Incorporating Consideration for Poor Households into U.S. Environmental 
Protection Agency Affordability Methodologies for Drinking Water Treatment 
 
by 
 
Daniel Beryl Gingerich 
 
 
 
 
A thesis submitted to the Graduate Faculty of 
Auburn University 
in partial fulfillment of the 
requirements for the Degree of 
Master of Science 
 
Auburn, Alabama 
August 3, 2013 
 
 
 
 
Key Words:  drinking water treatment, drinking water affordability, drinking water economics  
 
 
 
 
 
Approved by 
 
Mark O. Barnett, Chair, Malcom Pirnie Professor of Civil Engineering 
 Clifford R. Lange, Associate Professor of Civil Engineering 
Aditi Sengupta, Assistant Professor of Economics 
  
ii 
 
 
 
 
 
 
Abstract 
 
 
 In the 1996 Amendments to the Safe Drinking Water Act, Congress charged the U.S. 
Environmental Protection Agency (EPA) with developing affordability methodologies to 
determine if new drinking water treatment regulations will be affordable to consumers.  In 1998, 
the EPA introduced a methodology based on the median household income (MHI).  When the 
EPA proposed a new method in 2006, it kept the use of the MHI as an appropriate measure of 
household income but introduced additional criteria for identifying economically disadvantaged 
communities (EDCs) to also receive variances.  However, the EPA?s continued use of the MHI 
ignores that poor households are the most burdened by new drinking water regulations.  This 
thesis examines the use of a lower income threshold, the first quartile household income, and 
recommends that water treatment be deemed unaffordable if either the costs are predicted to 
exceed 2.5% of a community?s first quartile household income or if the system will be located in 
an economically disadvantaged community.  This proposed rule would potentially qualify 12,000 
drinking water systems.  The EPA?s definition of EDCs is also examined as the current definition 
fluctuates over the business cycle, producing a situation in which it is harder for systems to 
receive variances during economic recessions, when they are most needed.  Adjustments to the 
EDC criteria that set fixed thresholds based on MHI, poverty and unemployment rate in periods 
of economic growth or long-term averages are proposed to correct this.  EDC definitions that use 
consumption measures, i.e. county food stamp program participation rates and share of 
households spending more than 35% of income, are investigated as well.  
iii 
 
 
 
 
 
 
 
Acknowledgments 
 
 
 I am grateful to my committee chair, Dr. Mark O. Barnett, for his guidance.  I am 
incredibly thankful for the opportunity and direction he has given me in developing my 
intellectual curiosity into the researcher?s tools for designing, performing and communicating 
research and its results.  I would not be the developing scholar that I am today were it not for his 
advice.  I?m also appreciative of my other two committee members, Dr. Clifford R. Lange, for 
his assistance and support throughout the entire graduate school process, and Dr. Aditi Sengupta, 
for her help in learning how to think like an economist.   
 I am also thankful for my colleagues in Civil Engineering.  From my ?lab? mates, 
Hangping Zheng and Ryan Bock; fellow teaching assistants, Qiqi Liang, Xiao Zhao, Kristi 
Mitchell and Yanyan Gong; my lunch partner, Mengyuan Zheng; and my classmates from all of 
my classes, thank you for making the graduate school experience much more bearable and 
sometimes even quite fun. 
 I am also very appreciative of my ?interested? and educated proofreaders, Jessica Brewer 
and Emily Smith, for making sure that my work is readable to someone who hasn?t spent the past 
year researching it.  When it is your turn to defend, I will be more than happy to return the favor. 
 Finally, I am eternally grateful to my parents, Erin Holmes and Samuel Gingerich, for 
setting the bar so high.  The expectations may have been great, but the encouragement has been 
even greater. 
  
iv 
 
 
 
 
 
Table of Contents 
 
 
Abstract ......................................................................................................................................... ii 
Acknowledgments ....................................................................................................................... iii 
List of Tables ............................................................................................................................... vi 
List of Illustrations ...................................................................................................................... vii 
List of Abbreviations ................................................................................................................. viii 
Chapter 1:  Introduction   .............................................................................................................. 1 
 Background   ..................................................................................................................... 1 
 Objectives   ....................................................................................................................... 6 
 Overview   ......................................................................................................................... 8 
Chapter 2:  Literature Review   ................................................................................................... 10 
 Drinking Water Affordability at the U.S. Environmental Protection Agency   .............. 10 
 The Economics of Arsenic   ............................................................................................ 17 
Chapter 3:  EPA?s Revised Metrics and Lower Incomes for Drinking Water Affordability   ... 33 
 Introduction   ................................................................................................................... 33 
 Data and Methods   ......................................................................................................... 38 
 Results of the Affordability Methodologies   ................................................................. 43 
 How do These Evaluations Compare?   .......................................................................... 47 
 Discussion   ..................................................................................................................... 50 
 Conclusions   ................................................................................................................... 56 
 
v 
 
Chapter 4:  Adjusting the Definition of an ?Economically Disadvantaged Community?   ........ 60 
 Introduction   ................................................................................................................... 60 
 Data and Methods   ......................................................................................................... 64 
 Results and Discussion Using Proposed EPA Criteria   ................................................. 67 
 Sensitivity Analysis   ...................................................................................................... 70 
 Alternatives to the Current Ratios and Measures   ......................................................... 76  
 Conclusions   ................................................................................................................... 81 
Chapter 5:  Conclusions and Recommendations   ...................................................................... 83 
 Conclusions   ................................................................................................................... 83 
 Recommendations ........................................................................................................... 86 
References   ................................................................................................................................. 92 
Appendix A:  ORD Sites Included in this Study   .................................................................... 100 
Appendix B:  Cost Effectiveness of Arsenic Treatment for Each System Size Class  ............. 122 
 
 
 
 
 
 
 
 
  
vi 
 
 
 
 
 
 
List of Tables 
 
 
Table 2-1 Estimated cancer risks and avoided cases at or above MCL  ..................................... 19 
Table 2-2 Projected cost and population served by CWS size  .................................................. 26 
Table 2-3 Annual household costs per CWS size class .............................................................. 27 
Table 3-1 Affordability methods considered .............................................................................. 37 
Table 3-2 Arsenic and total water treatment costs in 2000 and 2010 ......................................... 46 
Table 3-3 Expenditure margin .................................................................................................... 56 
Table 4-1 Definitions of EDCs considered ................................................................................. 66 
Table 4-2 Counties and systems qualifying for variances under proposed EPA criteria ............ 67 
Table 4-3 Other regulatory definitions of EDCs ......................................................................... 74 
Table 4-4 Alternative criteria for EDC definitions ..................................................................... 77 
Table B-1 Cost effectiveness for arsenic treatment under low cancer occurrence estimates ... 123 
Table B-2 Cost effectiveness for arsenic treatment under high cancer occurrence estimates .. 124 
 
 
 
  
vii 
 
 
 
 
 
 
List of Figures 
 
 
Figure 3-1 Systems with arsenic treatment cost violating one of the incremental burden   
     standards ................................................................................................................................ 47 
 
Figure 3-2 Systems qualifying for variances based upon proposed EPA criteria in (a) 2000 and    
     (b) 2010 .................................................................................................................................. 48 
 
Figure 3-3 Systems with arsenic treatment in excess of 2.5% of the Q1HI in (a) 2000 and (b)      
     2010........................................................................................................................................ 50 
 
Figure 3-4 Number of different affordability rule scenarios under which a system qualifies for   
     variances ................................................................................................................................ 51 
 
Figure 3-5 Marginal benefits and costs for increasing regulated water quality .......................... 53 
Figure 3-6 Systems qualifying for variances under proposed rule in (a) 2000 and (b) 2010 ..... 58 
Figure 4-1 Counties qualifying for a variance under the EDC rule ............................................ 68 
Figure 4-2 Counties qualifying at various ratios of (a) MHI and non-MA MHI and (b) poverty  
     and unemployment rates for the EDC criteria during the sensitivity analysis ....................... 71 
 
Figure 4-3 Small systems qualifying at various ratios of (a) MHI and non-MA MHI and (b)  
     poverty and unemployment rates for the EDC criteria during the sensitivity analysis ... ?..72  
 
Figure 4-4 Customers potentially affected at various ratios of (a) MHI and non-MA MHI and  
     (b) poverty and unemployment rates for the EDC criteria during the sensitivity analysis ... .73 
 
Figure A-1 ORD sites included in Chapter 3 study..????????????????...100  
  
viii 
 
 
 
 
 
 
 
List of Abbreviations 
 
 
BLS United States Bureau of Labor Statistics 
CB United States Census Bureau    
CWS Community Water System 
EDC Economically Disadvantaged Community 
EEC Environmental Economics Committee to the Science Advisory Board of the U.S. 
 Environmental Protection Agency 
EPA  United States Environmental Protection Agency 
MCL Maximum Contaminant Level 
MHI Median Household Income 
NDWAC National Drinking Water Advisory Council 
ORD U.S. Environmental Protection Agency Office of Research and Development 
Q1HI First Quartile Household Income 
SDWA Safe Drinking Water Act 
SNAP       Supplemental Nutrition Assistance Program
1 
 
 
 
 
 
