Census Note 8: Advances in Homeownership Accross the States andFannie Mae Foundation Census Note 08 (October 2001)*Advances in Homeownership Across the States
and Generations: Continued Gains for the Elderly and
Stagnation Among the Young Summary of Findings Analysis of homeownership trends by age in the 50 states using new data from Census 2000 indicates that:
Introduction The decade of the 1990s witnessed the largest national gain in the homeownership rate since the 1950s. As detailed by Simmons (2001a), this increase of 2.0 percentage points was especially significant because it followed an actual decline in the homeownership rate during the 1980s. The increase was widespread, reflecting a "coast-to-coast expansion" in homeownership (Simmons 2001a). This Census Note probes beneath the surface of the nation's rising homeownership rate to inquire more about its distribution. One key question concerns what age groups benefited from the expansion of homeownership. During the homeownership slump of the 1980s, young adults of the baby boom generation experienced sharply lower homeownership rates relative to the preceding generation, while at the same time elderly groups enjoyed large increases in homeownership. Did this generation gap continue to grow in the 1990s, or did the increases in homeownership help young adults to reverse some of their previous declines? The overall national gain in homeownership may have been equally spread across all age groups, or it may have been concentrated in a single age group. Yet another possibility is that not a single age group experienced an increased homeownership rate; rather, the overall increase could have been generated by an age distribution that shifted toward older age brackets (the middle-aging of the baby boom generation) where homeownership is higher than at young ages. A formal compositional analysis can help us understand exactly how much of the overall national gain in homeownership stems from each possible factor. The pattern of generational gains across the 50 states is also of interest. The widespread prevalence of rising homeownership suggests the young and elderly may have faired equally well everywhere, but Simmons (2001a) shows that some states experienced especially sharp reversals in homeownership trends. Did the gains of the 1990s for young households occur in the states where they fared relatively the best in the 1980s, or did the gains of the 1990s pull up the states that were weakest in the 1980s? In addition, the trend in house prices was much more favorable in some locations than others, and this has a particular impact on young adults. How did prices relate to changes in homeownership rates, and was that relationship any different than what was observed in the 1980s? As a final question, one wonders about the neglected, hidden effect of household formation on trends in homeownership rates. All other things equal, states with rising household formation will tend to have more renters and a falling homeownership rate. Conversely, if renters drop out of the housing market, leaving relatively more homeowners, the homeownership rate will rise. Is it possible that the gains in homeownership of the 1990s partly reflect the hidden effect of losses in household formation? In what states might this be true? Data currently available from Census 2000 make possible a preliminary analysis that provides some initial answers to these questions. Pending the Spring 2002 release of household income data by age and race from Census 2000, we can gain useful insights by probing deeper into the limited data now available. This may help to inform the direction of future investigations. Previous Findings on Homeownership Trends The trend in homeownership during the 1990s was a welcome reversal of changes in the previous decade. During the 1980s, the national homeownership rate declined, albeit by only 0.2 percentage points, for the first time since the Great Depression. The declines in the 1980s were widespread; only 18 states experienced increases. In contrast, during the 1990s, the national homeownership rate rose by 2.0 percentage points and all states but one (Arkansas) registered an increased homeownership rate (Simmons 2001a). The 1980s also had witnessed a striking polarization of homeownership opportunity, with sharp declines in the homeownership rate for ages younger than 45 (declines of more than 5.0 percentage points) accompanied by sharp increases of comparable magnitude among the elderly (Myers et al. 1992). These divergent trends led to a greatly expanded generation gap in homeownership opportunity. It might seem surprising that the elderly could fare so well in the 1980s at a time when the young adult age groups, then occupied by the large baby boom generation, were experiencing such declines. The best explanation was that the elderly had purchased their homes in earlier decades when they were middle aged. As they aged into their elderly years, these cohorts carried with them much higher homeownership than was characteristic of the generation that preceded them. Previous research has shown that the homeownership rates of those in their 50s during the previous decade are far stronger determinants of homeownership rates among those who are currently elderly than are current income and prices (Pitkin 1990). According to this lagged cohort interpretation, elderly homeownership rates were likely to continue rising during the 1990s. Conversely, during the 1980s, young baby boomers were entering homeownership for the first time and faced market trends full square. Their purchase of homes could