Segregation by Design: Mechanisms of Selection of Latinos and Whites into Gated Communities

Abstract

Gated communities have been thought to contribute to urban inequality, but empirical evidence is limited. This study utilizes the American Housing Survey for 2001 to examine the differential access of Latinos and Whites to gated communities in metropolitan United States. The results show that education is the most important sorting mechanism: as education increases, so does the probability to gate. On one hand, education trumps the effects of social class for owners, leading to segmentation within each class category, regardless of race/ethnicity; on the other hand, Latinos with higher education tend to select gated residences more often than comparable Whites.

Introduction

The rapid increase in the number of gated communities is alarming to many scholars, because they see this process as an extension of residential segregation (Davis 1990; Blakely and Snyder 1997; Marcuse 1997; Connell 1999; Caldeira 2000; Low 2003a). In the pursuit of security from crime and increased social diversity, of higher property values, and a better sense of community, affluent Whites are said to have found yet another mechanism of seclusion. As a result, gated communities are thought to produce more social separation and fragmentation, which further leads to increased urban inequality. Since gated communities are having wider impact on cities, it is important to bring more evidence to bear on whether or not gated communities increase urban inequality. In this study, we focus on the differential access to gated enclaves, particularly by Whites and Latinos.

While the majority of scholars across disciplinary boundaries regard gated communities as wealthy enclaves, there is no clear conceptualization of gated communities from a sociological point of view. Scholars conceptualize gating by comparing it to similar social processes, depending upon disciplinary boundaries. Most prominently, McKenzie (1994, 2003) relates gated communities to Common Interest Developments and to privatization of government within the context of urban politics; Low (2003a) organizes her ethnographic study of gated communities around psychological and social explanations within the context of urban anthropology; Glasze (2005) and Le Goix and Webster (2006) frame gated communities as “club goods,” within the context of urban geography.

This is the first study that situates gated communities in the context of three theoretical models in the field of urban sociology: the spatial assimilation model, the place stratification model, and the ethnic community model. The first two complementary perspectives have generated a lot of research findings and explanations regarding the residential patterns of minorities and Whites in urban America. The third model is relatively more recent and therefore applicable to the more recent residential patterns reflecting the changes in immigrant streams. All three perspectives seem particularly useful in order to understand the mechanisms of selection into gated communities by Whites and minorities. Placing the study of gated communities within the context of spatial assimilation, place stratification and ethnic community models in turn helps to broaden the scope and explanatory power of these perspectives.

The place stratification model directs attention to the importance of racial and ethnic segmentation of housing markets and the unequal access to more affluent neighborhoods. Guided by the understanding that residential location is in some cases determined by race/ethnicity, we ask: Does the selection mechanism into gated communities vary by race/ethnicity? The spatial assimilation perspective posits that as minority members obtain higher levels of human and social capital they become more likely to access affluent suburban neighborhoods. This understanding informs our second research question: Are Latinos of higher socioeconomic standing equally likely to live in gated communities as comparable Whites (Whites of similar socioeconomic standing)? The ethnic community model sees minorities as no longer constrained to either assimilate or be denied access to affluent neighborhoods. Despite possessing the human and social capital to enter White affluent suburbs, minority members choose to establish their own affluent ethnic communities. Therefore, this model suggests that Latinos following recent trends in increased gatedness in the United States, as well as in Latin-American countries, may in fact tend to gate at a higher rate than Whites of comparably high socioeconomic status.

To adjudicate between these three research questions, we use the American Housing Survey National Sample data from 2001, which is the first representative housing data for the United States, containing questions about gated communities. We make two important choices: we study the selection mechanisms into gated communities only for Whites and Latinos and we limit our study to the South and West regions of urban America.

We choose Latinos for two reasons: one, since 2000 Latinos constitute the most numerous minority in the United States and the fastest growing minority as well. The 2000 U.S. Census revealed that Latino population has outnumbered Blacks as the largest minority and has grown from 21.9 million in 1990 to 35.3 million in 2000, a growth of 61% (Spatial Structures for Social Sciences 2001), as opposed to the total population growth of 13% over the same decade. According to the American Community Survey in 2004, the total Latino population has grown to 40.5 million people, which constitutes 14.2% of the total population (U.S. Census Bureau 2006).

The second reason comes form our preliminary work, which shows two processes: (1) Based on 2001 AHS data, gating is positively related to the number of Latinos, net of other characteristics; and (2) the increase of the Hispanic population in the South and the West is associated with an increase in gated communities between 2001 and 2003, controlling for other relevant factors. The increase in gating and the increase in Latinos are most prominent in the South and the West regions of America.¹ Therefore, to understand more clearly the change in residential patterns brought about by gated communities, we focus only on those two regions.

No studies to date have focused on the racial differential mechanisms of selection into gated communities. Sanchez, Lang, and Dhavale (2005) study the causal mechanisms of gating and while most prior research had focused on gated communities as oases of White, affluent homeowners, the authors show that renter gated communities seem to be more prevalent among Hispanics than among other minorities, including Blacks. At this point however the differences by race/ethnicity in the mechanisms of selection into gated communities are not clear. Our study is the first to address this issue and lead the way in the better understanding of minority incorporation into the new residential form of gated communities.

Theoretical Background

Until around 1970, gated communities (hereafter GCs) were still relatively rare consisting mostly of gated retirement communities in retiree-attractive climates such as California, Texas, Florida, and Arizona (Wilson-Doenges 2000). Following the lead of these master planned retirement developments, other GCs were constructed around country clubs and resorts as havens for the upper tiers of society (Blakely and Snyder 1997; Low 2003a). As these areas were also the first to experience large waves of Hispanic immigrants, fear and uncertainty led White residents to seek security behind the gates of housing developments (Low 2003a; Byers 2003).

Blakely and Snyder (1997) and Low (2003a) contend that Whites tend to fortify themselves behind gates for reasons such as fear of the increased diversity, fear of increasing crime rates, desire to protect their property values, and desire to create a sense of community. The fortification of Whites seems similar to the mechanisms of residential segregation. However, the gating of Whites is only part of the process; Latinos themselves are choosing gated residences (Sanchez, Lang, and Dhavale 2005; Vesselinov 2008). If GCs were only an extension of segregation we would not expect to find a lot of minority members residing in these communities.

The investigations of this issue led us further to McKenzie’s work (2005) of describing an established community in Las Vegas where some residents lobby to build a wall around the community. One of the arguments made by members of the homeowner’s association was that most respectable communities in the area were gated and to attract more prosperous, younger, and professional residents, this older community needed to build a wall and a gate itself. Then, is it the case that Latinos are in fact participating in a fashion trend carving their own “privitopias” in the urban space?

To adjudicate between the different explanations we turn to three theoretically complementary perspectives: spatial assimilation and place stratification (Alba and Logan 1991, 1992; Logan and Alba 1993; Charles 2003) and the ethnic communities model (Logan, Zhang, and Alba 2002). We first consider the definition of GCs and then focus in more detail on each of the theoretical perspectives.

