Misleading comparisons of homeownership rates between groups and over time: the effects of variable household formation

Despite ominous signs of housing market stress, the homeownership rate reached an all time high in 2006. We seek to understand whether the conventional definition of homeownership, which is based on the share of households and ignores the effects of variable household formation, has confounded the assessment of owner-occupied housing. We find that, from 1990 to 2000 and especially from 2000 to 2006, declining household formation led to an elevated homeownership rate in the U.S.

UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript Misleading Comparisons of Homeownership Rates between Groups and Over Time: the Effects of Variable Household Formation Abstract Despite ominous signs of housing market stress, the homeownership rate reached an all time high in 2006. We seek to understand whether the conventional definition of homeownership, which is based on the share of households and ignores the effects of variable household formation, has confounded the assessment of owner-occupied housing. We find that, from 1990 to 2000 and especially from 2000 to 2006, declining household formation led to an elevated homeownership rate in the U.S. The effect of household formation varies substantially between racial/ethnic groups. Asians, who achieve high homeownership rates, have the lowest propensity to form independent households and stand in sharp contrast to African Americans. Despite much higher homeownership rates, Asians do not necessarily have better access to owner-occupied housing than do blacks. The overall conclusion is that, without accounting for household formation, the conventional measure of homeownership is an incomplete, if not misleading, measure of homeownership opportunity.UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript "The more ownership there is in America, the more vitality there is in America, and the more people have a vital stake in the future of this country." -President George W. Bush, June 17, 2004 (The Bush Administration, 2004a) Homeownership is an integral part of the American Dream (Rossi and Weber, 1996; Rohe et al., 2002) and an important housing policy target (The Bush Administration, 2004a). Under the Clinton and Bush administrations, a goal was set to increase the homeownership rate to a record high, primarily by extending homeownership to previously underserved groups (Gabriel, 2001). The period from 1990 to the mid 2000s saw a concerted increase in homeownership rates across the country (Myers et al., 2005). The national rate increased by two percentage points in the 1990s and by another three points in the early 2000s and eventually reached a record high of 69.2 percent in 2006, reversing the decade of homeownership decline in the 1980s (Myers et al., 1992; Simmons, 2001; Woodward and Damon, 2001; U.S. Census Bureau, 2009c). The increase is remarkable given the large rise in immigration and housing prices in recent years. Rising homeownership has been touted as a sign of victory of the road to an "ownership society," in which individuals would take more responsibility through private ownership and individual assumption of risk (The Bush Administration, 2004b). The increase in homeownership has been explained by a number of factors, including the spread of innovative mortgage lending practices, strong incentive for 1 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript investment, the economic prosperity during the late 1990s, and declining interest rates (Eggers, 2001; Gabriel and Rosenthal, 2002). What has not been generally recognized is how much the common definition of homeownership rates may have distorted measurement of these trends and their interpretation. Conventionally defined as the share of households that are owner-occupied, analysis of homeownership rates may be incomplete because it ignores people who do not form independent households. The rate of household formation is highly variable over time and between groups, and so neglect of this factor can be misleading about the overall trend in housing achievement. The observed differentials or trends over time could derive from different causes and have different implications than generally assumed. Simply stated, if most foregone household formation is withdrawn from the rental category, then lower household formation creates an upward distortion of homeownership rates. In this view, rising homeownership rates might not indicate a trend of rising housing prosperity; rather, it could reflect the growing exclusion of renters from the market and falling housing prosperity. Another implication is that population groups with higher homeownership rates might achieve that result by cultural practices that encourage sharing of living quarters rather than emphasizing independent living and the renting of separate dwelling units. 2 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript Thus both the interpretation of trends over time and of differences between groups could be biased by reliance on the conventional homeownership rate calculated on a per household basis. The alternative is to define homeownership on a per capita basis so that all those who may have foregone household formation are included in the denominator. In fact, for improved analysis it is preferable to separately identify both renter householders and owner householders relative to all those who are non-householders1 (non-heads). This paper explores potential biases stemming from the conventional formulation of homeownership rates through a comparison of conclusions drawn from the per household and the per capita methods. Differences between groups and changes over time are explicitly modeled. This overall goal is pursued through two specific sets of research questions. The first seeks to discover whether or not recent rises in homeownership are exaggerated by declining household formation. We will evaluate whether or not the rise in homeownership rate which is taken as a sign of success of the ownership society is attributable to depressed household formation caused by high rents or other factors. We will also examine the extent the relationship has changed between two periods of time: the 1990s and the early 2000s. Our second set of questions asks how much variable rates of household formation have intruded into analytical conclusions about homeownership 3 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript differentials between groups. To what extent do variable rates of household formation contribute to the race/ethnic disparities in homeownership? In particular, Asian households have been identified as having unusually high homeownership rates, especially in the view of the fact that the majority are immigrants and newcomers to the United States (Painter et al, 2001a; Painter et al., 2003). This contrasts particularly with African Americans. Even though most of them were U.S.-born, they have very low homeownership rates (Bostic and Surette, 2001; Freeman, 2005). In the following section, we first review previous studies and look at alternative ways in which homeownership may rise. After a summary description of major trends in homeownership and household formation rates, we develop explanatory models of the propensities of people to form households and buy homes in the 100 most populous metropolitan areas. Particular attention is given to racial/ethnic differences and changes over the decades. We then simulate household formation and homeownership rates under different assumptions. Finally, we conclude with a discussion of the implications from our findings. Previous studies Demographic effects on housing There has been a long-standing recognition of demographic effects on housing (see e.g., Carliner, 1975; Borsch-Supan, 1986; Leppel, 1986; Hendershott, 1988; 4 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript Pitkin, 1990; Ermisch, 1991). Fresh realization of the growing importance of demographic effects has spawned a new wave of research that pays close attention to the demographic determinants of housing demand (see e.g., Green, 1996; Haurin et al., 1997b; Masnick et al., 1999; Skaburskis, 1999; Riche, 2003; Myers, 2004 ). There is a great diversity in the way different demographic groups adjust their household consumptions. For instance, the elderly who already own homes are largely insulated from market fluctuations, whereas young households