Environmental Inequality in Tarrant County: An Analysis of Public and Private Sector Waste

Abstract: 

Previous studies analyzing the relationship between environmental hazards and the socioeconomic characteristics of communities found that inequality in the distribution of hazards among people in the communities was determined by race, ethnicity, and social class. The purpose of this research is to examine the distribution of Superfund and municipal solid waste sites within neighborhoods differing on social, demographic, and economic characteristics of the inhabitants in Tarrant County, Texas. The hypothesis that socially vulnerable communities are more likely to contain solid waste sites than those that are not vulnerable was tested using data from the 2000 U.S. Census and the Texas Commission on Environmental Quality lists for Tarrant County. Results show that none of the hypotheses were supported. Population density of a census tract was found to be the influencing factor for waste site presence.

Table of Contents: 

    Introduction

    Distributional politics appear to have prevailed such that those segments of the population with fewer political, organizational, and technical resources have borne a disproportionate share of the society’s environmental burdens (Saha & Mohai, 2005, p. 639).

    Social inequality is inherent to the infrastructure of society so that even our interactions with our environment (specifically how we contaminate it) result in patterns reflecting a discriminatory distribution of effects. This unequal distribution has compelled researchers to examine society’s interaction with the environment. There are three ways to contaminate an environment: through the air, the water, and the ground. Water and air pollution are more difficult forms of pollution to examine for environmental injustice because of their transitory and migratory nature. Ground contamination usually remains in one location. Most research focusing on ground contamination uses various types of indicators for environmental hazards.

    The purpose of this study is to determine if there are any social, demographic, or economic correlations with the communities that live around two different types of waste sites: municipal solid waste landfills and Superfund sites in Tarrant County. Rodeheaver, Williams, and Cutrer studied environmental inequality through a variety of solid waste indicators in Dallas County in 1997, but no study on Tarrant County has been identified. This study seeks to contribute to waste distribution research in the rest of the Dallas–Fort Worth metropolitan area.

    Background

    In 2000, Tarrant County had a population of 1,446,219 people living in 863.42 square miles in the year 2000. It is one of the largest urban areas in the United States in terms of population, and is one of the fastest growing urban counties in the nation, with a 12.1% change in population from 2000 to 2005. In 2000, it had a population density of 1,675 people per square mile and was one of the top four most densely populated counties of Texas (USBC, 2000). The city of Fort Worth is the county seat and the largest city in the county, with a population of 534,694 in 2000. The other major city in Tarrant County is Arlington with a population of 332,969 (CIP, 2000). The median income for the county was $46,179. The population was 62% white (non-Hispanic) and 38% nonwhite in 2000 (USBC, 2000). Tarrant County provides an adequate area to study for environmental inequality research because it is a well-established area and has had a long history of development similar to the other areas in Texas that have been studied.

    Past Research

    Research on environmental inequality began in the 1960s as a joint product of the Civil Rights movement and the growing public interest in the health of the environment. Student environmental interest and activism was growing, and after Earth Day of 1970, started to concern itself with the health of the urban poor. The movements had started to grow together and over the decades evolved into a synthesized interest in environmental justice (Bullard, 1990). This environmental justice movement was spawned by the principle of equal rights as propagated by past social justice movements and by the emergence of environmental quality as a basic right (Bullard, 1990; Rodeheaver et al., 1997). A number of different issues came to the public’s awareness in the early 1970s, such as waste disposal, air and water pollution, population control, and natural resource protection. The “energy crisis” in the 1970s also challenged people to face up to environmental problems. In addition, poverty, both in the United States and around the world, also emerged as a significant social issue (Bullard, 1990).

    At the same time, Congress was passing legislation such as the Solid Waste Disposal Act of 1965 and the Resource Recovery Act of 1970 that shaped waste disposal practices. In a study of the historical context of hazardous waste facility siting, Saha and Mohai (2005) argue that these federal legislative acts were passed to encourage states and municipalities to take control of dumping by allocating sanitary landfills. This helped to curb a portion of unregulated dumping, but hazardous waste was still uncontrolled until the Resource Conservation and Recovery Act (RCRA) of 1976 and the Hazardous and Solid Waste Amendments (HSWA) of 1984. Saha and Mohai used Detroit as an example to show that strict regulations were lacking before this legislation when local governments were in charge of siting. What regulations did exist were weakly enforced. They argue that this was because society at the time was not aware of the true nature of hazardous waste, and that there was no infrastructure for information dissemination or for public education. By the time public awareness began to grow, the federal government passed the RCRA and HSWA, which moved the responsibility for waste regulation from local governments, where people would be able to participate at the community level, to the state and national levels, weakening “the political opportunity structure for public participation in siting decisions” (Saha & Mohai, 2005, p. 624).

    The public apprehension about waste facility locations in the 1970s led to community action organization (Saha & Mohai, 2005). Although there was concern about the environmental health of urban areas at this time, and subsequent community organizing to address the health issues, it did not trickle down to the poor or minorities right away. They did not yet have an effective voice in the political arena. With the nation undergoing an environmental revolution in terms of the public awareness of common polluting practices, and the media’s highlighting of big cases such as Love Canal in New York, Three-Mile Island, and Warren County in North Carolina, environmental concern grew rapidly. Those for whom these issues were of growing importance were, however, for the most part, middle class and white. Their attention was not on equitable distribution of locally unwanted land uses (LULUs). Low-income and minority people were also not well-represented in governmental bodies. All of this “majority” concern resulted in the minorities and the poor cohabitating with society’s negative environmental externalities. They made up the commonly referred-to path of least resistance for waste storage (Bullard, 1990; Saha & Mohai, 2005). These circumstances created the Not-In-My-Back-Yard (NIMBY) syndrome. Locally unwanted land uses have a history of being relegated to the socially vulnerable communities and this phenomenon further increased the disparity of facility siting. Bullard maintains that “public officials and private industry have in many cases responded to the NIMBY phenomenon using the place-in-blacks’-back-yard (PIBBY) principle” (1990, p. 4).