Definition of Gated Community

Gated communities have been defined spatially in two major ways, either as subunits within more general territories or as independent spatial units. One group of scholars considers GCs as a facet of large planned communities or Common Interest Developments (McKenzie 1994, 2003; Luymes 1997; Kennedy 1995; Gordon 2004). There is also a growing literature, which includes GCs in the category of private neighborhoods (Nelson 2005²; Glasze, Webster, and Frantz, 2006). Alternatively, others argue that the existence of fences and walls, and security features (guards, surveillance cameras), distinguish GCs as a residential setting that is significantly different from nongated enclaves (Blakely and Snyder 1997; Le Goix 2003; Low 2003a; Vesselinov, Cazessus, and Falk 2007).

The latter approach is more important if we want to understand the specific ways in which GCs change the residential patterns in urban America. On one hand, a very important institutional aspect of GCs is that, like all private neighborhoods, they are characterized by homeowner associations, where elected boards oversee the common property and establish covenants, conditions, and restrictions (CC&Rs) as part of the deed. However, on the other hand, unlike CIDs, planned communities, or even individual gated residences, GCs are surrounded by a secured barrier that denies all public access not only to personal residences but also to the area’s streets, sidewalks, and neighborhood amenities. Therefore, for the purposes of our analyses, we adopt Low’s definition of a GC: “[a] residential development surrounded by walls, fences, or earth banks covered with bushes and shrubs, with a secured entrance” (Low 2003a:12). The GC’s physical barriers (walls, gates, security guards) not only put them in a separate category of privatization, commodification, and seclusion of space. More generally, by analyzing gating as a separate urban process, it is possible to isolate the specific characteristics and the specific implications from this process for urban inequality.

Spatial Assimilation Perspective

The spatial assimilation perspective posits that as immigrant and nonimmigrant minorities accumulate more cultural, social, and economic capital, the chances that they will enter more affluent White neighborhoods increase. Since suburban areas have higher socioeconomic status (Logan and Schneider 1982) the process of spatial assimilation is most pronounced in these areas. As Alba and Logan argue (1991: 432) “suburbanization is an important indicator to the extent to which minorities are becoming integrated more fully into American society.” In many ways, choosing a GC residence parallels moving into a more affluent Anglo neighborhood because for a long time GCs have functioned as wealthy White residential enclaves (Blakely and Snyder 1997; Low 2003b).³

Massey and Mullan (1984) consider the process of spatial assimilation for Hispanics,⁴ especially in the southwestern region of the United States. Their research indicates that, like Blacks, when Hispanics move into an area that is predominantly White, succession sometimes occurs. However, when specifically considering Latino “invasions” into White neighborhoods, succession occurs in less than half of the cases. Instead of being based purely on race, the residential choice of Whites in these neighborhoods often depends on the characteristics of the Latinos moving into the area. As a result, assimilation more readily occurs for Latinos of higher socioeconomic status.

Similar conclusions have been arrived at by researchers who conducted studies in the 1990s. Some Latinos do follow the patterns of Whites moving into the suburban areas surrounding central city areas, particularly those with the most positive background characteristics, such as good English-speaking ability and high socioeconomic status (Logan, Alba, and Leung 1996). Alba et al. (1999) maintain that among Latinos and Caribbean Hispanic groups, being married but not having children makes you more likely to live in suburban areas. The authors believe that this pattern suggests the lesser ability of Hispanic families with children to have choices in regards to residential location.

At the same time Alba et al. (1999: 458) argue that “between 1980 and 1990 the neighborhoods occupied by middle-income Asian and Hispanic suburbanites became more ethnically and racially diverse (containing fewer non-Hispanic whites) than did those of their white counterparts.” Still, those neighborhoods were no less affluent. Therefore, according to the authors we may be witnessing a weakening in the core of the spatial-assimilation model.

The conclusion about racial and ethnic diversification of suburban neighborhoods is confirmed in another study. Alba, Logan, and Stults (2000) find that in three metropolitan areas, Los Angeles, Miami, and San Francisco, the majority status of Whites within the suburbs is diminishing, this time regardless of the area’s affluence. This trend may indicate future patterns; as the number of Latinos continues to increase rapidly, they may be living in suburban neighborhoods composed mostly of other Latinos, or possibly other minorities.

Based on the spatial assimilation model we expect that many affluent Latinos, better educated, with higher incomes and less children, would exhibit the same gating patterns as comparable Whites (Whites of similar socioeconomic standing).⁵

Place Stratification Approach

Even when Latinos do move into the more affluent suburban areas, there appears to be a socioeconomic gradient (Alba and Logan 1991). This gradient suggests that living in suburban areas comes at a far greater price for minorities than it does for Whites. The price differentials, the levels of residential segregation from Whites, and the issues of housing discrimination evoke the need to apply the place stratification approach. According to Alba and Logan (1991: 432), the stratification model is warranted by the racial segmentation of housing markets, which segmentation has been widely confirmed in the experiences of Blacks (Massey and Denton 1993; Oliver and Shapiro 1995; Briggs 2005) as well as Latinos (Betancur 1996; Logan, Alba, and Leung 1996; Alba, Logan, and Stults 2000; Kandel and Cromartie 2001).

Between 1970 and 1980 and again in the 1990s, Hispanic segregation was much below the level of Black segregation, but it has substantially increased in many urban areas where the population of Hispanic immigrants has been growing (Massey and Denton 1987; Logan, Alba, and Leung 1996). As with Blacks, rates of segregation are higher in regions where there are large minority populations. Therefore, as Latinos continue to migrate to the United States and often move to specific areas where there are a large number of other Latinos, they are facing issues of residential segregation in these areas more than in the greater United States (Logan, Alba, and Leung 1996). The latest trends indicate that Hispanic segregation increased in regions where the members of the group had declining incomes relative to Whites in addition to a growing proportion of Latinos (Logan, Stults, and Farley 2004).⁶

In three metropolitan areas, Los Angeles, Miami, and San Francisco, the majority status of Whites within the suburbs is diminishing, another conclusion drawn from the place stratification model. In these suburban areas, the percentage of Whites is slowly declining, regardless of the affluence of the area (Alba, Logan, and Stults 2000). The authors argue that this trend may indicate future patterns for Hispanic suburban dwellers. As the Latino population continues to rapidly increase, Latinos may be living, in the future, in suburban neighborhoods composed mostly of other Latinos, or possibly other minority groups, like African-Americans and Asians.

Residential segregation between Whites and Hispanics in many smaller geographic areas has also increased during the 1990s, again especially in those counties with rapidly growing populations. Kandel and Cromartie (2001) argue that the current distribution of Hispanics and Whites in high-growth Hispanic areas is similar to the pattern observed among Blacks and Whites from 1970 and 1990 in the areas of the South. Hispanics seem much more likely to live in larger cities, while Whites are moving increasingly into areas outside of the Census-defined city.

Therefore, based on the place stratification model it seems that while some of the more affluent Latinos may join Whites in GCs, this will not be the overriding pattern of gating for Latinos. Residential preferences and ethnic enclave experiences may lead to consolidation of their own GCs rather then sharing such communities with Whites.

Ethnic Community Model

Racial and cultural residential preferences.

