1. Introduction

This chapter has two objectives. The first is to summarize five major new publicly available databases released by the U.S. Environmental Protection Agency, the U.S. Centers for Disease Control and Prevention (CDC), the U.S. Council on Environmental Quality (CEQ), and the University of Wisconsin. These databases allow researchers and the public to examine a wide variety of demographic, environmental, and public health and service datasets at the census tract, local government, county, and state scales and compare results across scales. The second objective is to illustrate these tools with several current challenging opportunities faced by the U.S. Department of Energy (DOE).

2. Five databases

We begin by describing five databases that allow researchers and members of the public to gain a holistic view of communities in the United States. The U.S. government and others have been collecting data about people and businesses and publishing summaries of these since 1790. However, the 1950 census is the first one that is accessible to people looking for detained granular data. For many decades, researchers, including the first author, had to go to libraries and sort through thick books of data. Obtaining the data required hand copying it and then punching it on special cards before feeding it into a computer for analysis. To say it was a chore to secure the data from the paper files is an understatement. The new databases have created a much more user-friendly world.

2.1 EPA-EJScreen

As part of its obligations under President Clinton’s (1994) Environmental Justice Order 12898 [1, 2] in 2010, the U.S.EPA committed to developing a publicly accessible database that would allow people to see how their area compared to others and for scientists to screen areas for specific characteristics. First released in 2015, it has been updated, most recently in 2023. Users can find information such as proportion of people of color, low income, less than a high school education, unemployment, and the youngest and oldest population. EPA’s data is driven by the assumption that these are the most vulnerable populations. Also, indicators of air quality, automobile traffic density, location of hazardous materials and waste sites, and other environmental indicators are available.

The key strength of this database is having a built-in GIS tool that allows the user to draw circles, triangles, rectangles, and other polygons around a user-selected centroid. Hence, if the user does not choose to use census tracts, cities, or counties, s/he can explore other data collection shapes, which is critical because it allows comparison of places of equal size and shape. The tool comes with a technical manual [2]. However, many people do not have the patience to read all of EPA’s caveats about the data. For example, the air pollution data are only as good as the density of monitors, which are limited in places. Also, even though the number of hazardous waste, storage tanks, and other data are calculated, their presence does not mean that anyone is being exposed. The 2022 update added important information such as broadband access, unemployment, climate change, and medically underserved populations.

2.2 CDC-PLACES

Supported by the Robert Wood Johnson Foundation, PLACES is a CDC database organized around census tracts to help local health officials [3]. In its early days, PLACES focused only on the 500 most populated cities in the United States. Users could obtain data but also were able to view mapped census tract data for the period 2016–2019. Twenty-seven metrics are available mostly for outcomes and prevention, with a few for health risk behaviors and self-assessed health status. The database was expanded in 2020 to provide small area estimates for all counties, incorporated jurisdictions, census tracts, and ZIP codes in the United States. Users can download data as well as view data on an interactive map. Like the EJScreen data, the challenge is to find your way through the various screens and, most importantly, to understand what the data is telling you. This is a powerful set of tools to mix with the data in EJScreen. There will be a learning curve for those not familiar with public health methods, such as age adjustment.

2.3 CEQ-CEJST

In early 2022, the Council on Environmental Quality (CEQ) [4] released an initial version of its Climate and Economic Justice Screening Tool (CEJST) that was developed to provide a uniform definition of disadvantaged communities for federal agencies to use in targeting their allocations under President Biden’s Justice40 covered programs in the areas of climate, clean energy and energy efficiency, clean transit, affordable and sustainable housing, training and workforce development, the remediation and reduction of legacy pollution, and the development of clean water infrastructure. The tool uses datasets and methodologies to identify communities that are economically disadvantaged and overburdened by pollution and underinvestment in housing, transportation, water and wastewater infrastructure, and health care. CEJST relies on income, education, environmental burdens, health, and other economic and environmental factors at the census tract level. The user can compare the percent of the population in each tract relative to a nationally measured metric for each indicator. A community, as represented by a census tract, is deemed by the CEJST to be “disadvantaged” if (i) it is above the CEQ established threshold for one or more environmental or climate indicators and (ii) it is also above the threshold for all of the socioeconomic indicators. The tract must satisfy both criteria to be considered disadvantaged.