    Previous to this era of quasi-public watchdogism, facility siting was not as unequally distributed. That may be due to the fact that prior to the 1960s, there was no heavy-handed governmental regulation of waste. In 1979, following his visit to the Love Canal site, President Carter passed the “Superfund Act” (Comprehensive Environmental Remediation, Compensation, and Liability Act, or CERCLA) which listed uncontrolled sites where toxic waste had been dumped and abandoned. The most hazardous sites on this list were determined by the EPA to pose a substantial health threat to nearby residents. They were put on the National Priority List (NPL) to be funded in the future for cleaning and are known as Superfund sites (Saha & Mohai, 2005).

    A study conducted in Florida analyzed the characteristics of communities around 53 superfund sites (Stretesky & Hogan, 1998). Blacks, Hispanics, the poor, and the unemployed living in densely populated areas were more likely to live near a Superfund site. The researchers found that race was a stronger predictor than economic factors of living near a Superfund site and argued that this was a manifestation of indirect institutional discrimination This is indicative of the lack of regulation in the past. Laws regarding waste were usually left up to local governments, which were not stringent in their legislation or enforcement. Superfund sites are a manifestation of this unregulated waste accumulating over time. Such was the case in Chicago, a city with a long history of illegal dumping. Pellow (2004) conducted a qualitative study of cases of illegal solid waste dumping in the city, starting in the early twentieth century. In the 1910s the city was experiencing a fight over the dumping of waste in African American and European immigrant neighborhoods. Officials of the city of Chicago and private waste haulers were working together in open dumping. This alliance between the private and public sectors of the city seems to have been a constant even up into the nineties. Waste was still being dumped illegally into poor and minority neighborhoods, despite community organization and public outcry.

    Starting in the late sixties, as the nation started to examine these practices and to regulate them, the people who successfully protested the sites being located in their communities inadvertently pushed the sites into communities where the residents were least able to fight against the facility. Some sites were already slated for poor and minority areas of towns as a practice of discriminatory zoning. This was the case with Houston municipal landfills. In Bullard’s book (1990), landfills were found to be disproportionately located in black and poor communities. Nearly one-fourth of the city’s population was black, yet more than three-fourths of the solid waste facilities sat in these neighborhoods. Twelve out of thirteen of the city’s landfills and incinerators were located in lower-income areas.

    This social history of environmental conflict provides the context in which this study operates. The history of the distribution of environmental hazards is key to constructing a lens in which to view the research of this field. Knowing this historical context allows one not only to understand theory but to theorize as well, as to what factors had an influence in building this phenomenon that is now being studied.

    Research Methods

    Environmental inequality has been analyzed by looking at locations of specific environmental hazards in relation to characteristics of communities around them. Researchers use different variables that are observational measures of environmental equity, which specifically look at “the distribution of environmental resources and contaminants according to race and ethnicity, economic class, and other traditional sociological variables” (Rodeheaver & Cutrer, 1995, p. 8). The unit of analysis for which data are gathered changes from study to study. Towns, zip code areas, census tracts, and census block groups are all common units of analysis for the variables. Census tracts seem to be the most popular choice in the research because they are based on relative population homogeneity (Stretesky & Hogan, 1998). Environmental inequality studies vary widely in the proxies they use to measure each variable, whether it is percent nonwhite, percent black, or percent Hispanic for race/ethnicity or median household income, poverty status, or housing value for economic status. There are also a number of other factors at play, such as age, sex, educational attainment, and political voice, although race/ethnicity or economic status usually have the strongest correlation with measures of environmental inequality.

    The term environmental racism grew out of early studies of this type. Early environmental inequality studies frequently found environmental hazards located in minority communities. The term environmental racism was first used by the United Church of Christ Commission on Racial Justice in 1982, in a campaign to help the residents of the small, mostly black town of Afton in Warren County, North Carolina, receive aid in cleaning up a toxic waste site (Rodeheaver & Cutrer, 1995). Afton was designated as a state dumping site “for more than 32,000 cubic yards of soil contaminated with highly toxic PCBs (polychlorinated biphenyls), which had been illegally dumped along the roadways in fourteen North Carolina counties in 1978” (Bullard, 1990, p. 30). Nationwide protests about the situation spurred the U.S. General Accounting Office in 1983 to investigate community characteristics of hazardous waste landfill sitings in the South. Three of the four major hazardous waste landfills of the South were located in black majority communities, where more than one-fourth of all four communities had incomes below the poverty level (Bullard, 1990; Rodeheaver & Cutrer, 1995).