The ethnic community model is to a large extent based on residential preferences as well as level of resources. In a study of expressed neighborhood racial/ethnic composition preferences in Los Angeles, Clark (1992) found that all four ethnic groups studied showed strong desires to live in areas composed of own-race neighborhoods. Even households who suggested that they had “no preference” in regards to racial composition of their neighborhood generally made choices in favor of own-race neighborhoods. Fewer than half of Hispanics who expressed no real preference in regards to the racial makeup of their neighborhood chose neighborhoods that were composed of 80% Hispanics. And while 38% of Whites and 35% of Blacks stated they had “no preference” regarding this topic, only 19% of Hispanics reported having “no preference.”

Charles (2000) reports that Whites, as the most desired out-group, make up between 23% and 33% of minority’s ideal neighborhoods. White respondents’ ideal neighborhood approaches 50% White, compared to mean same-race preferences of 41% for Latinos and Asians, and 37% for Blacks. The majority own-group neighborhoods seem most attractive to Latinos, who prefer Whites as their neighbors the most, then Asians, and lastly Blacks (Charles 2005). These findings suggest a rather strong desire of many Latinos to live among those of their own racial/ethnic background. Based on the evidence about these residential preferences, it seems that part of Latinos will not only be excluded from White GCs, but that they willingly may be searching and building their own GCs.

Cultural preferences, based on the country of origin, may also influence the residential patterns of Latinos in the United States. The spread of GCs in Latin America is well documented in the literature (Caldeira 1996, 2000) and it could be argued that Latinos are used to the existence of GCs even before they immigrate to the United States. Living in GCs in Latin America is associated with high status, higher level of security, and certainly higher level of material resources. Having been accustomed to an environment of gated enclaves and finding similar enclaves in California, Florida, and other places with higher concentration of Latinos, some Latinos may be more inclined to emulate and seek this residential form.

Capitalizing on such cultural preferences, Dávila (2001) and Ramos (2004) argue that there are developers and marketing agencies that target specifically the Latino population living in the United States. Dávila (2001) argues that oftentimes developers even utilize the Spanish-language media outlets such as magazines, radio stations, and television channels, to advertise GCs specifically to Latinos.

Ethnic enclaves and ethnic communities.

Many Latino and Asian immigrants share the preference of congregating in ethnic enclaves—“El Barrio” (Freidenberg 2000), “Little Havana” (Portes and Stepick 1993), “Chinatown” (Chen 1992; Zhou 1992), or “Koreatown” (Waldinger 2001; Danico 2004). Sometimes such ethnic concentrations are interchangeably called either “ethnic enclave” or “ethnic community.” However, distinctions should be made between each based on the different mechanisms that bring minorities into these neighborhoods (Logan, Zhang, and Alba 2002).

Ethnic enclaves are concentrated in popular immigrant gateways, which are usually metropolitan areas with long history of accommodating immigrants (e.g., New York, Los Angeles, Miami). Nee, Sanders, and Sernau (1994) argue that newly arrived Asian immigrants tend to locate job opportunities in ethnic enclaves first because barriers like language and cultural differences limit these new arrivals’ social mobility in the host society. Similarly, other enclave studies (Portes and Jensen 1989; Funkhouser and Ramos 1993; Ebaugh and Curry 2000) find comparable dynamics among Latino immigrants’ selection of housing in these gateway cities. For the new immigrants, enclaves serve as a cultural buffer zone (Kramer 1970). At the same time, enclave residents are penalized by lower incomes, reduced health benefits, and poorer working conditions (Gilbertson and Gurak 1993; Waldinger 1996, 2001). Given the prevalence of the disadvantages, many immigrants tend to reach out for nonethnic job opportunities and residences after they acquire enough human and social capital (Zhou and Logan 1991; Nee, Sanders, and Sernau 1994).

Conversely, while ethnic enclaves are established because of economic and social necessity (Logan and Molotch 1987; Funkhouser and Ramos 1993; Nee, Sanders, and Sernau 1994), ethnic communities are based upon individual preferences and taste. Also, ethnic communities are often located in suburban areas and constitute high-status areas and are therefore less likely to be used as ports of entry by immigrants who have limited residential and job opportunity (Zhou and Logan 1991).

The trends in residential preferences, in establishing central cities’ ethnic enclaves and suburban ethnic communities lead to the conclusion that probably some GCs can be theorized as ethnic communities. Given the uneven history of Latino entry into more affluent Anglo neighborhoods, we may observe a gating pattern where minority members, who have obtained higher socioeconomic status, establish their own GCs adjacent to the ethnic enclaves where they have lived prior to obtaining such status. Betancur (1996) argues that higher-income Latinos often move to the fringes of Latino areas. This residential choice would allow some Latino immigrants to “upgrade” from living in inner-city enclaves yet to be close enough to their core culture. At the same time, better-off Latinos may also choose to locate in suburban areas and establish their own suburban ethnic GCs.

Research Design

Integrating the theoretical perspectives of the place stratification approach, the spatial assimilation approach, and ethnic community, we arrive at three research questions: (1) Does the selection mechanism of choosing residence in a GC vary significantly by race/ethnicity? (2) Are Latinos of higher socioeconomic standing equally likely to live in GCs as comparable Whites? (3) Are Latinos more likely than Whites to live in GCs?

The first question is based on the place stratification perspective and corresponds to our first hypothesis:

• Hypothesis 1: After controlling for relevant characteristics (social class, education, and nativity), Latinos will exhibit a lower tendency to gate than Whites.

The second question stems from the arguments based on the spatial assimilation model and yields two further hypotheses:

• Hypothesis 2: After controlling for relevant characteristics (social class, education, and nativity), Whites and Latinos will exhibit similar gating patterns.
• Hypothesis 3: Latinos with higher socioeconomic status and educational attainment will be more likely to gate than Latinos with lower socioeconomic status and educational attainment.

The third research question leads to the last two hypotheses:

• Hypothesis 4: After controlling for relevant characteristics (social class, education), Latinos exhibit a higher tendency to gate than Whites.
• Hypothesis 5: Latinos with higher socioeconomic status and educational attainment are more likely to gate than Whites with similar socioeconomic status and educational attainment.

If the place stratification model helps to better explain the relationship between race/ethnicity and gating then we would expect that when all relevant factors are held constant, Latinos will be significantly less likely than Whites to live in GCs. At the same time, the spatial assimilation model would predict that the effect of race/ethnicity on the propensity to gate should be conditioned by Latinos’ social class and education. Latinos of higher socioeconomic standing (social class and education) should be as likely to live in GCs as Whites of the same standing. In addition, Latinos with higher socioeconomic status will be more likely to gate than Latinos with lower socioeconomic status. The ethnic community model would predict that Latinos are more likely than Whites to live in GCs, given cultural patterns in their countries of origin (or ancestry) and the regions in the United States they are most concentrated in.

Data, Variables, Method

Data.

The data for this analysis come from the 2001 National Sample⁷ of the American Housing Survey (AHS). The AHS is the largest, regular national housing sample survey in the United States. It is the only representative survey in the United States that contains questions related to GCs. The U.S. Census Bureau conducts the AHS to obtain up-to-date housing statistics for the Department of Housing and Urban Development. The AHS is a panel sample of housing units, initiated in 1973, carried out annually through 1984, and biannually since then. The AHS contains both a national sample and a sample of selected metropolitan areas.