and new immigrants, as new housing market entrants, directly encounter the full market forces (Borsch-Supan, 1986; Chevan, 1989; Kendig, 1990). Demography has changed significantly over time. For instance, the married couple household share has declined to below 50 percent of total U.S. households in recent years, while the number of non-retired people who live alone has increased significantly over time (U.S. Census Bureau, 2009a). All these changes influence homeownership attainment. The role of household formation It is challenging to pinpoint exactly how variable rates of household formation affect homeownership attainment, because the conventional homeownership rate is not based on all members of the population but only on those who have formedUU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript households (or are householders). As a result of this challenge, only a few prior studies have addressed the relationship in any way. Borsch-Supan (1986) controlled for the endogeneity of household formation in estimating housing demand. A demand-side housing program was found to create a substantially more housing demand than originally anticipated, because it also encouraged the formation of independent households. Haurin, Hendershott, and Kim (1994) examined the tenure choice of American youth by controlling for possible sample selection biases associated with household formation and labor supply. Household formation was found to have a significant effect on youth's housing demand and the youth were particularly sensitive to market fluctuations. Haurin and Rosenthal (2007) treated the household formation decision as endogenous to the tenure choice decision and used Heckman's procedure to adjust for household formation in the selection equation. They found that lower headship rates were associated with lower homeownership rates and that declining household formation reduced homeownership rates in the recent two decades. This finding is the opposite of the reasoning and hypothesis in the present study, perhaps for the reason that their selection equation did not include personal income and housing price and rent, which are important to the decision of household formation (Borsch-Supan, 1986; Miron, 1988; Kent, 1992; Skaburskis, 1994; Ermisch, 1999). In addition, family 6 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript income rather than household income was used in the tenure choice equation, which could confound the estimates of non-family households and non-family members. The Haurin and Rosenthal study is further discussed for its methodology later in this paper. Puzzling racial/ethnic differences in homeownership The importance of addressing underlying variations in practices of household formation takes on even greater importance in the case of ethnic groups that have different cultural customs of living arrangements. In these cases, homeownership attainment becomes more behaviorally complex and assumes added theoretical meaning. Asian residents present a noteworthy case because they attain a very high homeownership rate, even though most of them are immigrants and came to the U.S. recently (Myers et al, 1998; Painter et al, 2003; Yu and Myers, 2007). Net of other relevant factors, their homeownership rate is almost on par with that of native-born, non-Hispanic whites. This has been widely touted as evidence of Asians' successful adaptation to the U.S. and as an emblem of their so-called "model minority" status. African-Americans or blacks, in contrast, have persistently low homeownership rates, even though almost all of them were bom in the U.S. The literature has repeatedly documented the persistent black-white homeownership gap, which can not be fully explained by blacks' low socioeconomic status (Bianchi et al, 7 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript 1982; Horton, 1992; Wachter and Megbolugbe, 1992; Flippen, 2001). Although blacks have significantly improved their homeownership rates in recent years, the black-white homeownership gap remains wide (Bostic and Surette, 2001; Freeman, 2005). Another distinctive feature of African-American living arrangements that is well-recognized is their much lower prevalence of married couple households (U.S. Census Bureau, 2009b). The implications for homeownership are not fully recognized because a lower incidence of married couple households leads to a higher probability of forming independent households. Contrary to our expectation, however, one study showed that blacks' high rates of household formation increased their homeownership rates. Black-white homeownership gaps would have been even larger had blacks had a similarly low propensity for household formation to whites (Haurin and Rosenthal, 2007). With the rapid changes in composition of the U.S. population, the relationship between household formation and homeownership is likely to have implications for trends in the overall rate of homeownership and housing consumptions in the future. To address these issues, we must first examine how the conventional homeownership measure is constructed.UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript Deconstructing the conventional measure of homeownership achievement There are two different ways in which the proportional achievement of homeownership can increase-either by transfers of renters to homeowner status or by elimination of renters from the housing market (Masnick et al, 1999). This can be understood from the conventional construction of the homeownership rate, as presented in the following equation: ^ _ Owner _HHs Ownership _ Rate =-------------------------------------------, Owner _ HHs + Renter _ HHs where the number of owner-occupied households is divided by the sum of owner {Owner_HHs) and renter-occupied households (Renter JiHs). The common assumption is that homeownership increases occur because renters change to homeowners. Given the same number of households in the denominator, the increase in the number of owners in the numerator would increase the overall homeownership rate. This is a favorable interpretation of what is indicated by homeownership increase, because homeownership opportunities are expanded to more households. In contrast, the removal of renters from the denominator (or a slower growth in the number of renters than the number of homeowners) will also cause the homeownership rate to rise even if none of those renters transfer to homeownership in the numerator. For instance, when housing prices increase rapidly and personal incomes stagnate, many vulnerable individuals may postpone household formation or 9 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript drop out of the housing market. They would live with their parents or other relatives, or they might double up with roommates. Instead of an increase in the numerator in the first scenario (more homeowners), there would be a decline in the denominator (fewer renters added). In this case, a rising homeownership rate has an unfavorable interpretation because it does not reflect better access to owner-occupied housing. Instead, this would suggest a decline in housing opportunities for some and greater polarization because would-be renters have dropped out of the housing market. The key assumption in this argument is that when the rate of household formation is rising, more people are establishing independent living quarters, typically as renters and, hence, the homeownership rate is falling. The most direct evidence to indicate that household formation operates through the rental tenure is provided by surveys of newly formed households. Although homeowners account for over two-thirds of all households, they account for less than one-quarter of newly formed households. Cursory examination of the American Housing Survey shows, in fact, that 77.1% of newly formed households have moved into rental units.11 Therefore, when the household formation rate is falling, it is reasonable to assume that it is the ranks of renters or would-be renters that are being reduced most greatly. As a result, the overall homeownership rate can increase because fewer people are forming independent households. Such an aggregate relationship was reported in a study of 10 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript changes from 1990 to 2000 in the 50 states with regard to rates of household formation and homeownership attainment. A substantial inverse correlation (-0.33) was found between changes in