    In these cases, race had a stronger correlation with waste site presence than economic measurements. In research on environmental inequality, especially waste distribution, much attention is given to determining the better predictor for site presence: race/ethnicity or economic status (environmental racism or environmental classism). Numerous studies find that either one or the other is a stronger predictor of site presence. A study done in Dallas County, Texas, used census tracts and toxic waste, solid waste, Superfund sites, and the EPA’s Toxic Release Inventory (TRI) data to determine if race and ethnic minority status is a more important factor than economic class. The study found that median house values as a proxy for economic class was the significant variable (Rodeheaver et al., 1997). In another study looking at the distribution of Superfund sites in the metropolitan areas of Texas and Louisiana, race/ethnicity was not found to be a significant predictor. Proxies measuring class, status, and power were found to be better predictors of site placement (Denq, Constance, & Joung, 2000). Another article examining the “race versus class” debate found that different methodologies yield different results. The article looked first at possible problems with different indicators of environmental hazards. TRI data is self-reported and thus risks being underreported. It also might not be relevant to a de-industrialized area that is being studied, such as an urban area where industrial sites are now abandoned. Abandoned wastes sites, like Superfunds, would be a better indicator for this kind of study. However, a problem with using the NPL list is that not all sites that need to be on the list are on it because they have either not been detected or a community has not had the ability to successfully organize and draw attention to the problem. Another popularly operationalized variable is Treatment, Storage, and Disposal Facilities (TSDFs). These facilities are often found in industrial parks where the surrounding population is small. TSDFs may also be inappropriate for measuring environmental inequality in a de-industrialized area (Krieg, 1998).

    Taking these issues into account, Krieg proposes studying multiple indicators to increase the accuracy of the studies’ results. Therefore, he operationalizes Superfund sites and TRI data as indicators for environmental hazards while looking at their correlation with race/ethnicity and median household income in the Boston area. Median household income was found to be negatively associated with the TRI data regardless of the area. Race had a stronger association with Superfund sites than TRI data inside Boston, and inversely, race had a stronger association with TRI data than Superfund sites in suburban areas. This is an important finding, as it illustrates that accurate findings hinge upon the relevancy of environmental hazard indicator in the study. For example, in this study, TRI data were not appropriate for the de-industrialized city where minorities live around old, abandoned waste sites (Krieg, 1998).

    The accuracy of findings is also dependent on a broader and more fundamental methodological consideration: the definitions employed for environmental inequality that shape what a study is looking for. In a 2000 study exploring environmental inequality in metropolitan areas across the United States, Downey used different definitions of environmental inequality to illustrate that the definitions employed for it affect the conclusions that can be drawn from a study’s results. According to Downey (2006), there are two broad categories of definitions: “those that focus on racially discriminatory intent and those that focus on inequitable environmental outcomes” (p. 24). He uses the definitions of disparate social impacts and relative distribution under the inequitable environmental outcomes category. Disparate social impacts inequality is defined as “when members of a specific social group are more likely to live in environmentally hazardous neighborhoods than we would expect if group members were randomly distributed across residential space” (Downey, p. 25). Relative distributions of burdens versus benefits is defined as “when those who receive greater benefits than others from the capitalist production and distribution process do not bear a greater share of the burdens of this process than do others” (Downey, p. 26). Thus, the study looks at whites, and the middle and upper classes versus minorities and the poor and working classes to see who is more burdened by industrial pollution. The disparate social impacts method of looking for environmental inequality in a given area is more frequently employed than the relative distributions of burdens. By using the Environmental Protection Agency’s 2000 Toxic Release Inventory with 2000 U.S. Census data at the census tract level across 14 major metropolitan areas, the study found that the definitions employed to measure environmental inequality in different ways did have a significant effect on whether or not it was determined in an area.

    These are some important arguments about different methodological concerns that researchers need to address in accurately studying environmental inequality. Understanding how this kind of research developed and what key concepts and factors researchers employ as a means to evaluate the circumstances of environmental equity is important for understanding this study.

    Hypotheses

    Before analysis of the distribution of waste sites in Tarrant County, the hypotheses for this study were based on a review of the literature and on looking for some predictors that the past studies had missed.

    • Hypothesis 1: Census tracts with a majority of nonwhite-headed households will be more likely to contain a waste site than Census tracts with a majority of white-headed households.
    • Hypothesis 2: Census tracts with a majority of female-headed households will be more likely to contain a waste site than Census tracts with a majority of male-headed households.
    • Hypothesis 3: Census tracts with a majority of household heads over age 65 will be more likely to contain a waste site than Census tracts with a majority of households under age 65.
    • Hypothesis 4: Census tracts with a majority of housing units being rented will be more likely to contain a waste site than Census tracts with a majority of housing units not being rented.
    • Hypothesis 5: Census tracts with a waste site will have lower median incomes than Census tracts without a waste site.
    • Hypothesis 6: Census tracts where the majority of the population has less than a high school degree will be more likely to contain a waste site than Census tracts where the majority of the population does not have a high school education.
    • Hypothesis 7: Census tracts with the majority of the population having less than a bachelor’s degree will be more likely to contain a waste site than Census tracts where the majority of the population has a bachelor’s degree.
    • Hypothesis 8: Census tracts with a majority of nonwhite-headed households will be more likely to be adjacent to a tract containing a waste site than Census tracts that have a majority of white-headed households.
    • Hypothesis 9: Census tracts with a majority of female-headed households will be more likely to be adjacent to a tract containing a waste site than Census tracts with a majority of male-headed households.
    • Hypothesis 10: Census tracts with a majority of household heads over age 65 will be more likely to be adjacent to a tract containing a waste site than Census tracts with a majority of households under age 65.
    • Hypothesis 11: Census tracts with a majority of housing units being rented will be more likely to be adjacent to a tract containing a waste site than Census tracts where the majority was not renters.
    • Hypothesis 12: Census tracts with the majority of the population having less than a high school degree education will be more likely to be adjacent to a tract containing a waste site than Census tracts where the majority of the population has at least a high school education.
    • Hypothesis 13: Census tracts with the majority of the population having less than a college degree education will be more likely to be adjacent to a tract containing a waste site than Census tracts where the majority of the population has a college degree.