Dependent variable.

The dependent variable in the analysis is a dichotomous variable capturing whether or not the household lives in a GC. In the AHS survey the variable is measured by a “yes/no” answer to the question: “Is your community surrounded by walls or fences preventing access by persons other than residents?”⁸

Independent variables.

The independent variables included in the analyses are those established in prior research on residential patterns. First, we incorporate a set of demographic characteristics: Hispanic or White, age, family size, nativity, sex, marital status, and whether or not the household has moved between 1995 and 2000. As “Whites” we have selected all non-Hispanic Whites combining two measures: the first measure captures the respondent’s answer to the question about race while the second measure captures whether or not the respondent is of Spanish origin by using two categories, yes or no. We define as Latinos all respondents answering “Yes” to the question about Spanish origin.

In residential research, the variable “age” is usually accompanied with its squared term, because it has been shown that age has a curvilinear effect particularly related to homeownership and mobility. Family size is an important indicator and for Latinos we expect that the higher number of family members will be linked with lower likelihood of living in a GC. Nativity is measured with a variable, which in AHS captures the “Country of birth of the householder.” As native born, we have designated all respondents being born in the United States or Puerto Rico.

The second set of independent variables included in the analyses is a group of socioeconomic characteristics measuring the respondent’s educational attainment and social class. Since in the AHS data, education is not a continuous variable that measures the years of schooling but a categorical variable, we have recoded the variable to capture four categories of respondents: people with less than high school, people with a high school diploma, people with an Associate degree or some college, and people with a college degree or more. The social class measure was constructed using the AHS variable “Family Income,” which is based on salary and wages, and it includes all other income. The other income usually comes from sources such as dividends from stocks and bonds, business, savings, etc. Therefore, our social class measure is based on both income and wealth capturing two central social characteristics. We have differentiated three categories: lower, middle, and upper class. Lower class is composed by respondents with family income equal to or less than $27,000, which is 150% above the official Census Bureau poverty threshold ($18,104) for a family of four. Middle-class respondents have reported income between $27,000 and $73,000, where the upper limit corresponds to 400% above the poverty line. The third group, the upper class, contains all respondents who have reported income above $73,000.⁹

The third set of variables includes two housing characteristics, number of bedrooms and age of structure, the latter addressing an issue closely related to GCs: scholars argue that recent housing, particularly in the South and West, is more likely to be within GCs than housing built in the past.

The last fourth set of independent variables consists of four metropolitan-level characteristics: urban versus suburban, percentage Latino, percentage units built in the 1990s and percentage of the labor force in active military service. The first variable is important, because rental residences tend to cluster in the central cities, versus suburban housing that tends to be composed mostly of single-family housing. The second variable, percentage Latinos, is necessary because Latinos are usually concentrated in few metropolitan areas and their access to GCs may be conditioned on their proportion of the population. Related to the possible population effect is also the fact that the concentration of Latinos is in areas of recent growth not only in population but also of new housing. Therefore, we include a control variable for the percentage housing units built after 1990. Lastly, the variable percentage of the labor force in active military service is used as a proxy to control for a possible effect of the disproportionate presence of military housing in some metropolitan areas.¹⁰

Methods.

Given our dependent variable, the best methodological approach is to use logistic regression analysis and estimate the probability of whether a household lives in GC versus does not live in GC. Since traditionally residential patterns differ between owners and renters (Ellen 2000), and that Sanchez, Lang, and Dhavale (2005) discovered that renters constitute an important aspect in the process of gating, we decided to look separately at the mechanisms of selection into owner GCs and rental GCs. We estimate two models for owners and two models for renters: the first model in each regression analysis includes all independent predictors as specified above; in the second model, we add several interaction terms between race and education, and race and class.

Building on the logistic regression analyses we further calculate the predicted odds and odds ratios for Latinos and Whites in 12 different categories for each group, based on social class and education. This allows us to study in more detail whether the effects of race are conditioned by class and education. According to Jaccard (2001), this is a more thorough way to understand whether there are important interaction effects between explanatory variables used in logistic regression models. The predicted log odds are calculated based on the two models—one for owners and one for renters—without the inclusion of product terms:

$$\text{logit}\left(\pi \right)=\alpha +\beta 1DR+\beta 2XA+\beta 3XFS+\beta 4DN+\beta 5DS+\beta 6DRM+\beta 7XMS+\beta 8DLHS+\beta 9DHS+\beta 10DSC+\beta 11DLC+\beta 12DMC+\beta 13XNB+\beta 14XAS+\beta 15DC+\beta 16XPH+\beta 17XPB+\beta 18XPM,$$

where DR is the dummy for race; XA is the predictor variable age and the values for age are the mean values for owners and renters respectively; XFS is Family Size for which we have taken again the mean values for owners and renters; DN is the dummy variable for nativity, which equals to 1; DS is the dummy for sex, equals 1; DRM is the dummy for being a recent mover, the value of 1 denotes a householder who changed residences in the past 5 years; XMS is Marital Status and equals to 0 because we have decided to use the category “Married,” the most common category for renters and owners; DLHS is the dummy for less than high school education, DHS is the dummy for high school, DSC is the dummy for some college, DLC is the dummy for lower class, and DMC is the dummy for middle class; XNB is Number of Bedrooms, where the mean value is used for owners and renters; XAS is Age of structure, where the last category is denoted in the calculation, unit built after 1990s, because the rise of GCs is particularly significant in that decade; DC is the dummy for urban versus suburban location, we have denoted an urban location (equal to 1) in the equation for renters and suburban (equal to 0) location for owners; XPH is the is the predictor variable for percentage Latinos at metropolitan level and the values are the mean values for owners and renters respectively; XPB and XPM are the variables percentage units built after 1990 and percentage in the military and both are measured as mean values for owners and renters.

Lastly, in order to analyze in more depth whether Latinos are as equally likely as comparable Whites to live in GCs, we study the predicted probabilities and plot them against three measures: income, class, and education. Income has been a traditional axis of inequality and there are significant income differences between Latinos and Whites (Spatial Structures for Social Sciences 2001); in addition, the first GCs were ones where the more affluent groups separated themselves. Education is one of the pillars of inequality and it captures a different aspect of inequality, related more to social status than to affluence. Social class is also a traditional measure for social divisions and in our case it captures the distinctions between groups based on income and wealth. Therefore, if there are differences between Whites and comparable Latinos in the process of gating, as expected, they should be captured along these three social dimensions.

Findings

The descriptive characteristics of the variables included in the analyses are presented in Table 1.¹¹ Latinos constitute a slightly higher proportion among gated owners compared to nongated owners, 0.21 versus 0.18, and also among gated renters compared to nongated renters, 0.35 versus 0.33; however, these differences are not statistically significant. On average, gated owners tend to live in metropolitan areas with higher percentage Latinos (about 23%) as opposed to nongated owners (16% Latinos); similarly, gated renters are found in metropolitan areas with an average of 22% Latinos, compared to 18% for nongated renters. The differences for owners and renters are statistically significant.

Table 1.