headship rates and changes in homeownership rates (Myers 2001). Further, an analysis of the baby boomer cohort in the 100 largest metropolitan areas found that, net of other factors, each one percentage point increase in the headship rate over the decade translated into a decrease of roughly one-half percentage point in the homeownership rate, although this varied by race/ethnicity (Myers et al., 2005). The association between household formation and homeownership is likely moderated by cultural differences. For instance, Latinos are more likely to live with extended family and have lower headship rates, while African-Americans are more prone to living in single-parent families and have high headship rates. The variable rates of household formation could be an important yet underappreciated factor in explaining racial/ethnic differences in homeownership rates. An alternative, population-based measure of homeownership opportunity One way to address the questions raised above is to look beyond the conventional definition of homeownership. An alternative to the almost-universal per household measure is the per capita measure of homeownership, which uses population instead of households as the denominator of homeownership (e.g., Myers 11 UU IR Author Manuscript UU IR Author Manuscript ■ ■ f : 1 University of Utah Institutional Repository Author Manuscript and Lee, 1998; Myers and Park, 1999). We can term this measure the "owner headship rate," which denotes the percent of a given population that are householders of owner-occupied units. The higher the rate, the more owner households are formed. Certain limitations to the dichotomous per capita homeownership measure should be noted. This method contrasts owners to all others and does not differentiate non-householders from renter householders. Therefore, it does not account for the large variations between racial/ethnic groups in their likelihood of forming renter households. An alternative adopted for the present research takes account of fuller information to define a per capita measure that subdivides non-homeowner heads into renter heads and non-householders. We treat household formation here as an individual's decision whether to be a renter household, an owner householder, or a non-householder. In the present research, this is estimated as a trinomial logistic regression that explicitly models both renter householders and owner householders relative to non-householders. People make those decisions based on a set of personal demographic or socioeconomic factors and based on the alternative costs of renting and owning. Accounting for these factors, we can then compare racial/ethnic differences in household formation and homeownership. This general approach has been used in previous studies of household formation and living arrangement (e.g., 1 2 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript Leppel, 1991; Hughes, 2003; Van Hook and Glick, 2007). More detailed discussion of the advantages of adopting the trinomial specification is reserved for the methods section. Sample and data The study uses the decennial censuses 5% Public Use Microdata Samples (PUMS) in both 1990 and 2000 from the IPUMS data base (Ruggles et al, 2003). We also retrieve the American Community Survey (ACS) Microdata Samples in 2006 from the same source. The sample for analysis includes all persons who are age 18 years and older. We limit our sample to the 100 most populous metropolitan areas, so that we might capture the housing market effects of housing prices and rents on the formation and tenure decision. Boundaries for the 100 metropolitan areas are specified in accordance with the geographic definitions used in the 2000 census. The names of the metropolitan areas are listed in Appendix 1. The areas are comprised of one or more whole counties, with the exception of the New England region where metro areas are built from aggregations of townships. Data from the 1990 census and 2006 ACS is re-arranged to conform to these 2000 definitions. For this study we do not use primary metropolitan statistical areas that are subsets of the larger consolidated metropolitan statistical areas. Instead, we use the whole CMS A as a unit to represent the 13 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript metropolitan housing market area. Thus, our set of 100 most populous metropolitan areas includes both CMSAs and freestanding MSAs. About 70 % of the total U.S. population lives in these areas. Descriptive findings In this section, we first report housing consumption patterns in the year 2000 and then track changes from 1990 to 2006. Four indicators are used in this analysis. The first indicator is the headship rate, which is used to measure household formation and measured for the population universe. The result is reported in Column A in Table 1. The headship rate is then partitioned into owner headship and renter headship, separately reported in Columns B and C. The last indicator (Column D) is the conventional homeownership rate measured with the household universe. Note that the conventional homeownership rate can also be estimated by the ratio of the owner headship to total headship (Column B/Column A). Table 1 about here Variable rates of homeownership and household formation Large racial/ethnic differences can be observed in the headship rates, owner headship rates, renter headship rates, and homeownership rates by racial/ethnic groups in the year 2000 (Table 1). Whites had the highest homeownership rate (71.4%), followed by Asians (53.5%), and the lowest homeownership rates are found among 14 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript Latinos (44.1%) and blacks (44.6%). Based on the conventional homeownership rate alone, one would conclude that both Asians and whites had better access to owner-occupied housing than blacks and Latinos. Once we take household formation into consideration, the conclusion becomes substantially different. First, we observe that whites had the highest headship rate of 50.1%. That is, there were roughly 50 householders or 50 households formed for every 100 whites. (See Table 1.) In contrast, Asians formed fewer independent households. There were only 38 households for every 100 Asians, 12 households fewer than for whites. More specifically, there were only 20.4 owner households for every 100 Asians, 15.4 fewer than for whites (Column C); however, the ratio of owners to the relatively low total number of householders yields a high homeownership rate (Column D). Thus it can be seen that the high homeownership rate alone does not necessarily suggest that Asians have superior access to owner-occupied housing. In contrast, our assessment of access to homeownership is much improved for blacks. Although this group had a very low homeownership rate and fully 26.8 renter households for every 100 blacks, the highest of the four racial/ethnic groups, as a proportion to their population group, black homeowners actually outnumbered Asian homeowners, 21.7 per 100 versus 20.4 per 100. As a percentage of households, the 15 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript black homeownership rate was lower than for Asians because blacks had many more renter households than did Asians, due to blacks' higher household formation rates. Rising homeownership amidst declining household formation With this understanding of the racial/ethnic differences, let us now address the changes from 1990 to 2006. We disaggregate homeownership rates and headship rates by three broad age groups and the four racial/ethnic groups (Figure 1). Homeownership rates (per household) are shown on the top of the figure. Graphs on the bottom show headship rates (per capita) presented in the same fashion. Figure 1 about here Figure 1 indicates that almost all demographic groups experienced increases in homeownership rates from 1990 to 2006. As expected, whites had the highest homeownership rates among all ethnic groups, followed by Asians who experienced the largest percentage point increase in homeownership rate over the 16 year period. Within each racial/ethnic group, the elderly (age 65 and over) had the highest homeownership rates of all age groups, except among elderly Asians. Figure 1 also shows that the large increases in homeownership rates were accompanied by declining headship rates in almost all age groups through the 16 year period. Minority groups and the elderly experienced a more pronounced decline in headship