 
CHAPTER 1 
INTRODUCTION 
BACKGROUND 
 Arsenic is a naturally occurring chemical that is found in groundwaters worldwide.  Well-
studied arsenic hotspots include the Southwestern United States, the Western portions of the 
United States? Rust Belt and Southeast Asia (Amini et al. 2008).  Geochemists, stymied in 
identifying areas with high arsenic concentrations by a lack of field data, also predict hotspots in 
Central Asia, Eastern Australia, Northern Africa and the Sahara and the Southern tip of South 
America, based on pH and electrochemical conditions of groundwaters which are favorable to 
the release of arsenic in drinking water (Amini et al. 2008).  These conditions include high pHs 
and the accompanying oxidizing conditions and low pHs created by mining and industrial 
activity.  The literature reports common arsenic concentrations in groundwaters ranging from 0.5 
?g/L to 5000 ?g/L, with only a few places reporting levels more than 10 ?g/L.  Surface waters 
have values that exceed 1 ?g/L only when affected by anthropogenic contaminants or under 
localized geologic influence (Smedley and Kinniburgh 2002).  This translates to arsenic levels in 
the wells for public drinking water supplies here in the United States ranging from an average of 
0.67 ?g/L in the Southeast to 6.59 ?g/L in the Southwest (Kumar et al. 2010).   
 Elevated arsenic concentrations in drinking water pose health risks to consumers; second 
to radon, inorganic arsenic greatest known risk due to a waterborne chemical  (Brown 2008).  
Arsenic?s health risk to humans was first noted in German scholar Georgias Agricola?s De re 
metallica, published in 1556, which documented that workers who handled arsenical cobalt 
2 
 
(CoAsS) developed skin conditions on their hands (Smith et al. 1992).  A 1637 Chinese 
manuscript called the Tian Gong Kai Wu discussed Pee-song, an arsenic based pesticide and 
fungicide used on the rice fields of China  and that workers that were involved in its production 
couldn?t work for more than two years without losing their hair and becoming ill (Wang and Wai 
2004).  Skin cancer was first linked to arsenic consumption based on studies of patients that had 
taken Fowler?s solution (a 1% arsenate solution used to treat skin conditions) in the late 19th 
century (Hutchinson 1887).  As metal smelting increased during the industrial revolution, more 
and more workers were exposed to airborne arsenic, and the connection between cancers and 
arsenic was quickly realized (Wilson 2001).  In the twentieth century, research on arsenic?s 
effect on drinking water rapidly increased, with studies in Taiwan, Japan, India, South America 
(Chile and Argentina), Europe (United Kingdom and Finland) and the United States, all pointing 
towards a relationship between the consumption of arsenic and increases in various cancer and 
non-cancer health effects (National Academy of Science 2001).  Exposure to arsenic has been 
linked to increased incidence of skin cancer, bladder cancer, lung cancer, liver cancer, kidney 
cancer, nasal cancer, prostate cancer and uterine cancer (Kapaj et al. 2006; National Academy of 
Science 1999, 2001).  Non-cancer effects of arsenic exposure include sensory-motor 
deterioration; behavioral changes; loss of long-term memory; reduced IQ; pregnancy 
complications in women; increased risk of thrombosis, atherosclerosis and other cardiovascular 
diseases; chronic respiratory difficulties; disruption of the endocrine system (including diabetes 
mellitus); a characteristic skin pigmentation change known as Blackfoot Disease; gangrene and 
spontaneous amputation; and reduced liver function (Kapaj et al. 2006). 
 In 1942 the United States Public Health Service, responding to the emerging body of 
evidence linking cancer and arsenic, set a limit for arsenic in drinking water of 50 ?g/L.  The 
3 
 
U.S. Environmental Protection Agency (EPA) adopted this limit on an interim basis in 1975 
when they were given the mandate for drinking water protection as part of the Safe Drinking 
Water Act (SDWA) (Cho et al. 2007).  The EPA faced pressure from the Congress, citizens and 
the courts to finalize an arsenic rule.  When Congress passed the SDWA Amendments of 1996, 
they mandated that the EPA propose a finalized arsenic rule by January 1, 2000 and a final 
regulation deadline of January 1, 2001.  The EPA proposed a 5 ?g/L regulation on June 22, 2000 
and a final regulation of 10 ?g/L was published in the Federal Register on January 22, 2001.  
 The arsenic rule was finalized on the last day of the Clinton Administration.  On January 
20, 2001, following the inauguration of President George W. Bush, his Chief of Staff Andrew 
Card published a memo entitled ?Regulatory Review Plan? which directed government 
regulatory agencies to temporarily withdraw or suspend regulations proposed in the past two 
months, including the arsenic rule (66 FR 7702).  The EPA complied, and on May 22, 2001 
announced a delay of the effective date of the regulation to February 22, 2002 in order to 
accommodate a review of the risks, benefits and costs associated with the arsenic rule (66 FR 
28342).  The final rule had an EPA  projected cost of $205.6 million annually (Abt Associates 
Inc. 2000) and other estimates, such as that by the American Water Works Association Research 
Foundation, projected it would cost over $550 million per year to treat arsenic to 10 ?g/L (Frost 
et al. 2002), Given these high costs, the EPA wanted to ensure that the benefit-cost analysis was 
correct.  In order to accomplish this, the EPA requested that the National Academy of Sciences 
review the health effects and risks associated with arsenic reduction, and directed that the EPA 
Science Advisory Board to review the benefits associated with reduction of arsenic in drinking 
water and the National Drinking Water Advisory Council (NDWAC) to review the costs 
associated with arsenic treatment (66 FR 28342).  The EPA also sought public comment on the 
4 
 
regulation, its benefits and costs and if the rule was justified (66 FR 42974).  The National 
Academy of Science (National Academy of Science 2001), the Science Advisory Board (EPA-
SAB 2001) and the NDWAC (National Drinking Water Advisory Council 2001) released their 
reports late in the summer of 2001, generally finding in favor of the 10 ?g/L rule.  Support from 
the public was mixed.  Many in the water supply industry that commented to the EPA criticized 
the rule as unjustifiably costly given the lack of solid data to support it and the general public 
responded in support of the original rule.  Polling data showed that a majority (56%) of 
Americans criticized the May extension of the arsenic rule, journalists questioned the decision 
and the U.S. House of Representatives voted to reinstate the 10 ?g/L limit (Sunstein 2002).  In 
this environment, the director of the EPA, Christine Whitman, announced the EPA?s decision to 
end the delay and to reinstate the 10 ?g/L rule on October 31, 2001 (Woods 2001). 
 Even after this announcement, the costs associated with the arsenic rule remained an 
issue.  The water treatment community, the EPA and Congress focused on the costs that small 
systems were facing to attain compliance with the arsenic regulation.  The regulatory burdens 
placed on small systems are of particular concern because they have a smaller customer base to 
defray capital costs when compared to larger systems. The EPA has created funding mechanisms 
for small systems (e.g. Drinking Water State Revolving Fund), but these funding sources remain 
under-funded by Congress and are often underutilized by systems (Levine 2012).  As part of the 
1996 SDWA Amendments, the EPA was specifically directed to establish criteria for assessing 
the affordability of drinking water treatment.  If a treatment was found unaffordable, systems 
would be granted permission to use approved variances to reduce the costs.   
 In 1998, the EPA announced it would be using a measure that limits the burdens placed 
on household incomes for affordability determinations, i.e. that if the total costs of water 
5 
 
treatment (including expenditures for the proposed regulation) for the average system within a 
system size class (e.g. 25-100 customers, 101-500, 501-1,000) more would exceed 2.5% of the 
MHI for that class, it would be considered unaffordable (63 FR 42032).  However, as part of the 
appropriations process for 2002, on November 8, 2001, Congress charged the EPA with 
reassessing affordability (Departments of Veterans Affairs and Housing and Urban 
Development, and Independent Agencies Appropriations Act, 2002).  The EPA tasked two 
groups, one of which was the Environmental Economics Committee to the Science Advisory 
Board (EEC) and the National Drinking Water Advisory Council (NDWAC), with reviewing the 
1998 affordability methodology.  The EEC returned with a report, that while supportive of the 
general methodology of the burden measure, found that ?significant levels of inequality both 
within and between systems argued for lower income levels? (the first quartile household 
income, or Q1HI, and 1.5 standard deviations below the mean household income were both 
suggested) and demonstrated the need for a more decentralized approach (U.S. Environmental 
Protection Agency-Environmental Economics Committee (EEC) 2002).   
 The EPA proposed a new, two-part affordability criterion in March of 2006 that 
implemented suggestions from both the EEC and NDWAC.  The first part was a test using 
nationwide averages to determine if a regulation would be unaffordable by comparing the 
compliance costs to the median household income (MHI) within a size class.  If compliance costs 
for a regulation exceeded a set percentage of MHI (0.25%, 0.50% and 0.75% were all 
considered), then the regulation would be determined unaffordable nationwide.  If the test did not 
qualify all small systems for a variance, then a more decentralized approach could be applied by 
the states in the second step.  This second step would be based on written testimony before 
Congress by Scott Rubin, an attorney for the National Rural Water Association, in which he 
6 
 