Descriptive Characteristics of Variables Included in the Analyses, 2001

	Owners				Renters
	Gated (n = 313)		Nongated (n = 4,027)		Gated (n = 406)		Nongated (n= 1,701)
Variables	Mean	SD	Mean	SD	Mean	SD	Mean	SD
Demographic characteristics
White	0.79	0.41	0.82	0.38	0.65	0.48	0.67	0.47
Hispanic	0.21	0.41	0.18	0.38	0.35	0.48	0.33	0.47
Age	55.40**	16.41	52.74	15.72	41.57	15.16	41.60	14.02
Family size	2.37***	1.39	2.71	1.52	2.16***	1.28	2.52	1.58
Nativity (native-born = 1)	0.80	0.40	0.85	0.35	0.72	0.45	0.72	0.45
Sex (male = 1)	0.59	0.49	0.63	0.48	0.55	0.50	0.57	0.50
Recent mover	0.37*	0.48	0.30	0.46	0.82***	0.37	0.74	0.44
Marital status
Married	0.58**	0.49	0.66	0.47	0.37	0.48	0.38	0.49
Widowed	0.15*	0.36	0.11	0.32	0.08†	0.26	0.06	0.24
Divorced	0.17	0.37	0.14	0.34	0.23	0.42	0.27	0.45
Never married	0.10	0.30	0.09	0.28	0.32	0.47	0.28	0.45
Socioeconomic characteristics
Education
Less than high school	0.10†	0.30	0.13	0.34	0.20	0.40	0.22	0.42
High school	0.17†	0.38	0.22	0.41	0.21	0.41	0.22	0.41
Some college	O.33†	0.47	0.28	0.45	0.30	0.46	0.29	0.45
College degree	0.40	0.49	0.36	0.48	0.29	0.45	0.27	0.44
Social class
Lower	0.25†	0.43	0.20	0.40	0.40	0.49	0.40	0.49
Middle	0.39	0.49	0.42	0.49	0.48	0.50	0.49	0.50
Upper	0.36	0.48	0.38	0.49	0.12	0.33	0.11	0.31
Housing characteristics
Number of bedrooms	2 79***	0.95	3.12	0.84	1.61***	0.73	1.92	0.89
Age of structure
Unit built before 1960s	0.27***	0.45	0.52	0.50	0.27***	0.44	0.56	0.50
Unit built 1970s	0.26†	0.44	0.22	0.42	0.30**	0.46	0.23	0.42
Unit built 1980s	028***	0.40	0.16	0.37	0.28***	0.45	0.16	0.37
Unit built 1990s	0.18***	0.39	0.09	0.29	0.15***	0.36	0.05	0.22
Metropolitan characteristics
City/suburb (city = 1)	0.50	0.50	0.48	0.50	0.59	0.50	0.47	0.50
Percentage Latinos	22.53***	18.36	15.88	15.09	22.12***	14.68	18.46	15.80
Percentage built, 1990s	18.64	9.33	18.75	7.16	18.30†	9.29	17.52	8.05
Percentage military	0.88	1.62	0.93	1.70	0.70***	1.40	0.90	1.62

^*p < 0.05;

^**p < 0.01;

^***p < 0.001;

^†p < 0.1.

At the same time, some other notable distinctions can be observed between gated and nongated owners, as well as between gated and nongated renters. On average, gated owners tend to be slightly older than nongated owners, which is to be expected given the long history of retirement GCs. The average family size is lower in owner GCs compared to nongated communities, and lower in renter GCs compared to nongated. This is also to be expected since families with more children can rarely afford the higher premiums associated with house prices and rents in GCs (Vesselinov 2008). A very high percentage of gated owners are native born (80%), but interestingly this percentage is still lower compared to native born within nongated places (85%). This means that 20% of gated owners are foreign born compared to 15% among nongated owners. Since the majority of foreign born are Latinos (93%), this is the first indication that foreign-born Latinos may prefer to live in GCs.

In addition, owner GCs differ from owner non-GCs by education: a relatively higher proportion of respondents with some college or college degree and beyond can be found in GCs than outside of them. Just under three quarters of the gated owners, 73%, have at least some college education and above. The same percentage for nongated owners is 64.

The age of structure also differs for gated versus nongated owners and renters. Almost half of all owner gated units, 46%, have been built since 1980, while only 25% of all owner nongated units have been for the same period. Among gated renters, 42% live in units built since 1980; at the same time, only 21% of nongated renters live in units built during the same time period.

Does the Selection Mechanism of Choosing Residence in a Gated Community Vary Significantly by Race/Ethnicity?

The results from the logistic regression analyses are presented in Table 2, which shows the factors influencing the propensity to gate for White and Latino owners and separately for White and Latino renters.¹² The analyses reveal that there are no statistically significant differences in the propensity to gate by race/ethnicity for owners or renters; the regression coefficient capturing the potential racial/ethnicity effect while positive is not statistically significant. For owners, being foreign-born constitutes an advantage in the likelihood to reside behind gates; more specifically, foreign born are 1.5 times more likely to choose a gated residence than native born.¹³ Being a recent mover exhibits large positive influence on the propensity to gate, where respondents who moved in the past 5 years are 1.4 times more likely than nonmovers to live in GCs.

Table 2.

Coefficients from Logistic Regressions of Living in a Gated Community versus Not Living in a Gated Community, for Latinos and Whites in the South and West, 2001

Variables	Model 1, Owners			Model 2, Renters
	PE	SE	OR	PE	SE	OR
Demographic characteristics
White	-	-	-	-	-	-
Hispanic	0.020	0.201	1.021	0.136	0.176	1.146
Age	0.022	0.019	1.000	−0.019	0.024	0.981
Age2	0.000	0.000	1.000	0.000	0.000	1.000
Family size	−0.025	0.061	0.976	−0.142*	0.065	0.867
Nativity (native-born= 1)	−0.382*	0.182	0.682	−0.028	0.173	0.972
Sex (male=1)	0.038	0.143	1.039	−0.104	0.125	0.901
Recent mover	0.354*	0.144	1.424	0.389*	0.164	1.476
Marital status
Married	-	-	-	-	-	-
Widowed	0.161	0.233	1.175	−0.135	0.232	0.873
Divorced	0.147	0.199	1.158	−0.443*	0.180	0.642
Never married	0.092	0.243	1.096	−0.314†	0.177	0.731
Socioeconomic characteristics
Education
Less than high school	−0.702**	0.171	0.496	0.021	0.210	1.021
High school	−0.497**	0.186	0.608	0.002	0.184	1.002
Some college	−0.096	0.151	0.909	−0.004	0.161	0.996
College degree	-	-	-	-	-	-
Social class
Lower	−0.049	0.201	0.952	−0.322	0.224	0.725
Middle	−0.154	0.151	0.857	−0.317†	0.201	0.728
Upper	-	-
Housing characteristics
Number of bedrooms	−0.506***	0.086	0.603	−0.451***	0.090	0.637
Age of structure
Unit built before 1960s	−1.525***	0.200	0.218	−1.981***	0.219	0.138
Unit built 1970s	−0.634***	0.198	0.531	−0.886**	0.212	0.412
Unit built 1980s	−0.132	0.196	0.876	−0.636**	0.213	0.530
Unit built 1990s	-	-	-	-	-	-
Metropolitan characteristics
City/suburb (city=l)	0.034	0.132	1.035	−0.080	0.125	0.923
Percentage Latino	0.024***	0.004	1.024	0.019***	0.004	1.019
Percentage built 1990s	−0.005	0.009	0.995	0.002	0.008	1.002
Percentage military	−0.005	0.040	0.995	−0.068	0.044	0.934
Intercept	−1.364*	0.595	-	1.213	0.746	-
Hosmer-Lemeshow goodness-of-fit test			p < 0.345			p < 0.211

^*p < 0.05;

^**p < 0.01;

^***p < 0.001;

^†p < 0.1.

