Wednesday, July 22, 2015

For Housing Demographers It’s All About the Data – But Sometimes the Data Come Up Woefully Short

by George Masnick
Senior Research Fellow
Housing demographers are often frustrated by data that range from inconsistent to totally unavailable when attempting to research demographic and housing trends. The inconsistencies between various data sources on estimates of household numbers and household growthvacancy rates, and homeownership rates are well documented and continue to be dissected and discussed, but there are other metrics that have been even more elusive to pin down that would help enormously to better understand today’s demographic/economic trends and their housing implications.

Two broad areas of housing consumption are particularly difficult to measure.  The first concerns the doubling up of generations living in a single residence.  The second is the opposite – when a single household lives in more than one housing unit on a regular basis. 

We would like to be able to answer many questions about the increasing trend of young adults who live with their parents. We have also identified a growing trend of grandparents who live with their grandchildren (and in many cases the grandchildren’s parent or parents as well) but we cannot identify grandparents who might not live with their grandchildren but live close by and provide support in childrearing.  We would like to know more about how delayed marriage/partnering, and divorce/remarriage, affect housing consumption of multiple units.

For the most part, existing data cannot tell us what actually takes place in the housing history of specific households over time as individuals age and change their household configurations, marital/partnership status, and employment/income profiles.  Most difficult to measure are the linkages between generations in structuring patterns of geographic mobility (affecting both those who move and those who do not move in order to be close to family), young adult household formation, and housing consumption across the age spectrum.

Data sets that do have information about life-course household transitions and some information about relationships between generations rarely have any data on housing.  Data sets that have housing information lack information about life-course changes preceding and during current occupancy.  Information about extended family members that do not reside in the household being interviewed is generally totally lacking.

We would like to know not only how many adult children presently live with their parents, but how many have boomeranged and the type of housing boomerang children moved out of when they moved back home.  How often does moving back home occur for specific households, and how long does it last?  Short spells of returning home presumably have much different consequences than long ones.  Chronic returns might have very different causes and consequences than one-off situations.  Returns to large houses with higher-income parents have different consequences than returns to small homes having low household income. 

We would like to know more about the background details of children when they leave a parental household – reasons for leaving, characteristics of housing (on both ends of the move), and, household size and composition (again on both ends of the move).  Are boomerang kids and their household/housing characteristics different from those who leave and do not return?  This information would be immensely helpful in better understanding the present and future housing consumption of those Millennials who have been slow to form independent households and become homeowners.  

Ideally, answers to the kinds of questions just raised require panel data that track individuals and their housing over time.  The few nationally representative panel surveys with public use micro data, such as the Panel Study of Income Dynamics (PSID) or the National Longitudinal Survey of Youth (NLSY), have limited housing data.  And even these surveys have historically been deficient on the collection of data on individuals and their extended families: the PSID has collected data at regular intervals since 1968, but only in 2013 added a Family Roster and Transfer Module in which respondents and their spouses are asked to enumerate all living parents and children over 18 and to report about recent and long-term transfers of time and money to these individuals.  The new PSID module is the first to fully enumerate all biological, adopted, and step-relationships of parents, parents-in-law, and adult children, and it is the first major data collection effort on transfers of time and money in the PSID since 1988.

Other efforts to assemble panel data to directly study co-residence patterns between adult children and parents, such as in a recent Federal Reserve Bank of New York Report (utilizing its own Consumer Credit Panel (CCP)) are neither a nationally representative sample of all households nor available to other researchers, raising concerns about the reliability of the data. For example, the CCP data set reports a much higher rate of co-residence than other data sources such as the Current Population Survey.  Still, because of the scarcity of panel data to answer some of our questions, data sets the FRBNY CCP cannot be entirely dismissed.

Another reason for growing inter-generational co-residence is the need to support and take care of grandchildren.  Increasing grandparent-grandchild co-residence certainly has important consequences for housing choices, further postponing independent household formation among some Millennials, and perhaps delaying downsizing among the Baby Boomer grandparents.  We would like to know if older Americans who live close to their grandchildren are different from grandparents who live with their grandchildren.  Do they also play financial and childcare roles with respect to their grandchildren?  Are retirees more likely to move to be close to their children if their grandchildren are young?  Does the existence of young grandchildren make retirement moves that put greater distance between them and their grandchildren, or that are to age restricted communities, less probable?  Do older empty nesters with young grandchildren actually downsize less than those with older grandchildren? 