proposed five criteria for determining affordability.  These five criteria are as follows (Rubin 
2002): 
1. Community MHI less than 65% of national MHI; 
2. Community poverty rate twice the national average; 
3. Community two-year average unemployment twice the national average; 
4. Total water bills in excess of 2.5% of community MHI; or 
5. Total water bills would double as a result. 
The EPA proposed the first three be used by the states in making affordability determinations at 
the local level (71 FR 10671).  Since the proposal of these criteria, the call for affordability 
considerations has quieted, but not completely disappeared, in the EPA and Congress.  The EPA 
hasn?t acted on these criteria since their proposal, and it does not seem like they will address it 
until the next regulation raises concerns about affordability and funding issues.  Problems 
surrounding affordability could arise soon. The EPA?s current drinking water policy agenda 
includes determinations of the contaminants listed in the third candidate contaminant list, small 
systems compliance with the Stage 2 Disinfection Byproducts Rule and a potential hexavalent 
chromium rule. 
OBJECTIVES 
 As the EPA addresses contaminants that are easily and cheaply removed from drinking 
water first, more and more regulations in the future will involve contaminants with costly 
removal techniques.  This poses a problem for the EPA and ensures that it is only a matter of 
time until it creates the ?next arsenic MCL?, an expensive regulation that places an immense 
regulatory burden on small systems.  Affordability concerns and funding issues will reappear on 
the EPA?s agenda when this happens.  A retrospective study of how the EPA determines if a 
7 
 
regulation is affordable, applied to one of the more costly drinking water regulations to date 
would be useful to the EPA in guiding decisions about affordability going forward.  Alternative 
funding mechanisms will have to be developed that can address the inequity of setting 
regulations for all systems based on the economic situations of large systems, as was the case for 
the arsenic MCL (Jones and Joy 2006; Oates 2002). 
 To study these problems, socioeconomic data and information on the cost of arsenic 
treatment were compiled and the affordability methods applied to the data.  Data about income 
levels and distribution, poverty rates and community populations was taken from the U.S. 
Census Bureau?s (CB?s) 2000 and 2010 decennial censuses and the 2006-2010 American 
Community Survey.  Data about unemployment came from the U.S. Bureau of Labor Statistics 
(BLS).  Cost data utilized the EPA?s Arsenic Demonstration Sites Project that included fifty sites 
that installed a variety of technologies throughout the United States.   These compiled data sets 
were then used to perform an affordability determination using the current and proposed EPA 
criteria and using the lower income level recommended by the EEC.   
 Median household income, poverty rate and unemployment data were then converted into 
a GIS database for use in a sensitivity analysis.  First, the number of counties nationwide that 
could potentially qualify for a variance under the proposed socioeconomic status criteria were 
calculated, as well as the share of the systems qualifying for each characteristic.  Second, the 
effect of adjusting the various ratios for the three EDC criteria (both making them more lenient 
and more strict) on the number of counties that qualify were determined.  The adjusted criteria 
were combined together into a few possible combinations to look at alternatives to the EPA?s 
proposed method.  Finally, two alternative measures that represent expenditures rather than 
income were analyzed as potential indicators of the need to grant variances. 
8 
 
 The main objective of this thesis is to review and propose policy solutions to the 
questions of affordability for small systems.  The hypotheses of this study are:  1) the EPA?s 
current and proposed methodologies are insufficient because the current methodology is too 
difficult to trigger and the proposed methodology is not strict enough when it comes to granting 
variances; 2) using the EEC proposed income level of the first quartile household income (Q1HI) 
as the income used for affordability determinations can successfully identify systems that 
deserve variances, without being too lenient; and 3)  that other definitions of EDCs will 
successfully identify these communities.  The specific objectives of this research are: 
1) Application of the EPA?s current and proposed criteria to make a determination about the 
affordability of the arsenic MCL for small water systems to study performance of EPA 
affordability methodologies; 
2) Use of the EEC?s recommendation for a lower income level, the first quartile, as the 
income for the burden measure for an affordability determination of the arsenic MCL for 
small water systems to demonstrate the ability of the first quartile household income ; and 
3) Performance of a sensitivity analysis on EPA socioeconomic status variance criteria for 
EDCs and testing of other definitions of EDCs. 
OVERVIEW 
 A review of the current literature on the arsenic MCL, its cost, the burden it imposed and 
funding problems is presented in Chapter 2 and precedes the results of this thesis, which are 
found in Chapters 3 and 4.  Chapter 3 is written in the format of a draft manuscript currently in 
preparation for submission to the Journal of the American Water Works Association.  This 
chapter discusses the application of the EPA?s current and proposed methods for determining 
affordability to the arsenic MCL.  It also outlines the development and application of an 
9 
 
alternative affordability methodology using an income level lower than the MHI used by the 
EPA.  The chapter concludes with a comparison of the two methods in identifying systems of 
deserving variances for arsenic treatment.  Chapter 4 turns to an analysis of the affordability 
criteria selected by the EPA by determining how many counties qualify for variances and how 
adjusting these criteria affect the final results.  This chapter is also formatted as a draft 
manuscript, and it is also currently in preparation for submission to a peer-reviewed journal.  
Chapter 5 concludes with a summary of the results, suggestions for research going forward and 
implications and suggestions for policy development and analysis when tackling the problems of 
the high cost of drinking water treatment.    
  
10 
 
 
 
 
CHAPTER 2 
LITERATURE REVIEW 
DRINKING WATER AFFORDABILITY AT THE U.S. ENVIRONMENTAL 
PROTECTION AGENCY 
 In debate over the initial passage of the SDWA in 1974, a congressman from Ohio asked 
the following question about a community facing a new drinking water regulation with a water 
system that had used up all its bonding authority: ?How can that community come up with the 
money to meet the standards that some EPA directive might set forth?? (Cong. Rec. 1974).  At 
the time no one had the answer to his question (Pontius 2008).  It would take the EPA a decade 
to include affordability as part of its considerations for SDWA regulations, and then only in 
determining best available technologies for meeting new regulations (Beecher and Shanaghan 
1998).  As part of the SDWA Amendments of 1996, the EPA was explicitly tasked with 
answering this question by addressing concerns over affordability of drinking water regulations 
(Beecher and Shanaghan 1998; Pontius 2008; Tiemann 2010). 
 The concept of affordability was developed by economists working on housing problems 
and draws heavily on the work of two statisticians in the 19th century, Ernst Engel and Herman 
Schwabe, studying how household budgets were spent in an attempt to uncover laws about social 
interactions (Hulchanski 1995).  These researchers developed a measure of burden for various 
items in a household budget, the expenditure-to-household-income ratio.  While this work to 
establish descriptive laws of household expenditures was eventually discredited, it is still 
11 
 
influential in determining affordability for housing (?one weeks pay for one month?s rent? is 
taken from the results by Engel and Schwabe), and the expenditure to income ratio is still used.  
Housing economists have six common uses for this measure:  (1) describing expenditures made 
by households, (2) analysis of these expenditures, (3) for government to help with administering 
subsidies, (4) defining households in need, (5) criteria for selecting households to receive items 
at free or reduced prices and (6) predicting the ability of households to pay for expenditures 
(Hulchanski 1995). 
 EPA drew on the expenditure to income ratio in developing its own affordability metrics 
for SDWA regulations.  The EPA first looked at affordability in the context of providing 
guidance to the state primacy agencies, the state-level departments tasked with overseeing the 
drinking water programs in the individual states under the SDWA.  The EPA identified six 
different characteristics that contribute to affordability for water treatment:  1) household 
affordability, the impact of rate increases on households; 2) financial capacity, how a water 
systems has its finances organized; 3) access to private capital, how easily a system can achieve 
funding through the private market; 4) eligibility for public capital, how easily a system can 
receive funding from public grants or other programs; 5) fiscal conditions, the stresses the local 
government faces in funding water systems (amid other infrastructure demands); and 6) general 
socioeconomic conditions, including income, poverty and unemployment (U.S. Environmental 
Protection Agency 1998). On August 6, 1998, EPA announced its affordability criteria for small 
system variances (63 FR 42032).  This affordability criteria joined the two dozen other 
affordability metrics the EPA had implemented for various programs in offices such as the 
Offices of Water; Solid Waste; Compliance Monitoring; and Policy, Planning, and Evaluation.  
Most of these other criteria were based on a two-step process.  The first step screened out 
12 
 
systems where the additional cost of treatment would not be a significant burden and a second 
step looked at the community?s financial capacity, as was recommended by a panel for the EPA 
Office of Police, Planning and Evaluation (Beecher and Shanaghan 1998; U.S. Environmental 
Protection Agency 1998).   
 In contrast to the other methods at the EPA, the SDWA affordability criteria had only the 
burden screening step, drawing exclusively on the household affordability component of 
affordability by including only a screening to identify systems that would not be significantly 
burdened by new regulations.  The EPA used an expenditure-to-income ratio and said that if it 
exceeded 2.5% the new regulation would be unaffordable.  The EPA would predict the average 
household cost for treatment for a series of water system size classes as part of its economic 
analysis for each new rule and compare these costs with the MHI for the size class (Abt 
Associates Inc. 2000; Baird 2010; Pontius 2008).  This process was simplified by the calculation 
of what is called an expenditure margin.  The EPA calculated that 2.5% of the U.S. MHI to be 
$750 (1998) and that currently an average of $250 (1998) was spent annually on household water 
supply.  The expenditure margin is $500 ($750-$250) and any new regulation cannot cost the 
median system in a size class more than this amount without being unaffordable (Abt Associates 
Inc. 2000; Cooke 2005; Hilkert Colby et al. 2010; Jones and Joy 2006). 
. In November 2001, EPA was tasked with revising its affordability criteria as part of 
Congress?s appropriations process for 2002 (71 FR 10671).  The EPA turned to two different 
groups for advice on how to review their rules on affordability:  the EPA?s Environmental 
Economics Committee of the Science Advisory Board (EEC) and the National Drinking Water 
Advisory Council (NDWAC). 
13 
 
