Superfund Sites and the Social Composition of the Communities that Surround Them

SUPERFUND SITES AND THE SOCIAL COMPOSITION OF THE COMMUNITIES THAT SURROUND THEM A PAPER SUBMITTED FOR COMPLETION OF SENIOR RESEARCH FOR THE COLLEGE OF ARTS AND SCIENCES STETSON UNIVERSITY BY Sompi Harmetz IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE(S) OF BACHELOR OF ART ENVIRONMENTAL STUDIES AND GEOGRAPHY BACHELOR OF ART POLITICAL SCIENCE ADVISOR Dr. Jason M. Evans, Ph.D. MAY 2017Harmetz i Table of Contents Table of Contents .......................................................................................................................................... i List of Illustrations ........................................................................................................................................ ii List of Tables ............................................................................................................................................... iii Acknowledgments ....................................................................................................................................... iv Abstract ........................................................................................................................................................ v Introduction ................................................................................................................................................. 1 Literature Review ......................................................................................................................................... 3 The History and Precepts of Environmental Justice ................................................................................. 3 Love Canal and the Creation of CERCLA ................................................................................................... 3 Toxic Waste and Race in the United States .............................................................................................. 5 Controversies in Environmental Justice .................................................................................................... 7 The Impact of Zoning and Environmental Justice ..................................................................................... 8 The Creation of Superfund Sites ............................................................................................................. 10 The Challenge of Public Participation ..................................................................................................... 11 The Danger of Superfund Sites ............................................................................................................... 12 Study Area .................................................................................................................................................. 14 Methodological Choices .......................................................................................................................... 15 Methods ..................................................................................................................................................... 16 Observations and Data Analysis .................................................................................................................. 25 Conclusions and Discussion ........................................................................................................................ 29 Appendix .................................................................................................................................................... 31 Bibliography ............................................................................................................................................... 39 Harmetz ii List of Illustrations Figure 1. Florida Superfund Sites, 2017 ...................................................................................................... 14 Figure 2. American FactFinder .................................................................................................................... 18 Figure 3. JMP Stepwise Regression ............................................................................................................. 24 Figure 4. Nominal Logistic Fit Effect Summary ........................................................................................... 27 Figure 5. Social Vulnerability by Census Block Group in Hillsborough County, FL ...................................... 33 Figure 6. Social Vulnerability by Census Block Group in Miami-Dade County, FL ...................................... 34 Figure 7. Z-Score Black by Census Block Group in Hillsborough County, FL ............................................... 35 Figure 8. Z-Score Black by Census Block Group in Miami-Dade County, FL ................................................ 36 Figure 9. Z-Score Hispanic by Census Block Group in Hillsborough County, FL .......................................... 37 Figure 10. Z-Score Hispanic by Census Block Group in Miami-Dade County, FL ........................................ 38 Harmetz iii List of Tables Table 1. T-Tests of Tracts within One Mile Containing One or More Superfund Site versus All Other Tracts ................................................................................................................................................................... 25 Table 2. Nominal Logistic Fit of Tracts within One Mile Containing One or More Superfund Site versus All Other Tracts ............................................................................................................................................... 26 Table 3. Stepwise Fit of Tracts within One Mile Containing One or More Superfund Site versus All Other Tracts .......................................................................................................................................................... 28 Table 4. Number of Superfund Sites in Florida Counties ............................................................................ 31 Table 5. Social Vulnerability Variables ........................................................................................................ 32 Harmetz iv Acknowledgments I would like to thank Dr. Evans for guiding me through this project, Dr. Abbott for inspiring my path in Environmental Justice, Dr. Hauer for helping me navigate Census Block Group Data, and Emily Niederman for always lending a helping hand in GIS. Harmetz v Abstract Florida Superfund Sites and the Social Composition of Communities That Surround Them Superfund sites can be seen as dangerous to those in the nearby area because of the large amounts of chemicals they might contain, even once remediated. This can lead surrounding neighborhoods to become exposed to a wide array of health defects and social problems such as the stigmatization of property henceforth and discriminatory housing practices. By analyzing the location of Superfund sites in Florida and the demographics of their Census Block Group, this paper sought to determine if there is a positive correlation between Superfund locations and several sociological variables. The results showed that Census Block Groups within a 1-mile radius of a Superfund site tend to have more people below the poverty line, are more