Available data on the rise of co-resident grandparents and their grandchildren are mostly from cross-sectional surveys, like the Current Population Survey, and are biased toward intergenerational families where those in the first wave of Millennials had their children relatively quickly, often while teenagers or still in school, or when not yet absorbed into the labor force.  Such early births are more likely to be to parents without a college education and to be non-marital.  The large and growing share of Millennials who pursue higher education are more likely to postpone childbearing (thus postponing grandparenthood for many Baby Boomers) and their births are more likely to be marital.  Will first grandchildren who come along later in life, when their parents are older and more economically secure and their grandparents are more likely to be retired, be more or less likely live with grandparents?  Live close to their grandparents?   

Unfortunately, nationally representative data that allow us to identify who is even a non-coresident grandparent are practically non-existent.  The sole exception is from the Survey of Income and Program Participation (SIPP), a longitudinal survey following panels of households for 2½-to-4 years, which for three of its panels has asked if a person has any biological children and if those children, in turn, have any biological or adopted children.  The SIPP data overlook persons who are not biological grandparents but are grandparents through marriage, either as stepparents themselves or who became a grandparent when their children partnered with someone who already has children, but has not adopted them.

Analyses of SIPP data on grandparents have been published for the 2001 panel, the 2004 panel, and the 2008 panel.  A 2014 panel is now in the process of data collection.  These data estimate that there were 64 million grandparents in 2009 (second wave questionnaire of the 2008 panel), of which one-in-ten lived with their grandchildren. According to these data, only 22 percent of co-resident grandparents were over the age of 70 compared to 34 percent of grandparents who did not live with grandchildren. We have little insight into proximity of these non-coresident grandparents to their children and the housing choices they have made if they have recently moved.

One additional panel survey that collects data to answer questions about grandparents is the University ofMichigan’s Health and Retirement Study (HRS).  It does allow identification of all grandparents, co-resident grandparents, reasons for moving, and does have some housing data, but it is somewhat limited by a sample design that selects particular birth cohorts.  Still, more analysis of these data, collected annually from 1992 to 1996 and on alternate years since, can help us better answer some of our questions.  

Shifting gears, how people utilize multiple housing units (their own and/or other’s) at different times during the week, month, or year is almost a complete mystery.  We would like to know more about middle-aged and older people who sometimes dwell in two or more housing units while maintaining control of each.  There is a catch-all category of households in some data sets identifying people who have a primary or usual residence elsewhere, and this category has been growing in recent years.  However, households interviewed at their primary residence are not asked if they sometimes live in another home, and data are not collected about the characteristics of that home and the reasons for living in it.  Are grandparents spending some time living with their grandchildren on a regular basis, or buying or renting a residence that they occupy occasionally to be close to their young grandchildren?  Are retirees who once lived close by their young grandchildren retaining a previous residence for a longer period of time to facilitate occasional visiting after retirement migration?  Are more adult children still living in retirees’ previous homes after they retire and move elsewhere?

We would also like to know how long individuals maintain an active consumption of multiple housing units when they change jobs or form new relationships.  Is “living together” increasingly less a status and more a process that could involve two or more housing units over an extended period of time?  Is the rise of long-distance telecommuting predicated on being able to spend some time in two or more locations, and is it the case that housing units in multiple places are owned or rented to facilitate this?  If people are now better able to rent out or share housing with others on a part-time basis (for example, through VRBO and Airbnb), are they more likely to maintain multiple units for their own occasional use? 
Until panel surveys from nationally representative samples collect data on life-course transitions, on intergenerational relationships, and on housing consumption more broadly defined, analysts will continue to try to research trends with data that are usually inadequate to the task, and to have questions that simply cannot be answered.

1 comment:

  1. Do you work with the Boston Research Data Center? Seems like the geocoded SIPP would be helpful to you. People at the census are currently working on connecting that to tax records, which also would help you to create a panel. And I agree - as a housing economist we need better data.