    Methods and Data

    This study examined the distribution of two different types of waste sites in Tarrant County: municipal solid waste sites (landfills) and Superfund sites. As was discussed earlier, there are many different indicators researchers may use to measure environmental inequality, and some researchers use more than one (Krieg, 1998; Rodeheaver et al., 1997). Krieg makes the argument that different sites have different community characteristics, and thus, using more than one as an indicator yields more accurate results. This study examined municipal solid waste sites and landfills to see how waste is distributed across communities. In addition, this study included Superfund sites as a second indicator. Their nature and point of origin is much different than that of landfills; they are the result of private sector activities. Landfills are not necessarily toxic, but the Superfund sites are severely toxic. The goal of this research, however, was not to examine the exposure of communities to toxicity, but to examine a full spectrum of locally unwanted land use distributions. Landfills, regardless of their toxicity, are a hazard around which most people do not wish to live, which is why most landfill developers typically follow the path of least resistance.

    The municipal solid waste site data was provided by the Texas Commission on Environmental Quality through a list of all open, closed, active, and inactive landfills in Texas. The Superfund sites were found through the Texas Commission on Environmental Quality as well. There are 20 sites in total. Five are Superfund sites, and 15 are municipal solid waste landfills. To analyze the distribution of these sites, population data were gathered from the 2000 U.S. Census at the census tract level. Rodeheaver et al. used this unit of analysis in their 1997 study of Dallas County. Both Rodeheaver’s (1997) and Stretesky’s (1998) studies reasoned that despite past research relying on zip code areas as a unit of analysis, they prefer census tract level because zip code boundaries ignore major demographic changes in communities. According to Rodeheaver et al. (1997), “the census tract as a unit of analysis (of aggregated data) is…probably the closest unit of analysis that appears as a ‘natural’ area (or neighborhood)” (p. 11). Census tracts are a popular unit, as can be seen in several past studies of environmental inequality (Denq et al., 2000; Downey et al., 2005; Downey, 2006; Saha & Mohai, 2005; Stretesky & Hogan, 1998; Rodeheaver et al., 1997). A problem with using census tracts that researchers (Krieg, 1998; Stretesky & Hogan, 1998) acknowledge is the variation in size between urban and non-urban census tracts. This is the case especially in Tarrant County; the non-urban census tracts’ physical boundaries are much larger than the urban tracts’. Despite the problems with the census tract level data, for the purpose of this research census tracts were deemed the most appropriate geographic unit.

    The dependent variable was whether or not a census tract contained a Superfund or municipal waste site. As stated earlier, this study does not concern itself with a community’s exposure to toxicity. Instead the focus is on the presence of a locally unwanted land use. Tracts with sites present were coded as 1, and those without were coded as 0. The list of all the municipal solid waste landfills in Tarrant County included the physical addresses and coordinates of the sites listed as well. These location descriptions were used in conjunction with physical site visits, the program Google Earth™, and Tarrant County census tract maps acquired via the U.S. Census website to determine their tract locations. The Superfund sites had similar information as well, and the same methods were used for determining their tract locations. There are 310 census tracts in Tarrant County. Seventeen of those had at least one site present. Only two tracts had more than one site in them. One of these tracts had four sites in it, and another tract contained two sites. These tracts were still coded as 1 because, for the purpose of this study, the interest is only in presence, not quantity. One site fell on the border of two tracts, so both were included. A second group was made up of tracts that included the 17 tracts with at least one site present and tracts immediately adjacent to those. This is illustrated in Figure 1.

    The adjacent tracts were identified by using Geographic Information Systems to indicate all the tracts with sites and then to designate those tracts that shared a border as adjacent. Tracts that simply met at a corner were not deemed adjacent. The total number of tracts in this group was 85. Stretesky and Hogan (1998) used this grouping method in their study on Superfund sites in Florida. They found that some variables were more statistically significant with the second group.

    The data for the independent variables came from the 2000 U.S. Census, in the Summary 3 File, acquired through the Texas State Data Center website. The researchers gathered Census tract data on race and ethnicity of the household head, female-headed households, household heads age 65 and older, housing units rented, people with less than a high school degree and people with less than a college bachelor’s degree, and median income. The researchers created nominal dummy-coded variables for all the data except median income, using a threshold of 50%. For race and ethnicity the researchers coded percent nonwhite-headed households (including Hispanic ethnicity) as more than 50% of a tract a 1 and less than 50% a 0. The same was done for percent of female-headed households, percent of household heads age 65 and over, percent of renters, percent of people with less than a high school degree education, and percent of people with less than a college bachelor’s degree education in a tract. These variables, including median income, serve to represent traditionally socially vulnerable groups of the population.

    The statistical analysis started with running frequencies on all of the variables. The nominal dummy-coded independent variables were run through cross-tabulations and chi-square tests, comparing characteristics of tracts with a site versus those without sites. The second grouping of 85 adjacent tracts was also run through cross-tabulations and chi-square tests. Variables measured at an interval-ratio level were run through t-tests exploring differences in means for both groupings of tracts with sites versus those without sites.

    Data Analysis and Results

    First, frequencies for all the variables are presented in Table 1. Next, bivariate analyses were employed in determining results. The dummy nominal variables, nonwhite-headed households, female-headed households, household heads of age 65 and over, rented housing units, less-than-high-school degree education and less-than-college degree education were cross-tabulated with site presence in contingency tables with chi-square as the test of significance. The results of these tests can be viewed in Table 2 and Table 3.