urban, contain more renters, and have more Hispanic and Black residents, than other Census Block Groups. The racial indicators percent Black and percent Hispanic showed the strongest effects when analyzed with logistic and stepwise regression, suggesting that race was the highest indicator of proximity to Superfund sites. These findings are consistent with prior research suggesting that racial minorities and those below the poverty line are systemically more likely to have exposure to chronic and acute environmental hazards. Harmetz 1 Introduction Superfund sites can be seen as dangerous to those in the nearby area because of the large amounts of chemicals they might still contain even once remediated. Hazardous chemicals from these sites such as arsenic, asbestos, beryllium, and cadmium often leach into the soil and move to ground water, air, or surface water (EPA 1994). The chemicals can put nearby residents at risks for various chronic diseases and health impacts. Due to the diverse nature of Superfund sites, ranging from closed dry cleaning facilities to former chemical landfills, the risk of exposure to said chemicals varies depending on the degree of cleanup, chemicals at the site, and distance from people. In addition to the wide array of health defects that may come from living near a Superfund site, there is also a degree of social problems that may influence the individual such as the stigmatization of the property henceforth, creating a low property value in the immediate area. Authors such as Maantay (2002) suggest this leads to discriminatory housing practices that pipeline minorities into these communities. In this study, I analyzed where hazardous waste sites are located in Florida and the demographics of the Census Block Groups in which they are contained and any Census Block Groups within a one-mile radius of the Superfund sites. The demographic variables included percent impoverished, percent over 65, percent urban, percent rich, percent renters, percent female, percent Hispanic, percent black, percent Native American, and percent in nursing homes. These variables were chosen after looking at a study by Evans et al. (2014) that looked at social vulnerability. To answer these questions, I used Census Data to physically map Census Block Groups with information pertaining to demographics to see if certain groups were Harmetz 2 disproportionately located within Census Block Groups containing Superfund sites. I used the methodology established by Stretesky and Hogan (Stretesky and Hogan 1998) that includes performing T-tests on variables within the study to illustrate this. I hypothesized that there would be a positive association between areas of poverty, minorities (Black and Hispanic), and higher population density with Superfund locations. I also used SoVI-Lite, a methodology of indicating the social vulnerability of a group to environmental hazards, throughout the state of Florida (Evans et al. 2014.). Social vulnerability is important because it shows not only which groups are able to survive environmental hazards but also those who can recover (Cutter 1996). By using both of these analyses, I was able to see if those communities located near Superfund sites also fall into socially vulnerable Census Block Groups according to SoVI-Lite. By using T-tests I was able to further analyze variables used in the SoVI-Lite methodology. I used logistic regression and stepwise regression to decipher which variables in the model held the most significance. I used bi-variate analysis to look at the relationship between voter turnout and Superfund sites. I hypothesized that there would be a negative association between Superfund sites and voter turnout per county. The entirety of the project is framed through an environmental justice lens which emphasizes the fair treatment of all people so that no one group shall receive a disproportional amount of environmental burden (Bullard et al. 2008). By analyzing various demographics I was able to see if these groups exist largely next to Superfund sites, therefore holding a higher environmental burden. I was also able to see if those within these areas are able to recover and seek remediation should they be affected by these Superfund sites by employing SoVI-Lite. Harmetz 3 With this procedure, Florida legislature, permitting groups, and local land use groups may see if the unfair amount of hazardous environmental burden on communities who may find it difficult to recover. The legislature may consider future legislation to compensate these communities or prevent future hazardous waste sites from being built in areas that cause them harm. Literature Review The History and Precepts of Environmental Justice Love Canal and the Creation of CERCLA In the late 1970s, Lois Gibbs began noticing that her sons had a series of disorders as well as many other people in their community ranging from skin rashes to miscarriages to other birth defects. In this same area, Hooker Chemical and Plastics Corporation had buried approximately 21,000 tons of toxic waste 40 years’ prior (Thomson 2016). Research indicated that a high number of residents had chromosomal breakage and because of this they expected quick remediation from the government- but they were mistaken and were instead told to take what they now knew and merely exercise caution in their everyday activities. A tension between community, state, and the federal government now existed and each was unsure of who was to blame and who was to be responsible. Fliers littered the streets with phrases like “We are fighting for the rights of everyone to live in a healthy environment. We are fighting for the rights of our little ones to live a long and healthy life!!” and a call for the “right to health; bear normal and healthy children; to a safe home (Thomson 2016).” Community inhabitants claimed that these were part of the inalienable rights within the United States Constitution and that the government was failing to protect their best interest. It was not until 1980 during the 96th Congress that congress would pass the Harmetz 4 Comprehensive Environmental Respond, Compensation, and Liability Act (CERCLA or Superfund), P.L. 96-510 to address hazardous waste dumps, provide emergency response, and give liability to parties involved (Switzer and Bulan 2002).Prior to the passage of CERCLA, many bills were considered by Congress including H.R. 85, S. 1480, H.R. 7020, and S. 1341 (Switzer and Bulan 2002). H.R. 85 went through the Committee on Public Works and Transportation and then the Committee on Ways and Means during 1979. This bill would use taxes on petroleum products, oil, chemical feed stocks and inorganic compounds to address oil spills and hazardous substances in navigable waters. Owners of these companies that released these toxins held strong liability putting the burden both on the government and to private owners. The full house did not consider this bill allegedly due to strong opposition from oil companies (Switzer and Bulan 2002). H. R. 7020 went through the Committee on Interstate and Foreign Commerce, the House Committee on Ways and Means and then was passed through the House on September 23, 1980. It gave the government permission to respond to dangerous hazardous waste at hazardous waste sites. The bill, like H.R. 85, created a fund through taxes on products meaning cleanup would be paid