CEQ’s approach elicited hundreds of suggestions, including from these authors [5]. When CEQ revised their approach, many of the objections were addressed. The advantage of the CEQ’s approach is users can look directly at census tract data, which is close to a neighborhood of 4000–8000 people in most cases. The main disadvantage is that a census tract can be declared eligible for assistance by crossing one disadvantaged category. In the U.S., this means that about one-third of the U.S.’s 84,000+ census tracts are “disadvantaged.” Realistically, there are insufficient resources to help all of these places, and we suggest using county and statistical tools to group these disadvantaged places (see below).

2.4 University of Wisconsin: County Health Rankings & Roadmaps

Supported by the Robert Wood Johnson Foundation, the University of Wisconsin Population Institute [6] through County Health Rankings & Roadmaps (CHR&R) provides public health-related data for almost every county in the United States. Users can secure county-scale data, guidance on how to use it through emails, webinars, and podcasts. Most importantly, the user can see the overlap between health behaviors, health outcomes, social indicators, and services.

For example, the first author lives in Middlesex County, NJ. In less than one minute, he found that Middlesex is ranked in the 50th–75th percentile in health outcomes in New Jersey as well as in health factors that influence health outcomes. The database has 14 health outcome measures and 60 health factors including behaviors, clinical care attributes of the area, socioeconomic factors, and physical environmental factors. Many of these are used to rank each county in the context of its host state and the United States. The data can be downloaded by year back to 2010. The only important limitation is that the data stops at the county scale, and it does not provide data at the municipal or census tract levels.

2.5 CDC, ATSDR, and The Department of Health and Human Services: Cumulative environmental justice

In August 2022, the Environmental Justice Index (EJI) [7] was released to capture what these several organizations consider the cumulative impacts of environmental burdens, that is, the total impact from a combination of environmental factors. The objective is to provide a single environmental justice score for local areas. The database contains 36 indicators divided into environmental burden, social vulnerability, and health vulnerability.

Cumulative impacts are the total harm to human health that occurs from the combination of environmental burden such as pollution and poor environmental conditions, preexisting health conditions, and social factors. It is the first national, geographic-driven tool designed to measure the cumulative impacts of environmental burden through the lenses of human health and health equity. The authors were briefed about this latest database and tested it with some sample data. However, we have not personally used it in for a major study. Hence, we have no insights based on in-depth work with the database.

3. Two case studies

3.1 Prioritizing places that need EJ program assistance: County scale

The U.S. federal government has a long history of helping distressed places and populations. The Marshall Plan and Japan after World War II are illustrative [8, 9]. Within the U.S., the Appalachia area has benefited from ongoing support [10], as well as federal Superfund and brownfield programs, and grants to help build sewage treatment facilities have made a major difference, and many others.

The Biden Administration created a federally funded Justice40 Initiative [11] that requires federal agencies to provide 40% of the benefits of their climate, clean energy, clean water, affordable and sustainable housing, and several other investments to disadvantaged communities that are marginalized, underserved, and stressed by pollution. The Department of Energy (DOE), indeed every federal agency, faces a challenging opportunity to meet this objective. DOE’s environmental management program is addressing the large environmental legacy of its former nuclear weapons research and manufacturing operations. It is the single largest long-term federal program and its cost was estimated at over $500 in the year 2020 and growing [12]. DOE’s funds are heavily concentrated at a few locations, unlike federal money for social security, Medicare, highways, and other programs that spread dollars across the country. Third, because of the huge cost, DOE sites are perceived as places to find budget reductions.