    In examining the chi-square statistics, none of the variables were found to be significant in predicting site presence. This was true for both grouping of tracts with sites and tracts with sites plus adjacent tracts. In fact, in only two instances was the relationship between the predictor variable and the presence of a site even in the right direction. These two instances will be noted in the discussion of the hypotheses in the next section.

    Table 2Figure 2, and Figure 3 show how the two most significant factors in the literature, race/ethnicity and economic status, interact with waste site presence.

    Not one of the variables was significant enough to reject the null hypotheses. Thus, none of the hypotheses were supported in Tarrant County, Texas, as I will discuss next.

    Hypothesis 1 stated that Census tracts with a majority of nonwhite-headed households will be more likely to contain a waste site than Census tracts with a majority of white-headed households. Refer to Table 2 and Figure 2. Table 2 shows that 4.1% of the tracts with a majority of nonwhite-headed households contained a waste site, while 6.7% of the tracts without a majority of nonwhite-headed households had a site. Thus the relationship was not in the predicted direction and was not statistically significant (chi-square = .316, p > .05).

    Hypothesis 2 states that Census tracts with a majority of female-headed households will be more likely to contain a waste site than Census tracts with a majority of male-headed households. Refer to Table 2. Table 2 shows that none of the tracts where female-headed households were in the majority had a waste site compared to 5.8% of the tracts where female-headed households were not in the majority. This result was not in the predicted direction and was not statistically significant (chi-square = .339, p > .05).

    Hypothesis 3 states that Census tracts with a majority of household heads over age 65 will be more likely to contain a waste site than Census tracts with a majority of households under age 65. Refer to Table 2. Table 2 shows that none of the tracts where the majority of the household heads were older than age 65 had a waste site compared to 5.5% of tracts where the majority of household heads were not older than age 65. This result was not in the predicted direction and is not statistically significant (chi-square = .809, p > .05).

    Hypothesis 4 states that Census tracts with a majority of housing units being rented will be more likely to contain a waste site than Census tracts with a majority of housing units not being rented. Refer to Table 2. Table 2 shows that 7.7% of the tracts where the majority of the houses are rented had a waste site compared to only 3.9% of the tracts where the majority of the houses are not rented. Although these results were in the predicted direction, the results were not statistically significant (chi-square = .147, p > .05).

    Hypothesis 5 states that Census tracts with lower median incomes will be more likely to contain a waste site than Census tracts with higher median incomes. Refer to Table 4 and Figure 3. The average median income for tracts without a site was $48,179, and for tracts with a site, $42,024. The t test for the median income for the differences of the means resulted in a nonsignificant finding, t(df = 308) = 1.169, p > .05.

    Hypothesis 6 states that Census tracts where the majority of the population has less than a high school degree will be more likely to contain a waste site than Census tracts where the majority of the population does not have a high school education. Refer to Table 2. Table 2 shows that 4.5% of the tracts where the majority of the population does not have a high school education have a waste site compared to 5.8% of the tracts where the majority does have a high school degree. These results were not in the predicted direction and were not statistically significant (chi-square = .683, p > .05).

    Hypothesis 7 states that Census tracts with the majority of the population having less than a bachelor’s degree will be more likely to contain a waste site than Census tracts where the majority of the population has a bachelor’s degree. Refer to Table 2. Table 2 shows that 3.8% of the tracts where the majority of the population has less than a bachelor’s degree compared to 6.1% of the tracts where the majority does not have a bachelor’s degree. These results were not in the predicted direction and were not statistically significant (chi-square = .446, p > .05).

    Hypothesis 8 states that Census tracts with a majority of nonwhite-headed households will be more likely to be adjacent to a tract containing a waste site than Census tracts that have a majority of white-headed households. Refer to Table 3. Table 3 shows that 26.0% of tracts where the majority of the household heads were nonwhite were adjacent to a tract with a waste site compared to 28.7% when the majority of the household heads were white. These results were not in the predicted direction and were not statistically significant (chi-square = .604, p > .05).

    Hypothesis 9 predicts that Census tracts with a majority of female-headed households will be more likely to be adjacent to a tract containing a waste site than Census tracts with a majority of male-headed households. Refer to Table 3. Table 3 shows that 33.3% of the tracts with a majority of female-headed households had a waste site in an adjacent tract compared to 27.1% of the tracts where a majority of household heads were not female. These results were in the predicted direction but were not statistically significant (chi-square = .599, p > .05).

    Hypothesis 10 states that Census tracts with a majority of household heads over age 65 will be more likely to be adjacent to a tract containing a waste site than Census tracts with a majority of households under age 65. Refer to Table 3. Table 3 shows that none of the tracts where the majority of the heads of household were age 65 or older had a waste site in an adjacent tract compared to 27.5% of the tracts where the majority of the heads of households were not age 65 or older. These results were not in the predicted direction and were not statistically significant (chi-square = .538, p > .05).

    Hypothesis 11 states that Census tracts with a majority of housing units being rented will be more likely to be adjacent to a tract containing a waste site than Census tracts where the majority was not renters. Refer to Table 3Table 3 shows that 27.7% of the tracts where the majority of the houses were rented were adjacent to tracts with a waste site compared to 41.1% of the sites where the majority of the houses were not rented. These results were not in the predicted direction and were not statistically significant (chi-square = .927, p > .05).

    Hypothesis 12 states that Census tracts with the majority of the population having less than a high school degree education will be more likely to be adjacent to a tract containing a waste site than Census tracts where the majority of the population has at least a high school education. Refer to Table 3Table 3 shows that 23.9% of the tracts where the majority of the population does not have a high school degree were adjacent to a tract with a waste site compared to 28.4% of the tracts where the majority of the population did have a high school degree. These results were not in the predicted direction and were not statistically significant (chi-square = .463, p > .05).