for partially by the government and also responsible parties who caused the toxic release (Switzer and Bulan 2002). The bill then moved to the Senate’s Environment and Public Works Committee. This bill crafted by the House was not considered in Senate and failed but compromise from it would eventually be the basis for CERCLA. S. 1341 was a bill in 1979 and went through the Senate Committee on Environment and Committee on Public Works. It attempted to create a 1.6 billion dollar fund with $325 million appropriated by the government and the rest coming from taxes on oil, chemical feedstocks, and Harmetz 5 inorganic substances to prevent toxic harm in navigable waterways (Switzer and Bulan 2002). It never went to full committee in the Committee on Environment and Public Works. S. 1480 was introduced to the Committee on Environment and Public Works. It established a $200 million "post-closure liability fund" to fix hazardous from closed hazardous waste disposal facilities and $4.085 billion for other cleanups. It then went to the Senate Committee on Finance but did not pass (Switzer and Bulan 2002). The 96th Congress was in its last two months during the Carter Administration. Democrats worked to ensure that some form of a hazardous waste bill would be passed because, after this session, the White House and Senate would be under Republican Control (Switzer and Bulan 2002). CERCLA was a compromise created due to the emergency of Love Canal and the Valley of the Drums (a toxic waste site in Kentucky). Due to the quick nature of its passing, it was based on a bill presented only days before its final passage and was hastily rushed through passage with neither House nor Senate committee consideration (Switzer and Bulan 2002). Love Canal is commonly cited as the spark that created the environmental justice movement. It was a time where the community came together to protest the environmental burden placed on them by their government. To other scholars, this event does not inspire romantic notions of the founding of the movement because it was a predominately white community with some noted instances of racism during their protests. While scholars are divided on Love Canal’s overall relationship to the environmental justice movement, it did indeed create the momentum that would later force the government to create CERCLA. Toxic Waste and Race in the United States Shortly after, other notable environmental justice studies began. The United Church of Christ’s Commission for Racial Justice released the Toxic Waste and Race in the United States Harmetz 6 report (1987) which studied the relationship between race and commercial hazardous waste facilities. The results showed that a high amount of Hispanics and Blacks were at a risk of unfair exposure. This research is still relevant today and sparked a wave of literature studying the relationship between race and socioeconomic status and distance from hazardous waste facilities. From this point forward a series of precepts were created to guide environmental justice work. The framework of environmental justice comes from precepts of the Civil Rights act of 1964, the Fair Housing Act of 1968, and the Voting Rights Act of 1965 (Bullard et al. 2008). These elements combined create an environmental justice is defined as the "fair treatment and meaningful involvement of all people regardless of race, color, national origin or income with respect to the development, implementation, and enforcement of environmental laws, regulations, and policies. Fair treatment means that no group of people, including racial, ethnic or socioeconomic groups, should bear a disproportionate share of the negative environmental consequences resulting from industrial, municipal and commercial operations or the execution of federal, state, local and tribal programs and policies (Bullard et al. 2008).” Environmental justice seeks to be precautionary and does this by focusing not only on direct environmental harm but on other practices that may lead to environmental harm such as “housing, land use, industrial planning, healthcare and sanitation services (Bullard et al. 2008)." These factors, in turn, may impact "redlining, economic disinvestment, infrastructure decline, deteriorating housing, lead poisoning, industrial pollution, poverty, and unemployment (Bullard et al. 2008).” While these elements may seem disconnected, to those living in urban ghettos or poor communities, each element can be seen as a compounding factor that may place someone at risk to vulnerability and environmental harm. By assessing these variables, it is possible to untangle how marginalization Harmetz 7 occurs and ensure procedural equity, geographic equity, and social equity to all groups (Bullard et al. 2008). Still, there is controversy in our polarized politics as to how much regulation our government should have in these instances. Initially developed as a regulatory agency, some argue that the EPA is meant to regulate the environment and was not derived as a health agency. While grassroots agents and lobbyists have fought for the acknowledgment over the years for environmental justice within our government, there has been a fine line as to whether or not marginalization or pure coincidence has caused these groupings of hazardous exposure. Controversies in Environmental Justice A nuance of the environmental justice field today is its relatively young age in comparison to other frameworks. With this young age has come a series of challenges including its methodological standards and justification. Environmental justice’s focus on equity has attracted scholars in many fields such as sociology, health, law, geography, political science, economics, and environmental studies meaning that it has been studied with many different methodologies and viewpoints (Douglas 2008). Certain fields may be unaware of important literature established in environmental justice for this reason. An economist may focus more on welfare status than a lawyer who focuses on zoning laws. Therefore, it is necessary to speak about these controversies to better understand certain complications that may arise while studying anything in the environmental justice perspective. A large criticism of environmental justice is that what is true of one case study, for example, a city in Georgia, is not true of all studies. Each jurisdiction is so unique that often times in environmental literature, they can be overgeneralized (Anderton 1996). It is important to note that what is discovered during my research will not be a clear indicator of what is true in Harmetz 8 another state. Another large criticism is the lack of standardization in methodology and databases of hazardous waste facilities (Maantay 2002). This is due to the relatively new field and the wide variety of those who study environmental justice. Scientists are divided on the "proper" unit to assess environmental justice. Liu (2000) calls for what the scientist deems as appropriate for their own study and Anderton et al. (1994) seek the smallest possible scale for the project. Baden et al. (2007) argue that throughout environmental justice literature that there is little justification for why scientists choose their unit size and that it's all arbitrary (Douglas 2008). There is an importance in choosing the correct unit of analysis because choosing too large of a unit may create