householders in each and every age bracket. This drop-off has been significant enough to help reverse a longstanding trend: the steady, decades-long decline in the average number of U.S. adults per household.
Average adults per household rose from 1.92 in 2007 to 1.95 in 2018. That may not seem like much. But when put into context, it stands out as a historical anomaly. Over the prior half-century, adults per household had been on a steady decline, dropping from a high of 2.48 in 1948 (the first year for which we have data) to a low of 1.91 in 2003. The ensuing plateau from 2003 to 2007—and the rise ever since—poses a serious long-term challenge to the housing industry. More people per household, of course, translates directly into less demand for housing units.
The demographic challenges facing the housing industry are well known: Slowing growth in the adult population over the coming years and decades means slowing growth in the number of households. This demographic outlook is virtually set in stone. Even with no change in household behavior, population growth is projected to generate 32% fewer new households in the year 2030 (barely 900,000) than were created in 2018 (nearly 1.4 million). The projection is lower, even, than the very disappointing yearly average over the last five years (just over 1.0 million). Now imagine a future in which behavior increasingly favors “togetherness” lifestyles. The outlook darkens further.
In recent years, the housing industry’s defenders have argued that we just need to be patient. As the economy improves, Millennials will start getting married and having kids earlier. They will move apart from their parents, friends, and relatives. They will start renting or buying more homes. And in so doing they will trigger a resurgence in homebuilding.
But ten years has passed since the last recession, and Millennials show no sign of doing any of these things. More seriously, as we suggested earlier, Millennials and their parents may be ushering in a whole new set of social norms favoring co-living and multigenerational resource sharing. The “nuclear family” was a newfangled fad of the early post-World War II era. No one decreed that it must last forever.
The housing industry’s defenders also sometimes argue that Americans would buy or rent more homes if they became more affordable. But national indexes do not show any dramatic change in affordability for homebuying or for renting. (If anything, homebuying is more affordable now than in earlier postwar decades—even after recent mortgage rate hikes.)
To be sure, these are national indexes. Regionally, there are large and growing gaps between low affordability in coastal and Sun Belt urban centers and high affordability in the heartland. No question: Millennials living and working in San Francisco and Boston and Seattle would probably buy more housing there if prices and rents came down. But these higher prices seem to have reached a fairly stable supply-and-demand equilibrium. Millennials aren't moving out. And homebuilders aren't installing many new dwellings—constrained as they are in these regions by politically popular zoning and regulatory regimes that discourage construction. Where is the evidence that such regimes will weaken in the foreseeable future?
For the construction industry as a whole, the long-term future holds plenty of promise. Corporate plant capex remains strong. Retail will eventually rebound. Aging-in-place Xers and Boomers will keep spending heavily on home remodeling for many years to come. And the medium-term political prospects are improving for an infrastructure boom. But a positive sea change in the construction of new housing units, either for sale or for rental? Don’t bet on it.