    Hypothesis 13 states that Census tracts with the majority of the population having less than a college degree education will be more likely to be adjacent to a tract containing a waste site than Census tracts where the majority of the population has a college degree. Refer to Table 3Table 3 shows that 22.8% of the tracts where the majority of the population does not have a bachelor’s degree were adjacent to a tract with a waste site compared to 29.0% of the tracts where the majority of the population did have a bachelor’s degree. These results were not in the predicted direction and were not statistically significant (chi-square = .285, p > .05).

    These were surprising results, as the literature researched for this study has found that solid waste sites are not distributed equitably. Granted, the indicators that researchers use to measure environmental inequality differ, along with the unit of analysis, but both landfills and Superfund sites have been found to be inequitably distributed through the population, especially in the South (Bullard, 1990; Denq et al., 2000; Rodeheaver et al., 1997; Stretesky & Hogan, 1998). The use of census tracts as the unit of analysis has been a common approach in this type of study. In their study of toxic waste sites in Dallas County using census tracts, Rodeheaver et al. (1997) found that tracts with higher median home values were less likely to have toxic waste sites located in the tract or nearby, yet those tracts that were predominantly minority were more likely to have these sites. Stretesky and Hogan’s (1998) study of Superfund sites in Florida also showed that predominantly minority and lower income census tracts were much more likely to have toxic waste sites. Studies employing these methods have focused on areas as close to Tarrant County as the city of Houston, Dallas County, and the Arlington/Fort Worth area (Bullard, 1990; Denq et al., 2000; Rodeheaver et al., 1997; Stretesky & Hogan, 1998).

    To determine the cause of these negative results, the researchers examined a third variable. Knowing that Census tracts do differ in size considerably throughout Tarrant County and in population densities, a post hoc hypothesis was developed that population density might be affecting the results. A t test on Census tract population density was analyzed by using data pulled from the same census dataset. The average population density in the 293 tracts that did not contain a site was 3,507 people per square mile, versus 808 people per square mile for the 17 tracts that contained a site. The t test of the difference of the means resulted in a significant finding, t(df = 28.142) = 11.295, p < .001. Refer to Table 5 and Figure 4.

    Discussion

    The result for the population density is a very reasonable explanation for the results of the original analysis. The larger the population density, the less likely a census tract was to have a site. Even without conducting a social historical analysis of the development of these sites and their surrounding communities, it is clear that Tarrant County has developed quite differently than did Dallas or Harris counties. Tarrant County has many large, rural, and/or undeveloped census tracts near the outer edges of the county. This is basically illustrated in all the figures, but Figures 1 and 4 are more important in terms of visualizing development. Figure 1 does this via the road networks. In contrast, Dallas County is nearly completely developed. According to the 2000 U.S. Census, the population density of Dallas County was 2,522 people per square mile, whereas the population density of Tarrant County was 1,675 people per square mile. Also notable is the fact that the land areas of the two counties are very similar. Tarrant’s land area is 863.42 square miles while Dallas’ is 879.6 square miles, although their population densities vary highly.

    Tarrant’s low population density is magnified in its large Census tracts. As Stretesky and Hogan (1998) pointed out, the main problem with using census tracts as the unit of analysis is the considerable difference in size between the urban and non-urban tracts. In 2000, these large tracts were relatively undeveloped, as is illustrated in Figure 1. Almost all the green and yellow coded tracts are devoid of the massive road networks that are indicative of population. Anderton, Oakes, and Egan (1997) argue that the boundaries of large, non-urban census tracts obscure community differences rather than reflecting them. Census tracts work well when an area is populated because they follow community lines, but their relevancy in a model falls away when there is not enough population.

    The intersection of these problems with low population density tracts was the key factor in this study. As the figures illustrate, it seems almost all the sites ended up being located away from the population. Due to the size of the county and the concentration of the population living in the more central areas, it seems that the cities and businesses could afford to locate waste outside of areas where people were living. As Tarrant County grows, however, the researchers anticipate that these tracts with sites and their adjacent tracts will reflect the population disparity trends found in past studies as more socially vulnerable people will move into these areas because they cannot afford to live in other areas.

    After understanding this finding, it seems that there might exist other and perhaps better ways of analyzing the distribution of these sites within the population. Tarrant County’s boundary lines are just a square designed to contain a group of areas. City boundaries are a more natural phenomena in describing an area. Looking at tracts within cities might be more accurate in representing the distribution of sites, but there may also be a more accurate unit of analysis than the census tract. This may not entirely be the case, as the nature of the size of sites in this study would possibly affect other units of analysis. These waste sites typically take up a considerable amount of space and may have buffer areas as well, contributing to the large size. It is understandable that they might be located in larger census tracts and in areas that might not be immediately surrounded by residential zoning, regardless of the unit of analysis used.

    Two other less substantial explanations for these results pertain to methodological considerations. One possible contributing factor is the number of sites that this study used, which is also tied to the type of sites this study used. There were only 20 sites distributed within 17 census tracts out of a total of 310 Census tracts in Tarrant County. This might be indicative of a methodological error. Through replicating methods used in past studies, Tarrant County was discovered to have unique features that became problematic in the analysis. The chi-square statistic is sensitive to small numbers. In the case of this study, the chi-squares were showing very small expected frequencies. To remedy this problem, other indicators of environmental inequality for solid waste, such as Treatment, Storage, and Disposal Facilities, have sometimes been included in other communities, which could possibly result in more data (Krieg, 1998).