DOE’s largest sites have been called “state-anchored regions” made dependent and left with a negative legacy [13]. Brauer [14, 15] asserted that DOE-centered regions had a bifurcated labor market and that private industry would not move to these areas because if DOE chose to create new missions the companies would not be able to match DOE’s economic packages, and they would leave to work for DOE. Hooks and Getz [16] observed that from 1970 into the 1990s the large DOE-site regions were unique among federal agencies in not attracting private business to their regions. DOE’s smaller Brookhaven (New York region), Rocky Flats (Denver), Fernald (Cincinnati), Argonne (Chicago), and Lawrence Livermore (San Jose and San Francisco) sites are in or close to major metropolitan regions that can attract new activities to empty space. However, most of these sites have been closed. Savannah River, Idaho, Hanford, Portsmouth, and Paducah are not. While some have argued for economic assistance for these dependent regions, others asserted that steel mills, chemical plants, mines, and other private enterprise facilities had closed, and no such direct aid programs were created for them [17].

The DOE’s large and ongoing operations in erstwhile rural regions represent one important piece of historical context; the second is the argument that DOE owes the large rural site areas additional support. The federal government purchased and/or confiscated large parcels of land at Savannah River, Oak Ridge, Hanford, and elsewhere that had been at least partly occupied. Residents were given a short time to leave. Some relocated outside the area, but others relocated nearby. Economist Milton Russell [18] made a case that the federal government should split allocations to these sites into operations and additional payments to account for the legacy. For example, Aiken (SC) and Benton’s (WA) demographic profiles are different than its neighbors, partly shaped by DOE operations and by its pre-DOE history [19, 20]. In short, what is a reasonable amount of aid for places that were taken over by the federal government leaving long-term environmental and political legacies? In the case of DOE’s large rural sites, federal investments in energy science and engineering, infrastructure, environmental justice, and climate change are pertinent to the ongoing DOE activity and historical legacies.

Given this legacy, we focus the illustrations on answering two questions:

Regarding economic, demographic, environmental, and health indicators, how much difference is there between the core counties hosting DOE’s 10 major sites and the surrounding counties?
How can the data be used to prioritize the counties into having a need for EJ-related assistance?

Ten DOE sites accounted for over $6.5 billion (82%) of $7.9 billion allocated across the complex [21] for environmental cleanup. We focused on the 54 counties surrounding these DOE EM’s sites having the most expensive legacy (Table 1). DOE will be at the sites for the foreseeable future. Accordingly, comparative studies are valuable for tracking the evolution of on-the-ground conditions in the surrounding regions that may influence the DOE’s decisions. The geographical focus of this study are the ten counties that host these DOE site operations, the ten host counties adjacent to them, and the 34 adjacent to the host counties. The DOE core host counties (group 1) are small urban nodes with major operations offices. The DOE has brought high wages to these areas, and we assumed that the ten core host counties would have relatively high wages per employee. However, other characteristics should be associated with these core host functions. One of these is more people, higher density of activity, more traffic, and hence more air emissions. The other ten host DOE counties (group 2) are part of the DOE-site region, but they are not the places where large sums of federal resources are received before being reallocated. The group 3 counties are adjacent to the Group 1 and 2 counties and should be less directly impacted by DOE activities.

Site (n = number of counties)*	Core host county	Adjacent DOE host counties	Counties adjacent to host counties
Current DOE site region (n = 54)	Group 1 10 inner host counties	Group 2 10 other host counties	Group 3 34 adjacent to host counties
1a- Hanford Office of River Protection and 1-b Hanford Operations Office, WA (n = 5)	Benton	Franklin	Grant, Walla Walla, Yakima
2- Idaho National Laboratory, ID (n = 4)	Bonneville		Bingham, Jefferson, Madison
3- Los Alamos National Laboratory, NM (n = 8)	Los Alamos	Rio Arriba, Santa Fe, Taos	Colfax, Mora, Sandoval, Torrance
4- Moab, UT (n = 4)	Grand		Emery, San Juan, Uintah
5- Nevada National Security Site, NV (n = 4)	Nye		Esmerelda, Eureka, Lincoln
6-Oak Ridge, TN (n = 6)	Anderson	Roane	Cumberland, Knox, McMinn, Morgan
7- Paducah KY (n = 4)	McCracken		Ballard, Graves, Marshall
8-Portsmouth, OH (n = 7)	Pike	Jackson, Ross, Scioto	Gallia, Highland, Pickaway
9-Savannah River, SC (n = 8)	Aiken	Allendale, Barnwell	Bamberg, Edgefield, Hampton, Lexington, Orangeburg
10-WIPP, NM (n = 4)	Eddy		Chaves, Lea, Otero

Table 1.