rates. Thus the decline in household formation occurred at the same time as 16 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript the rise in homeownership, possibly because renter households were being eliminated at a rapid pace. Recall that 77.1% of new household formations occur in rental units and we presume that the bulk of reductions in household formation also come from this category. Thus under the observed conditions the rise in homeownership might reflect growing polarization rather than growing opportunity. The foregoing descriptive analysis is very broad and does not control for other factors such as racial/ethnic differences in demographic composition and socioeconomic status, detailed age, incomes, or rents and housing prices. We proceed now to a more detailed examination of multiple determinants that shape household formation and tenure choice in the population. Individual-level analysis of household formation Methods We employ the multinomial logistic regression used in previous studies, such as Clark and Mulder (2000) and Leppel (1986), to estimate the probability of an individual being a non-householder (non-head), a renter householder (renter head), or an owner householder (owner head). These three categories in the dependent variable account for the housing status of all people who are 18 and older. Renting or owning is simultaneously compared to the status of non-household headship and we examineUU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript the effect of different determinants in increasing or decreasing the probability of renting or owning relative to not forming a household. Previous studies have examined tenure choice while adjusting for the endogeneity of other decisions with the Heckman's sample selection model (e.g., Borsch-Supan, 1986; Haurin et al, 1997a; Painter et al, 2001b). The Heckman-style analysis has been very useful in many instances. However, research results based on the Heckman procedure are sensitive to the choice of variables in the selection model (Manski, 1989 ; Newey, 1999). The anomalous results of Haurin and Rosenthal (2007) might be due to such problems in the selection model which they did not explicitly present. As previously discussed, their selection equation does not appear to include personal income and housing price and rent, which are important to the decision of household formation. Nor was it clear how the parameter estimates were different from the tenure choice model that ignored household formation. It is also unclear whether the sample selection model fulfills the assumption that the error terms in the regression equation and in the selection equation follow a bivariate normal distribution. The assumption is critical to the validity of the research findings (Greene, 1997: 714 ; Cameron and Trivedi, 2005: 549). The multinomial modeling procedure is appropriate for assessing the determinants of household formation, which is an outcome with three or more 18 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript unranked categories. Rather than assume that household formation is a mere background condition that introduces error into understanding our main topic of interest, as does the sample selection correction procedure, we judge household formation to be of equal interest to tenure choice and we consider it to be a jointly determined decision (Borsch-Supan 1986). The multinomial specification allows us to look specifically at changes in household formation over time by examining the coefficients on key variables that influence people's decisions to rent or own relative to being a nonhouseholder. Given that this examination of change is our key purpose, the Heckman procedure is less appropriate. We specifically test the assumption of the independence of irrelevant alternatives (IIA) using the Small and Hsiao's test111 (1985). The test statistics show that the assumption of IIA is accepted and the three outcomes in the dependent variable are sufficiently dissimilar from each other. Therefore, the multinomial logit model is appropriate in this analysis. Multinomial logit regression yields relative risk ratios, which are the exponentiated values of multinomial regression coefficients. The interpretation of relative risk ratios is similar to odds ratios in a logistic regression. While it is appropriate to use multinomial logit regression in this analysis, the method has its disadvantages. First, multinomial logit regression produces multiple comparisons and 19 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript a large number of parameters, which could hamper interpretations. Second, relative risk ratios can not be easily compared and understood. As a partial remedy to these, we graph the relative risk ratios and predict probabilities through simulations. The model used in this analysis is specified as follows: H = RACE +AGE + MG + GENDER + X+Y H = householder status (Non-head or non-householder = 0, renter householder = 1, and owner householder = 2), RACE = racial/ethnic group, AG = age group, MG = immigrant group, GENDER= individual's gender, X = individual's socioeconomic characteristics, and Y = metropolitan housing context. Variables H is the outcome variable of interest. For the present analysis, we pay particular attention to the effects of race/ethnicity. RACE includes four groups which are non-Hispanic whites, blacks, Asian and Pacific Americans, and Latinos (reference group = non-Hispanic whites). The behavior of racial groups in the sample is expressed as a deviation from the reference group. AG or age group is coded as 18-24, 25-34, 35-44, 45-54, 55-64, or 65-74 (reference group = 35-44). MG is the duration in the U.S. based on immigrant year of arrival, coded as immigrants who came in last 10 years, in last 10-20 years, in last 2 0 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript 20-30 years, and more than 30 years ago (reference group = the native-born). Immigrant status and age are especially important dimensions of household formation, because headship rates vary predictably by age and immigrant status (Smith et al., 1984; Skaburskis, 1994). Gender is also a key demographic determinant of householder status. We use it as a control variable in the analysis. Before 1980, men were automatically designated as householders in married couple households. Since the 1980 census, either husband or wife could be designated householder. Consequently, there has been a gradual shift in the gender of householders. Women have become much more likely to be householders over the years (Myers, 1992). There are also large racial/ethnic variations in the likelihood that women are householders. For instance, 73 percent of Asian householders were male in 2000. In comparison, only 46 percent of black householders were male. Therefore, it is necessary to control for gender in the estimation of household formation. Female will be the reference group in the model. Individual characteristics The second model controls a set of individual characteristics (X), which include personal income, educational attainment, martial status, and English proficiency. It also controls metropolitan context (Y), which includes metropolitan housing prices (observed at the lower quartile), median rent, and percent changes in 2 1 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript housing prices by metropolitan areas in the last 5 years. (See Table 2 for a full list of the variables). These variables are added to the first model. Income is an important factor in the decision of forming independent households. Rising real income has increased the real affordability of housing and resulted in a steady increase in household formation after WWII (Carliner, 1975; Hendershott, 1988; Miron, 1988). However, real income has stagnated in recent years, which puts a damper on household formation. Previous studies have shown that household income may be endogenous to the household formation decision among young people (Haurin et al, 1994). Therefore, we use personal income in the estimation. This will serve to highlight the individual decisionmaking involved. Although personal income does not represent all the resources required to rent or buy homes, and pooling incomes with multiple earners remains an option; nonetheless, an individual with higher personal income is substantially more advantaged in housing consumption. Educational attainment is the principal measure of human capital, serving as a proxy