 The EEC was charged with examining several questions by the EPA including:  questions 
about the basic method, appropriate income limits and shares, the level at which the 
determination should be made and whether there should be a separate affordability method for 
groundwater-sourced systems and surface-water-sourced systems.  The EEC concluded that the 
basic method the EPA was using was appropriate as it was practical while still addressing 
questions over efficiency and equity.  The EEC recommended that the EPA consider an income 
level below the MHI because of inequality between water system customers and between the 
systems themselves.  The EEC also concluded that the 2.5% of MHI threshold was too high and 
that it should be lowered with the EPA providing clear guidance to the states.  Other conclusions 
the EEC made were to recommend the EPA consider a more regional (or ideally, a local) 
approach with any national method serving as a screen and that the affordability methodology 
encourages the consolidation of smaller systems together should be considered  during the 
affordability process (and encouraged by the agency to increase the economic efficiency of water 
supply) (U.S. Environmental Protection Agency-Science Advisory Board 2002). 
 The NDWAC Work Group on the National Small Systems Affordability Criteria 
recommended a much more dramatic departure from the EPA?s original methods.  The NDWAC 
recommended dropping the expenditure margin approach the EPA had adopted and implement 
an incremental approach.  Rather than looking at how any new regulations would add to the 
burden systems and households already face, the NDWAC recommended that the EPA only 
consider regulations one at a time.  If the regulation exceeds 1% of MHI, then it should be 
determined to be unaffordable and the EPA should allow the granting of variances.  An 
incremental approach was recommended because it could avoid the ?adding up? problem, the 
problem that occurs when later rules are found unaffordable because of earlier, expensive 
14 
 
regulations that consumed the expenditure margin.  The NDWAC did not completely ignore 
cumulative cost burdens in their methods, but left these considerations to the state primacy 
agencies that would have better access to this data.  The 1% MHI threshold was selected because 
it represented a doubling of the water bills in 2003 (based on the 2000 EPA Community Water 
System Survey).  This new method was not unanimously endorsed by the working group?s 
membership; a representative from the National Rural Water Association believed that the new 
method was still not sufficiently strict enough to identify systems (National Drinking Water 
Advisory Council 2003). 
 The EPA considered the EEC and NDWAC proposals in revising its affordability criteria.  
The NDWAC?s recommended incremental approach addressed many of the EPA?s significant 
concerns.  Because of this the EPA adopted the NDWAC?s approach in March 2006.  The EPA 
then announced that it was seeking feedback on using an incremental burden approach with a 
threshold of either 0.25%, 0.50% or 0.75% of MHI.  The EPA announced that it was also 
considering adding a second step in the affordability process.  The purpose of this second step 
would be to identify economically disadvantaged communities (EDCs).  EDCs were mentioned 
in Section 1452 of the SDWA as an area ?of a public water system that meets affordability 
criteria established after public review and comment by the State in which the public water 
system is located?. This language was considered too broad to be of use to state primacy 
agencies and before this 2006 proposal, the EPA had not defined an economically disadvantaged 
community (EDC) in a regulatory context for the entire nation.  In an evaluation of the Drinking 
Water State Revolving Fund?s impact on EDCs, the EPA referred to a general definition in 
socioeconomic characteristics like MHI, the poverty level and unemployment rates.  However, 
this definition did not impact regulatory enforcement (U.S. Environmental Protection Agency 
15 
 
2000).   The EPA identified three potential criteria that could be used for this purpose:  i) a 
county with a MHI of less than 65% of the national average MHI, ii) a county with two-year 
average unemployment rate that is twice the national average and iii) a county with a poverty 
rate that is twice the national average.  Systems in counties that meet one of these criteria could 
be granted variances even if the national screen found the regulation affordable (71 FR 10671).  
These criteria draw on the ability of households within the system to pay for drinking water.  
Ability to pay depends heavily on the level of a household income and employment, so these 
proposed measures would at least measure the correct factors that contribute to affordability 
(U.S. Environmental Protection Agency 1998). 
 The criteria used by the EPA in identifying EDCs were developed by Scott Rubin at the 
National Rural Water Association.  After the EPA proposed the expenditure margin approach 
based on MHI, Rubin studied the relationship between the poverty rate in an area, a factor crucial 
in systems being able to get funding from their customers, and an area?s MHI.  He found that 
there is only a moderately strong relationship between the two variables (R2=0.7609), and that 
the poverty rate can range anywhere from 0% to over 20% when a community?s MHI is equal to 
$30,000.  The correlation between MHI and the percent of households living in poverty areas is 
even worse with an R2 of 0.4707 (Rubin 2001).  Rubin examined various criteria that states and 
federal agencies use to identify EDCs.  Many states selected criteria that drew on MHI 
comparisons, and the state of Washington used a criterion based on unemployment rate.  The 
federal agency criteria that Rubin analyzed included the Department of Commerce?s Economic 
Development Administration (average twenty-four months of unemployment greater than 1% 
above the national average, per capita income that is no more than 80% of national average 
income), the Treasury (30% of the population live below the poverty line, average 
16 
 
unemployment rate 1.5 times the national unemployment rate) and the Department of 
Agriculture?s Rural Utilities Service (MHI of 80% of the state?s non-metropolitan MHI, or an 
MHI below the poverty level) (Rubin 2002). 
 Even with the EPA?s proposed changes to the affordability guidelines, some still have 
found issue with the EPA incremental criteria.  Fred Pontius (2008) argues that the EPA?s 
affordability criteria still need work and that they should remain focused on the issue of 
household affordability.  Pontius also discusses how the Small Business Administration had 
nominated the EPA?s affordability policy for revision under the Regulatory Review and Reform 
Initiative in 2007 and as a top ten rule for revision in February of 2008.  Gregory Baird shows 
that in 2020 and 2030, due to dramatic increases in water rates to keep up with the need for 
infrastructure maintenance and upgrades, water and wastewater treatment will take up 5% of 
household income (Baird 2010).  The 2013 Infrastructure Report Card released by the American 
Society of Civil Engineers in March of 2013 highlighted both the need for water revenues to 
increase to pay for infrastructure maintenance and increasing regulatory burdens (American 
Society of Civil Engineers 2013).  If this is to be the case there is little reason to believe that 
affordability will be going away soon. 
 A second line of criticism the EPA?s affordability metrics is subject to is that they do not 
capture the factors that really affect whether water treatment will be affordable or not.  While 
MHI, poverty and unemployment rate can serve as indicators of a household?s income resources, 
they do not serve as a measure of their actual behavior when it comes to consumption.  
Households do not always spend in accordance with their income, especially when they 
experience negative shocks they expect to be temporary (e.g. unemployment) and can tap into 
other resources like savings or access to credit (Citro and Michael 1995).  Furthermore, annual 
17 
 
incomes as reported by the Census Bureau (CB) do not match up well with a measure of the total 
income of the nation produced by the Commerce Department.  For example, in 1996 the 
difference between the CB?s total national income and Commerce?s was two trillion dollars or 
about $20,000 per household.  A better measure of the resources a household has available to 
them and how they spend their resources can be determined by studying their consumption 
patterns and behavior (Citro and Michael 1995; Meyer and Sullivan 2003; Rector et al. 1999).  
Poor households are usually identified by this method as those experiencing material hardships 
in funding the necessities of life.  These material hardships are often categorized into four main 
classes, food insecurity, housing insecurity, inability to obtain medical care because of cost and 
inability to pay household bills (Heflin et al. 2009).  This option is more advantageous to 
researchers because it considers hard and fast options that allow researchers to evaluate the 
actual choices made by consumers who can?t afford everything they need (Pilkauskas et al. 
2012).  The general public and policymakers also appear to favor material hardship measures 
since officials and the populace concern themselves not with income, but if consumers can afford 
their necessities (Mayer and Jencks 1989).  These advantages make measures of material 
hardship more desirable than the income definitions of need. 
THE ECONOMICS OF ARSENIC 
 Arsenic was first regulated in the 1940s by the U.S. Public Health Service.  The limit was 
set at 50 ?g/L and was designed for arsenic not as a carcinogen, but as a toxin (Scheberle 2004).  
In the 1960s, the U.S. Public Health Service recommended the standard be set at 10 ?g/L, but 
never acted to adopt this lower standard (Reimann and Banks 2004).  It was the 50 ?g/L rule that 
the EPA adopted on an interim basis in 1975 following the passage of the SDWA (Tiemann 
2007).  It would take evidence that arsenic was a carcinogen, and the accompanying assumption 
18 
 
