Real Estate Data for the Rest of Us

articles about “Demographics

The Recession’s Lost Generation of Homeowners Isn’t Millennials – It’s the Middle-Aged

Though the published homeownership rate for young adults is still falling, true homeownership among young adults started rising in 2013. Adjusted for longer-term demographic shifts, young-adult homeownership is now at pre-bubble levels, but middle-aged homeownership is lagging.

The latest Census data shows homeownership is still falling for young adults, and the National Association of Realtors (NAR) reports that the share of first-time home-buyers is slipping. While the housing market is clearly improving, with four of the five key indicators of the housing recovery from our Housing Barometer at least halfway back to normal, it looks like the recovery is happening even without much improvement in first-time homeownership. Does that mean the housing recovery isn’t for real?

Not so fast. The official homeownership rate published by the Census gives a misleading picture of homeownership trends. In fact, homeownership among young adults is both on the rise and not too far off from where demographics say it should be. To see this, we did two things in this analysis: (1) account for changes in household formation to get a true measure of homeownership, and (2) adjust for longer-term demographic shifts to compare homeownership levels today with pre-bubble levels.

The answer: our “true” homeownership rate disagrees with the published homeownership rate, and shows that homeownership among young adults increased between 2012 and 2013 after hitting bottom in 2012. However, once we adjust for the huge demographic shifts among young adults – far fewer young adults are married or have kids than two or three decades ago – homeownership in 2013 was roughly at late-1990s levels. That means that the demographic shifts among young adults account for the entire decline in homeownership for 18-34 year-olds over the last twenty years. In other words, if the pre-bubble years of the late 1990s can be considered relatively normal, than today’s lower homeownership rate for young adults might be the new normal, thanks to demographic changes.

But that doesn’t mean all’s well. There may be longer-term damage to homeownership from the recession – but to the middle-aged, not millennials. Homeownership among 35-54 year-olds is lower today than before the housing bubble, even after accounting for demographic shifts. Here’s why.

Young Adult Homeownership Actually On the Rise
The published homeownership rate equals the share of households that own their home instead of rent. It does not, however, capture changes in whether people are dropping out of the housing market to live under someone else’s roof, like those millennials in their parents’ basement, who – in case you missed it – are for real. But if, say, people move out of their parents’ homes and into their own rental apartments, the published homeownership rate would still be falling even if the share of young adults who own remains the same.

Instead, we looked at the true homeownership rate, which equals the number of owner-occupied households divided by the number of all adults; in contrast, the published homeownership rate equals the number of owner-occupied households divided by the number of all households. Of course, the true homeownership rate is always going to be much lower – by half or more – than the published homeownership rate because there are roughly twice as many adults as there are households. The key point, though, is that the published and true homeownership rates can move in different directions if the number of adults per household is changing. That is, in fact, what happened during the recession and recovery (see note #1).

During the recession, as more young people moved in with their parents and fewer headed their own households, published homeownership rate fell from 44.1% in 2005 to 36.8% in 2012 – the 7-point decline was a 17% drop in homeownership. (What we’re calling “published” numbers actually differ slightly from the quarterly and annual homeownership estimates by age group published by the Census – see note #2.) However, the published rate understated the decline: the true homeownership rate for young adults fell from 17.2% in 2005 to 13.5% in 2012 – a drop of 22%.

Then, during the recovery, more young people started to form their own households, primarily as renters. The additional renters pushed the published homeownership rate for 18-34 year-olds down further in 2013 to its lowest level since our analysis begins in 1983:


But our true homeownership rate, which takes household formation into account, turned up slightly in 2013. It’s still the second-lowest year historically, but the tide has turned.


The true and published homeownership rates diverged in 2013 because the headship rate  – the share of young adults who head their own household – rose in 2013 back to the highest level since 2010. The headship rate is still low compared with during and before the bubble, but its recent increase must be taken into account in order to get the trend in homeownership right.

Is Today’s Millennial Homeownership Rate the New Normal?
With our true homeownership rate, we can now determine how much of today’s low homeownership level among millennials is due to longer-term demographic shifts or due to the recent recession. The demographics of 18-34 year-olds have changed dramatically over the past 30 years, between 1983 and 2013, such as:

  • The percent married fell from 47% to 30%
  • The percent living with their own children fell from 39% to 29%
  • The percent non-Hispanic white fell from 78% to 57%

Each of these demographic shifts is a headwind for homeownership. Young people who are married, have children, or are non-Hispanic white are more likely to own a home than among young people who aren’t.