for future earnings. Therefore, the more educated should have a higher propensity to form independent households and to buy homes than the less educated. The three education categories are (1) no high school diploma, (2) high schoolUU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript diploma or some college, and (3) college degree or better. Those who have high school diploma or some college education will be the reference group. Marital status is also a major determinant of household formation (Sweet, 1990). Married couples are more likely to form independent households and also buy homes, but the married partners do this jointly, which constrains their individual headship to no more than 50%. In contrast, previously married individuals have acquired housing experience living as a married couple but need no longer share their headship with a partner. As a result their individual household formation and hoemownership may exceed that of married persons. The three categories representing marital status are (1) never married, (2) currently married, and (3) formerly married. Those who are currently married will be the reference group. English proficiency, which is particularly important for immigrants, also is included in the model. English proficiency supports socioeconomic incorporation of immigrants in the U.S. Those who speak English well are expected to have higher headship rates, because they are more adapted to the U.S. and less likely to have large households than do their compatriots (Myers et al, 1996). Moreover, English use in the home (i.e. speaking only English) is the foundation of acculturation (e.g.Portes and Hao, 2002), which may additionally enhance the prospects of household formation. Once human capital and income are introduced, it is not clear how much 23 UU IR Author Manuscript UU IR Author Manuscript L 1 ^ ' ' • . : University of Utah Institutional Repository Author Manuscript the differences in household formation due to English ability will remain. The three categories of English proficiency are: (1) Speak only English, (2) Speak English well but not exclusively, and (3) Speak English poorly or speak no English. Those who speak English well but not exclusively will be the reference group. Finally, we include three variables to reflect the relative costs of renting or owning in each metropolitan area. Housing price is measured as the 25th percentile home price and rent as the median gross rent in each metropolitan area. 1V The use of these proxies follows Gyourko and Linneman (1996). Previous studies have shown that both housing price and rent level affect household formation (e.g., Borsch-Supan, 1986; Kent, 1992; Skaburskis, 1994; Ermisch, 1999; Hughes, 2003). In addition, we also include a variable that measures the recent rate of price appreciation in the metropolitan area, i.e. the percent change in housing price over the five years preceding the census observation, using the OFHEO Housing Price Index (HPI) (Office of Federal Housing Enterprise Oversight, 2007). v While high housing prices at one point of time may reduce the demand for owner occupied housing, rising housing prices may generate an expectation of future appreciation. As a result, rising prices serve to lower the expected user cost of housing and may encourage people to form owner households and abandon both renting and non-householder status. The reverse may be true in housing market downturns. Despite improved affordability, 2 4 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript people are reluctant to form new households or buy homes fearing further declines in the economy and in housing prices (Myers et al., 2005). Income, housing price, rent, and HPI are all adjusted for inflation to constant 2005 dollars. Summary statistics Summary statistics are presented in Table 2, which also shows the variables used in the multivariate analysis. The mean values are computed and reported in three separate columns for the years 1990, 2000, and 2006. The percent share of each attribute is reported under each variable, excepting personal income, housing prices, rent, and changes in housing prices. Table 2 about here In the section of household membership, we can see that the percent share of owner householders has increased slightly, while the percent share of renter householders had declined from 18% to 16.1%. This corresponded to the increase in homeownership over the years. The percent shares of whites and the youth declined over time, so did the shares of the native-born and the currently married. Whites were still the largest racial/ethnic group. Because of the gradual increase in immigrant population in the U.S., the overall level of English proficiency had declined over the years. Most of the observations were born in the U.S., though the share of newUU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript immigrants who came to the U.S. in the last 10 years had increased over time. Average personal income reached its peak in 2000 and declined slightly from 2000 to 2006. Housing prices and rent changed little between 1990 and 2000, and increased significantly from 2000 to 2006. In particular, housing prices index from 2001 to 2006 increased by 53.4%, which is much larger than before. Coefficient estimates Table 3 reports the relative risk ratios (RRR) of the determinants for 1990, 2000, and 2006 respectively. Sections I, III, and V present results from the "demographic only" models of the three years, while Sections II, IV, and VI report estimates from the models that introduce covariates for socioeconomic factors, education, English proficiency, and metropolitan housing prices. Each reported coefficient reflects the effect of a particular characteristic on one of the three types of household status, relative to the probability of being a non-householder. There are two columns for each model. The left column reports the probability of being a renter householder, while the right column shows the probability of being an owner householder. The baseline group is the probability of being a non-householder, which is omitted from the table. Table 3 about hereUU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript Interpretation of the coefficient estimates is straightforward. The status of racial/ethnic, age, and immigrant groups observed in each year (1990, 2000, or 2006), relative to male native-born whites who aged 35-44, is given by the relative risk ratios for GENDER, RACE, AG, and MG. The reference group is given the value 1.000. Ratios of less than 1.000 indicate a reduced likelihood of being a renter householder or being an owner householder, whereas ratios greater than 1.000 indicate an increased likelihood. Demographic variables Let us first examine the roles of demographic variables in household formation first. Tables 3 also report the odds ratios of gender, age groups, and immigrant status. Males had a higher probability of forming independent households in general, and forming owner households in particular. With respect to age groups, household formation increased with age. More specifically, people increased their likelihood of forming owner households steadily until age 65-74. In contrast, the likelihood of forming renter households followed an "M" shape and gradually declined after age 25-34 until age 65-74, when retirement began to affect household formation. Consequently, homeownership rates increased through the adulthood. Young people had not only the lowest probabilities of household formation, but the lowest likelihood of forming owner households. 