«Contesting the streets Volume 18, number 1 • 2016 U.S. Department of Housing and Urban Development | Office of Policy Development and Research ...»
housing market that has experienced limited changes during the past two decades. The homeownership gap between White households and Black and Hispanic households remains at more than 20 percentage points. Using 1989 data, Wachter and Megbolugbe (1992) estimated that about 80 percent of the gap between White households and Black and Hispanic households can be explained by differences in endowment (including differences in income, education, age, gender, and family type). Gabriel and Rosenthal (2005) found relatively constant compositional differences over time.7 In summary, the decades since the 1960s have seen a stable homeownership rate in the aggregate and across demographic groups, until recently. The recent increase and decrease have brought us back to the homeownership rate of the 1960s. The question is whether this rate heralds a new period of stability or whether fundamental forces are at work that will drive homeownership lower in the long term.
The Future of Homeownership When examining the 15-year period from now to 2030, Goodman, Pendall, and Zhu (2015) predicted a decreasing U.S. homeownership rate to 61 percent. We extend this methodology based on demographic forecasts to 2050 and estimate further declines. The methodology that we use is based on historical decennial census data and projected population by age, race, and ethnicity provided by the U.S. Census Bureau. We capture the potential impact of differing borrowing constraint regimes based on recent headship and homeownership transition rates. Changes in the headship rate and homeownership rate by racial/ethnic and age groups combined enable us to predict the overall homeownership rate based on individual group population projections. We model lending and economic conditions based on two scenarios: (1) a slow transition scenario to headship and homeownership is calculated using the transition rates for 2000 to 2010, while (2) the fast scenario is Coulson (2002) estimated the impact of differences in housing market characteristics (housing value to rent, density, vacancy, share of the population living in suburbs) and socioeconomic characteristics (income, household type, educational attainment, number of children, race and ethnicity, immigration status) in explaining the difference in the homeownership rate across these four regions—Midwest, South, Northeast, West—and across states.
Gabriel and Rosenthal (2005) estimated that differences in credit constraints account for about 5 percentage points of the homeownership gap for Black households and are nearly nonexistent for Hispanic households. Gyourko, Linneman, and Wachter (1999) also found that compositional differences explain a large share of the homeownership gap but that differences reappear when differences in location between White and minority households within metropolitan regions are taken into account.
150 Point of Contention: Declining Homeownership A Renter or Homeowner Nation?
calculated using the transition rates based on the average of the 1990-to-2000 and 2000-to-2010 transition rates. The slow scenario assumes the continuation of the relative difficulty of young households attaining homeownership in the decade of 2000 to 2010, while the fast scenario uses the average of the past two decades, which includes the 1990 to 2000 decade, in which transitions to headship and homeownership were faster (Goodman, Pendall, and Zhu, 2015). These two scenarios reflect different credit conditions resulting from lending practices and economic circumstances that particularly affect the transitions of young households.
Headship and homeownership figures are estimated for nine age groups (each decade from 15 to 24 through 75 to 84, and 85+) and four racial/ethnic groups (White non-Hispanic, Black non-Hispanic, other non-Hispanic, and Hispanic), using the census population projections (U.S. Census Bureau, 2015c). In summary, the projected age- and race-specific headship and homeownership rates from 2020 to 2050 are calculated using the following formula.
Yarst = Yarst-1 + Transitionars,
with Y being the homeownership or headship rate for age group a of group r in scenario s (fast or slow) in decade t (2020, 2030, 2040, and 2050) and Transition being the change in the headship rate for a cohort of a given subgroup between 2000 and 2010 or the average of that transition rate between 1990 to 2000 and 2000 to 2010. The predicted headship rates for each subgroup and scenario in a given decade, Yarst, are then multiplied by the census population projections for each subgroup, Xart, to estimate the projected number of households by subgroup, scenario, and decade.
These estimates are then used as weights to estimate the overall homeownership rate by scenario and decade.
The slow scenario (exhibit 5) predicts that the homeownership rate will decline 4.8 percentage points between 2010 and 2030 (from 65.1 to 60.3 percent) and the fast scenario predicts the rate will decline 2.9 percentage points (to 62.2 percent), with the average scenario predicting the rate will decline 3.85 percentage points (from 65.1 to 61.25 percent). This projected decline Exhibit 5
is driven by a decline in homeownership for most racial/ethnic and age groups, particularly for younger households. It is accentuated by the projected increase in diversity as the homeownership gap remains large, even when considering that the Hispanic homeownership rate is expected to increase slightly in both slow and fast scenarios. These effects more than offset the positive effects on homeownership of an aging population.
Applying this methodology to 2050, the homeownership rate is predicted to decline to 57.9 percent in the average scenario (a 7.2-percentage-point decline from the 2010 level). This prediction is based on the average of a decline to 55 percent by 2050 in the slow scenario, in which the transition rate into household formation and homeownership access remains similar to what has been observed from 2000 to 2010 and a decline to 61 percent in the fast scenario based on the average of the transitions from 1990 to 2000 and from 2000 to 2010 (exhibit 5). Although not insignificant, these projected declines are not consistent with estimates of a 20-percent decline.
