«AmericAn neighborhoods: inclusion And exclusion Volume 16, Number 3 • 2014 U.S. Department of Housing and Urban Development | Office of Policy ...»
Social Effects of HOAs Much of the controversy over HOAs relates to issues of exclusion or fragmentation (socially, racially, and economically). These topics have received much less attention than the fiscal ones. Meltzer (2013) offers the most comprehensive analysis of how HOAs can affect racial/ethnic and income segregation. Unlike previous studies, she observed jurisdictions over multiple decades in an attempt to better identify whether the growth in HOAs is driving changes in segregation. Results from ordinary least squares and instrumental variable regressions indicate that an increase in HOA presence exacerbates Black-White and Hispanic-White residential segregation. Any segregation, however, is tempered by the concentration of HOA units in larger communities. On the other hand, no significant effect of HOAs affecting income segregation exists, which suggests that HOAs do not intensify existing tendencies toward income sorting.
Gordon (2004) made one of the first empirical contributions by examining the residential composition of PUDs in California in 1990 and their association with overall metropolitan segregation.
Gordon used the entropy index of segregation to measure diversity among several races and income groups at the block group and metropolitan level. She found that PUD block groups are less racially diverse than other block groups in central city and suburban areas. She also found that PUD block groups are more diverse with respect to income, but this heterogeneity is largely because PUDs include more households in relatively higher income brackets. At the metropolitan level, the difference between PUDs and other block groups explains a very small share of total segregation. Gordon suggests that the lack of an effect at the metropolitan level is not surprising, given the small proportion of the population that lived in PUDs as of 1990, but she cautions that residential segregation will become more pronounced as HOA membership increases over time (which it certainly has).
Also studying California, Le Goix (2005) executed a neighborhood-level analysis of gated communities and segregation in Los Angeles. He measured segregation by comparing the level of socioeconomic differentiation between gated communities and their neighboring areas and the differentiation between any other two adjacent neighborhoods; if the former differentiation is higher, then he concludes that gated communities are associated with increased segregation. Similar to Gordon, Le Goix did not find evidence to support an association between gated communities and segregation at the level of the municipality. He also observed that gated communities tend to exist in ethnically homogeneous neighborhoods (which are observed at the census block group) and are themselves homogeneous in terms of age and socioeconomic status.
Vesselinov (2008) was the first to test segregation and gated communities for multiple cities in the United States. Using data from AHS on membership in gated communities, as of 2001, Vesselinov found that segregation and the number of gated communities are associated with higher proportions of recent immigrants. She also found that although gated communities are prevalent in the southern and western regions of the country, segregation is less prevalent in these regions. Because the analysis is contemporaneous (she uses 2000 Census data), the implications of her findings
Cityscape 75Cheung and Meltzer
are ambiguous—it is not clear whether gated communities are simply tempering segregation or whether they have simply emerged within less segregated metro areas. Vesselinov also noted that a number of characteristics often associated with segregation, such as proportion of the population that is Black or college-educated, are not associated with gated communities.
Because of their exclusivity, RCAs can fragment communities not only demographically, but also civically (or politically). Gordon (2003) empirically tested the validity of such claims by operationalizing social capital by residents’ voting behavior. She specifically analyzed the effects of PUDs in California on voting behavior in statewide general elections during the 1990s. Results indicate that areas with PUDs do not exhibit significantly different voter turnout, registration, and party affiliation after potential selection bias is taken into account. These findings call into question the popular view that private governments crowd out participation in traditional public government.
In sum, HOAs do create value for their owners, as evinced by their properties’ sales price premiums relative to non-HOA properties. HOAs can also affect the quality of life, however, for nonmembers in a municipality. The nature and degree of public services are influenced by HOA presence, as are segregated living conditions.
Predicting HOA Formation Developers are intentional and strategic in building HOA-governed housing; in other words, the emergence of HOAs is not a random phenomenon. The nonrandom nature of their growth has both policy and methodological implications. If it turns out that HOAs and other private governments are beneficial for their members, then any disparities in access to these associations (and the services they provide) raise questions of equity. Is it appropriate for the public sector to support and facilitate the formation of these private institutions? On the other hand, the efficiency gains from their localized service provision could bestow benefits for members and nonmembers alike, and this outcome may be more politically (and socially) appealing. As demonstrated previously, sophisticated empirical efforts have started to answer many of these questions. Ignorance of the nonrandom nature of HOA formation could bias the estimates of their financial and social effects, however. For example, if we do not account for the fact that HOAs tend to locate in the outskirts of municipalities, where not only is more land available, but also more money is required to build because of new infrastructure requirements, we could be observing inflated price premiums. This error falsely informs not only policy decisions but also consumer decisions.
In this analysis, we propose a three-pronged framework for considering HOA formation, which we will implement in the estimation strategy that follows. The likelihood of HOA formation should depend on (1) demand-side factors, (2) supply-side factors, and (3) institutional factors. We focus on within-municipality formation and consider the likelihood of any neighborhood receiving an HOA. This scale of analysis is compelling, because HOAs are in fact experienced at the community level, and the prevalence of HOAs among submunicipal neighborhoods has implications for the residential and service composition of the host municipality overall.