Prominent in the analyses for owners are the effects of education. Net of all other explanatory factors, household heads with the highest education are more likely to reside in a GC compared to people with lesser education. College graduates are more than two times more likely to live in a GC compared to people with less than high school education, and they are about 1.6 times more likely to be found in a GC compared to high school graduates.

The effect of age of structure on the propensity to gate fits the image of GCs as a fairly recent phenomenon. Contemporary GCs are usually found in the areas with newest housing buildings. If a housing unit is built in the 1990s, it almost two times more likely to be found within a GC compound compared to units built in the 1970s, and it is 4.5 times more likely to be within a GC compared to units built in the 1960s and earlier.¹⁴

Interestingly enough, the percentage Latinos at a metropolitan level has a positive and statistically significant effect on the propensity to gate (0.024). Although the effect size is small, the analysis clearly shows that the higher the percentage of Latinos in metropolises, the higher the individual propensity to gate. This alludes to structural effects related to the relative population weight of Latinos in urban areas; race/ethnicity does not have an effect on the propensity to gate at the individual level; however, it does have an effect as a contextual condition.

The analyses of the propensity to choose gated residences among renters (Table 1, Model 2) reveal similar effects of race/ethnicity as found among owners: the individual-level variable of being a Latino does not significantly affect the propensity to gate. At the same time, the percentage Latinos at metropolitan level has a positive and statistically significant effect on the outcome variable (0.019). Again, the size effect is not large; however, the results show that as the number of Latinos grows as a percentage of the metropolitan population, the individual likelihood to gate increases, too.

Some of the other predictors impact the propensity to gate among renters in similar ways as they affect owners, as well. A recent mover is 1.5 times more likely to live in a GC compared to a nonmover; the higher the number of rooms in a residence, the lower the likelihood to gate; a housing unit built after 1990 is more than 7 times more likely to be in a GC than a unit built during the 1960s or before that, 2.4 times more likely to be in a GC than a unit built in the 1970s, and about 2 times more likely to be in a GC than a unit built in the 1980s.

There are some differences between renters and owners. Divorced owners are not significantly different compared with married owners, but divorced renters are less likely than married renters to reside in GCs. One explanation for this difference is that once people are married and become owners, subsequent divorce settlements, particularly if there are children involved, may enable them financially to select housing within a GC or remain in a GC. However, this cannot be the reality for all divorcees, many of whom in fact begin to rely on less financial resources compared to married couples (Zagorsky 2005; Hoffman and Duncan 1988). For those people, especially ones having to rent their dwellings, living in a GC is not a viable option.

Another difference between owners and renters is the effect of education. Whereas education affects the propensity to gate among owners in significant ways, it has no effect on the likelihood to gate among renters. The next section studies in further detail the possibility that the effects of race/ethnicity are conditioned by education and class. What comes through the analyses in this section is that Latinos are not disadvantaged compared to White in accessing GCs. Yet we find some support for the place stratification model, because there is differential access to GCs. In this case, the segmentation comes through the different proportions of Latinos in metropolitan areas. The separate regression equations for owners and renters show similar effects related to percentage Latinos in urban areas: as the percentage Latinos increases so does the individual likelihood to gate. Taking this effect into consideration, we look more closely at the impact of race/ethnicity on one hand and education and class on the other.

Are Latinos of Higher Socioeconomic Standing as Equally Likely to Live in Gated Communities as Comparable Whites?

As described in the methods section, we investigate the possibility of interaction effects between race/ethnicity, and education and class based on Jaccard’s (2001) suggestion. We calculate the odds and odds ratios for Latinos and Whites in 12 different categories for each group, based on social class and education. The predicted log odds are based on Model 1 for owners and Model 2 for renters (Table 2). The predicted odds then are obtained as exponent of the predicted log odds and are reported in Table 3 together with the odds ratios. For example, the predicted odds for lower-class White renters with some college education are 0.566, which means that the probability of living in a GC for this group is a little more than half of the probability of not living in a GC. The odds ratios for all renters indicate that the effect of race/ethnicity on the propensity to gate not only is statistically insignificant but furthermore is not conditioned by class or education (the odds ratio for all renter categories is 1.146).

Table 3.

Predicted Odds and Odd Ratios for Latinos and Whites

	Lower Class				Middle Class				Upper Class
	Less than High	High School	Some College	College Degree	Less than High	High School	Some College	College Degree	Less than High	High School	Some College	College Degree
Owners
Latinos	0.082	0.101	0.151	0.166	0.089	0.109	0.163	0.180	0.110	0.135	0.201	0.222
Whites	0.103	0.126	0.188	0.207	0.092	0.113	0.169	0.186	0.108	0.132	0.197	0.217
Odds ratio	0.801	0.801	0.801	0.801	0.965	0.965	0.965	0.965	1.020	1.020	1.020	1.020
Renters
Latinos	0.665	0.653	0.649	0.652	0.669	0.656	0.652	0.655	0.918	0.901	0.896	0.899
Whites	0.581	0.570	0.566	0.569	0.584	0.573	0.569	0.572	0.801	0.786	0.782	0.785
Odds ratio	1.146	1.146	1.146	1.146	1.146	1.146	1.146	1.146	1.146	1.146	1.146	1.146

There is evidence of interaction effect between race/ethnicity and class for owners; the odds ratios for Latinos and Whites are the lowest for the lower class (0.801), slightly higher for the middle class (0.965) and highest for the upper class (1.020). Therefore, it seems that the effect of race/ethnicity on the propensity to gate for owners, to a small extent, is conditioned by class. More specifically, the odds ratios for Latinos and Whites in the middle and upper classes we would interpret as practically equal to 1; that is, there is no significant difference in the propensity to gate between Latinos and Whites. Therefore, we can conclude that middle-class Latino and White owners are equally likely to gate, as well as that the upper-class Latinos and Whites are equally likely to gate. The significant difference comes for the lower-class owners, where lower class White owners are about 1.25 times more likely to gate compared with Latinos.

If we look at the predicted odds in each cell, we notice that there is a horizontal increase for both Latino and White owners, from the first category to the last, and an increase within the lower-class, middle-class, and upper-class categories. Looking at lower-class Latino owners, we observe that the predicted odds start at 0.082 for lower-class Latino owners with less than high school education, then increase to 0.101 for lower-class Latino owners with high school diploma, then increase to 0.151 for lower-class Latino owners with some college, and again increase to 0.166 for lower-class Latino owners with college degree. Similar incremental increase in the predicted odds from lower to higher levels of education is observed within all other groups of owners: middle-class Latino owners, upper-class Latino owners, lower-class White owners, middle-class White owners, and upper-class White owners.