Fifty-four counties included in the study.

*

Source: [22].

More than 100 metrics were available for this analysis. We picked six each to represent economic, environmental, public health, and demographic attributes. Eleven are from the EPA’s EJScreen database, 10 from the University of Wisconsin’s, two from the U.S. Census Bureau, and one from a private source. Please note that we make no assertion that these shed light on every issue related to the sites. Our goal was to pick representative metrics that are consistent with massive literature and available in federal databases. Table 2 lists the 24 metrics used and summarizes the results.

Metric	Major host counties, adjacent counties, and other sets of counties (n = 54)	Group Means*	(95 confidence limits)*
Economic (n = 6)
Unemployment, 2019 (state test value =100)	Major host (n = 10) Non-major host (n = 10) Adjacent (n = 34)	112 133 116	(93,132) (110,156) (104,128)
Home value (state test value =100)	Major host (n = 10) Non-major host (n = 10) Adjacent (n = 34)	81 78 62	(64,98) (53,103) (55,70)
Severe housing cost challenge (state test value =100)	Major host (n = 10) Non-major host (n = 10) Adjacent (n = 34)	85 96 88	(69,102) (83,108) (76,100)
Broadband access (state test value =100)	Major host (n = 10) Non-major host (n = 10) Adjacent (n = 34)	100 89 92	(93,107) (81,97) (87,96)
Wages (test value =100) [values adjusted to median of each state’s counties]	Major host (n = 10) Non-major host (n = 10) Adjacent (n = 34)	130 105 109	(109,150) (89,121) (100,118)
Population changes 2010–2020 (census counts) (test value =0)	Major host (n = 10) Non-major host (n = 10) Adjacent (n = 33)	3.4 −0.9 1.1	(0.5,6.4) (−5.9,4.1) (−1.3,3.5)
Environmental (n = 6)
Pm 2.5 (state test value =100)	Major host (n = 10) Non-major host (n = 10) Adjacent (n = 34)	86 48 68	(30,142) (6,90) (46,90)
Ozone (state test value =100)	Major host (n = 10) Major adjacent (n = 10) Adjacent (n = 34)	78 50 64	(36,118) (10,90) (42,84)
Diesel particles (state test value =100)	Major host (n = 10) Non-major host (n = 10) Adjacent (n = 34)	54 42 44	(28,82) (14,70) (30,58)
Sites with chemical risk management plans (state test value =100)	Major host (n = 10) Non-major host (n = 10) Adjacent (n = 34)	98 96 94	(54,144) (56,136) (72,116)
Less than ½ mile from park (state test value =100)	Major host (n = 10) Non-major host (n = 10) Adjacent (n = 34)	92 58 75	(67,117) (34,83) (61,89)
Food environment (state test value =100)	Major host (n = 10) Non-major host (n = 10) Adjacent (n = 34)	118 119 113	(90,145) (97,141) (102,123)
Public Health (n = 6)
Premature death rate (state test value =100)	Major host (n = 10) Non-major host (n = 10) Adjacent (n = 34)	117 114 115	(93,141) (79,149) (105,125)
Personal assessment of health as fair or poor, (state test value =100)	Major host (n = 10) Non-major host (n = 10) Adjacent (n = 34)	103 129 121	(87,119) (110,147) (114,128)
Adults lacking insurance (state test value =100)	Major host (n = 10) Non-major host (n = 10) Adjacent (n = 34)	94 103 107	(72,116) (77,129) (94,119)
Had flu shot, 2019 (state test value =100)	Major host (n = 10) Non-major host (n = 10) Adjacent (n = 34)	94 88 86	(72,116) (74,102) (80,93)
Covid-19 mortality rate, 2020 (state test value =100) Note only 93 counties had rates	Major host (n = 8) Non-major host (n = 10) Adjacent (n = 28)	99 110 120	(65,133) (67,154) (97,143)
Health factors, 2022 (test value =50, number in two best quartiles =1, others =0) Only 106 counties had rates	Major host (n = 10) Non-major host (n = 10) Adjacent (n = 33)	70 30 58	(35,105) (0,65) (40,75)
Demographic (n = 6)
Low income, % (state test value =100)*	Major host (n = 10) Non-major host (n = 10) Adjacent (n = 34)	103 132 122	(78,134) (110,156) (112,134
Population of color, % (state test value =100)	Major host (n = 10) Non-major host (n = 10) Adjacent (n = 34)	82 112 94	(54,112) (76,148) (78,112)
Linguistic isolation, % (state test value =100)	Major host (n = 10) Non-major host (n = 10) Adjacent (n = 34)	124 130 132	(104,144) (114,146) (124,142)
Less than high school graduation, % (state test value =100)	Major host (n = 10) Non-major host (n = 10) Adjacent (n = 34)	112 134 126	(84,140) (110,160) (114,138)
Population < 5 years old, % (state test value =100)	Major host (n = 10) Non-major host (n = 10) Adjacent (n = 34)	102 100 100	(80,124) (80,122) (90,110)
Population > 64 years old, % (state test value =100)	Major host (n = 10) Non-major host (n = 10) Adjacent (n = 34)	130 124 124	(110,152) (100,148) (112,138)