have been depressed by lagging incomes, but the dramatic increase in paid employment among women created a preponderance of two-earner couples able to compete well in the housing market. In addition, the upward trend in prices during the 1980s may have discouraged home purchases by reducing homeownership affordability, but that explanation also was not supported by the evidence. Perhaps the most remarkable finding about homeownership trends in the 1980s was that homeownership rates among the young declined the most in places where prices fell most sharply. In contrast, in states where house prices escalated the most, young adults' homeownership rates held to previous levels or even increased (Myers and Wolch 1995). The direction of causation is worth debate. On the one hand, growth in the number of home buyers relative to supply might have led to an increase in home prices. On the other hand, falling prices might have reduced the number of willing home buyers. Theory and evidence both suggest that home buying was depressed in the 1980s by the poor investment prospects in many states. The largest component in the user cost of housing is anticipated appreciation, not the monthly mortgage payments. When prices are declining, young households would be rational to wait before committing to a home purchase, and, hence, homeownership rates will decline. The prospects for young adults in the 1990s were uncertain. Prices trended upward in many more states than in the 1980s, suggesting that the investment incentive for home purchase would lead to increased homeownership rates. However, the effect of investment incentives may have been offset by growing numbers of young Latino, black, or immigrant households that historically have lower homeownership rates (Simmons 2001b). At the same time, new outreach and homeowner education programs and new mortgage instruments might have reduced prior disadvantages of those groups. The net effect on homeownership rates of the young are uncertain for the 1990s. Methods Defining and Measuring Homeownership and Household Formation Homeownership is one of a handful of variables for which the Census Bureau seeks to obtain information from a 100 percent sample of residents during the census. The resulting data are free of sampling error and are among the first released from Census 2000. Summary File 1 (SF1) of Census 2000 and similar tabulations from the 1980 and 1990 censuses are used to measure homeownership trends. A housing unit is owner-occupied if the owner or co-owner lives in the unit, even if it is mortgaged or not fully paid for (U.S. Bureau of the Census 2001). All other occupied housing units are classified as renter-occupied. Traditionally, the homeownership rate is defined as the percentage of all occupied housing units that are owner-occupied. The ownership rate is extended to apply to people as well as housing units in the following manner. Every occupied housing unit contains one household, and every household has exactly one householder. The latter is the person in whose name the housing unit is owned or rented.1The characteristics of the householder can then be used to categorize households. In this Census Note the homeownership rates of different age groups are compared. Individuals can hold one of three statuses: owner householder, renter householder, or nonhouseholder. The ownership rate is measured by the ratio of owner householders to all householders. This is an incomplete measure of housing status, however, because nonhouseholders are not part of the equation. A supplementary measure is the ratio of total householders to all people. Often termed the headship or householder rate, this measure reflects the rate of household formation. Homeownership and headship rates of different age groups can be compared, and the changes in homeownership and headship rates can be juxtaposed within a single age group. Falling headship alters the household base on which homeownership rates are computed and thus has potential to increase homeownership rates. 1 In cases in which multiple parties hold that status, the household designates only one person to hold the status of the reference person (householder) for the household. Decomposing Aggregate Changes in Homeownership The overall changes in homeownership rates can be decomposed into portions due to increases of rates within age categories and, alternatively, portions due to the effects of changes in the distribution of households across age groups. Here we describe a method also employed in Simmons (2001b). This decomposition method is used to answer the questions: (1) How would the aggregate homeownership rate have changed during the 1990s had the age structure remained as it was in 1990? And, (2) How would the aggregate homeownership rate have changed during the 1990s had homeownership rates of each age group remained as they were in 1990 (but age distributions shifted)? Answering these questions is important because shifting age structures during the 1990s placed more households in brackets in which higher homeownership was likely. The following mathematical relationships are used to answer
the questions:
Where the variables are defined as follows:
We will compute the third equation for each separate age group, and then sum across all age groups to yield results for total U.S. households.2 Correlation Trends for 50 States Useful inferences can be drawn by studying the variation of different trends across the set of 50 states. Correlation analysis reveals the extent to which larger values of one variable are associated with larger values of another variable, a positive correlation, or when two variables are inversely related, a negative correlation. When the variables in question reflect changes over time -- time trends -- the direction and magnitude of changes can be correlated for two different variables. 2 The equation is not a full decomposition because an additional term is needed to represent the interaction of changes in rates with changes in composition. As now stated, the equation folds the interaction in with the overall composition effect, making the estimate of the aging effect conservative. (When summed across age groups, the interaction term is negative.) Nonetheless, in this case, the interaction effect is very slight and its contribution to the overall decomposition is minimal. Observations used in computing correlation coefficients are the 50 states. The District of Columbia is excluded because it is a central city, not a state, and functions as an anomalous outlier. The correlation coefficient can be supplemented by graphic displays in the form of scatterplots that show the two dimensions being associated. A regression trend line superimposed on these scatterplots reflects the central tendency of the association computed by the correlation coefficient. Results Age Effects on the National Increases in Homeownership Homeownership rises markedly over the life cycle. The differences recorded between age groups far exceed those recorded between major racial or ethnic groups. For example, in 2000, the homeownership rate of all Hispanics was 45.7 percent, while that of non-Hispanic whites was 72.4 percent, a difference of 26.7 percentage points (Simmons 2001b, table 1). Much larger differences are found between age groups, as shown in table 1. In 2000, even if ages 15-24 is ignored, the difference in homeownership rates between ages 25-34 and 55-64 amounts to 34.2 percentage points. A similar difference pertained in 1990, but the gap was substantially smaller in 1980.
During the 1990s, despite the nation's gain of 2.0 percentage points in the overall homeownership rate, virtually no change was recorded in any age group between 25 and 64 (table 1). This lack of change is an improvement relative to the sharp declines in homeownership rates registered in the 1980s. The relative stability is also notable in light of ongoing shifts in composition among young adults that are not favorable to the homeownership rate. These shifts include fewer married couples, increased prevalence of immigrants, and higher shares of minority households. Among the elderly, homeownership rates did increase substantially -- by 2.9 percentage points. During the preceding decade of the 1980s, elderly homeownership increased even more, while an opposite trend of large declines was recorded for age groups under age 55. This polarization of housing status between the generations during the 1980s was disguised by the minimal decline of only 0.2 percentage points in the overall homeownership rate (Myers and Wolch 1995). How exactly was the gain of 2.0 percentage points achieved in the overall homeownership rate during the 1990s? Given little change in homeownership rates among middle-aged and younger adults, it would appear that the elderly were the sole group to account for the overall increase. And yet the elderly account for only 21.0 percent of the nation's households in 2000 and are not sufficiently numerous to single-handedly drive the nation's homeownership rate. The sharp differences between homeownership rates of successive age groups also creates the potential for major impacts due to changing age distribution of the population. Most prominently, 45 million of the 105 million households in 2000 were headed by members of the massive baby boom generation (born 1946-1964). Positioned in the 25-44 age bracket in 1990, those households marched ahead to ages 35-54 in 2000. Did their shift into higher homeownership age groups have something to do with the national increase in homeownership? The overall 2.0 percentage point change in homeownership rate can be decomposed into the portion due to changing rates at each age and the changing distribution of households across the successive age groups. For example, with the arrival of the front half of the baby boom generation in the 45-54 age bracket, that age category's share of all households increased from 15.6 percent of all households in 1990 to 20.2 percent in 2000. The high homeownership rate at ages 45-54 thus assumed greater weight in the overall national homeownership rate in 2000 than in 1990. Conversely, with the departure of the back half of the baby boomers from the 25-34 age bracket, that age category's share of the nation's households fell from 21.6 percent in 1990 to 17.3 percent in 2000. As a consequence, that age group's lower homeownership rate assumed less weight in the national total. A systematic decomposition of these effects -- changing homeownership rates at each age and the changing share of all households located in each age group -- is provided in table 2. Of the overall increase in the nation's homeownership rate of 2.0 percentage points, 0.8 percentage points is attributable to increases in homeownership rates within age groups, while 1.2 percentage points is attributable to changing shares of households in different age groups. Of the total gain due to changing homeownership rates, virtually all of it is confined to ages 65-74 and 75 or older. Of the total gain due to the changing share of households located in different age groups, the vast majority is located at ages 45-54, which was entered by the baby boom generation between 1990 and 2000.