2 7 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript With respect to immigrant status, newly arrived immigrants had very low probabilities of owner household formation. However, immigrants' likelihood of renter household formation declined gradually over time. Comparing the results between the estimates of renter householders and owner householders, we can conclude that homeownership is determined by the probabilities of both owner and renter household formations. Covariates We compare the results between the first and the second models in each year and examine how the inclusion of socioeconomic status and metropolitan context affects household formation. Results in Sections II, IV, and VI show that higher income, living in married or formally married households, having higher levels of education are all positively associated with household formation. Speaking only English reduces the likelihood of forming renter households, while speaking English poorly or not at all reduces the likelihood of forming owner households. With respect to metropolitan context, higher housing prices discourage the formation of owner households but encourage renter household formation. The opposite is true for changes in housing prices. Rising housing prices over time encourage the formation of owner households. Meanwhile, higher rent retarded bothUU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript types of household formation, expect for the year 1990. The results are largely consistent in the three years. While these results are expected, accounting for other demographic variables and the covariates only slightly attenuates the racial/ethnic differences in household formation. In comparison with the odd ratios reported in Sections I, III, and V, odds ratios shown in Sections II, IV, and VI are smaller. The basic relationship across demographic groups remains largely unchanged. To better understand variable household formation, we graph the relative risk ratios of racial/ethnicity just discussed. Figure 2 shows the results in 1990, 2000, and 2006. There are two sets of bars in the figures. The dark bar reports the relative risk ratios of being an owner householder relative to being a white owner householder, while the lighter bar shows the odds ratios of being a renter householder relative to being a white renter householder. Figure 2 about here Evidenced in Figure 2, all dark bars have negative values indicating that whites have the highest probability of being owner householders. This finding is not surprising and consistent with previous studies, which show minorities have lower homeownership probabilities than whites. What's interesting is that there are major 2 9 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript differences in the probability of being renter householders. All lighter bars have positive values except for Asians. Asians were the least likely to form renter households, while blacks had the highest likelihood. After controlling for all other confounding factors, we can conclude that the large racial/ethnic differences in homeownership stem mostly from variable rates of renter household formation. Despite blacks' low homeownership rates, they were as likely to form owner households as Asians in 2000 and 2006. Simulations The results reported in the tables are odds ratios which are not easily compared. The results are for three points of time but do not directly reveal how changing household formation has affected homeownership probabilities over time. In this section, we simulate the probabilities of household formation and use the predicted value to calculate homeownership rates to address three specific questions. We follow a simulation method used in previous studies (e.g., Wachter and Megbolugbe, 1992; Bostic and Surette, 2001), which is a variation of the decomposition approach pioneered by Oaxaca (1973) and Blinder (1973). First, what would Asian homeownership rates be in 2000 if they had formed households at the same rate as blacks? We first use the logit estimates of the parameters for the year 2000, assume the individual characteristics of Asians, and 3 0 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript predict the homeownership rates of Asians. We then compare the actual rates with the simulated or predicted rates. This is to quantify the extent to which Asian and black homeownership disparity is the result of their differences in household formation. Second, what would homeownership rates be if each minority group formed households as whites did in 2000? This allows us to examine the extent to which the white-minority homeownership gaps can be explained by their different propensities of household formation. We use the same procedure discussed above and compare actual rates with simulated or predicted rates. This is to quantify the extent to which Asian and black homeownership disparity is the result of their differences in household formation. Third, what would the homeownership rates be in 2000 and 2006 if people had formed households in the same way they did in 1990? We use the logit coefficient estimates for the year 1990 and individual characteristics in 2000 and 2006 to simulate, or predict, the homeownership rates for each of the four racial/ethnic groups and also for the total. We then compare predicated rates with actual rates. In so doing, we study how much homeownership would have changed had all racial/ethnic groups formed households the way they did in 1990. In other words, we assume the same household formation behavior in 1990 while using the individual characteristics in 2000 and 2006 in the simulation. 31 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript The answer to the first question is that, in the year 2000, Asians' homeownership rate would only be 37.7 percent had Asians followed the same propensities for household formation as blacks. That would be 15.9 percentage points lower than the observed homeownership rate in 2000 and 7.4 percentage points lower than blacks' observed homeownership rate. Under this assumption, Asians would have formed more households, or more specifically, far more renter households and fewer owner households. Asians' headship rate would have been 43.7 percent, 5.3 percentage points higher than Asian's actual headship rate in 2000. Said alternatively, the black-Asian homeownership gaps could be largely explained by their different propensities of household formation. In fact, blacks would have a higher rate of homeownership had Asians followed the same propensities of household formation as blacks. We address our second and third questions respectively in Figures 3 and 4. Figure 3 illustrates the homeownership rates of all racial/ethnic groups had they had the same propensities of household formation as did whites in 2000. Results show that all minority groups would have had higher homeownership and headship rates. In particular, blacks would have had the largest increases in homeownership rates (11.8 points), followed by Latinos (2.9 points). Asians would have only seen an increase of 1.0 point over their actual rate. Assuming whites' propensities for household 3 2 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript formation, minority-white homeownership disparity would significantly attenuate. The overall homeownership rate would have increased by about 2.0 percentage points. Minorities' headship rate would also have been higher. In other words, there would have been more minority households had minorities shared the same propensity for household formation as whites. Figure 3 about here Figure 4 illustrates that had all racial/ethnic groups in 2000 and 2006 followed the same propensities of household formation as they did in 1990, homeownership rates would have been lower than the observed rates. In fact, on average, the rates would have dropped by 0.7-3.8 percentage points relative to the observed rates in 2006. The overall homeownership rate would have only increased slightly from 1990 to 2000 and decreased by 0.8 percent points from 2000 to 2006. Figure 4 about here Most groups would have formed more households, had they had the same propensity of household formation in 1990; the overall headship rates would have been about 1.0 point higher than the actual rates in both 2000 and 2006. From 1990 to 2006, all racial/ethnic groups experienced declines in the probability of renter household formation. In other words, people became less likely to form renter households over time. This would be good news if all the renters were converted to 33 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript owners and homeownership rates increased accordingly. In reality, this analysis shows that few groups experienced any increases in the probability of owner household formation. Therefore, homeownership increases in recent years were largely an artifact of declining renter household formation. Discussion Overall, what is the pattern of household formation and how has changing household formation affected homeownership rates over time? Clearly, there are large variations in household formation across racial/ethnic groups, even after controlling for other confounding factors. Asians have high homeownership rates, due in large part to their very low rates of renter household formation. In fact, on a per capita basis, Asians have a number of homeowners