Demand-Side Factors The likelihood of HOA formation will depend on the preferences of existing (and potential) residents. The preferences of potential HOA homeowners matter because they are the ones purchasing
the housing; the preferences of existing residents matter in so much as they can influence the successful completion of any particular HOA development. A long line of research on housing segregation also suggests that households typically choose to (or are encouraged to) locate near other households of similar socioeconomic positions (Bayer, McMillan, and Ruben, 2004; Ellen, 2006;
Yinger, 1995). Therefore, we would expect to see the socioeconomic characteristics of existing residents positively correlate with those of new HOA homeowners simply because residents prefer familiar neighbors. On the other hand, if the HOA serves as a mechanism to retain homogeneity within an otherwise diverse community, the two may be negatively correlated. We rely on this correlation (whatever direction it may be) to model HOA formation.
We specifically hypothesize that the preferences of potential HOA owners should be correlated with the economic and demographic characteristics of current residents. Most obviously, we would expect to see an increase in the likelihood of HOA formation among more affluent residents, because they have the means to pay for the housing and the additional association fees. In addition, preferences for HOA membership (and more specifically, the services they provide) could be correlated with demographics, such as race/ethnicity and age.5 For example, communities with golf courses are more likely to attract more affluent households comprising older, White individuals, who are statistically more likely to play golf (Strahilevitz, 2005). HOAs also presumably offer a more controlled or exclusive residential community, and preferences for this type of living environment may also fall along demographic lines.
Supply-Side Factors Because HOAs typically accompany new housing developments, the likelihood of their formation should be correlated with factors that facilitate the physical production of the homes they govern.
The availability of land is paramount, and, specifically, enough consolidated land to build often large or sprawling developments. All else being equal, HOAs should be more likely to form where it is easier to build new, sizable housing developments. Thus, distance to the central city should be negatively correlated with the location of HOAs. In addition, the vacancy rates, homeownership rates, and age of the local housing stock capture the composition and tightness of the existing housing market.
Institutional Factors Finally, we consider broader, what we term institutional, factors that can affect the likelihood of HOAs at the neighborhood level, across municipalities. Existing empirical evidence suggests that HOAs do interact with the public sector in their service provision (Cheung, 2008b; Cheung and Meltzer, 2013). Therefore, the likelihood of HOA formation could also be a function of municipalitywide fiscal and regulatory conditions. For example, HOAs could be more likely to form in municipalities with lower per capita spending on services (especially services that tend to overlap with HOAs’ responsibilities); in this case, the HOA is forming in response to some underprovision by the public sector.
This correlation is in addition to any correlation between income and race, ethnicity, and age.
Model Because we are interested in the conditions that correspond with HOA formation in a particular census tract over time, we take a duration analysis approach. This analytical approach enables us to include a set of temporally changing covariates, and we can eliminate from the “eligible” tracts the ones that already have an HOA. Therefore, we are really getting, at any point in time, the likelihood of the first HOA adoption. We follow Florida census tracts from 1970 to 2008 and relate the time that passes before an event (“failure”) to time-varying demand-side, supply-side, and public finance (institutional) covariates. A tract experiences failure when the first HOA incorporates within its boundaries. This observation represents an uncensored observation. If a tract never has an HOA form, it is a censored observation.
We fit a Cox proportional hazards model with time-varying covariates. The hazard function, which describes the instantaneous risk of an HOA forming at a point in time, is assumed to take on the following form—
where λ0(t) is the baseline hazard function and X is the covariate vector. By assuming proportional hazards (that is, that the covariates are multiplicatively related to the hazard), it is possible to estimate the β (the coefficients on the covariates) with the baseline hazard unspecified. The exponentiated coefficients can be interpreted as multiplicative effects on the hazard.
It is also possible to stratify the baseline hazard functions across a particular set of categories. We stratify the hazards by counties, because counties in Florida can differ substantially in demographics, economic makeup, and government (all of which could be correlated with the likelihood of HOA formation at the neighborhood level). The stratified Cox model thus fits the following model—
Although the coefficients β are the same for each county, the baseline hazard functions are allowed to be different for each county. We first present unstratified and then stratified estimation results in the exhibits that follow.
Data In this section we describe the data sources for our analysis and present an overview of the data in our sample.
HOA Data Our duration variable is identified off of the time until a particular census tract obtains its first HOA. Therefore, we need to know the precise location of each HOA in the state. Florida has obvious advantages for such an analysis: it has one of the highest numbers of HOAs in the United States (more than 16,000 as of 2010), and its municipalities are relatively diverse in terms of density and demographic and economic composition. Information on Florida HOAs was obtained 78 American Neighborhoods: Inclusion and Exclusion Why and Where Do Homeowners Associations Form?
from Sunshine List, a private, Florida-based corporation that has compiled the most comprehensive and up-to-date list of HOAs in the state. This dataset includes information on the location and creation date of every active HOA in Florida as of 2008 (the first HOA was incorporated in 1959).6 This company compiles a list of all the HOA officers in the state for the purposes of marketing to service providers (lawyers, accountants, landscapers, and so on). Each entry includes information about an officer who sits on the board of the HOA, a unique HOA identification number, the officer’s address, and the incorporation date of the HOA.
Using Geographic Information System (GIS) software, we geocode the reported addresses of the officers onto an electronic parcel map of the state obtained from the Florida Department of Revenue.
Because HOA officers generally live in the HOA they serve, we overlay a census tract map on the parcels, and we assign to each census tract the year of incorporation for the first HOA in that tract.