In sum, the national gain of 2.0 percentage points in homeownership rate during the 1990s was heavily shaped by age effects. Only the elderly enjoyed an increase in homeownership rates, and that was less than in the 1980s. Among younger age groups, previous declines of the 1980s were stemmed, but no increases were achieved, save among the small group under age 25. Nonetheless, it is remarkable that no further declines were experienced in the 1990s, given the growing presence among young adults of nontraditional families, immigrant groups, and members of racial and ethnic minorities, all of which traditionally have lower homeownership rates. Instead, the bulk of the national increase came from the simple middle-aging of the baby boom generation as it passed into age brackets characterized by higher homeownership rates. Age Group Trends in Homeownership at the State Level The national pattern of homeownership change by age group was not followed in all 50 states, despite the widespread gains in the overall homeownership rate during the 1990s. All states but one (Nebraska) experienced gains in both the 65-74 and 75 and older age brackets (figure 1). However, more than half the states experienced declines in the homeownership rate in the 35-44, 45-54, and 55-64 age brackets. At the same time, 32 states experienced gains in homeownership in the 25-34 age bracket. Nonetheless, this is a huge turnaround from the experience of the 1980s, when the vast majority of states suffered losses of homeownership in age groups under 55 (figure 1).
A key question is the degree to which states' homeownership rate trends for each age group during the 1980s continued during the 1990s. Overall, states' homeownership gains of the 1990s were inversely related to those of the 1980s (Simmons 2001a). This is signified by a negative correlation coefficient (r=-0.32) in table 3. The strongest negative correlations are for households under age 55. This signifies that states that suffered the deepest losses in homeownership among young and middle-aged households during the 1980s tended to enjoy the strongest rebounds in the 1990s.
Illustrating this point, figure 2 portrays the changes observed at ages 25-34, locating each state's percentage point change in homeownership rate in the 1990s (vertical axis) versus the change recorded in the 1980s (horizontal axis). States that recorded losses in both decades are in the bottom left quadrant (15 states); those that recorded gains in the 1980s, followed by losses in the 1990s, are in the bottom right quadrant (3 states). In contrast, 32 states rebounded with gains in the 1990s after experiencing a declining homeownership rate at ages 25-34 in the 1980s (upper left quadrant). No state experienced gains both decades. What figure 2 makes clear is that states Number of States with the largest relative gains in the 1980s tended to have the deepest losses in the 1990s. Conversely, states with the deepest losses in the 1980s tended to rebound with the strongest gains in the 1990s. This inverse relationship is signified at ages 25-34 by a negative correlation (r=- 0.59).
Among the elderly, the correlations of homeownership trends indicate a positive association between trends of the 1990s and trends of the 1980s (table 3). In other words, in states where the elderly were more advantaged in the 1980s, they tended to experience greater gains in the 1990s. This exacerbates the polarization between generations that set in during the 1980s. In contrast, among the young, negative correlations between trends of the 1990s and 1980s suggest the possibility of catch-up as age groups in previously lagging states did relatively better. Correlation with House Price Trends The positive association between advances in homeownership rates and rises in house prices was one of the more remarkable findings from the 1990 census. Contrary to assumptions based on nominal affordability, in states with rising house prices, young adults achieved rising homeownership, most likely because they perceived greater incentive to invest in Did this pattern continue into the 1990s? The state pattern of house price appreciation virtually reversed in the nation between the two decades.3 During the 1980s, the northeastern states, plus California, led the nation in price increases. During the 1990s, the situation reversed, with those same states leading the nation in price decreases. Overall, only 14 states experienced price gains in the 1980s, with the highest appreciation rate in Massachusetts (72.3 percent increase). The state with the worst price performance was Wyoming, which experienced a 52.8 percent decrease. During the 1990s, 32 states experienced price gains, led by Oregon with a 57.7 percent increase. The weakest price trend occurred in Connecticut, which had a 31.5 percent decline. Overall, there was much less dispersion in price trends for states in the 1990s than occurred in the 1980s. During the 1980s, the overall correlation between homeownership rate changes and price trends across the states was 0.71. During the 1990s, this overall correlation was reduced to 0.30 (table 4). Among age groups younger than 55, however, the homeownership-price correlation did not diminish nearly so much, falling at ages 25-34, for example, only from 0.75 in the 1980s to 0.65 in the 1990s. This remains a strong association between price trends and trends in homeownership rates. How much this might be related to trends in incomes within each age group will not be known until the release of sample, "long form" data from Census 2000.