Substantial uncertainty surrounds both the census population projection to 2050 and the projection of the headship and homeownership rates for specific groups. The methodology developed by Goodman, Pendall, and Zhu (2015), however, provides some sense of what the homeownership rate might be, based on projected demographic changes and recent historical trends in transition rates from renting to owning that reflect lending condition. These or similar scenarios would play out if, for a given age and racial subgroup, households have similar outcomes in the future as they have had in either the past 10 or 20 years.
One risk to this outcome is an increase in rental and housing costs relative to income. While rents and house price increases historically have tracked inflation, more recently they have exceeded both inflation and median income. Commentary on the likelihood that rents and house prices in the future may increase, relative to inflation and income, points to increases in supply inelasticity in many housing markets as a potential cause for these shifts.
A possible way to incorporate the effect of potential changes beyond variations in demographic and transition rates is to find a case that can serve as a counterfactual if the United States evolves to have consistently more expensive housing costs (both rental and owner) relative to income as a result of increasing housing supply inelasticity. In such a scenario, the homeownership rate is expected to be significantly depressed, because households would be less able to accumulate the downpayment needed to become homeowners.
California today provides an example of what the U.S. housing market might look like in 2050, both in terms of demographics and housing costs if current trends continue over the long run. As of 2010, California’s homeownership rate was 55.9 percent, when the U.S. rate was 65.1 percent.
Coulson (2002) shows that this low level of homeownership is a function of housing market determinants. The rent-to-income ratio is particularly high in California, where, as of 2010, households spent 33.4 percent of their income on rent on average compared with 29.5 percent nationwide.
Also, as of 2010, the median house value in California was $370,900 compared with $179,900 nationwide and the median rent was $1,066 compared with $713. For the United States to reach California levels, house rent would need to increase 1.2 percent on average in real terms during the next 35 years. For perspective on the possibility of this scenario, during the past 5 years, rental
costs, in fact, increased at an annual rate of 1.2 percent in real terms, while increasing at an annual rate of 0.7 percent since 1981 (ratcheting up in the latter half of the 1990s), according to the Consumer Price Index (Bureau of Labor Statistics, 2015).
In the United States as a whole, household growth has been much more rapid than the construction of new housing units, driving up both rents and home prices. These trends seem likely to continue, at least for the near term. The latest numbers for the 2014-to-2015 period, averaging data from the American Community Survey and the Housing Vacancy Survey, indicate household formation of 1.16 million units. The 2014 starts for private single-family and multifamily units were 1.00 million. If we add to those units government-subsidized housing (100,000 units) and manufactured housing (75,000 units), and if we assume a reasonable obsolescence rate (300,000 units), we find that nearly 300,000 fewer units are being produced than the rate of household formation. Although we do not know whether there has been a long-term sectorial shift in housing supply elasticity, if there has been, there will be significant consequences for homeownership outcomes. The baseline scenario assumes an increase in rents or house prices at historic rates, which have been near zero in real terms over the very long term, but which have increased since the second half of the 1990s. We develop the “California scenario,” assuming that these more recent rent and house price trends continue.
The reduction in discretionary income associated with higher rental costs relative to income observed in California can contribute to the development of a “discouraged renter effect.” Households in metropolitan areas with high incomes, high amenities, and high housing costs (Diamond,
2015) pay a larger share of their income on rental costs, reducing their discretionary income and limiting their ability to save for a downpayment. In addition, these households face high housing prices (and requisite downpayments) if they want to purchase in the location where they work.
That is, in order to access the location they value at that point in their life cycle (access to job, consumption amenities), households can either share a dwelling with other family or nonfamily members to limit the share of their income spent on rent and/or save less, taking longer to save for a downpayment. Even if the household accesses one of the low-downpayment programs and can save for that, they may not have the income to qualify for the mortgage at current prices. Thus, the combination of high rental and purchase costs can lead households to remain renters longer, delaying both household formation and homeownership and potentially precluding their ever becoming homeowners. Although California is currently a relative outlier in terms of homeownership, with only the state of New York having a lower homeownership rate, by 2050 the United States is expected to have a similar demographic makeup to the one found in California today. We have taken account of the changes in demographic makeup into our projections, but we have not taken into account a sharp increase in rent-to-income and house price-to-income ratios. If the United States experiences the same housing cost-to-income ratios as California, the nation might reach similar or lower levels of homeownership—what we call the California scenario.
For the California scenario, we combine the California headship and homeownership levels by age and racial groups and its historical transition rates (the same variables used at the national level discussed previously, but for California) with the projected individual population data for the United States to predict the U.S. homeownership rate. In these scenarios, in which the United States experiences headship, homeownership, and transition rates similar to California’s, the national
Cityscape 153Acolin, Goodman, and Wachter
homeownership rate is projected to decrease to 52.6 percent by 2050 in the fast scenario and to 47.7 percent in the slow scenario. In this case, the United States would be a nation of renters by 2050.
Although not a 20-percentage-point decrease from the 2015 homeownership rate, this decrease does represent a more than 20-percentage-point decline from the highs of the early 2000s. Such a scenario would create a drastically different housing landscape in the United States from today’s.
Factors That May Affect the Predicted Homeownership Rate Projection of homeownership rates requires making predictions about the effects of many uncertain parameters. The standard user cost model provides a framework to identify household tenure decision based on maximizing utility, but it still requires making assumptions about the evolution of variables such as the relative price of housing services. In addition, the user cost only partially predicts household decisions, which are also influenced by credit availability.