We can also observe that the effects of class are to some extent conditioned by education, because belonging to the middle class does not automatically mean higher propensity to gate compared to lower class; rather, while middle-class Latino owners with less than high school education exhibit slightly higher propensity to gate compared to their counterparts in the lower class, they are less likely to gate compared to lower-class Latino owners with some college and with college degree. In turn, while the upper-class Latino owners with less than high school education are more likely to gate than middle-class Latino owners with the same level of education, they are less likely to gate than middle-class Latino owners with some college and college degree. The same pattern is repeated across all levels of education for Latino and White owners. The highest propensity to gate among owners is recorded for the upper-class Latinos with college degree, 0.222, and for the same category Whites, 0.217. A close third are lower-class White owners with college degrees, with probability to gate of 0.207.

For renters, although we do observe a pattern of increase comparing the same educational level within lower, middle, and upper class, we do not observe an increase in the propensity to gate across the educational categories within each income class. For example, the propensity to gate for lower-class Latino renters with less than high school education (0.665) is higher not lower than the propensity for the lower-class Latino renters with high school diploma (0.653), and in fact, it is the highest for all educational groups in this class category. The propensity to gate for Latino and White renters is the highest for the groups with less than high school education within each income class categories; however, the differences are quite small. Therefore the more important overall pattern for renters is that the upper-class Latinos and Whites are much more likely to gate compared to the lower- and middle-class. The highest propensity to gate in the entire Table 3 is recorded for upper-class Latino renters with less than high school education, 0.918, followed by upper-class Latino renters with high school diploma, 0.901.

The study of the possibly differential effect of race/ethnicity based on education and class shows that the race/ethnicity influence on the likelihood to gate is not conditioned by education or class for renters. The effect is however conditioned by class for owners; in particular, lower-class Whites are more likely to gate compared to lower-class Latinos. There also seem to be additional segmentation within the class categories based on education for Latino owners and White owners. Within each owners’ class category, the propensity to gate increases from lower to higher levels of education. Therefore the study finds partial support for the first hypothesis associated with the spatial assimilation model: middle- and upper-class Latinos exhibit similar likelihood to gate compared to middle- and upper-class Whites but only for owners. For renters, there are no differential effects by class and education. The support we find for the second hypothesis is also complicated; Latinos of higher income class and educational level are more likely to gate than Latinos of lower income class and educational level; however, the increase in the propensity to gate from lower- to upper-class owners is conditioned by education. Among Latinos (and Whites) the members of the lower class with some college and college degrees have higher propensities to gate compared to members of the middle class with less than high school and high school diploma. Similar is the comparison between middle- and upper-class Latino owners. For Latino renters, there is really not much difference between lower and middle class; only the likelihood to gate for the upper class is clearly higher compared to the two other class groups.

Are Latinos More Likely Than Whites to Live in Gated Communities?

The last two hypotheses tested in the study are related to the ethnic community model and suggest that Latinos may be more likely to gate than Whites, and that Latinos of higher socioeconomic status may be more likely to gate than comparable Whites. These hypotheses we investigate further by calculating and plotting mean predicted probabilities separately for Latinos and Whites, and owners and renters, by income group, education, and class (Figure 1). We plotted the predicted probabilities, based on Models 1 and 2 (Table 2), along three axes of differentiation: income, class, and education.

Figure 1.

Predicted probabilities to gate, for Whites and Latinos by income, in the South and West, 2001. Predicted probabilities to gate, for Whites and Latinos by education, in the South and West, 2001

The first graph in Figure 1 shows the mean predicted probabilities by income group, where the blue line represents Latino owners, the red line represents White owners, the green line—Latino renters and the purple line—White renters. We separate Latinos and Whites into eight income groups.¹⁵ The mean predicted probabilities for each income group are higher for renters, both Latinos and Whites, compared to owners. This may be the case because rental GCs are more accessible in terms of rents than owner GCs, which include much higher premiums such as mortgages and maintenance fees. The commitment is also much more involved on the part of a gated homeowner, compared to a renter, who can easily move if not satisfied with the conditions.

For two income groups, the group with incomes from $27,000 to $39,999, and the group with incomes from $60,000 to $72,999, the graph clearly shows that Latino renters are more likely than White renters to gate, all other conditions being equal. It is difficult to contemplate why these two particular groups would differ from Whites. It may be that as Latinos are breaking into the middle class (the income group $27,000 to $39,999 is the first to denote middle class), they feel more compelled, as members of a minority population in the United States, to solidify that social status by entering into a GC. Therefore for them, it may be more a question of status rather than fear of crime. For the second group, which denotes the highest income group within the middle class (from $60,000 to $72,999), it may be more important to find protection and they buy into the myths that GCs provide better security against crime.

The second graph on Figure 1 shows the mean predicted probabilities by class for Latino owners, White owners, Latino renters, and White renters. Interestingly enough, among Latino owners, the lower class is most likely to gate, and among White owners again the lower class is most likely to gate. It seems that the mechanism at work here is the interaction between education and class. As shown on Table 3, each income class is also composed by owners and renters with different levels of education. Since education seems to be the stronger factor influencing the propensity to gate for owners (education also exerts statistically significant effect for owners, Table 2), we believe that it influences the predicted probabilities to gate by income class, as well. In this case what determines the higher predicted probabilities for lower class is the presence of householders with some college and college degree; of the lower-class Latino owners 15% have some college and 9% college degree, while among the lower-class White owners 26% have some college and 20% have college degree. Therefore, the higher probability to gate is more a function of educational degree than belonging to the lower class.¹⁶ In contrast, for Latino and White renters, the progression from lower to higher probabilities to gate follows the progression from lower to upper class.

The most consistent and clear patterns in the likelihood to gate are associated with the effects of education (last graph shown on Figure 1). First of all, as is the case on all three graphs, it is clear that overall renters have higher propensity to gate compared to owners. Second, Latino owners with less than high school education are slightly more likely to gate (0.067) compared to White owners with less than high school education (0.043). Indeed, the differences between the two groups are small, but these differences are reproduced across all categories of comparison, where Latinos are consistently more likely to gate than Whites. One possible explanation for this effect may be that while the individual effect of race/ethnicity is not statistically significant, the effect of nativity is (Table 2). Specifically, foreign born are more likely than native born to select residences in GCs. Since more than 90% of foreign born are Latinos, this effect works in combination particularly with education; as noted, in the scholarly literature it is well known that GCs are quite well spread in Latin America. Therefore, educated foreign-born Latinos would be most susceptible to cultural influence in their residential choices because living in GCs in their countries of origin is related to higher social and economic status.

The slight gating advantages of Latinos over Whites increase from lower to higher educational categories. For example, the difference in the mean propensity to gate between Latino and White renters with less than high school is 0.009; with high school diploma is 0.025; with some college it increases to 0.035 and for renters with college degrees and higher comes to 0.052. A similar pattern is repeated for owners.

Third, there is a steady incremental increase in the probability to gate from lower to higher levels of education for Latino owners, White owners, Latino renters, and White renters. The only exception is that White owners with college degree have lower probability to gate (0.075) compared to White owners with some college (0.078); still, the difference is quite negligible. Therefore, based on the analysis so far, we can argue that controlling for all relevant factors, the group most likely to gate is the group of college-educated Latino renters (0.251), followed by Latino renters with some college (0.226) and Latino renters with high school diploma (0.205). The fourth place is taken by White renters with college degree (0.199).