that there is no ?safe? amount of arsenic below which there is no observable adverse effects, for 
regulators to revise the rule downward. 
  The evidence that arsenic has negative health consequences has existed since the 16th 
century (Smith et al. 1992).  The first link between arsenic exposure and cancer was made by 
Hutchinson in 1887 as he observed skin tumors developing on patients who had used arsenic-
containing medications (Hutchinson 1887).  Since the 1960s, epidemiologists have linked arsenic 
to skin, bladder, lung, kidney, nasal, liver, prostate, urinary tract and uterine cancers (Bates et al. 
1995; Chen and Wang 1990; Chen et al. 1985; Chiou et al. 1995; Cuzick et al. 1992; Guo et al. 
1997; Hopenhayn-Rich et al. 1996; Mushak and Crucetti. 1995; Tseng et al. 1968; Tseng 1977; 
Tsuda et al. 1995; Wu et al. 1989).  Non-cancer effects of arsenic include damage to the 
peripheral sensory nervous systems, reduced intellectual functions in children, increased rate of 
miscarriage, lowered birth weight and increases in other birth defects (if the mother consumes 
arsenic), enlargement of the liver, blood clots and associated cardiovascular conditions, reduced 
blood supply to the heart (ischemic heart disease), thickening of the carotid artery, chronic cough 
and bronchitis, disruption of hormone production, type II diabetes and skin discoloration in the 
extremities (Blackfoot?s Disease) (Kapaj et al. 2006).  For a more in-depth (but out of scope of 
this thesis) review of studies investigating the health effects of arsenic, the reader can turn to the 
reviews by the National Research Council conducted during the arsenic regulatory process 
(National Academy of Science 1999, 2001)). 
 The EPA took data from epidemiological studies of cancer risks in developing the risk 
assessment models they used to calculate the benefits of reduced exposure.  The data for risk 
assessment came from a study of the arsenic risk faced by villagers in Taiwan obtaining drinking 
water from wells with high arsenic concentrations (Morales et al. 2000).  This data was paired 
19 
 
1From Exhibit 5-4(c) (Abt Associates, 2000). 
2From Exhibit 5-9(c) (Abt Associates, 2000). 
with EPA data on the exposure faced by American consumers.  The EPA was the able to predict 
the cancer risks faced by American consumers.  These predicted risks are presented in Table 2-1 
(Abt Associates Inc. 2000).  For reference an ?acceptable? cancer risk is set by the EPA as  
Table 2-1 Estimated cancer risks and avoided cases at or above potential MCLs 
MCL (?g/L) 
Cancer Risks for U.S. 
Populations At MCL1 
Annual Total Cancer Cases 
Avoided per Year2 
3 0.11-1.25 x 10-4 57.2-138.3 
5 0.27-2.02  x 10-4 51.1-100.2 
10 0.63-2.99  x 10-4 37.4-55.7 
20 1.1 ? 3.85  x 10-4 19.0-19.8 
 
  
1x 10-4 (Gurian et al. 2001b).  These risks were applied to the exposed populations in order to get 
the number of cancer cases prevented (Abt Associates Inc. 2000).  Other risk estimates show 
much greater risks associated with arsenic concentrations.  The California Office of 
Environmental Health and Hazard Assessment (COEHHA) determined a unit cancer risk, the 
risk of developing a cancer per additional unit of carcinogen exposed to, of 2.7x10-4 (?g/L)-1 and 
the National Research Council (NRC) unit cancer risk was 2.3x10-4 (?g/L)-1, ten times greater 
than the EPA?s unit cancer risk of 2.41x10-5 (?g/L)-1 (Brown 2008). 
 Putting a monetary value to prevented cancer cases requires another pair of values, 
willingness to pay and the value of a statistical life.  The willingness to pay is the amount a 
consumer would be willing to pay to avoid having to experience a risk (in this case the risk of 
cancer).  The value of a statistical life is the willingness to pay to avoid a fatal risk, divided by 
the risk (Kolstad 2000; Viscusi et al. 2000).  The EPA obtained a willingness to pay to avoid a 
non-fatal case of cancer of $607,162 (1999; from Viscusi, Magat and Huber?s study of 
willingness to pay to avoid chronic bronchitis (1991)) and a value of a statistical life of $6.1 
20 
 
million (1999; a meta-analysis performed by the EPA of wage risk-premiums identified that 
most workers exposed to fatal risks value their life at a median value of $6.1 million) (Abt 
Associates Inc. 2000). 
 The EPA identified thirteen different treatment technologies that could be used to remove 
arsenic from a drinking water supply.  The EPA ran a Monte Carlo simulation in order to 
determine the national costs based on a decision tree that considered the necessary removal 
efficiency and costs.  These costs were summed over all of the systems to develop national 
treatment costs.  State regulatory enforcement costs were also calculated.  Treatment and 
regulatory costs were added together to come up with national costs (Abt Associates Inc. 2000). 
THE ACCURACY OF THE COST-BENEFIT ANALYSIS 
 When establishing the arsenic MCL at 10 ?g/L, the EPA used its discretionary authority 
to establish a MCL higher than the minimum feasible level of 3 ?g/L and well above the health 
protective concentration of 16-40 ng/L (COEHHA estimated a health protective concentration of 
4 ng/L and NRC 1.5 ng/L) (Brown 2008).  This was done because the EPA found that the 
benefits of a 3 ?g/L MCL did not outweigh the costs, but (when unquantifiable benefits, i.e. not 
benefits from lung or bladder cancers, were considered) that a 10 ?g/L rule would meet this 
requirement.  The costs of implementing the 10 ?g/L rule was found to be $205.6 million (1999) 
annually with an upper-bound estimate of quantifiable benefits  of $197.7 billion (1999) 
annually.  The benefit-cost ratio for the 10 ?g/L rule was found to be approximately 0.96 (the 
ratio for the 3 ?g/L rule was 0.62; 5 ?g/L rule, 0.75; and 20 ?g/L rule, 0.98) (Abt Associates Inc. 
2000).  While the 10 ?g/L failed the quantifiable benefit-cost test, many prominent 
environmental economists have written in support of the idea that agencies should not have their 
21 
 
hands tied by failed benefit-cost calculations when other factors can come into play (Arrow et al. 
1996). 
 This finding of arsenic treatment?s affordability was highly controversial and received 
much criticism in the literature.  The most prominent critique among these was that of Burnett 
and Hahn (2001).    Burnett and Hahn found several questionable assumptions underpinning the 
EPA?s analysis, including faulty assumptions about risk, using upper-bounds instead median or 
best estimates, and a failure to take into account that cancers occur in the future, which suggested 
to Burnett and Hahn that discounting of monetized health benefits should be performed.  By 
correcting for these, Burnett and Hahn adjusted the value of the statistical life to $1.8 million.  
Using a sublinear dose-response curve and reasonable estimates for non-quantifiable benefits, 
Burnett and Hahn develop a best estimate of 11 lives saved for a monetized benefit of $20 
million and a cost effectiveness per life saved of $65 million per life.  Burnett and Hahn 
concluded that the 10 ?g/L rule was unsupported by a cost-benefit analysis.  Some have found 
flaws with the idea of discounting the value of a statistical life, most notably Lisa Heinzerling 
(2002) who pointed out that the EPA values the reduction in risk (and not the loss of life), and 
that this reduction in risk occurs when the regulation is promulgated (and not twenty years 
hence) and Richard Wilson (2001), who argues that if discounting is to take place it should be 
done not to lives lost but years of life lost. 
 Another set of authors, Frost et al. (2002), take a different track in analyzing the EPA?s 
cost-benefit analysis.  Although the authors criticized the EPA?s application of the scientific data 
on the risks associated with arsenic, their main critique was with the costs associated with 
bringing water systems up to the arsenic MCL.  The authors used the cost data from the 
American Water Works Association Research Foundation (AWWARF), which was significantly 
22 
 
higher than the EPA?s data, from $55 million (1999) for a 20 ?g/L rule to well over a billion 
dollars a year for a 3 ?g/L rule.  With these costs a 10 ?g/L rule would cost upwards of between 
three and ten million dollars for each year of life saved, well above the EPA?s calculated value of 
about $370,000 (1999).  This led the authors to conclude that the arsenic rule was severely 
economically inefficient at an MCL of 10 ?g/L. 
 Modeling work by Gurian et al. (2001a) also examined how much the arsenic regulation 
would cost.  The authors used the same cost curves as Frost et al. (2002) and incorporated this 
into their treatment technology selection algorithm.  This algorithm was then used to model if the 
selection of arsenic removal technology would be necessary, and if necessary, which treatment 
technique would be used for the systems in the Safe Drinking Water Information System (this 
universe of systems was the same one used by the EPA in doing their estimates, albeit with a 
different set of cost curves).  In doing so the authors calculated that a 10 ?g/L rule would cost 
nationwide $294 million (1997) per year.  The authors also performed an uncertainty analysis on 
their results by propagating through the errors resulting from the various assumptions and 
variations in their data.  With this uncertainty analysis performed, their low estimates are similar 
to the EPA?s high estimates and their high estimates are similar to the AWWARF?s low 
estimates.  Gurian et al. (2001a) predict that in total Americans nationwide would be exposed to 
300 g of arsenic less each day, and there would be 55 fewer deaths annually from bladder cancer 
each year.  Considering only fatal bladder cancer, the arsenic rule would save a life at the price 
of about $17 million each.  This is almost three times higher than the value of a statistical life 
used by the EPA, and so the arsenic rule was inefficient. 
 An important assumption in these algorithms is that they are based on systems selecting 
the least-cost method for arsenic removal; however, this may not be the case.  First, such an 
23 
 