One way to quantify the total effect of these demographic factors on homeownership is to predict what might have happened to homeownership with these demographic shifts if none of the changes in behaviors or circumstances that evolved during the bubble, recession, and recovery took place. In other words, pick a baseline time period for a regression analysis – such as the years just before the bubble started (1994-1999) – and estimate what would have happened to homeownership if the likelihood of owning a home for someone with particular demographics in the baseline time period stayed constant and were applied to the actual demographic shifts that really happened (e.g., fewer people getting married, having kids, being non-Hispanic white; see note #3). This “demographic baseline” shows that demographics alone would have predicted a steady decline in the true homeownership rate over the past two decades, from 16.0% in 1994 to 13.1% in 2013 – a drop of 18%.


As we saw above, the true homeownership rate among young adults rose from 1994 to 2005 – bucking the demographic baseline – then fell sharply until 2012 before turning up slightly in 2013. Surprisingly, this means that homeownership in 2013 was still actually a bit above the demographic baseline – which implies that that the overall drop in homeownership for young adults in the past 20 years can be explained by demographic trends alone. Of course, within those two decades the true homeownership rose thanks to easy credit during the bubble and then plunged during the recession, but the bigger point is that it plunged back down only to just above the underlying demographic baseline (see note #4).

In short: although the share of young adults who actually own a home remains considerably lower today (even with the uptick in 2013) than at any time since 1983, it is roughly at late 1990s levels after taking demographic shifts into account. Unless those long-term demographic trends reverse, there might be little room for young-adult homeownership to increase. You’d have to ignore demographic trends that pre-date the bubble to believe that young-adult homeownership will eventually return to its unadjusted pre-bubble levels.

This also implies that there probably hasn’t been a huge shift in millennials’ attitudes toward homeownership, either, since today’s millennials are roughly as likely to own homes as people with similar demographics two decades ago (see note #5).

Worry Instead About Middle-Aged Homeownership
Here’s the surprise: it’s the middle-aged, not millennials, whose homeownership rate today looks lower than before the bubble. Using the same demographic-baseline analysis, the 2013 homeownership rate for 35-54 year-olds is below the “demographic baseline” (which barely budged over the past 20 years for this age group). Furthermore, homeownership for the middle-aged has not yet begun to turn around as of 2013, unlike for millennials:


Whereas the 2013 homeownership rate for millennials, after adjusting for demographics, is at 1997 levels, the 2013 demographics-adjusted homeownership rate for the middle-aged is at its lowest level in at least two decades (and probably in almost four decades: see note #6).

To see why homeownership is now lagging more among the middle-aged, we repeated the demographic-baseline analysis for five-year age bands. In 2005, the year when the true homeownership rate peaked for most age groups, 25-29 year-olds were the age group for which homeownership was highest relative to the demographic baseline, followed by 30-34 year-olds. These were first-time home-buyers getting easy credit for overpriced homes; then, they bore the brunt of the foreclosure crisis, losing their homes and wrecking their credit history. Eight years later, in 2013, 35-39 year-olds were the age group where homeownership was lowest relative to their demographic baseline; homeownership among 40-44, 45-49, and 50-54 year-olds was also low relative to baseline. Of course, 25-29 or 30-34 year-olds in 2005 grew into 35-39 year-olds in 2013: they are essentially the same people, eight years older.

And that’s the point: the rise and fall of homeownership during the housing bubble and bust is about people who are middle-aged today. The millennial generation was still in their early 20s or younger in the mid-2000s – too young to have bought during the bubble and then to have suffered a foreclosure: Only the oldest among the 18-34 year-old group in 2013 would have been of home-buying age during the bubble.

Turning more millennials into homeowners, therefore, may not be the missing piece of the housing recovery after all. Long-term demographic changes mean that homeownership among young adults is roughly where it should be. The real missing homeowners are the middle-aged.