Table 2.

Comparison of twenty-four metrics in fifty-four counties.

*

Numbers are rounded off to nearest whole number, with the exception of population change.

The numbers in Table 2, with one exception, are standardized to host states, that is, test values of 100. For example, suppose state A’s places had a median unemployment rate of 5%. Study area A5 had a value of 7.5%. By examining all the values in the state, we find that A5’s value places it at the upper quartile in the state. The value in the table for area A5 would be 150 In other words, a value of 150 means that the site had a value 50% higher than the state’s places, whereas a number of 50 means that that site has a much lower value than the host state’s places. The numbers could theoretically range from 0 (lowest in the state) to 200 (highest in the state). We used one-way ANOVA tests and Tukey’s-b post hoc test to measure statistical significance. However, given the small number of cases, the large size of the table, and to simplify the presentation, we present the 95% confidence limits for the numbers. The original reports present all the statistical significance tests [5, 23, 24].

The exception in the table is population change. Colleagues suggested that readers want to see the actual values. Hence, no change (0.0) is the test value.

The answers to question one are not subtle. The ten core host counties manifest stronger economic outcomes than the non-major host counties and the adjacent counties for all six indicators, better health outcomes for five of the six, and less vulnerability for four of the six demographic measures. The exceptions are the premature death rate and the age groupings (>64 and < 5 years). The differences are small (5–25%) and not significantly different in post hoc tests because of small numbers of cases. Yet the consistency of these observations should not be ignored.

Regarding the environmental indicators, the inner core counties have poorer metrics regarding the three air pollutants and the presence of sites with risk management plans. Yet they have a better outcome for distance from parks. Overall, these results are consistent with the idea that the DOE’s inner core counties have measurable benefits that are generally associated with an urban environment. Regarding environmental and social justice, these findings suggest that a good deal of resources are needed outside the host counties, indeed in some cases 100 miles away, which implies management and potentially political challenges.

We created a tool to score each of the 54 DOE counties and classify them into four “need” groups that approximate quartiles (Note that they are not quartiles because there are many ties). We used all 24 metrics in Table 2.

The simplest grouping was to assign each of the 24 variables an equal weight (4.168 out of 100 points). This is reported as the equal weight score (Table 3). We recognize that preferences vary. Accordingly, we created weighted cumulative scores around economics and demographics, which as noted earlier have been the two historic themes for the study areas. The weighting was done by valuing the economics metrics as being twice as important as any of the other metrics. In other words, the six economics metrics received 40 points, whereas the environment, public health, and demographic sets each received 20 points. The demographic weighted results weighted demographics as 40 points, and each of the three other sets were weighted 20 points. Also, we created a “rapid fix” score by focusing on five metrics: seasonal flu vaccination, obtaining health insurance, supporting those with severe housing cost needs, lack of broadband access, and lack of healthy food environment. These five could be addressed with short-term support but are not part of the historic DOE cleanup mission.