assumption requires perfect knowledge of the true long-term costs faced by water supply systems 
when they make their decisions related to arsenic removal, which is highly unlikely to be the 
case.  Second, many of the more conventional arsenic removal techniques (e.g. filtration in a 
traditional water treatment system, coagulation/flocculation, activated alumina adsorption) pose 
the possibility of assisting systems in meeting not only primary MCLs but also secondary MCLs 
(for contaminants like iron, manganese, sulfate and nitrate) and other goals like hardness 
reduction.  If this is the case for a particular plant, it may select a more expensive treatment 
option and just pay a ?premium? for the assistance in meeting the optional targets.  Source water 
composition can also drive systems to select more expensive technologies, since waters high in 
sulfate are unsuitable for arsenate removal by anion exchange due to sulfate?s higher selectivity 
for conventional anion exchange resins.  Finally, considerations of residuals management can 
force utilities to select more expensive treatment techniques in order to avoid the burden of 
disposing potentially hazardous wastes (Chen et al. 1999). 
 Cass Sunstein (2002) developed a third argument against  the EPA?s cost-benefit analysis 
of the 10 ?g/L arsenic rule, an argument that questioned neither the EPA?s cost calculations nor 
benefit estimates.  Sunstein criticized cost-benefit analysis as a whole for its creation of an 
illusion of certainty, when in fact the best that can often be done is creating a range of potential 
values for the benefits.  Sunstein does not believe that this should be used to fundamentally 
undermine the use of cost benefit analysis by regulators, but kept in mind by policy practitioners, 
legislators and the judiciary should not view the results of benefit-cost analyses as highly 
accurate, simply because numerical values are assigned to benefits and costs.  For the arsenic 
rule, Sunstein found that the benefits could range anywhere from $0-$560 million (1999) 
depending on a whole host of assumptions made in interpreting the risk information available.  
24 
 
Because of this wide benefits range, Sunstein said it is incorrect to believe the EPA?s 10 ?g/L 
rule was the right or wrong thing to do based on the science it had at the time.  There simply was 
no way to finalize the benefits in a conclusive way. 
 The biggest factor in explaining the variability of the cost estimates is the treatment 
technology cost curve utilized by the different studies (EPA, AWWARF and Gurian et al.).  
When they adjusted their national cost model to use the treatment technology cost curves from 
the EPA and AWWARF studies, Gurian et al were able to obtain a reasonably close match with 
the costs predicted by the other studies.  The primary reason for the difference in technology cost 
curves comes from differences in residual management used by the different studies.  The EPA 
study assumes ion exchange brines can be disposed of simply by use of sanitary sewers.  The 
AWWARF study required that arsenic containing brines would have to undergo arsenic 
removing coagulation, evaporation and then disposal in a landfill.  The difference between these 
two methods can lead to a doubling or even tripling of the costs for the anion exchange process.  
For activated alumina process, AWWARF included regeneration and residual disposal costs 
(which contributes to 20-40% of total costs for an activated alumina process) whereas EPA did 
not.  Gurian et al.?s model (based on the National Arsenic Occurrence Survey) had questionable 
assumptions about acceptable disposal processes, but Gurian et al. (2001b)believed it represented 
a happy medium as far as costs were concerned.  A second (and of smaller importance) 
explanation for the variability in the cost estimates is a 20-30% variation in the number of 
systems that will be required to adopt arsenic treatment.  In general the EPA predicted the fewest 
systems would need to install arsenic treatment and AWWARF predicted the most - with Gurian 
et al. (2001b) striking a middle value  
25 
 
 Wilson (2001) argued that the EPA had a history of overpredicting compliance costs and 
so that their cost estimates were higher than they would be in reality.  A report by the EPA in 
2012 confirmed that the EPA did overpredict the costs associated with the arsenic rule (National 
Center for Environmental Economics (NCEE) 2012).  This overprediction was due to the 
innovation in the technology that would become available to remove arsenic from drinking 
water.  Iron adsorption was developed after the Best Available Technologies (BAT) were 
identified and included in the cost estimates.  Iron adsorption was shown to be a cheaper 
technique than many of the previously identified options, which led the EPA to designate it as a 
BAT.  This designation led to its widespread adoption, which contributed to the development of 
cheaper adsorbents that followed.   The use of iron adsorption for arsenic removal by water 
systems meant that costs had been overestimated since it was not considered during the cost 
analysis (NCEE 2012). 
 EPA has also taken a look at the accuracy of its cost-benefit analysis ? with a focus on 
the costs, not just for the arsenic rule, but for all its rules.  In general, the EPA has concluded that 
it overestimates the costs for its rules (National Center for Environmental Economics 2012).  
This error is not exclusively due to the regulators making errors as part of their simulations, 
incomplete information about regulations, or from accurate cost information being withheld by 
industry, but a mathematical principle known as Jensen?s inequality.  The ratio of ex ante to ex 
post costs for regulations is a convex function when ex ante costs are held constant and ex post 
costs allowed to vary.  Applying this to the situation of regulatory costs provides an explanation 
for the overestimation of the costs related to regulations (Simpson 2011). 
 
 
26 
 
        1 From Abt Associates Inc. (2000). 
           2 From U.S. Environmental Protection Agency (2002). 
THE COSTS OF ARSENIC FOR DIFFERENT SYSTEM SIZES 
 Lost in the focus on the big picture of arsenic?s total costs and total benefits to society 
was the important question of how these costs and benefits would be distributed.  The 
distribution of costs and benefits to society would become a major flash point for critics of the 
arsenic regulation shortly after its finalization in October of 2001.  The EPA?s analysis analyzed  
System Size Class Projected Costs 
(in millions $, 1999)1 
Percentage of U.S. 
Population Served2 
<500 $18.4 1.4% 
501-3,300 $33.2 5.3% 
3,301-10,000 $27.9 8.0% 
10,001-100,000 $72.9 26.5% 
>100,000 $40.8 58.8% 
the distribution of costs to community water systems (CWS) of various size classes.  Table 2 
shows the projected cost for various size classes (Abt Associates Inc. 2000) as well as the share 
of the U.S. population served by CWSs that are served by a system of that size class (U.S. 
Environmental Protection Agency 2002a).  As can be seen, there is little correlation between 
total costs for a size class and the number of people served by the size class.  This has significant 
implications for household water bills, as presented in Table 3.  A customer in the smallest 
system pays over ten times the average cost for arsenic treatment and almost 400 times as much 
as customers in the  largest systems (Abt Associates Inc. 2000).  The difference between these 
Table 2-2 - Projected costs and population served by CWS size 
27 
 
1From Exhibit 6-17 (Abt Associates Inc. 2000). 
costs caused significant controversy over whether or not customers in small systems could afford 
arsenic treatment for their water supplies. 
Table 2-3 ? Annual household costs per CWS size class  
System Size Class Costs per Household 
$(1999)/yr.1 
<100 $326.82 
101-500 $162.50 
501-1,000 $70.72 
1,001-3,300 $58.24 
3,301-10,000 $37.71 
10,000-50,000 $32.37 
50,001-100,000 $24.81 
100,001-1,000,000 $20.52 
>1,000,000 $0.86 
All Systems $31.85 
 
 
 Oates (2002) interpreted these results to argue for a decentralized regulation process.  
Given that arsenic treatment is a ?local public good?, in that the level of the good (protection 
from arsenic) is the same for all consumers, that all the consumers are relatively concentrated in 
a single area and that the distribution of costs are wildly divergent, it makes sense for regulations 
to be established at a local level.  For consumers in large systems paying less than a dollar a year 
for the relatively small risk reduction is relatively rational, but Oates argued that it does not make 
28 
 
sense for customers in a small system to pay well over $300 annually to experience the same 
small risk reduction.  As such, Oates recommended that the EPA should have kept the MCL at 
50 ?g/L and set a recommended limit at 10 ?g/L for systems that could afford it. 
 Analyses of the costs of arsenic treatment conducted even before a 10 ?g/L rule was 
proposed also found that a significant proportion of the national costs would be faced by small 
systems.  Simulations by Frey et al. (1998) concluded that for a 10 ?g/L rule, small systems 
would pay (on an average run of the author?s model) 38.5% of the $708 million national annual 
compliance costs.  Small groundwater systems would pay between 30 and 40% of the total 
national compliance costs for arsenic MCLs between 0.5 and 10 ?g/L, with large groundwater 
systems facing 39-45% of the national compliance costs.  
 Using their treatment cost model, Gurian et al. (2001b) also analyzed the impact of the 
arsenic regulation on small water systems and found that about 1.5% of water systems will face 
total water bills in excess of 2.5% of MHI (serving 0.03% of the US population).  If one assumes 
that a system that cannot afford treatment will not undertake it, the number of systems taking no 
treatment action rises from 96.4% to 97.8% of systems with a corresponding 34% reduction in 
aggregate costs and a 4% reduction in aggregate benefits.  The authors examine several 
techniques that could increase compliance.  The first is subsidizing treatment costs; however, this 
method was disregarded as an undesirable disincentive to increasing efficiency in the water 
supply market.  They also examined the potential for point-of-use (POU) technology to assist 
and found that not only does allowing POU technology reduce national costs by 19%, but that 
aggregate benefits are increased by a slight 0.2%.  This increase in benefits will probably be 
reduced due to imperfect maintenance on individual POU units. 
29 
 