    In addition to addressing the amount of data, another possible problem with this study is how variables were constructed. This study dummy-coded all the independent variables, except median income. In constructing these dichotomous variables, “majority tracts” coded as 1 were based on a population amount over 50% of the tract’s population. This method could have possibly overlooked important differences in the population distribution relative to Tarrant County. In a study done on Superfund sites in Texas and Louisiana metropolitan areas, Denq et al. (2000) introduced the idea of relative deprivation as a way to account for the significance of demographic differences in tracts that do not have absolute group majorities. A way to employ this idea in this study would be to define majority as a tract’s population being over the average population for that group in all of Tarrant County. This method would then illustrate the relative disparities in distribution within Tarrant. Also possibly overlooked due to dichotomous variable construction are the detailed interactions of race and ethnicity throughout the tracts. The literature has shown that race/ethnicity is one of the strongest, if not the strongest, predictor for environmental hazard presence. This study combined race and ethnicity in its operationalization of the measure. To obtain a more detailed observation of this variable’s influence, separating blacks from Hispanics, and both from whites may yield a more accurate way of looking at what is happening with race and ethnicity in regards to distribution.

    Conclusion

    The notion of environmental justice comes from a disproportionate incidence of society’s waste burdening vulnerable social groups (Pellow, 2004). It was not until the late sixties and early seventies that, with the help of highly publicized cases of environmental contamination, the public started to develop awareness of the situation and the government began to get involved. As the government took a more active role in the regulation of waste, it was left to the public to organize and represent their concern about regulation. This shifting of power from local to state or federal governments allowed the opportunity for social inequality to arise and caused disparate waste regulation, site development, and site allocation (Saha & Mohai, 2005).

    In this study, the researchers examined a small segment of this research and applied it to an area that has not been studied before, Tarrant County. Areas near Tarrant County, such as Dallas County and Houston, have been studied in the past to examine environmental inequality. These areas were found to have various forms of environmental racism or environmental classism (Bullard, 1990; Rodeheaver et al., 1997). Instead of only trying to determine if either race and ethnicity or class are the main factors in predicting site presence or not, a list of sociological variables that traditionally reflect the socially vulnerable were studied. In finding that none of the hypotheses were supported and that population density was the influencing factor, this study can offer up a series of conclusions that add to the body of research on environmental inequality in terms of solid waste.

    First, it is important to understand how an area has developed in terms of zoning and population location when considering methods for measuring environmental hazard distribution. The results of this study do not explicitly support that environmental inequality is absent from Tarrant County, but rather that the measures used might not be the most relevant for studying it in this area. As Krieg’s (1998) research showed, different types of indicators for environmental inequality are more relevant to certain areas, depending on the social history of their development.

    This is also true when evaluating units of analysis. Census tracts are a proven acceptable means of looking at population distribution as they tend to adhere to community characteristics. This fails, however, when an area has large differences in urban and non-urban tract size as well as the occurrence of large differences in tract population densities. This problem may be due to the boundaries of the chosen study area. If the boundaries, like those of Tarrant County, do not reflect a natural progression of community distribution and development, analysis can be thrown off by taking into consideration outlying areas that are negligible to the study’s model.

    Secondly, variable construction needs to consider the nature of how the demographic and social aspects of a population are distributed, especially relative to that area. Differences can be highlighted or ignored in the analysis of variables that combine too much of the population or ignore the relative distribution of groups throughout the area. If the goal is to understand how different groups of society are affected by waste, then constructing variables that stay the most true to real-world circumstances are in the study’s best interest.

    The data of this study did not allow for a precise assessment. Future research should address these methodological points in the search to conduct a more accurate study. Researchers may need to think about dropping relatively insignificant non-urban tracts or looking at city boundaries instead of county lines as well as operating off of Krieg’s (1998) findings by varying the indicators for environmental hazards, while making sure the indicators are the most relevant to the area.

    References

    • Anderton, D., Oakes, J., & Egan, K. (1997). Environmental equity in Superfund: Demographics of the discovery and prioritization of abandoned toxic sites. Evaluation Review, 21, 3–26.
    • Bullard, R. D. (1990). Environmentalism and social justice. In Dumping in Dixie. Boulder: Westview Press, Inc.
    • CIP (County Information Project). (2000). Tarrant county information profile. In Texas Association of Counties: Online Resources. Retrieved March 8, 2007.
    • Denq, F., Constance, D. H., & Joung, S. (2000). The role of class, status, and power in the distribution of toxic Superfund sites in Texas and Louisiana. Journal of Poverty, 4, 81–100.
    • Downey, L. (2006). Environmental inequality in metropolitan America in 2000. Sociological Spectrum, 26, 21–41.
    • Downey, L., & Willigen, M. V. (2005). Environmental Stressors: The mental health impacts of living near industrial activity. Journal of Health and Social Behavior, 46, 289–305.
    • Krieg, E. J. (1998). Methodological considerations in the study of toxic waste hazards. Social Science Journal, 35, 191–202.
    • Krieg, E. J. (1998). The two faces of toxic waste: Trends in the spread of environmental hazards. Sociological Forum, 13, 3–20.
    • Pellow, D. N. (2004). The politics of illegal dumping: An environmental justice framework. Qualitative Sociology, 27, 511–525.
    • Rodeheaver, D. G., & Cutrer, J. G. (1995). Environmental equity and the case of West Dallas. An Aging Population, An Aging Planet, and a Sustainable Future, 1, 203–223.
    • Rodeheaver, D. G., Williams, J. L, & Cutrer, J. G. (1997). Environmental equity in Dallas County, Texas, neighborhoods and communities. Sustainable Communities Review, 1, 7–13.
    • Saha, R., & Mohai, P. (2005). Historical context and hazardous waste facility siting: Understanding temporal patterns in Michigan. Social Problems, 52, 618–648.
    • Stretesky, P., & Hogan, M. J. (1998). Environmental justice: An analysis of Superfund sites in Florida. Social Problems, 45, 268–287.
    • TCEQ (Texas Commission on Environmental Quality). (2005). Superfund sites in Tarrant county. In Texas Commission on Environment Quality. Retrieved March 8, 2007, from http://www.tceq.state.tx.us/remediation/superfund/sites/county/tarrant.html.
    • TCEQ (Texas Commission on Environmental Quality). Texas MSW sites. In Texas Commission on Environment Quality. Retrieved March 8, 2007.
    • USBC (U.S. Bureau of the Census). (2000). Geographic comparison table for the counties of Texas. Retrieved March 8, 2007.
    • USBC (U.S. Bureau of the Census). (2000). State and county quick facts: Tarrant county. In U.S. Census Bureau. Retrieved March 8, 2007.
    • USBC (U.S. Bureau of the Census). (2000). American fact finder: Census 2000 summary file 3 detailed tables. Retrieved March 8, 2007,.