  1. The true homeownership rate equals the published homeownership rate times the headship rate, which is the number of households relative to the adult population. This post explains in more detail why the published homeownership rate can be misleading.
  2. All data in this post are based on the Current Population Survey’s Annual Social and Economic Supplement (ASEC). What we call the “published homeownership rate” in this post differs from published homeownership data in the Census Homeownership and Vacancy Survey (HVS) because the ASEC and HVS cover slightly different time periods and slightly different survey samples, even though both are based on the CPS. More importantly, we used a different approach for identifying head-of-household than published Census reports do. Whereas Census homeownership rates by age (and other demographics) reflect the demographics of the individual identified in the survey as the householder/head-of-household, we treated married couples where one is identified as head-of-household as co-heads, each weighted at 50%. This does not affect the published or true homeownership rate for all adults since each household still has one head, but it does affect the published and true homeownership rates for specific age groups. Until the early 1990s, the woman in a married-couple homeowner household was identified as the head-of-household less than 10% of the time; in 2013, it was nearly 40%. Because men are, on average, older than women in married couples, the male bias in head-of-household systematically understates the homeownership rate for young adults (e.g. if a 33-year-old woman is married to a 36-year-old man, and he is recorded as the head of household, the homeownership rate among 35-54 year-olds is overstated while the homeownership rate of 18-34 year-olds is understated). We estimate that this bias toward assigning head-of-household to husbands resulted in an understatement of the true homeownership rate among young adults by 10% in 1983 and 3% in 2013.
  3. The analysis requires making some judgment calls about which factors are longer-term demographic trends that would probably have happened regardless of the recession versus recession effects that are likely to reverse in the recovery. We include sex, age sub-groups, marital status, presence of children, race, ethnicity, and nativity (i.e., born in U.S. or abroad) as our demographics. We also include school enrollment and educational attainment as demographic factors, though that’s a tougher call since school enrollment tends to rise in recessions because people can’t find jobs. We did not include employment status or income because those are clearly affected by the recession. Some clear-cut demographic factors – like marital status and having children – are also affected by economic conditions, but the longer-term shifts pre-date the recession. We regress true homeownership on demographics for a single model covering 1994-1999; calculate predicted homeownership based on those coefficients applied to the entire sample 1994-2013; and report annual averages of predicted homeownership (the “demographic baseline”) and the true homeownership rate. The demographic-adjustment analysis begins in 1994, which is the earliest year that the ASEC includes all of the important demographic variables.
  4. Determining whether true homeownership is above or below the demographic baseline depends on which years are chosen as the baseline. As an alternative to using a baseline, we used all years 1994-2013 to estimate the trend in homeownership while stripping out the effect of changing demographics: it shows that young-adult homeownership in 2013 is back roughly to 1997 levels.
  5. The fact that the true homeownership rate for young adults is near the demographic baseline mean, more precisely, that the cumulative, aggregate effect of all non-demographic factors on homeownership is minimal. These non-demographic factors include changing economic circumstances, credit conditions, attitudes, and more. It’s possible that separate non-demographic factors moved in opposite directions, cancelling each other out. In reality, it’s unlikely that attitudes toward homeownership, say, improved while economic circumstances worsened, though possible.
  6. The full demographic adjustment for creating the demographic baseline is possible for years starting with 1994, when all demographic variables were available. The same analysis can be extended back historically if school enrollment and nativity are omitted from the list of demographics. It shows that the demographics-adjusted true homeownership rate for 35-54 year-olds was lower in 2013 than any other year since 1976.
  7. The data analyzed in this post were downloaded from the IPUMS CPS website, which requests to be cited as: Miriam King, Steven Ruggles, J. Trent Alexander, Sarah Flood, Katie Genadek, Matthew B. Schroeder, Brandon Trampe, and Rebecca Vick. Integrated Public Use Microdata Series, Current Population Survey: Version 3.0. [Machine-readable database]. Minneapolis: University of Minnesota, 2010.

Basement-Dwelling Millennials Are For Real

Although a key Census survey counts students in dorms as living with parents, increased college enrollment and dormitory living do not account for the increase in millennials living under their parents’ roofs. Alternative analyses confirm that in the recession millennials have been much more likely to live with their parents than in the past.

The share of millennials – that is, 18-34 year-olds – living with their parents reached a many-decade high during the recession. Last week, an article suggested that these statistics are “criminally misleading” in overstating the increase in millennials actually living with parents because (1) they count dorm-dwelling college students as living with their parents and (2) college enrollment among young people has risen significantly.

Both these points are true: the Current Population Survey’s (CPS) Annual Social and Economic Supplement (ASEC) counts college students who are living in dorms as living with their parents, and college enrollment has indeed gone up. But it does not follow that basement-dwelling millennials are a myth. The ASEC and other Census data show that after adjusting for college enrollment and for dormitory living, millennials were more likely to live with parents in 2012 and 2013 than at any other time for which a consistent data series is available (1986 or 1990, depending on the data source).

Revisiting the ASEC Data
The ASEC counts college students who are living in dorms as living with their parents, so it’s impossible to separate out the dorm-dwellers who were reported as living with parents from college students actually living with their parents full-time. But the survey also reports whether 18-24 year-olds are enrolled in college (in survey years 1986 onward). As a first step, we can exclude all full-time college students from the analysis to make sure we’re not including any dorm-dwellers. Excluding full-time college students, the share of millennials living with parents is still far higher during the recession than at any other time since 1986.


… continue reading


Millennials Are Suburbanizing, While Big Cities Are Having a Baby Boom

Between 2012 and 2013, population growth for 20-34 year-olds was highest in Colorado Springs and San Antonio, while Austin and Raleigh were tops for 50-69 year-olds. But New York, Washington D.C., and Boston all had among the highest growth for 0-4 year-olds.

This morning the Census released its 2013 population estimates by age group for counties, which reveals which local areas are gaining or losing millennials, boomers, and other age groups. Earlier this year, the Census released 2013 population estimates for the overall population – not broken out by age group: at that time we pointed out that the most urban counties had slower population growth than the more suburban counties, even though the most urban counties were growing faster than they did during the housing bubble. (This post and this article explored the broad urban versus suburban trends.)