 Even with the passage of the arsenic rule, size differences have continued to play a role in 
compliance issues.  A study of the violations that was conducted in fiscal year 2011 found that 
no violations of the arsenic MCL were found in very large systems (serving greater than 100,000 
people) (Rubin 2013a).   A statistical analysis of the remaining systems found that while a larger 
percentage of very small systems (<100 customers) reported violations of the arsenic MCL, they 
were statistically no more likely to experience violations than other systems serving less than 
100,000.  A further analysis of states with elevated concentrations of arsenic also failed to find 
that very small systems were statistically more likely to have violations of the arsenic MCL 
(Rubin 2013a).  
 A 2013 study by McGavisk et al. (2013) analyzed a possible explanation for whether or 
not a system was in compliance with the arsenic rule by 2011, the MHI of the community.  The 
authors used an ANOVA test procedure to determine if there was a statistical difference in the 
community?s MHI between systems that were in compliance, were not in compliance and 
systems that closed during the interval between the finalization of the arsenic rule and the study 
period.  They concluded that there was a statistical difference between the compliance statuses: 
those systems that were compliant had higher incomes per capita than those systems that were 
not in compliance. Systems that closed and replaced in the study period were in areas with higher 
incomes per capita than systems that were either compliant or not compliant.  These differences 
were explained by arguing that systems in poorer areas would be unable to raise the funds to 
install treatment and those in richer areas could close one treatment plant and open another 
elsewhere that used a different source that would be in compliance.  The authors warned against 
treating MHI as the only factor in compliance and identified several case studies that illuminated 
MHI as only one factor in system compliance, other factors including operator inability to 
30 
 
upgrade systems, the community?s resistance to pay for system upgrades and miscommunication 
between regulators and those designing the systems. 
 Cho et al. (2007) analyzed the willingness to pay for consumers in several of Minnesota?s 
rural communities in order to bring their water supply into compliance with the arsenic rule. 
Using survey data from a contingent valuation study, the authors determined that consumers with 
drinking water contaminated only occasionally containing arsenic in excess of 10 ?g/L were 
willing to pay between $8 and $9 each year, while customers with water that regularly exceeded 
10 ?g/L of arsenic were willing to pay between $15 and $17 each year to bring their arsenic 
concentrations down to the MCL.  These willingness to pay values were significantly less than 
how much actually bringing systems into compliance would cost, and as such the authors 
concluded that the arsenic MCL was not reasonable for small water systems.  Cho et al. 
suggested that POU/POE technologies could be used instead, but should be installed, paid for 
and maintained by the customers and not the water system as required under SDWA.  Instead the 
system should make homeowners aware of the risks they face from arsenic in their drinking 
water and subsidize the costs of installation and operation for homeowners that choose to install 
them. 
 A 2006 study by Jones and Joy looked at fourteen small systems (6-95 system 
connections) on New Mexico tribal reservations and the measures they had taken in order to 
implement the revised arsenic MCL.  The authors ran several different scenarios analyzing the 
usage of different treatment technologies and compared these to the benefits the communities 
would receive. For 13 of the 14 systems, the benefit-cost ratio is never any greater than 0.29 (for 
home delivery of bottled water).  Jones and Joy (2006) concluded that the arsenic rule was 
inequitable for these 14 tribal communities and that these cases were similar to the problems 
31 
 
faced by small, rural populations nationwide in meeting the arsenic MCL.  The lack of adequate 
financial resources provided by the federal government meant that available funding for risk 
averting measures would have to be shifted from other non-mandatory health expenditures (e.g. 
doctor visits, nutrition) to meet the arsenic standard, meaning that in all likelihood small CWS 
users would face at best no change in their overall health risk level due to the new regulation (if 
not an increase to their risks). 
ENVIRONMENTAL JUSTICE AND ARSENIC 
 Even before the arsenic regulation was finalized, small water systems were preparing 
themselves to face extremely high costs for drinking water treatment.  For the first two decades 
of the 2000s, the EPA predicted that per capita water costs would be four times higher for small 
systems than for large systems due to needs like infrastructure replacement and capital 
improvements in order to bring systems into compliance with SDWA regulations (including 
arsenic) (Scheberle 2004).  Normative considerations about the inequity of the arsenic MCL (and 
drinking water regulations in general) for small systems tends to draw on considerations of 
environmental justice.  Environmental justice is defined by EPA as ?the fair treatment of all 
people?with respect to the development, imp lementation or enforcement of environmental laws, 
regulations and policies.?  Under this definition of environmental justice, those who live in small 
systems need to be given the same level of consideration that residents in large water systems 
receive during the regulatory process.  Other definitions of environmental justice (such as that of 
environmental justice theorist, Robert D. Bullard) outline a component of geographical 
inequality of the burdens.  Geographical inequality of the burdens holds that a situation where 
some areas receive benefits and others receive the costs and negative health impacts is 
inequitable and should be addressed.  This situation applies for the arsenic rule in particular 
32 
 
where benefits are equally distributed, but (because of the cost differences) the costs are not 
remotely equitably distributed between all recipients of the benefits (Rosenbaum 2008).  Cory 
and Rahman (2009) argue that there are two ways SDWA can raise environmental justice issues, 
unequal enforcement of regulations that increase the risk to minority communities and forcing 
systems into (often prohibitively) costly compliance in order to obtain the health benefits.  Of 
particular significance for small system operators is the latter. 
 In order to address environmental justice concerns the EPA is mandated by SDWA to 
consider affordability of their proposed regulations.  For the arsenic rule, the EPA determined 
that arsenic would not be unaffordable for small systems.  The expenditure margin for new 
drinking water regulations at the time the arsenic regulation was proposed was $500 
$(1999)/household/year.  Since no system size class had average household costs in excess of 
this expenditure margin (see Table 2-2), the rule was found to be affordable (Abt Associates Inc. 
2000). 
 As with most of the other facts surrounding the cost of the arsenic regulation, this finding 
was highly controversial.  With small water systems worried about how much the rule would 
cost the customers they challenged the finding (Hogue 2001).  Reviews conducted by the EEC 
found that systems were indeed having a hard time raising the capital necessary to meet the 
arsenic rule (EPA-EEC 2002).   
 The conclusion that obtaining the funding to pay for arsenic treatment could be difficult 
for some systems was supported by two different studies.  The Jones and Joy (2006) study of the 
benefits and costs on Native American communities located in New Mexico also analyzed 
whether arsenic treatment was affordable for those on the reservations.  One of these systems had 
water costs following installation of the least-cost treatment alternative in excess of the $500 
33 
 
expenditure margin.  All of the systems had water costs in excess of 2.5% of their local MHI.  
Jones and Joy concluded that arsenic treatment for water was unreasonable for customers on 
these 14 reservations.  The second was a study of the costs faced by 35 water systems in 
California by Hilkert Colby et al (2010).  About a fifth of the systems in the study sample had 
treatment costs for arsenic removal in excess of the expenditure margin for a regulation, and a 
seventh of the systems paid more solely for arsenic removal than the EPA allowed for the entire 
water bill.   
 A finding that required water treatment is unaffordable for systems is the first step in 
granting systems variances and other selective enforcement measures.  The concern exists that 
selectively enforcing a regulation like the arsenic rule would raise environmental justice 
concerns if minority communities were likely to be the ones receiving the variances.  Cory and 
Rahman (2009) set out to examine if selective enforcement of the arsenic rule for systems would 
cause environmental injustice in the state of Arizona.  Analyzing the 1,006 public water systems 
in the state, the researchers compared the racial make-up and socioeconomic class of the 
customers in the systems affected by arsenic to those that receive water from systems that do not 
have a problem with arsenic.  The authors found that racial minorities or lower income 
households are not more likely to be found in areas with arsenic contamination issues.  The only 
racial group that is disproportionately negatively affected by arsenic is whites (African-
Americans are actually a significantly smaller share of the population in arsenic-affected 
communities).  All of the measures of wealth (per capita income, per household income and 
household value) are higher in arsenic affected areas than in non-affected areas.  Cory and 
Rahman conclude therefore that selective enforcement of the arsenic regulation does not appear 
34 
 
as if it would produce environmental injustice against minority and poor communities in 
Arizona. 
 Cory and Rahman?s results may have been a unique feature of the systems in Arizona, as 
this was not the case for the residents of the San Joaquin Valley of California.  A study by 
Balazas et al. (2012) found that there was a significant environmental injustice posed by arsenic 
in community drinking water systems for residents in systems with more racial minorities and 
fewer homeowners.  Customers of these utilities were more likely to have higher concentrations 
of arsenic in their finished water than customers of more affluent systems.  Systems with more 
?persons of color? and fewer homeowners were also more likely to have been found in violation 
of the arsenic MCL, although a significant number of systems that had average arsenic 
concentrations above the MCL were not officially found to be in violation.  These problems were 
worse for small systems (defined by the authors as having fewer than 200 connections), where a 
10% decrease in the home ownership rate was predicted to correspond to an increase in 4.3 ?g/L 
of arsenic in the water.  The authors attribute these trends of smaller systems and systems serving 
low socioeconomic status customers to the reduced capacity of these systems to finance, install 
and operate arsenic removal processes.  The authors propose that the regulatory burden that gets 
passed onto these customers could force poorer households to adopt expensive alternatives (like 
the purchase of bottled water) or adopt treatment only imperfectly which would increase their 
related risks.  The authors conclude that the solution is not variances and exemptions, but for 
these systems to partner with state and federal agencies to seek financial and technical help in 
securing affordable arsenic treatment. 
  