    Table 1: Frequencies for All Variable

        Frequencies (in tracts) Percent Mean
    Sites 0
    1
    293
    17
    94.50%
    5.50%
    -
    -
    Sites + adjacent sites 0
    1
    225
    85
    72.60%
    27.40%
    -
    -
    Female household head 0
    1
    295
    15
    95.20%
    4.80%
    -
    -
    65+ yrs of age household head 0
    1
    309
    1
    99.70%
    0.30%
    -
    -
    Nonwhite household head 0
    1
    164
    146
    52.90%
    47.10%
    -
    -
    Rented housing units 0
    1
    180
    130
    58.10%
    41.90%
    -
    -
    Less than H.S. education 0
    1
    243
    67
    78.40%
    21.60%
    -
    -
    Less than bachelor's degree 0
    1
    231
    79
    74.50%
    25.50%
    -
    -
    Median income in tracts without sites
    Median income in tracts with sites
    0
    1
        $48,178.97
    $42,023.71

     

    Table 2: Relationship of Predictor Variables to Site Present in Tract

    Tracts with:   f % n x2
    Majority nonwhite-headed households Non-majority 11 6.7 164 .316
    Majority 6 4.1 146
    Majority female-headed households Non-majority 17 5.8 295 .339
    Majority 0 0 15
    Majority 65+ yrs of age headed households Non-majority 17 5.5 309 .809
    Majority 0 0 1
    Majority rented housing units Non-majority 7 3.9 180 .147
    Majority 10 7.7 130
    Majority less than high school education Non-majority 14 5.8 243 .683
    Majority 3 4.5 67
    Majority less than bachelor‟s degree education Non-majority 14 6.1 231 .446
    Majority 3 3.8 79

     

    Table 3: Relationship of Predictor Variables to Site Present in Tract or Adjacent Tract

    Tracts with:   f % n x2
    Majority nonwhite-headed households Non-majority 47 28.7 164 .604
    Majority 38 26.0 146
    Majority female-headed households Non-majority 80 27.1 395 .599
    Majority 5 33.3 15
    Majority 65+ yrs of age headed households Non-majority 85 27.5 309 .538
    Majority 0 0 1
    Majority rented housing units Non-majority 74 41.1 180 .927
    Majority 36 27.7 130
    Majority less than high school education Non-majority 69 28.4 243 .463
    Majority 16 23.9 67
    Majority less than bachelor‟s degree education Non-majority 67 29.0 231 .285
    Majority 18 22.8 79

     

    * p < .05

    Table 4: Independent Samples t-test for Median Income

    Group Statistics
    Independent Samples Test
    Site N Mean Std. Deviation Std. Error Mean
    Median Income 0
    1
    293
    17
    48178.969
    42023.706
    21494.01065
    12833.15992
    1255.6935
    3112.4985
      Levene‟s Test for Equality of Variances
      F Sig.
    Median Income Equality of variances
    assumed
    2.385 .124
    Equality of variances
    not assumed
    Independent Samples Test
      t test for Equality of Means
    t df Sig. (2-tailed) Mean Difference
    Median Income Equality of variances assumed 1.168 308 .244 6155.26340
    Equality of variances not assumed 11.295 28.142 .080 6155.26340

     

    Table 5: Independent Samples t-test for Population Density

    Group Statistics
    Site N Mean Std. Deviation Std. Error Mean
    Population Density 0 293 3506.621 2037.7880 119.0488
    1 17 807.712 854.2396 207.1835
    Independent Samples Test
      Levene‟s Test for Equality of Variances
      F Sig.
    Population Density Equality of variances assumed 8.846 .003
    Equality of variances not assumed
    Independent Samples Test
      t test for Equality of Means
      t df Sig. (2-tailed) Mean Difference
    Population Density Equality of variances assumed 5.426 308 .000 2698.9091
    Equality of variances not assumed 11.295 28.142 .000 2698.9091
     

     

    Figure 1: Tarrant County: Tracts with Sites and Adjacent Tracts

    Figure 1. Tarrant County: Tracts with Sites and Adjacent Tracts

    Figure 2: Nonwhite Status and Site Locations

    Figure 2. Nonwhite Status and Site Locations

    Figure 3: Median Income and Site Locations

    Figure 3. Median Income and Site Locations

    Figure 4: Population Density and Site Locations

    Figure 4. Population Density and Site Locations