Site & County (number of counties)	Equal weighting		Demographic weighting (40%)		Economic weighting (40%)		Rapid fix metrics
	Number		Number		Number		Number
	most needy	least needy	most needy	least needy	most needy	least needy	most needy	least needy
All ten sites (54)	10	13	12	13	12	12	9	20
Hanford (5)	3	0	4	0	3	0	1	0
Idaho National Laboratory (4)	0	1	0	1	0	1	0	2
Los Alamos National Laboratory (8)	0	3	0	3	0	3	0	5
Moab (4)	1	0	2	0	1	0	3	0
Nevada National Security Site (4)	0	2	0	2	0	2	1	0
Oak Ridge (6)	0	2	0	3	0	2	1	4
Paducah (4)	0	1	0	1	0	0	0	3
Portsmouth (7)	3	1	3	1	5	1	3	1
Savannah River (8)	3	1	3	1	3	1	0	2
Waste Isolation Pilot Plant (4)	0	2	0	1	0	2	0	3

Table 3.

Grouping of site areas into most and least in need groups.

Table 3 lists the set of counties most and least in need. Hanford, Portsmouth, and Savannah River have nine of the ten most needy counties. Several observations are in order. One is that the groups do not change much with the weightings. The correlations between the equally weighted scores and the demographic, economic, public health, and rapid fixes scores were r = 0.81, 0.69, 0.83, and 0.73, respectively (all significant at P < 0.01).

Clearly, need is clustered at three of the ten sites and within the sites tends to be in areas adjacent to the major site headquarters.

3.2 Upgrading educational programs

States, counties, cities, townships, and boroughs are the political units assigned to receive federal support. However, programs are typically implemented at neighborhood scales. In the United States, the census tracts are the closest units to neighborhoods. They are created by the U.S. Census Bureau and typically have 1200 to 8000 people. The United States has over 84,000 census tracts. The ten major DOE-EM sites have 770 census tracts in the 54 host counties. Data are available on more than 100 variables at the tract level from EPA, CEQ , US Census, CDC, and other reputable sources.

The empirical test that follows is a pilot restricted to eight counties and 150 census tracts that are around the Savannah River site (SRS). As the previous section shows, SRS had three of the ten counties most in need of assistance. The authors chose SRS because it is a large site (310 sq. mi./802.8 sq. km.) that is geographically surrounded by counties and communities with often markedly different socioeconomic and physical characteristics. Additionally, although active cleanup of the site provides employment and other economic benefits to these counties, SRS is also an active operating facility housing important DOE-Science and national security (NNSA) facilities. Thus, the site offers long-term economic opportunities and benefits to the surrounding region. Likewise, DOE will be reliant on these communities to provide an educated and trained workforce, specialized service business support, and overall political support for its continued presence in South Carolina.

In this illustration, we concentrate on improving educational achievement, but we also worked on efforts associated with expanding broadband access and assisting communities where many of their employees and families live in increasing the resilience of their vulnerable agriculture, buildings, and population to natural disasters [5, 23, 24].

South Carolina’s educational achievement rates are lower than the U.S., ranking 36th of the 50 states (i.e., only 14 states have lower educational achievement rates) regarding people 25 years and older with high school degrees, 35th for those with college degrees, and 31st with post 4-year college graduate degrees. Six of the eight counties surrounding the Savannah River Site report lower educational achievement levels than the state, with only Lexington and Aiken counties having a higher proportion of high school graduates than the state.

As an illustration, we use two indicators to measure public educational base at the county scale: (1) proportion of population 25 and older without a high school education and (2) Niche’s [25] rating of each county school system. The first is derived from the U.S. Census and published in EPA’s EJScreen database [1]. The actual proportion is standardized to the state, with 100 being the state median rate. A number more than 100 means a poorer performance than the state. The second metric is a rating by Niche, Inc. [25], which rates communities on a scale from D+ to A+. The table also includes county-scale data regarding population change and broadband access. Table 4 shows that Hampton, Allendale, Orangeburg, Bamberg, and Barnwell have high proportions of people without a high school degree and low Niche public school quality ratings, as well as marked population decreases and relatively low broadband access.