an “ecological fallacy” that is not true of smaller units (Anderton et al. 1994). In addition to the controversies of methods to study environmental justice, there are also controversies on how to define the field. There is difficulty in defining "injustice" because there is no formal definition of what is "just." A large portion of environmental literature speaks about marginalized communities but there is no established picture of what a "just" community would look like (Douglas 2008). There is also controversy in deciding if there is a precedent for environmental justice. Studies suggest that race will be the highest indicator of proximity to waste facilities (Mohai et al. 2009); United Church of Christ’s Commission for Racial Justice 1987). Others suggest that this relationship is weak (Yandle and Burton 1996) or does not exist (Anderton et al. 1994). The Impact of Zoning and Environmental Justice Land use laws and zoning have an impact on where certain buildings are allowed to be placed within a state. They were created to control land to help protect “public health, safety, and welfare within an existing legal concept of the police powers (Maantay 2002).” Residential Harmetz 9 zoning, commercial zoning, industrial zoning and agricultural zoning may all have an impact on hazardous waste in Florida and the likelihood that a site becomes Superfund because of the power it has a tool to shape a geography. While zoning was not created as a tool to create inequity it has the power to be used as such. Residential zoning can decide whether things like mobile homes may be allowed on property impacting the overall value of the land. Commercial zoning may impact whether or not warehouses are allowed within an area. Industrial zoning can allow different manufacturing plants and environmental factors associated with them such as an increased risk of air pollution. These industrial zones are also associated with high levels of adjacent minorities (Maantay 2002). Agricultural zoning may make an area more susceptible to runoff from pesticides. Land use can play a part in the distribution of toxic land use and its exposure to nearby residents. By moderating land use to ensure certain property values, zoning can also create the exclusion of undesirable land use and consequentially the exclusion of certain sociodemographic populations as well. A famous historical example is the 1885 law against laundromats in residential San Francisco which was written as an attempt to exclude Chinese people from living in white areas (Maantay 2002). These intentional exclusions continue today as zoning may prohibit things associated with low property values (mobile homes and factories) and systemically use legislation to impact the income level of a community (Maantay 2002). In addition to streamlining certain sociodemographic groups to certain areas, zoning may also impact the amount of environmental legislation in these areas. In a docket by the EPA, a National Law Journal investigation claims that areas with the highest white population experienced 500 percent higher hazardous waste penalties compared to sites with the greatest minority populations (Maantay 2002). These penalties were also 46 percent higher when laws Harmetz 10 that protect citizens from air, water, and waste pollution were broken in areas of majority white residents (Maantay 2002). Zoning may have a direct impact on how communities have been formed throughout Florida and which areas may be negatively impacted by environmental risk because of them. The Creation of Superfund Sites When looking at which chemicals are important, there is an emphasis on taking action against chemicals with the highest levels of maximum individual cancer risks (Vscusi and Hamilton 1999). The EPA decides which sites to remediate first by using the Hazard Ranking System that looks at contamination levels and how likely they will be to effect the surrounding area and then places these sites on the National Priorities List (NPL). A Record of Decision (ROD) is then released that announces the plan to remedy said site and the risk levels of chemicals post remediation. Liability can fall on: “current owners of a facility, person who owned a facility when the hazardous substances were disposed of, person who arranged for the disposal or treatment of hazardous substances at a facility, and any person who selected such facility and transported hazardous substances to it for disposal or treatment (Switzer and Bulan 2002).” When the level of risk is high in a densely populated area, the EPA typically sets stricter regulatory standards opposed to when the area is less densely populated (Vscussi and Hamilton 1999). This is possibly because fewer people are exposed. Conversely, if there are low individual risk levels in a densely populated community, as population density increases, there are less strict regulations put in place. This is possibly because as population increases, there are fewer people paying attention to these regulations. Harmetz 11 Higher percentages of minorities were less likely to have cleanup decisions and slower remediation (Vscusi and Hamilton 1999). According to the EPA, Superfunds in areas of minorities could take 20 percent longer to be put on the National Priority List (Maantay 2002). High voter turnout, states with constituents in environmentalist organizations, and senators with stronger environmental voting records within communities is correlated with lower environmental risks after remediation and stricter environmental cleanup standards (Vscusi and Hamilton 1999). When the risk is high, political activity has little impact according to Vscussi and Hamilton (1999) but when the risks are lower, political activity is highly important to convince agencies to enact regulatory standards. For this reason, voter turnout was tested in my study. The Challenge of Public Participation Up until this point, primarily outcome justice has been discussed. This is the belief that no disproportionate environmental burden should fall on one community. Yet, there are two forms of justice: process and outcome. In process justice, equity is allowing structural equity to all populations. It allows all populations to participate in politics and allows access to things such as adequate information, public meetings at times and places that are easily accessible, the language being spoken, having technical language reiterated in an easy to understand way, access to debate and speak and influence over decisions (Maantay 2002). Should procedural equity be compromised, it is often difficult to ensure outcome equity. Today, the process of public participation may be seen as difficult to understand and very time-consuming. Authors such as Maantay believe procedure has caused public processes to be ruled by the wealthy, educated and non-minority community (2002). Many barriers exist for the average person to participate publicly. Harmetz 12 First, technical expertise is not always presented in a way that is understandable to all. Oftentimes this means that the public is unable to access or understand information concerning their community. This is particularly relevant when highly technical Environmental Impact Statements (EIS) are used as sources in public hearings and community statements are disregarded as “ill-informed” or “emotional (Maantay 2002).” These technical