similar to blacks and Latinos and far fewer than whites. For Asians, their high ownership rate comes as an outcome of low renter household formation. In contrast, blacks and Latinos seem to have formed far more renter households, which have led to their relatively low homeownership rates. While we can not completely separate the cultural preferences of low household formation from structural barriers, the low rate of household formation has clearly inflated the homeownership rates of Asians. The differences between blacks and Asians would be much smaller had we measured homeownership on a per capita 3 4 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript basis or taken household formation into consideration. Minority-white homeownership disparity is largely attributable to variable rates of household formation. From 1990 to 2006, most demographic groups had seen a gradual decline in the rates of household formation. Such decline occurred mostly among renters. Therefore, declining household formation may have inadvertently increased the aggregate homeownership rates by as much as 3.8 percentage points in 2006 (Figure Conclusions Research findings show that the low rate of household formation artificially inflated Asians' homeownership rate. Decreasing household formation has contributed to rising homeownership from 1990 to 2006. Declining household formation bumped up the homeownership rate, largely through the elimination of renters from the housing market or the exclusion of would-be renters from joining the market. Therefore, rising ownership rates does not necessarily equal to more access to owner-occupied housing. Once we take household formation into consideration, Asians do not seem to be much better adapted to the U.S. housing market than Latinos and minority-white homeownership gaps are not as large as the observed rates suggest. These results are in stark contrast to those reported in Haurin and Rosenthal (2007). 35 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript Research findings also indicate that the current homeownership measure, defined as the percent of households that are owners, is a deficient, if not flawed, indicator of access to owner-occupied housing. It is therefore important to take household formation into consideration, when policymakers propose new housing policy. For this it would be useful to track homeownership on an alternative measure that is expressed relative to the entire population of each age group and not simply relative to the number of householders. The preceding analysis has introduced a method that separately identifies household formation and homeownership attainment. The procedure is an improvement over the currently used per-household based homeownership measure. Once we examine homeownership through the lens of household formation, we have a better understanding of the dynamics of homeownership attainment. The findings reported here do not nullify the advantages of homeownership or, more broadly, of the concept of ownership society. However, they do expose hidden faults in the presumed advantages of a high homeownership rate. A rise in that indicator can reveal two contradictory trends-either the success in the ownership society, or the elimination of households from the housing market. The risk of failure that is implicit in the ownership society deserves to be recognized on equal terms withUU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript the joys of homeownership. Only in this way can the public make a more informed choice about policies giving even greater weight to the ownership society. We discovered diverging trends between homeownership rates and household formation among Asians and Latinos. There has been a growing immigrant population and increasing ethnic diversity in the U.S. The concentration of immigrant population and declining affordability and rising cost of household formation have also made the accessibility of homeownership more illusive to measure. Per capita homeownership or owner headship measure may be a useful compliment to the current household based homeownership measure, particular in housing market downturn when headship rates may decreased much further than homeownership rates. Because the household based homeownership measure is a deficient indicator, there is a need for more comprehensive evaluations and see whether the "ownership society" agenda has fulfilled its promises. In addition, further research should look at whether declining household formation is a long term trend and whether housing formation has further declined in time of market downturn.UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript Reference: BIANCHI, S.M., FARLEY, R. and SPAIN, D. (1982) Racial Inequality in Housing: An Examination of Recent Trends. Demography 19, 37-51. BLINDER, A.S. (1973) Wage Discrimination: Reduced Form and Structural Estimates. Journal of Human Resources 8, 436-55. BORSCH-SUPAN, A. (1986) Household formation, housing prices, and public policy impacts. Journal of Public Economics 30, 145-64. BOSTIC, R.W. and SURETTE, B.J. (2001) Have the Doors Opened Wider? Trends in Homeownership Rates by Race and Income. Journal of Real Estate Finance and Economics 23, 411-34. CAMERON, A.C. and TRIVEDI, P.K. (2005) Microeconometrics : methods and applications. Cambridge University Press, New York, NY. CARLINER, G. (1975) Determinants of household headship. Journal of Marriage and the Family 37, 28-38. CHEVAN, A. (1989) The Growth of Home Ownership: 1940-1980. Demography 26, 249-66. CLARK, W.A.V. and MULDER, C.H. (2000) Leaving home and entering the housing market. Environment and Planning A 32, 1657-72. EGGERS, F.J. (2001) Homeownership: A Housing Success Story City scape 5, 43-56. ERMISCH, J.F. (1991) An ageing population, household formation and housing. Housing Studies 6, 230-9. ERMISCH, J.F. (1999) Prices, parents, and young people's household formation. Journal of Urban Economics 45, 47-71. FLIPPEN, C.A. (2001) Racial and Ethnic Inequality in Homeownership and Housing Equity. Sociological Quarterly 42, 121-49. FREEMAN, L. (2005) Black Homeownership: The Role of Temporal Changes and Residential Segregation at the End of the 20th Century. Social Science Quarterly 86, 403-26. GABRIEL, S.A. (2001) Opening the Doors to Homeownership: Challenges to Federal Policy. Cityscape 5, 31-45. GABRIEL, S.A. and ROSENTHAL, S.S. (2002) The Causes of Increase in Homeownership in the 1990s. Research Institute for Housing America, Washington D GREEN, R.K. (1996) Should the Stagnant Homeownership Rate Be a Source of Concern? Regional Science and Urban Economics 26, 337-68. GREENE, W.H. (1997) Econometric Analysis. Prentice Hall, Englewood Cliffs, N.J. GYOURKO, J. and LINNEMAN, P. (1996) Analysis of the Changing Influences on Traditional Households' Ownership Patterns. Journal of Urban Economics 39, 38 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript 318-41. HAURIN, D.R., HENDERSHOTT, P.H. and KIM, D. (1994) Housing Decisions of American Youth. Journal of Urban Economics 35, 28-45. HAURIN, D.R., HENDERSHOTT, P.H. and WACHTER, S.M. (1997a) Borrowing Constraints and the Tenure Choice of Young Households. Journal of Housing Research 8, 137-54. HAURIN, D.R. and ROSENTHAL, S.S. (2007) The Influence of Household Formation On Homeownership Rates Across Time and Race. Real Estate Economics HAURIN, R.J., HAURIN, D„ HENDERSHOTT, P. and BOURASSA, S. (1997b) Home or Alone: The Costs of Independent Living for Youth. Social Science Research 26, 135-52. HENDERSHOTT, P.H. (1988) Household Formation and Homeownership: Impacts of Demographic, Sociological, and Economic Factors. Housing Finance Review 7, 201-24. HORTON, H.D. (1992) Race and Wealth: A Demographic Analysis of Black Homeownership. Sociological Inquiry 62, 480-9. HUGHES, M.E. (2003) Home economics: Metropolitan labor and housing markets and domestic arrangements in young adulthood. Social Forces 81, 1399-429. KENDIG, H.L. (1990) A Life Course Perspective on Housing Attainment. In D. Myers (ed.), Housing Demography: Linking Demographic Structure and Housing Markets, University of Wisconsin Press, Madison, WI, 133-56. KENT, RJ. (1992) Household Formation by the Young in the United States. Applied Economics 24, 1129-37. LEPPEL, K. (1986) A Trinomial Logit Analysis of Household Composition. AREUEA Journal 14, 537-56. LEPPEL, K. (1991) Demographic Effects on Household Formation Patterns. Journal of Real Estate Research 6, 191-206. LONG, I.S. and FREESE, I. (2006) Regression models for categorical dependent variables using Stata. StataCorp LP, College Station, Tex. MANSKI, C.F. (1989) Anatomy of the Selection Problem. Journal of Human Resources 52, 343-60. MASNICK, G.S., MCARDLE, N. and BELSKY, E.S. (1999) A Critical Look at Rising Homeownership Rates in the United States Since 1994. loint Center for Housing Studies, Harvard University, Cambridge, MA. MIRON, I.R. (1988) Housing in postwar Canada : demographic change, household formation, and housing demand. McGill-Queen's University Press, Kingston, [Ont.]