County	% Less than high school graduation, (state rate = 100)	Average school ratings by Niche	County population change, 2010–2020, %	proportion without broadband access, 2020, %
Hampton	168	C	−14.3	27.5
Allendale	166	D+	−19.4	33.1
Orangeburg	152	C+	−7.6	25.1
Bamberg	144	C+	−12.9	26.5
Barnwell	132	C+	−8.1	26.6
Edgefield	122	B	0.6	22.5
Aiken	108	B+	7.7	16.8
Lexington	82	A	15.4	12.8

Table 4.

Educational achievement and public school ratings grouped by South Carolina County.

(Data organized by county proportion of residents with less than high school education).

Regarding public education, in 2017, the State of South Carolina took over control of Allendale’s public schools because of poor academic performance as it had done earlier from 1999 to 2007 [26]. Behind the terse summary that follows is the reality of massive differences in resources between the urban and rural areas; poor and more affluent areas; and areas with many and few people of color. The state of South Carolina ranked all its school districts [27]. The Superintendent of Education for South Carolina stated: “A millage tax in some of our poorest counties only brings in $20,000, and in our richest counties, it brings in $2 million.” She then noted that it is so difficult to build a new school. At the top of the 10 school districts in South Carolina in need of support were Allendale, two in Bamberg, two in Barnwell, and one in Hampton. The Superintendent added that it would cost $1–2 billion to address the infrastructure needs of South Carolina’s schools. We add that the five counties with the largest gaps in broadband access have the lowest public education rankings in Table 4.

Table 5 examines relationships between low educational achievement and demographic, health outcome, and population burdens for the 150 census tracts in the Savannah River region. The census tracts with the lowest educational achievement and low educational system ratings report higher asthma and heart disease rates, and a higher proportion of low income and people of color. They have higher unemployment rates and are stretched to pay their energy bills.

Variable	Zero-order correlations, R-values with low educational achievement
Energy burden	0.898
Heart disease	0.684
Asthma rate	0.853
Low income population	0.769
Population of color	0.689
Unemployment rate	0.487

Table 5.

Correlates of educational achievement in the study area (data represent calculations from census tracts).

The public school system of this state needs to make a major investment in the infrastructure of its public schools, which might not be acceptable to some community school systems and the state government even if resources were to be offered. However, these data show the need. Earlier, we pointed out that it is to DOE’s advantage to provide educational opportunities that will lead to potential employees and increasing local services for its employees and the site.

A crucial step in this direction would be for DOE to help provide increased science, technology, engineering, and mathematics (STEM)-related training in the counties with existing two-year technical Colleges and four-year colleges in the Savannah River region. This would require a smaller investment and would be focused on helping those having an interest and desire for a higher level education and/or technical training. Unfortunately, geographic disparities in terms of the locations of these higher education opportunities create greater travel burdens on many of those that would benefit the most. Aiken, for example, has a two-year technical college and a branch campus of the University of South Carolina. The Denmark Technical College, a Historically Black College or University (HBCU), is in Bamberg County but must also serve Allendale and Barnwell counties. It is the smallest of the schools serving the Savannah River site community with only 489 students. The City of Orangeburg in Orangeburg County has both a two-year technical college and two four-year universities. Both are HBCUs, one being the oldest private and the other being the oldest public HBCU in the state.

We calculated the number of higher education seats that are available per 1000 residents in each host county. We assumed that the current enrollments represented the number of potential seats available. In several instances, a regional technical college was designated to serve more than the county it was in, so we assumed that the total seats were available to each county.

Please note that our use of these data should not be taken at face value, because many of the residents are not going to go to college because of preference, family cost, and other personal reasons. Second, the seats numbers could change. For example, educational programs are increasingly delivered via the Internet, which may overcome physical seat shortages, assuming the availability of broadband, which as noted above is an issue in these counties. The point of this exercise is to demonstrate that there is a technical educational base in these areas for job training that could be delivered at the area’s 2- and 4-year colleges.

Table 6 shows that Allendale, Edgefield, and Hampton counties have higher available seat rates per 1000 people than Lexington and Aiken counties. Table 5 also shows higher available seat rates tend to be in the counties with lower rates of high school graduation and higher unemployment rates. Assuming that the potential seats estimates are reasonable, they indicate that there is existing capacity for an expansion of job-training and higher education opportunities in Allendale, Hampton, Orangeburg and elsewhere in this area.