documents are more difficult to debate against unless by another team of scientists and specialist, therefore, putting the government in a higher position of control over the case. Next, the government controls the timeline of which an environmental issue is discussed. If a community member works full time it is often impossible to attend these public meetings without taking off work. According to Maantay, the public typically has 10 days to respond to a public hearing concerning the environment (Maantay 2002). Without hiring outside assistance, 10 days for a layperson to respond meaningfully is very short and nearly impossible. While higher levels of public participation are correlated with better environmental regulation, this may also be because the communities with the highest levels of public participation are communities where their participation can make an impact. Structurally many communities face the challenges listed above and this may limit their political efficacy. The Danger of Superfund Sites Just because a community is near a Superfund site does not mean that they are exposed to harmful chemicals (Johnson 1995) and there is no standardized model to explain how to measure exposure (Maantay 2002). For this reason, some argue that this shows a limitation in environmental justice research. While the literature suggests that there may be no connection, the entirety of the framework of environmental justice suggests that these sites still be researched to Harmetz 13 ensure equality regardless. Even if there is no chemical harm done, there are also certain social problems that come with living near a Superfund site. Some authors suggest that there is little economic benefit to cleaning Superfund sites (Gallagher et al. 2007) while most suggest that property values are less following the creation of Superfund sites (Greenberg and Hughes 1992). In a study from Houston, Texas, Recai and Smith (2008) found that it took a decade and a half for property value to increase to what it had been before remediation. The authors cite that in many instances the drop in prices creates an influx of those with lower incomes to move into these communities. By creating a pipeline of people into these communities, it suggests a discriminatory practice. While the land for these houses may be cheaper, the probability of the surrounding areas flourishing economically is unlikely meaning that there will likely be fewer jobs available in these areas. Other research shows that individuals are willing to pay more for homes in environmentally safe areas away from hazardous sites (Thayer et al. 1992). This suggests that if people were able to afford these homes, they would. Harmetz 14 Figure 1. Florida Superfund Sites, 2017 Study Area This study took place in the State of Florida. According to Societal Vulnerability to Environmental Hazards (Cutter 1996), the most vulnerable counties in the United States are located in South Florida, I feel that this creates a just reason to use Florida as a case study. With 78 Superfund sites, as illustrated in Table 2, Florida has the sixth highest amount of Superfund sites in the United States (Kirpich and Leary 2017). Figure 1 illustrates how the location of Superfund sites in Florida seemingly clumps together in certain counties raising the question of why this is. Harmetz 15 Florida’s subtropical climate and sandy soils allow for a higher potential for hazardous waste to travel through water (Kirpich and Leary 2017) suggesting that there is a potential higher risk to those who live near Superfund sites in Florida. In a recent study by Kirpich and Leary (2017), it was found that there may be an association between cancer rates and Superfund sites in Florida (2017). My hypothesis states that minorities are more commonly located in Census Block Groups with a Superfund site than those who are not a minority. Florida has the 4th highest amount of immigrants (4.1 million) in the United States. There are approximately 4,223,806 Hispanic people and 2,999,862 Black or African American people located in Florida (US Census Bureau 2010). This means that the results of this case study may be affecting millions. Methodological Choices I chose to use Census Block Groups because I felt that they would create the highest spatial resolution. In addition to this, it is the same unit of analysis used in Evans et al. 2014, the project I am using as a guide to SoVi-Lite. By employing SoVI-Lite I was able to test for social vulnerability or the "characteristics of a person or population that influence how they will be affected by a natural hazard or disaster event (Evans et al. 2014)." This is important because it does not only show geospatial data but also the likelihood that a person will be able to escape or recover from a disaster. I chose to use SoVI-Lite because it is a simplified model in comparison to its sister model SoVI. Due to SoVI-Lite’s flexible nature, I was also able to apply it to different spatial scales. This way I knew for certain I could apply it to group blocks (Evans et al. 2014). Harmetz 16 Yet there are limitations to SoVI-Lite. For example, communities with a lot of nursery homes and retirement communities might register as highly vulnerable on this scale due to the disproportionately high amount of elderly residents. While age can be a factor in vulnerability, many retirement homes have wealthy inhabitants who would be able to fiscally survive after a disaster. This, therefore, may skew the results. Because Florida fosters many retirees this is highly possible. For this reason, I chose to emulate Stretesky and Hogan’s (1998) analysis of Superfund sites in Florida using T-Tests as well. I hypothesized that there will be a positive association between poverty, minorities (Black and Hispanic) and higher population density compared with Census Block Groups containing and within a mile of Superfund sites. Methods In order to determine if minority and impoverished communities hold a substantial amount of Superfund Sites, I used Census Block Group data from the state of Florida and replicated the methods of Evans et al. (2014) and Stretesky and Hogan (1998). Evans et al. (2014) outlined a method of determining social vulnerability which helped determine if communities are able to survive and recover from a disaster. I used these methodologies to calculate a “SoVi-Lite Score” which indicated social vulnerability. In order to determine if individual variables within Evans et al. (2014) have a significant correlation with Superfund site location, I used methodology employed by Stretesky and Hogan (1998). This methodology uses T-Tests to compare population means of the demographics of Census Block Groups to see if certain variables were more predominant in Census Block Groups that have Superfund sites and those that don’t. Harmetz 17 Through this methodology, I employed the usage of Z-Scores, which are the number of standard deviations from the mean that a data point is within a given set of data. To calculate Z-Scores, data is compared to one another and put along a normal distribution curve by subtracting the percentage value of a demographic variable by the mean percentage of that variable across all geographies then dividing this number by the standard deviation of the percentage for that variable (Evans et al. 2014). To put this in perspective, if this were replicated