To illustrate the association of homeownership and price trends, a scatterplot shows the changes in homeownership rate observed at ages 25-34 (vertical axis) versus the house price trend (figure 3). This is a double plot: one set of points represents 50 state observations for the 1990s (dark dots), the other, state observations for the 1980s (light dots). The two superimposed trend lines have virtually the same slope, but the entire distribution for the 1990s is shifted upward with about five percentage points greater increase in homeownership rates than was typical in the 1980s. As mentioned, the distribution of price trends is much less dispersed in the 1990s, and many fewer states fell in the lower left quadrant characterized by both falling prices and falling homeownership. 3 Constant quality prices are based on a repeat sales index maintained by the Office of Federal Housing Enterprise Oversight. These are expressed in 1989 dollars by using the shelter component of the Consumer Price Index.
Once again, the pattern observed for the elderly is very different relative to younger households and relative to the 1980s. Among the elderly in the 1990s, the correlation with house price trends actually turned negative: smaller gains are found in states with larger increases in prices. This is surely tied to the earlier reported finding that the correlation between state homeownership trends for the young and old broke down in the 1990s. The reversal in price trends between the 1980s and 1990s affected young and old very differently. Those who were just starting out in their housing careers would be most affected by the 1990s trend. However, those ages 65-74 in 2000 would have been strongly impacted by the trends of the 1980s and earlier. When that cohort entered this age group in 2000, they brought the benefits of higher homeownership accrued in the past. Thus, for the elderly the negative correlation between homeownership rate changes and current price trends may well be a spurious effect of a lagged relationship. Hidden Effect of Household Formation Trends on Homeownership Homeownership status is an incomplete, and potentially misleading, indicator of overall housing fortunes in a state. Usually, we assume that a rising homeownership rate signals improving wellbeing with regard to housing, but it is possible that rising homeownership could signal growing housing problems for those least able to afford homeownership. Recall that the homeownership rate (percent of households that are homeowners) is only computed for households. If some people are discouraged by economic or other conditions from forming households, they are excluded from this calculation. The problem is that new or forgone household formation primarily affects the number of renter households. Thus, shrinking household formation could raise the homeownership rate by eliminating renters. It is one thing if the reduced number of renters reflected transfers into homeownership; it would be quite another if the reduction in renters reflects drop outs from the housing market who are now forced to double up with friends and relatives. To test for this potential hidden effect, we measure the propensity for household formation by the per capita headship rate, i.e., the percent of people in an age group who are householders. In fact, during the 1990s, gains in homeownership were negatively associated with gains in household formation for the 50 states (r= -0.33). A stronger negative relationship prevails among younger and early middle-age adults (r= -0.40 at age 25-34 and r= -0.50 at age 35-44). Nationwide, the largest declines in household headship occurred at ages 35-44, falling from 54.3 percent of people in 1990 to 53.1 percent in 2000. Nineteen states, including many of the nations' largest -- California, New York, Texas, Illinois, and Michigan -- experienced declines in household formation at this age that were even greater than the national decline. Only one state (Maine) experienced an increase, albeit by a very slight 0.15 percent, in household formation. What is noteworthy is how closely the declines in household headship correspond to increases in the homeownership rate (figure 4). The effect is striking. Although 25 states experienced both declines in household headship and in homeownership rates (the lower left quadrant), the states with the deepest declines in headship tended to have the steepest increases in homeownership, and vice versa. Two states appear to deviate from the overall pattern. The outlier in the far lower left is Hawaii, the state that generally has the highest housing prices and greatest housing unit overcrowding problems in the nation. The other outlier is California. Rapid racial and ethnic change, fueled by the highest immigration in the nation, may be driving down both household formation and homeownership in that state. The reason for this strong relationship between rising homeownership and falling household formation is unclear. One suspects that would-be renters are dropping out of the housing market. Once more detailed data are released from Census 2000, we can investigate changes in household and family structure for each age group, and each race, to learn whether the ongoing changes may be driven by other social changes.