County	Potential seats per 1000 residents	Population, 2022	Less than high school education, average of census tracts in each county	Unemployment rate, average of census tracts in each county
Allendale	177	7980	19.1	12.1
Edgefield	162	27,644	16.7	7.4
Hampton	119	18,844	17.9	8.5
Orangeburg	81	83,661	14.5	10.2
Bamberg	77	13,448	20.4	5.0
Aiken	34	175,141	12.4	7.3
Lexington	32	311,950	10.7	5.3
Barnwell	24	20,101	17.2	5.0

Table 6.

Estimated seats per 1000 county residents in DOE-site region Counties.

(Data organized by estimated seats per 1000 county residents).

Reliable and affordable high-speed internet connectivity is essential for economic activity throughout the US. Access to high-speed internet is vital for a diverse set of industries, including education, agricultural production, manufacturing, mining, and forestry, and it acts as a catalyst for rural prosperity by enabling efficient, modern communications between rural American households, schools, and healthcare centers as well as markets and customers around the world [28]. We believe that increasing entry-level higher education and technical training at the two- and four-year colleges in this area requires improved broadband and increased computer access. The Biden Administration has made improving access to broadband a key element of its infrastructure programs [29]. In addition, the COVID-19 pandemic demonstrated the urgency of Internet access, as people relied on remote connections for medical visits, school, and work. Unfortunately, 22.3 percent of Americans in rural areas and 27.7 percent of Americans in Tribal lands lack coverage from fixed terrestrial 25/3 Mbps broadband, as compared to only 1.5 percent of Americans in urban areas. Approximately 182,000 households in South Carolina have no Internet access [30].

The lack of broadband access has disproportionally impacted lower income individuals who could improve their job prospects by taking remote courses to earn their GED or higher education credentials, as well as identify employment opportunities outside their local community. This also implies additional computers and tablets may be needed to take advantage of increased broadband access. The challenge of providing and maintaining Internet access in portions of this region is increased by the shrinking population in the rural and poorer communities noted above in Table 4. The cost of installing the necessary cable over large rural areas exceeds the potential revenue that can be earned by private operators.

The federal and state government will need to provide subsidies in these places. South Carolina’s Office of Regulatory Safety [31] reported on a grant program to increase broadband access in rural Allendale County. Allendale, Bamberg, Hampton, Barnwell, and Orangeburg all have broadband gaps and declining population. Realistically, they need help from external sources to upgrade access and maintain it.

4. Conclusions

The current databases are an amazing improvement over what existed a decade ago. Barring a major change in government desire for information, we are confident that the databases presented here will be substantially updated. We expect the U.S. Congress to pressure U.S. federal agencies to work cooperatively and share software, hardware, and data. We predict that the data delivery systems featured here will be considered obsolete in a decade. A great deal of effort will be focused on quick access and the display of data.

These illustrations highlighted the needs of rural, poor, and underserved areas. Areas southeast of the Savannah River site have limited broadband access and a poor public school educational base yet possess the potential for educational outreach to disadvantaged communities through their existing two- and four-year local colleges. Also, while not addressed in the illustrations, these same areas have much higher natural hazard-related agricultural and population risk rates, along with higher levels of social vulnerability. The databases allow users to quickly go beyond anecdotal information and bring a good deal of credible data to identify and prioritize policy options.

Acknowledgments

Funding for the case studies summarized in this paper was from the Consortium for Risk Evaluation with Stakeholder Participation (funded through the Department of Energy (DOE-FC01-06EW07053) and Vanderbilt University. The authors are solely responsible for the contents of the chapter.

Conflict of interest

The authors declare no conflict of interest.

Pei Fu

Chapters authored

Related collaborators

Carlos Nieto De Castro

Zhi-Cheng Tan

Urs Welz-Biermann

Pei-Fang Yan

Qing-Shan Liu

Da-Wei Fang

Ana P. C. Ribeiro

Salomé I.C. Vieira

João M. França

Carla S. Queirós