using a larger sample area, perhaps Florida and Georgia, these scores would be different. I used Z-Scores so that I could see if a score was equal, above or below the mean and how far from the mean it was. I could also use Z-Scores to standardized different data with a score so that they may be compared to one another within the geography of Florida. This methodology felt appropriate so that I could see if a group was marginalized within Florida as a whole. I chose to use nominal logistic regression to analyze the Z-Scores of Census Block Groups with and without Superfund site. This allowed me to see when all of the Z-Scores were compared among different variables, which ones were still significant. The more variables you run in a Logistic Regression, the higher chance of “overfitting” the model because with each variable added to the model there will be an increase in statistical validity. Because 12 variables were included in the nominal logistic regression, I also ran a stepwise regression. The stepwise regression adds and drops covariates based on Bayesian Information Criterion (BIC) which introduces a penalty for the number of variables in a model. By doing this, I could identify which variables were the most significant when a penalty was introduced. Step 1- Organized Data A) Used American FactFinder (http://factfinder2.census.gov) to collect all data. Harmetz 18 Figure 2. American FactFinder a. Logged onto the website and click advanced search and was taken to the website shown in Figure 2. b. Selected “Geographies” from the left index. i. Selected from “Most Requested Geographic Types” ii. Selected “Block Group-150” as geographic type. iii. Selected “Florida” as the State. iv. Selected every Florida County. There was 67 Census Block Groups. c. Selected “Topics” from the left index. i. Selected the variables from the Table 3. Harmetz 19 Step 2- Deleted Interfering Data A) Copy and pasted the “Total” column from the Table ID’d as P2 into each excel sheet. This gave each dataset a total population to refer to when calculating percentages. B) Clicked the “Sort and Filter” function within excel. C) Filtered selecting the P2 “Total” column for all Census Block Groups “less than 200” to eliminate all Census Block Groups that have less than 200 inhabitants. Step 3- Calculated Percentages A) Within their own excel spreadsheets did the simple division for variables: urban, renters, Hispanic, people living in nursing homes, Black, and Native Americans. (Total population count of individuals classified as x divided by the total population for the Census Block Group) B) Calculated percentage of population over 65 by adding both males and females to one cohort. Then used simple division as in step A. C) Calculated percent female by adding the all of the age group columns for females. This number is then divided using simple division as in step A. D) To calculated percent households earning over $100,000 all columns that have household earnings higher than this had to be summed and then this number was divided using simple division as in step A. Harmetz 20 E) “Calculated percent poverty by summing two values from the ratio of income to poverty tables: 1)”0.0-0.49”; and 2) “0.5-0.99” (Evans et al. 2014).” These summed values were then divided using simple division as in step A. Step 4- Calculated Z-Score A) Calculated population means for each variable. B) Calculated the standard deviation for each variable. a. Used stdev.p for data within the 2010 Census SF1. b. Used stdev.s for data within the 2007-2001 ACS. C) Calculated Z-Score using Equation 1. Z-Score q = (q – μ) / σ "Where q is the percentage value of a demographic variable in a given geography; μ is the mean percentage for that same variable across all considered geographies, and σ is the standard deviation of the percentage for that same variable across all considered geographies." (Evans et. al 2014). Step 5- SoVi-Lite Scoring & Tiered Classification A) Calculated SoVi-Lite raw values for each Census Block Group using Equation SoVI-Lite = (Zqpoverty) + Abs(Zqover65) + Abs(Zqurban) –(Zqrich100k) + (Zqrenter) + (Zqfemale) + (Zqhisp) + (Zqnursinghome) + (Zqnativeamerican) “Where Abs = Absolute value; and Z = Z-Score” (Evans et. al 2014) B) Transformed raw SoVi-Lite values into SoVi-Lite scores using Equation 3. Harmetz 21 Z-ScoreS = (S – μ) / σ C) Assigned vulnerability into tiers: limited social vulnerability (Z-Score <-0.5), moderate social vulnerability (-0.5< Z-ScoreS<0.5), and elevated social vulnerability (0.5 <Z-ScoreS). Step 6- Saved Appropriate Information A) Combined data in a large excel sheet that includes all Z-Scores. B) Saved both as a CSV (Comma delimited) files. Step 7- Located Superfund Data A) Located a list of all registered locations in the state of Florida on the Environmental Protection Agencies website. B) Used the "Sort and Filter" function to select only sites labeled as "Superfund NPL" and deleted all others. C) Saved as a CSV (Comma delimited) file. Step 8- Began Preliminary Mapping A) Added data of a Shapefile of Florida Census Block Groups into ArcGIS. B) Added the Superfund NPL CSV into ArcGIS and turned it into Shapefile. Step 9- Calculated Population Density A) Added data of a Shapefile of Florida using Albers Conical Equal Area. B) Uploaded the CSV of all Percentages Calculated and saved as a database file. C) Uploaded the CSV of all Z-Scores and saved as a database file. Harmetz 22 D) Joined the Percentages Calculated CSV to a copy of the Florida Census Block Groups shapefile. E) Used the “project” function to re-project the Florida Census Block Group map with Albers Conical Equal Area. This projection allowed for a more accurate measure of area. F) Opened the attributes table and added a new “float” field. G) In this new field, I calculated the geometry of the Census Block Group in square miles. H) Added a new field and used the field calculate to divide the total population by the geometry of the Census Block Group to attain population density. I) Exported this new table into excel. J) Calculated the Z-Score of population density and added this data to my Z-Score CSV file. Step 10- Isolate Census Block Groups with Superfund Sites A) Used the join function to join the Z-Score CSV to a Shapefile of Florida Census Block Groups. B) Used the “Select by Location” function to select features from the Census Block Group layer that intersect from the Superfund layer. a. Exported these Census Block Groups into database files to create Census Block Groups with Superfund sites. b. Selected the inverse to create Census Block Groups without Superfund sites. Harmetz 23 C) Used the "Select by Location" function to select features from the Census Block Group layer that intersect from the Superfund Layer within a one-mile range. a. Exported these Census Block Groups into a database file to create Block Groups within One Mile of Superfund sites. b. Selected the inverse to create a database file to create Census Block Groups Not within One Mile of a Superfund site. D) Opened these within Excel to save as Excel Spreadsheets. These spreadsheets now included all variables that would be tested statistically. Step 11- Statistical Analysis A) Conducted an F-test in excel to ensure that variance was not equal to the means of individual Z-scores with and without Superfund sites. Then conducted T-Test Assuming Unequal