. MYERS, D. (1992) Analysis with Local Census Data: Portraits of Change. Academic 39 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript Press, Boston. MYERS, D. (2004) Cohorts and Socioeconomic Progress. Russell Sage Foundation and Population Reference Bureau, New York City and Washington, DC. MYERS, D„ BAER, W.C. and CHOI, S.-Y. (1996) The Changing Problem of Overcrowded Housing. Journal of the American Planning Association 62, 66-84. MYERS, D. and LEE, S.W. (1998) Immigrant Trajectories into Homeownership: A Temporal Analysis of Residential Assimilation. International migration review 32, 593-625. MYERS, D., MEGBOLUGBE, I. and LEE, S.W. (1998) Cohort Estimation of Homeownership Attainment Among Native-Born and Immigrant Populations. Journal of Housing Research 9, 237-69. MYERS, D., PAINTER, G., YU, Z., RYU, S.H. and WEI, L. (2005) Regional Disparities in Homeownership Trajectories: Impacts of Affordability, New Construction, and Immigration Housing Policy Debate 16, 53-83. MYERS, D. and PARK, J. (1999) The Role of Occupational Achievement in Homeownership Attainment by Immigrants and Native Borns in Five Metropolitan Areas. Journal of Housing Research 10, 61-93. MYERS, D„ PEISER, R„ SCHWANN, G. and PITKIN, J. (1992) Retreat from Homeownership: A Comparison of the Generations and the States. Housing Policy Debate 3, 945-75. NEWEY, W.K. (1999) Consistency of two-step sample selection estimators despite misspecification of distribution Economics Letters 63, 129-32. OAXACA, R. (1973) Male-Female Wage Differentials in Urban Labor Market. International Economic Review 14, 693-709. OFFICE OF FEDERAL HOUSING ENTERPRISE OVERSIGHT (2007) House Price Index (HPI). Office of Federal Housing Enterprise Oversight, Washington D.C. PAINTER, G., GABRIEL, S.A. and MYERS, D. (2001a) Race, Immigrant Status, and Housing Tenure Choice. Journal of Urban Economics 49, 150-67. PAINTER, G., YANG, L. and YU, Z. (2001b) Heterogeneity in Asian American Homeownership: The Impact of Household Endowments and Immigrant Status. Lusk Center Working Paper Series 2001-1010. PAINTER, G., YANG, L. and YU, Z. (2003) Heterogeneity in Asian American Homeownership: The Impact of Household Endowments and Immigrant Status. Urban Studies 40, 505-30. PITKIN, J. (1990) Housing Consumption of the Elderly: A Cohort Economic Model. In D. Myers (ed.), Housing Demography: Linking Demographic Structure and Housing Markets, University of Wisconsin Press, Madison, WI, 174-99. PORTES, A. and HAO, L. (2002) The price of uniformity: language, family and 4 0 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript personality adjustment in the immigrant second generation. Ethnic and Racial Studies 25, 889-912. RICHE, M.F. (2003) How Changes in the Nation's Age and Household Structure will Reshape Housing Demand in the 21st Century. In J. Johnson and J. Cigna (eds.), Issue Papers on Demographic Trends Important to Housing, U.S. Department of Housing and Urban Development, Office of Policy Development and Research 125-47. ROHE, W., VAN ZANDT, S. and McCarthy, G. (2002) Home Ownership and Access to Opportunity. Housing Studies 17, 51-61. ROSSI, P.H. and WEBER, E. (1996) The Social Benefits of Homeownership: Empirical Evidence from National Surveys. Housing Policy Debate 7, 1-36. RUGGLES, S., SOBEK, M., ALEXANDER, T., FITCH, C.A., GOEKEN, R., HALL, P.K., KING, M. and RONNANDER, C. (2003) Integrated Public Use Microdata Series: Version 3.0. Historical Census Projects, University of Minnesota, 2003, Minneapolis, MN. SIMMONS, P. A. (2001) A Coast-to-Coast Expansion: Geographic Patterns of U.S. Homeownership Gains During the 1990s. Fannie Mae Foundation Census Notes 05. SKABURSKIS, A. (1994) Determinants of Canadian Headship Rates. Urban Studies 31, 1377-90. SKABURSKIS, A. (1999) Modeling the choice of tenure and building type. Urban Studies 36, 2199-215. SMALL, K.A. and HSIAO, C. (1985) Multinomial Logit Specification Tests. International Economic Review 26, 619-27. SMITH, L.B., ROSEN, K.T., MARKANDYA, A. and ULLMO, P.-A. (1984) The Demand for Housing, Household Headship Rates, and Household Formation: An International Analysis. Urban Studies 21, 407-14. SWEET, J.A. (1990) Changes in the Life-Cycle Composition of the United States Population and the Demand of Housing. In D. Myers (ed.), Housing Demography: Linking Demographic Structure and Housing Markets, University of Wisconsin Press, Madison, WI. THE BUSH ADMINISTRATION (2004a) President Bush's Policies Support America's Small Businesses The White House, Washington DC. THE BUSH ADMINISTRATION (2004b) President George W. Bush: a remarkable record of achievement. The White House, Washington DC. U.S. CENSUS BUREAU (2008) American Housing Survey for the United States: 2007 Current Housing Reports, Series H150/07 U.S. Government Printing Office, Washington D.C. U.S. CENSUS BUREAU (2009a) America's Families and Living Arrangements: HH-1. Households, by Type: 1940 to Present. U.S. Census Bureau, Housing and 41 UU IR Author Manuscript UU IR Author Manuscript L 1 £ ' ■ • . : University of Utah Institutional Repository Author Manuscript Household Economic Statistics Division, Fertility & Family Statistics Branch, Washington DC, Current Population Reports. U.S. CENSUS BUREAU (2009b) America's Families and Living Arrangements U.S. Census Bureau, Housing and Household Economic Statistics Division, Fertility & Family Statistics Branch, Washington DC. U.S. CENSUS BUREAU (2009c) Housing Vacancies and Homeownership (CPS/HVS): Historical Homeownership Rates for the U.S. and Regions U.S. Census Bureau, Housing and Household Economic Statistics Division, Washington DC. VAN HOOK, I. and GLICK, J.E. (2007) Immigration and living arrangements: Moving beyond economic need versus acculturation. Demography 44, 225-49. WACHTER, S.M. and MEGBOLUGBE, I. (1992) Racial and Ethnic Disparities in Homeownership. Housing Policy Debate 3, 333-70. WOODWARD, I. and DAMON, B. (2001) Housing Characteristics: 2000. Census 2000 Brief, U.S. Census Bureau, Economics and Statistics Administration, Washington DC, 8. YU, Z. and MYERS, D. (2007) Convergence or Divergence in Los Angeles: Three Distinctive Ethnic Patterns of Immigrant Residential Assimilation. Social Science Research 36, 254-85UU IR Author Manuscript UU IR Author Manuscript ■ ■ B : ' ' : University of Utah Institutional Repository Author Manuscript NOTES I According to the Census Bureau, A householder in the U.S. census refers to the person (or one of the people) in whose name the housing unit is owned or rented (maintained). If the house is owned or rented jointly by a married couple, the householder may be either the husband or the wife. There is only one householder per household. The Census Bureau formerly used the term head of the household to describe householders. Renter householders refer to householders who live in rental households, while owner householders refer to householders who live in owner households. II Authors' calculations from the 2007 American Housing Survey found that homeowners were 68.3% of all households in 2007 and renters accounted for 31.7% of households, but rental units were home to 77.1% of all householders who had moved from housing units rented or owned by someone else or whose previous residence was in institutional quarters U.S. CENSUS BUREAU (2008) American Housing Survey for the United States: 2007 Current Housing Reports, Series H150/07 U.S. Government Printing Office, Washington D.C.. III The test was described in Scott Long and Freese (2006).. The test results are available upon request. IV To make the housing quality consistent across metropolitan areas, we restrict the sample to rental units that only include two or three bedrooms, and owner units that only include three or four bedrooms. y In 1990, the change in Housing Price Index was measured from 1985 to 1990; in 2000, it was from 1995 to 2000; in 2006, it was from 2001 to 2006. 4 3