Today’s new data tell us whether key demographic groups – like millennials (20-34 year olds), boomers (50-69 year olds), and young kids (0-4 year-olds) – might be bucking the broader trend of more suburban counties growing faster than the most urban counties. To measure this, we use the same approach of dividing all U.S. counties into four quartiles based on their household density so that each quartile includes around one-fourth of the total population (see note on county definitions and age groups). Going from the highest to lowest density, the four categories correspond roughly to (1) big, dense cities; (2) big-city suburbs and lower-density cities; (3) lower-density suburbs and small cities; and (4) smaller towns and rural areas.

The punchline: millennial population growth in 2012-2013 in big, dense cities was outpaced by big-city suburbs and lower-density cities and even by lower-density suburbs and smaller cities. Boomer growth in big, dense cities also fell just short of growth in the big-city suburbs and lower-density cities. But the population of kids under the age of 5 grew fastest in big, dense cities. Let’s take a look at each of the age groups.

Millennials Not Flocking to Big Cities
From 2012 to 2013, population growth for millennials (20-34 year-olds) was highest outside big cities. The fastest growth was in the second quartile of counties ranked by density (big-city suburbs and lower-density cities). Furthermore, the third quartile (lower-density suburbs and smaller cities) edged out the top quartile (big, dense cities) for millennial population growth:


… continue reading


Baby-Boomer Downsizing? Perhaps Not So Fast

While the share of 50-to-69 years-olds living in multi-unit buildings rose slightly in 2012 and 2013, the long-term trend among older households shows downsizing getting rarer and happening later in life.

Throughout the recession and recovery, Millennials have hogged the attention: they suffered a particularly bad recession, which delayed their launch into the housing market, slowed overall household formation, and lowered first-time homeownership. But they’re hardly the only demographic that matters for housing. Baby Boomers will help determine the demand for different types of housing and the supply of homes for sale when – and if – they downsize.

This morning, Fannie Mae released a note on boomer downsizing, showing that the share of baby boomers in single-family detached homes has been roughly stable from 2006-2012 (rising slightly on a per-capita basis and falling slightly in the most recent years on a per-household basis). The big question is what happens longer term: are we about to hit a wave of baby boomers selling their single-family homes and moving into apartments and condos? It’s unlikely, for two reasons: baby boomers are still years away from the age of downsizing, and the long-term trend shows that older households today are less likely to downsize than older adults in the past.

Let’s start by looking at the age when older households move from single-family homes to multi-unit buildings. Based on the 2013 Current Population Survey’s Annual Social and Economic Supplement (CPS ASEC) – the most recent detailed demographic data available – baby boomers (born between 1946 and 1964, which means 50-68 years old in 2014) are less likely than almost any other age group to live in multi-unit buildings as opposed to single-family homes. The only age group less likely to live in multi-unit buildings is 70-74 year-olds, which is the age group that baby boomers will start to enter in the coming years.


… continue reading


Over the River and Through the Wood?

Although the recession has forced more young people to live with their parents, demographics – not the economy -- explain why more seniors are living with relatives. The share of foreign-born seniors is rising, and they’re four times more likely than native-born seniors to live with family.

Jed Kolko, Chief Economist
November 19, 2013

The last few years have seen an increase in multigenerational living. Young adults became far more likely to live with their parents during the recession than before and haven’t really started to move out. On the other side of the life cycle, seniors – specifically adults 65 and older – are also more likely to live with relatives than in the recent past. That means fewer Americans today need to go “over the river and through the wood” to see Grandma and Grandpa for Thanksgiving than they did 20 years ago. (Yes, the original poem is actually “through the wood,” not “woods,” and to grandfather’s house. We’ve been singing it all wrong. Have you?) But the reasons that seniors are increasingly likely to live with relatives are totally different from those that lead young adults to live with their parents.

Why are More Seniors Living with Relatives?
According to the Census (2012 ACS), 9% of seniors live in a household headed by their children, children-in-law, or other relatives (other than their spouse). Another 2% live in a household headed by people they aren’t related to, and 3% live in “group quarters” like nursing homes. The other 85% live in their own home.

Another survey, the Current Population Survey’s American Social and Economic Supplement (ASEC), shows the share of seniors living with their children or other relatives has grown over the past 20 years, from an average of 6.6% over the years 1994-1998 to 7.3% in 2013.  These data are volatile year to year, but the overall trend is clearly upward, as the “unadjusted” line in the chart below shows. (Confusingly, different government surveys report seniors’ living arrangements differently. See note at end of post for all the details. But don’t worry: all of the trends and comparisons in this post are based on apples-to-apples analyses.)


Note: based on CPS ASEC 1994-2013, via IPUMS site. See longer note at end of post.

… continue reading