Variance for those without equal variance and T-Test Assuming Equal Variance for those with equal variance. B) In Excel, I added a column to the Z-Score spreadsheet and put a ‘0’ next to Census Block Groups without Superfund Sites and a ‘1’ next to Census Block Groups with Superfund sites. C) Opened the Z-Score spreadsheet in JMP as an excel file. D) Changed the column that denoted Census Block Groups as having or not having Superfund sites to nominal data. E) Checked that all other columns were numeric data. F) Selected fit model under the analyze tab. Harmetz 24 G) Used the column that denoted Census Block Groups as having or not having Superfund sites as the y-column. H) Ran a nominal logistic fit. I) Ran a stepwise fit with the stopping rule, direction, and rules as shown in Step 12- Analyzed Voter Turnout A) Used STATA to run a bi-variate analysis of Superfund sites and 2010 Voter Turnout by county Step 13- Complete Final Maps A) Used ArcGIS to create a map showing all of Florida and its Superfund sites to illustrate where they cluster. B) Used ArcGIS to create maps showing SoVi-Lite scores by Census Block Group of the counties where there are the most Superfund sites: Hillsborough, Miami-Dade, Duval, and Broward County. Figure 3. JMP Stepwise Regression Harmetz 25 Observations and Data Analysis Table 1. T-Tests of Tracts within One Mile Containing One or More Superfund Site versus All Other Tracts Table 1 illustrates the results after T-tests were given for different variables Z-Scores in Census Block Groups within one mile of a Superfund site and those that were not. The variables for poverty, over 65, urban, rich, renter, female, Hispanic, black, population density, and social vulnerability all showed significance with P values less than .05. Poverty showed results in the direction I expected with those who lived below the poverty line having a higher percent living within a mile of Superfund sites than not in a Superfund site. The p-value showed strong statistical significance. Both urban and population density held expected results with those living in highly urbanized areas and with higher population density being more likely to be within a mile of a Superfund site than those who were Variable Superfund Mean Non-Superfund Mean Difference t Stat P-Value Zqpoverty 0.4336729 -0.025637487 0.4593103 10.3442966 1.98E-23 *** Zqover65 -0.294338 0.017397795 -0.3117354 -11.71441021 1.77E-29 *** Zqurban 0.1614763 -0.009495031 0.1709713 5.565009725 3.61E-08 *** Zqrich100k -0.328411 0.019424016 -0.3478348 -9.244176763 2.58E-19 *** Zqrenter 0.364296979 -0.02154 0.385835534 9.40715178 6.883E-20*** Zqfemale -0.084124 0.004979298 -0.0891037 -2.003267029 0.045534 ** Zqhispanic 0.3009958 -0.017712262 0.318708 6.291282278 5.67E-10 *** Zqblack 0.7388969 -0.043683748 0.7825806 13.19033621 1.81E-35 *** Zqnativeamerican 0.011537847 -0.00066 0.012200971 0.593994797 0.552642095 Zqnursinghome -0.04014 0.002378445 -0.0425184 -1.253056273 0.210574 Zq_popdensity 0.1644696 -0.009393224 0.1738628 4.975950301 8.07E-07 *** Z-Score SOVI 0.180006982 -0.01064 0.190648694 4.206827695 2.92859E-05 *** Harmetz 26 not. Hispanic held significant results showing a high amount of Hispanics living in Census Block Groups within a mile of a Superfund site than those who do not. Black held significant results showing a high amount of residents Black who live in a Census Block Group within a mile of a Superfund site than those who do not. Residents over 65 showed significant results for those being over the age of 65 living within a mile of Superfund sites less and in Census Block Groups not in these areas more. Those who were rich showed significant results for having fewer people in Superfund site locations and were more likely to not be in these areas. Amount female showed that women were less likely to live in Superfund areas albeit the significance of the p-value was much lower when compared to other p values in this study. Neither residents living in nursing homes nor residents Native American showed significant results in this study. Table 2. Nominal Logistic Fit of Tracts within One Mile Containing One or More Superfund Site versus All Other Tracts Variable Estimate Chi Square Prob>ChiSq Intercept 2.635022 446.75 <.001* Zqpoverty -0.42051 15.74 <.001* Zqover65 -0.16665 4.15 .0415* Zqurban 0.197899 4.52 .0334* Zqrich100k 0.528839 21.76 <.0001* Zqrenter -0.44723 14.64 .0001* Zqfemale -0.27191 7.51 .0062* Zqhispanic -0.81309 58.21 <.0001* Zqblack -1.00419 86.25 <.0001* Zqnativeamerican -0.39826 13.64 .0002* Zqnursinghome -0.34773 10.52 .0012* Zq_popdensity 0.081173 2.4 <.0001 Z-Score SOVI 0.160322 16.85 <.0001* Harmetz 27 Table 2 illustrates the results of the nominal logistic regression of Census Block Groups within one mile containing one of more Superfund sites versus all other tracts. In this regression, all variables showed statistical significance. Figure 4 illustrates that Percent Black and Percent Hispanic were the most significant variables in this model. Figure 5 and Table 3 illustrate the results of the stepwise regression. Here we see that after adding penalties for each variable added, percent black and percent Hispanic were the two most significant variables in the model. Figure 5. Stepwise Fit Effect Summary Figure 4. Nominal Logistic Fit Effect Summary Harmetz 28 Table 3. Stepwise Fit of Tracts within One Mile Containing One or More Superfund Site versus All Other Tracts The analysis of number of Superfund sites per county with the voter turnout in 2010 (the date of the last Census) revealed that there is a negative correlation with the number of Superfund sites in a county and voter turnout. The county with the highest amount of Superfund sites (Hillsborough with 16) had 47.7 percent voter turnout with a county that had none (Sumter) having 65.2 percent voter turnout. Figure 6. Superfund Sites and County Voter Turnout Variable Estimate Chi Square Prob>ChiSq ZQHispanic 0.43495 130.93 <.0001* ZQBlack 0.62285 376.8 <.0001* Harmetz 29 Conclusions and Discussion Consistent with my hypothesis, the racial indicators Black and Hispanic showed a large statistical difference in Census Block Groups within a mile of a Superfund site as compared to those without Superfund sites. The variable Black showed the largest amount of difference as well as statistical significance. This is consistent with the results of Mohai et al. (2009) which also found race as the most important indicator of Superfund sites. The logistic regression and stepwise regression showed that when penalties are added, percent Hispanic and percent Black was the most significant variables ran in the model. Poverty and population density also showed a significant positive association, similar to Mohai et al. (2009). In this study, I was surprised to see that those over the age of 65 were less likely to live in Superfund areas suggesting that they had all moved out as these areas became contaminated or lost value. When analyzing voter turnout, counties with Superfund sites on average experienced lower levels of turnout. This suggests that communities that would benefit from public participation in order to protect themselves from environmental injustice are going to the polls less. Future studies should look at the political efficacy of these communities and whether they find that their vote matters. Future studies should also use this correlation