Thus, the findings from our analyses show partial support for the ethnic community model. The first hypothesis is not supported because overall Latinos are not more likely than Whites to gate (as shown in the regression analysis, Table 2). However, the second hypothesis finds some support, because analyzing four educational categories separately for Latino owners, White owners, Latino renters, and White renters, we find that for each educational category Latinos are more likely than Whites to choose gated residences. In addition, the differences in favor of Latinos increase as educational levels increase. While the actual differences are small, there is a tendency, which is consistent with the expectation that Latinos with higher education are more likely to gate than comparable Whites. And while the overall effect of race/ethnicity is statistically insignificant, what the above analyses show is that the effect of race/ethnicity on the propensity to gate is clearly segmented by tenure, education, and class.

Conclusion and Discussion

In this study, we aimed to understand the mechanisms of incorporation of Whites and Latinos in GCs, and based on three sociological perspectives—the place stratification model, the spatial assimilation model, and the ethnic community model—we find several mechanisms to influence the choice in favor of selecting gated residential living. The results can be generalized for the southwestern region of the United States and as GCs continue to proliferate, it will be interesting to investigate the extent to which similar mechanisms are found in the rest of the country. Beyond the statistical generalizability, we think that the conclusions presented in the study are quite significant and are not bound to the southwest. We do need to think about what our theoretical models mean and how they reflect the changing reality. Based on the results from the study, we posit a very important issue about what it is that minority groups are expected to assimilate to, and whether we should interpret such process as integration.

At the individual level, the effects of race/ethnicity is conditioned by class for homeowners; in support of the spatial assimilation model, lower-class Whites are more likely to gate compared to lower-class Latinos. Middle- and upper-class Latinos and Whites are equally likely to do so. For renters, the effects of class are not conditioned by race/ethnicity; rather, the upper class is clearly more likely to reside in GCs. We find that education is the most important sorting mechanism in the story. On one hand, it supersedes the effects of social class for owners, leading to segmentation within each class category, regardless of race/ethnicity; on the other hand, it propels Latinos with higher education to select gated residences more often than comparable Whites. Finally, at the metropolitan level, as the proportion of Latinos in the population increases, so does the overall propensity to gate.

What conclusions can we draw from our findings for the relevance of the theoretical perspectives and regarding the process of gating in the larger context of urban inequality? As with prior research investigating the segmentation of residential patterns based on race/ethnicity, we find that all three perspectives work in combination. While the access of Latinos to GCs does not seem inhibited compared to Whites, the increased proportion of Latinos, as a group, in urban areas leads to increase in the individual likelihood to gate. For Whites, this may be the case because the increase in the relative number of Latinos (about a half of which are foreign born) around them leads to heightened fears of crime, dislike of diversity, and desire to establish more homogeneous communities. For Latinos, it may mean that they have the critical mass to form their own GCs, because of cultural preferences and/or the overall context of more GCs in the area. The results certainly show that more educated Latinos are more likely to gate than comparable Whites, which supports the ethnic community model. As recent studies have shown, in contrast with traditional models of minorities’ residential disadvantage, ethnic enclaves, and slow progression toward spatial assimilation, better-off minority members opt for establishing their own higher status communities. Gated communities seem to be some of these higher status communities, chosen by some better educated Latinos more often than the same choice would be made by comparable Whites.

Spatial assimilation perspective, however, is still relevant, particularly for owners. We may also argue that the perspective should be more relevant for owners because being a renter is a more transitory status and not specifically related to the American dream. Still, in light of the current housing crisis and the findings in this study, the spatial assimilation perspective may need to be redefined. With the foreclosure crisis still looming large despite the government efforts to help homeowners and the foreclosures affecting predominantly minority populations (Gerardi and Willen 2009; Been, Ellen, and Madar 2009), the attractiveness of owning a home may suffer. As a result, what would it mean for minorities to spatially assimilate? It seems that striving to own a home in a majority-White, affluent suburb is no longer the only avenue to residential success. Instead of comparing minorities (including immigrants and their offspring) to average measures of native born, we should be looking at the diversifying avenues of residential integration, incorporating at the minimum race/ethnicity, tenure, and group residential preferences.

It seems that one of the new avenues of residential integration may be related to GCs. Could entering a rental or owner GC be considered as a measure of successful residential integration? To the extent to which, on average, GCs are more affluent compared to other neighborhoods, in the traditional arguments of the spatial assimilation model, the answer is yes. At the same time, scholars argue that GCs are places that contribute to residential segregation (Le Goix 2005; Caldeira 2000; Vesselinov 2008, 2009) and thus lead to increased urban inequality. Therefore, we may ask in what specific way entering a GC (whether rental or homeowner) contributes to the incorporation of minorities into the American dream? Overall, GCs jeopardize the American dream, because they impose a new layer of differentiation, concentrated affluence, racial homogeneity, and therefore inequality. Consequently, on the surface, the pretty equal access to GCs means that Latinos are not disadvantaged compared to Whites and that Latinos are spatially assimilating. However, a more important question looms: To what exactly are they assimilating? Should gated communities and residential segregation in general be considered a part of achieving the American dream?

What is the most striking about the results in the study, and also quite problematic, is the role played by education in the propensity to gate. Given the historical and continuous struggles in the United States related to the causes and consequences of residential segregation, one would think that the more educated people become, the more aware they would be about their personal residential choices. Instead, in this study we find that the higher levels of education are consistently associated with higher probabilities to gate. This finding also points to the possible concentration of privilege in GCs. If mostly better educated individuals select gated residences it means that they also bring variety of other resources with them.

While the findings related to education are at first startling, they are also quite reasonable given the history of GCs and the extremely low awareness of their impact on urban areas. Gated communities came about as mostly exclusive residences for the very rich and subsequently gained popularity as retirement communities. Until recently, not too many scholars paid attention to these retirement communities because they were located mostly in Florida, Arizona, and California. In the past two decades, however, the building of GCs has increased exponentially, they have attracted many other groups (besides the retired), and in many other places in the United States and the world. Therefore, not only GCs now warrantee a much stronger line of research, but the results from the present study show the need for wider publicity related to the role of GCs. A more concerted effort is needed to inform city governments, community organizations, and individual citizens about the polarizing effect of gated enclaves and their close association with residential segregation. Hopefully, this study will be the first of many to come to investigate the mechanisms of selection into GCs and to increase readers’ awareness of the possible impact and meaning of these enclaves for urban America.

Biography

Elena Vesselinov is an Assistant Professor of Sociology at Queens College and the Graduate Center, City University of New York. Her research interests include gated communities, residential segregation, the link between social and spatial inequality, the mortgage foreclosure crisis, and comparative urbanization. Recent publications include “Gated Communities and Spatial Inequality,” in the Journal of Urban Affairs, “Members Only: Gated Communities and Residential Segregation in Metropolitan U.S.,” in Sociological Forum, “Gated Communities in the United States: From Case Studies to Systematic Evidence,” in Sociology Compass.