Conclusion Large differences in homeownership between age groups remain one of the sharpest divides in housing well-being in this nation. The passage of the large baby boom generation into middle age has carried many more households into age brackets where higher homeownership is common. This "middle aging" process accounts for most of the national increase in homeownership during the 1990s. Nonetheless, the polarization between age groups continued to increase during the 1990s. Although the 1990s witnessed a cessation to the steep declines in homeownership rates that were recorded for younger adults in the 1980s, ages between 25 and 54 did not recover any of their former losses. At the same time, homeownership among the elderly continued to surge ahead. The age pattern of homeownership attainment varied considerably across states, particularly for the young. Many more states achieved increases in homeownership among young households in the 1990s than in the 1980s. An important finding is that among young and middle-aged adults, increases in homeownership in the 1990s tended to be highest in states that had suffered the greatest losses during the preceding decade. Among the elderly, however, states with larger gains in the 1980s tended to have larger gains again in the 1990s. Two additional associations were uncovered by this analysis. One revealed the continued effect of house price increases on gains in homeownership rates for the young and middle-aged. Apparently for reasons of investment motivation, homeownership rates rise the most in states where prices trend upward. The second finding discloses the hidden effect of household formation on homeownership trends. Homeownership rates are boosted when fewer households are formed in an age group, apparently because the foregone households would have been renters. Thus, a rising homeownership rate could signal both success at the top and distress at the bottom of the housing market. This Census Note provides only a summary sketch of many important differences in homeownership rates between the generations and the states. As more detailed data are released from Census 2000, it will be possible to delve more completely into relationships reported here. Author Dowell Myers is Director of the Master of Planning Program in the School of Policy, Planning, and Development at the University of Southern California. He also heads the School's population dynamics research group. Excellent research assistance was provided by Zhou Yu. The author gratefully acknowledges the helpful advice and comments of Patrick Simmons. *About the Census Notes Series The Fannie Mae Foundation's Census Notes series provides timely analyses of Census 2000 data to stimulate discussion and further research. Although Census Notes are reviewed internally and on an informal basis externally, they have not been subject to the formal process of external peer review that is commonly used for the Foundation's research publications. Therefore, they should be viewed as works in progress and their findings considered preliminary. References Joint Center for Housing Studies of Harvard University. 2000. State of the Nation's Housing. Cambridge, MA: Joint Center for Housing Studies. Myers, Dowell, Richard Peiser, Gregory Schwann, and John Pitkin. 1992. Retreat from Homeownership: A Comparison of the Generations and the States. Housing Policy Debate 3(4):945-75. Myers, Dowell, and Jennifer R. Wolch. 1995. The Polarization of Housing Status. In State of the Union: America in the 1990s: Volume One: Economic Trends, Ed. Reynolds Farley, Chapter 6, pp. 269-334. New York: Russell Sage Foundation. Pitkin, John R. 1990. Housing Consumption of the Elderly: A Cohort Economic Model. In Housing Demography: Linking Demographic Structure and Housing Markets, ed. Dowell Myers. Madison, WI: The University of Wisconsin Press. Simmons, Patrick A. 2001a. A Coast-to-Coast Expansion: Geographic Patterns of U.S. Homeownership Gains During the 1990s. Fannie Mae Foundation Census Note 05, June 2001. Washington, DC: Fannie Mae Foundation. Simmons, Patrick A. 2001b. Changes in Minority Homeownership During the 1990s. Fannie Mae Foundation Census Note 07, September 2001. Washington, DC: Fannie Mae Foundation. U.S. Bureau of the Census. 2001. 2000 Census of Population and Housing: Summary File 1 Technical Documentation. SF1/01. Issued May 2001. |
| All Content © 2007 Fannie Mae Foundation. All rights reserved. |