to guide them in looking at other forms of public participation such as attendance at local government meetings. Should these counties have less public participation as low voter turnout suggest, they are less likely to receive the representation they need when harm arises from Superfund sites. A criticism of the method I used to calculate SoVi-Lite, developed by Susan Cutter (1996), is that while the lite version of her model was able to afford simplicity by only using a Harmetz 30 few variables, it also left out many important variables that may suggest what makes a community "socially vulnerable." For example, in this study, I added percentage black as a variable although they weren’t contained within the original methods by Cutter. This variable was particularly important to the demographics of Florida and in this instance, had the strongest results. SoVi-Lite, while a useful tool, appears to not be well rounded to the needs of specific areas and should be adjusted accordingly. While my results show that Superfund sites are highly correlated with certain sociodemographic variables, it is unclear whether this is causal. Future scientists may consider looking at some case studies and seeing why the Superfund sites have been chosen for remediation. There is the possibility that the EPA chooses sites with higher levels of minorities, poor and urban thus causing this large correlation in my study. Overall, an overwhelming amount of the Z-Scores of variables that indicate social vulnerability also shows a correlation with higher populations in Census Block Groups within a mile of Superfund sites than those who do not. As shown in Figures 3, 4, 5, 6, and 7. This means, that those who are most susceptible to social vulnerability and difficulty remediating after a natural disaster (including toxic hazards) currently live in these areas. The data suggests that groups—below the poverty line, renters, Black and Hispanic—are systemically more likely to have exposure to chronic and acute environmental hazards. In the future, the Florida legislature should consider compensation to these communities or possibly set aside emergency funds incase these areas are further investigated and chemical hazards are still persistent. Education and awareness campaigns should be created so residents in these areas are aware of the risks these Superfund sites may hold and ways that they can prevent harm to themselves and their families. Harmetz 31 Appendix Table 4. Number of Superfund Sites in Florida Counties County Number of Superfund Sites Alachua 1 Bay 1 Brevard 1 Broward 8 Duvall 9 Escambia 7 Gadsen 1 Hillsborough 16 Indian River 1 Jackson 2 Lake 1 Madison 1 Martin 2 Miami-Dade 14 Orange 3 Palm Beach 1 Pinellas 1 Polk 3 Santa Rosa 1 Seminole 2 Suwannee 1 Volusia 1 Total 78 Harmetz 32 Table 5. Social Vulnerability Variables Variable Data Source Table ID qpoverty 2007-2011 ACS C17002 qover65 2010 Census SF1 P12 qurban 2010 Census SF1 P2 qrich100k 2007-2011 ACS B19001 qrenter 2010 Census SF1 H11 qfemale 2010 Census SF1 P12 qhispanic 2010 Census SF1 P5 qblack 2010 Census SF1 P5 qnativeamerican 2010 Census SF1 P5 qnursinhome pop_density 2010 Census SF1 2010 Census SF1 P42 B01003 Harmetz 33 Figure 5. Social Vulnerability by Census Block Group in Hillsborough County, FL Harmetz 34 Figure 6. Social Vulnerability by Census Block Group in Miami-Dade County, FL Harmetz 35 Figure 7. Z-Score Black by Census Block Group in Hillsborough County, FL Harmetz 36 Figure 8. Z-Score Black by Census Block Group in Miami-Dade County, FL Harmetz 37 Figure 9. Z-Score Hispanic by Census Block Group in Hillsborough County, FL Harmetz 38 Figure 10. Z Score Hispanic by Census Block Group in Miami-Dade County, FL Harmetz 39 Bibliography Anderson, R. J. and T. D. Crocker. 1971. Air Pollution and Residential Property Values. Urban Studies 8:171-180. Anderton, D. L. 1996. Methodological issues in the spatiotemporal analysis of environmental equity. Social Science Quarterly 77:508. Anderton, D. L., A. B. Anderson, J. M. Oakes, and M. R. Fraser. 1994. Environmental Equity: The Demographics of Dumping. Demography 31:229-248. Bullard, R. D., P. Mohai, R. Saha, and B. Wright. 2008. Toxic Wastes and Race at Twenty: why race still matters after all of these years. Environmental Law 2:371. Cutter, S. L. 1996. Societal Vulnerability to Environmental Hazards. Progress in Human Geography 20:529-539. Daley M. D. and D. F. Layton. 2004. Policy Implementation and the Environmental Protection Agency: What Factors Influence Remediation at Superfund Sites? Policy Studies Journal 32:375-392. Douglas, N. 2008. Evidence of Environmental Justice: A Critical Perspective on the Practice of EJ Research and Lessons for Policy Design. Social Science Quarterly. 89:1153-1174. Evans, J. M, M. Hauer, D. Hardy, and J. S. Pippin. 2014. Demonstration Study at Glynn County, GA. Athens: Carl Vinson Institute of Government. University of Georgia. Evans, J. M, J. S. Pippin, D. Hardy, and M. Hauer. 2014. Social vulnerability and local hazard mitigation planning: Application and evaluation of a “SoVI-Lite” approach for Glynn County, Georgia. Athens: Carl Vinson Institute of Government. University of Georgia. Gallagher, J. and M. Greenstone. 2007. Does Hazardous Waste Matter? Evidence from the Housing Market and the Superfund Program. SSRN Electronic Journal 53: 951-1003. Greenberg, M. and J. Hughes. 1992. The Impact of Hazardous Waste Superfund Sites on the Value of Houses Sold in New Jersey. The Annals of Regional Science 26: 147-53. Johnson, B. L. 1995. Nature, extent, and impact of superfund hazardous waste sites. Chemosphere 31:2415-2428. Juliana, M. 2002. Zoning Law, Health, and Environmental Justice: What’s the Connection? Journal of Law, Medicine & Ethics 30:572. Kirpich, A. and E. Leary. Superfund Locations and Potential Associations with Cancer Incidence in Florida. Statistics and Public Policy 4:1-9. Maantay, J. 2002. Mapping Environmental Injustices: Pitfalls and Potential of Geographic Information Systems in Assessing Environmental Health and Equity. Environmental Health Perspectives Supplements 110:161. Harmetz 40 Mohai, P., P. M. Lantz, J. Morenoff, J. S. House, and R. P. Mero. 2009. Racial and Recai, Aydin and B. Smith. 2008. Evidence of the Dual Nature of Property Value Recovery Following Environmental Remediation. Real Estate Economics 2008:777-812. Stretesky, P. and M. J. Hogan. 1998. Environmental Justice: An Analysis of Superfund Sites in Florida. Social Problems 45:268-287. Switzer, C. and L. A. Bulan. 2002. CERCLA: Comprehensive Environmental Response, Compensation, and Liability Act (Superfund). Chicago, III.: Section of Environment, Energy, and Resources, American Bar Association. Chicago, Illinois. Thayer, M., M. Rahmatian, and H. Albers. 1992. The Benefits of Reducing Exposure to Waste Disposal Sites: A Hedonic Housing Value Approach. The Journal of Real Estate Research 7:265-281. Yandle, T. and B . Dudley. 1996. Reexamining Environmental Justice: A Statistical Analysis of Historical Hazardous Waste Landfill Siting Patterns in Metropolitan Texas. Social Science Quarterly 77:477-492. United Church of Christ’s Commission for Racial Justice. Toxic Waste and Race in the United States. 1987. Commission for Racial Justice. New York, New York. United States Environmental Protection Agency. 1994. Common Chemicals Found at Superfund Sites. Washington, DC.