35 
 
  
 
 
 
 
CHAPTER 3 
 
EPA?S REVISED METRICS AND LOWER INCOMES FOR DRINKING WATER 
AFFORDABILITY 
 
Introduction 
 Arsenic has been linked to increased risks of lung, bladder and skin cancer worldwide for 
many years now and its presence in drinking water was first regulated in the 1940s by the U.S. 
Public Health Service.  The EPA adopted U.S. Public Health Service?s 50 ?g/L limit on an 
interim basis and spent nearly three and a half decades collecting data about the risks associated 
with arsenic-induced cancer before finalizing the 10 ?g/L rule in 2001 (Tiemann 2007).  This 
arsenic rule was set to impose a significant burden on the smallest of drinking water systems.  
Although the EPA determined that the arsenic rule would be affordable for all system sizes, 
many criticized both the conclusion and the method the EPA used to make its affordability 
determination.  These criticisms prompted the EPA to reconsider their affordability methods for 
drinking water, culminating in the proposal (but not yet the finalization) of a new set of 
affordability methods. 
 The controversy over drinking water affordability and potentially excessive regulatory 
burdens placed on small systems has diminished as the arsenic rule nears a decade of 
enforcement.  However, the EPA is currently moving to make another round of regulatory 
determinations of the contaminants listed on the third contaminant candidate list (Roberson et al. 
2009) and other potentially forthcoming rules, such as a hexavalent chromium rule with costs 
projected at a minimum of $500 million (C. J. Seidel et al. 2013), mean that affordability issues 
may reappear as a factor in evaluating potential regulations.  The potential benefits of future 
36 
 
regulations could be dwarfed by a host of other factors that have led the EPA and others in the 
water supply community to predict that the cost of water and wastewater services could 
quadruple in the coming decades (Baird 2010).  If questions about affordability return, the EPA 
could find itself unprepared to address these concerns. 
 Concerns over the affordability of drinking water regulations have been part of the 
national discussion since the SDWA was first passed in 1974.  As part of the SDWA 
Amendments of 1996, Congress authorized the State Drinking Water Revolving Fund to help 
with problems of system affordability and mandated that the EPA develop a method for 
assessing affordability (Pontius 2008).  In response, the EPA proposed a method, called the 
expenditure margin approach, for this purpose in 1998 (63 FR 42032).  The EPA took the ratio 
of the total cost for water treatment (a baseline cost and the projected cost for the treatment 
required to meet the proposed standard) divided by the MHI in each of its small system size 
classes (i.e.., <100, 101-500, 500-1,000, 1,000-3,300 and 3,300-10,000 households).  If this ratio 
exceeds 2.5%, the regulation is considered to be unaffordable for that size class, and permission 
to use variances would be considered by the states for the systems (Method 1 in Table 3-1).  An 
alternative way of conceptualizing this method is by calculating 2.5% of the MHI and 
subtracting the median household water bill, this difference is termed the expenditure margin.  If 
the cost of the considered regulation is greater than this difference, then the regulation will be 
unaffordable.  For example in 1998, 2.5% of the MHI was $750 (1997, throughout this thesis 
prices are nominal prices in the year that follows the price) and the median water bill in the small 
systems was $250 (1997).  Therefore the expenditure margin was $500 (1997) and a rule could 
not cost more than this without being deemed unaffordable. Permission to use variances allows 
water systems to implement more affordable treatment techniques that provide some protection  
37 
 
Table 3-1 ? Affordability methods considered 
 
to consumers at a more equitable cost.  This method, and other affordability methods, does not 
consider the opportunity costs associated with the increased risk faced by consumers when 
variance technologies are used when a decision is made on whether or not systems will be given 
permission to use variance technologies. This method faced a test in 2001 during the revision of 
the arsenic MCL.  Contrary to the EPA?s finding that the revised arsenic MCL was affordable for 
all systems, Hilkert Colby et al (2010), Jones and Joy (2006) and the Environmental Economics 
Method  
Number 
Name Criteria Reference 
1 Expenditure Margin 
Approach 
Proposed total water costs (baseline plus 
proposed regulation costs) cannot exceed 
2.5% of national MHI (i.e. $750 (1999)) . 
(63 FR 
42032) 
2 Incremental Burden 
Approach 
(2a) Proposed regulatory costs cannot 
exceed one of three considered limits:  
0.25% MHI, 0.50% MHI, 0.75% MHI.  
(2b) If regulatory costs do not exceed the 
final limit, systems can still qualify if 
they are located in an economically 
disadvantaged community (county MHI 
less than 65% national MHI or poverty or 
unemployment rate twice the national 
average). 
(71 FR 
10671) 
3 Low-Income Expenditure 
Margin National Screen 
Proposed total water costs (baseline plus 
proposed regulation costs) cannot exceed 
2.5% of national first quartile household 
income (i.e. $555 ($2000) or $661 
($2010)). 
Proposed 
based on 
(63 FR 
42032) 
4 Low-Income Incremental 
Burden National Screen 
Proposed regulatory costs cannot exceed 
one of three considered limits:  0.25%, 
0.50% or 0.75% first quartile household 
income.  If a system does not exceed the 
final limit, systems can still qualify if 
they are located in an economically 
disadvantaged community. 
Proposed 
based on 
(71 FR 
10671) 
5 Low-Income Expenditure 
Margin Local Test 
Systems can qualify for extension if the 
system?s costs to meet the regulation 
exceed 2.5% of the municipality or 
county?s first quartile household income. 
 
38 
 
Committee (EEC) (2002) found that small systems ?genuinely struggled with costs? in meeting 
the EPA?s arsenic MCL. 
 Within days of the finalization of the arsenic rule (and as part of the appropriations 
process for fiscal year 2002), the EPA was asked to reevaluate its affordability methodology by 
Congress (EPA 2002b).  The EPA conducted a review and sought input on its affordability 
methodology from the water supply community.  Among those the EPA requested advice from 
was the EEC and the NDWAC.  The EEC found that the basic method for evaluating 
affordability was sound and justifiable on the basis of efficiency and equity (2002).  However, 
the EEC found several problems that the EPA should address in their revised criteria, such as the 
use of a median income and national income data, and offered proposals that would address these 
concerns.  The NDWAC recommended a more dramatic solution, that the EPA only consider the 
proposed regulation?s burden and not the total cost for drinking water treatment on water 
systems.  Using an incremental method addressed the ?adding up problem?; the ?adding up 
problem? describes the situation that occurs when early and expensive regulations use up the 
entire expenditure margin, leaving all later regulations to be found unaffordable no matter their 
cost (NDWAC 2003). 
 Based on EEC and NDWAC recommendations, the EPA proposed a new two-part 
methodology in 2006 that included a screen that used predicted cost for the average system in 
each size class in the US and guidance for state agencies to make affordability determinations.  
In the nationwide screen (Method 2a in Table 3-1), the ratio of the cost for installing treatment to 
meet the proposed regulation to the MHI for the system size classes would be determined, in a 
method known as the incremental burden approach (71 FR 20671).  If the incremental burden 
ratio exceeds one of three proposed threshold levels (0.25%, 0.50% and 0.75% of MHI), the 
39 
 
regulation is unaffordable for that size class.  In addition to this rule, and as part of its guidance 
for the states, the EPA proposed three criteria that would determine if an individual system 
would likely face undo difficulties in raising funds for necessary capital improvements:  (i) a 
MHI less than or equal to 65% of the national MHI, (ii) CB poverty rate at least twice the 
national average, (iii) or a two-year average unemployment rate of greater than twice the national 
average (Method 2b in Table 3-1, this method is examined in greater detail in the next chapter).   
  Because the arsenic rule is the regulation that triggered the debate over affordability, this 
chapter discusses the performance of the EPA?s expenditure margin and incremental burden 
affordability methodologies in the context of the arsenic rule.  The EPA originally projected the 
arsenic rule would cost $205.6 million (1999) nationwide annually (Abt Associates Inc. 2000).   
Quantifiable health benefits from lung and bladder cancer reductions came to an upper-end 
estimate of $197.7 million (1999) annually, with the unquantifiable benefits (from other cancers 
and non-cancer health effects linked to arsenic consumption, but without dose-response data 
available for the EPA analysis) estimated to be sufficient to make up the difference between 
costs and quantified benefits (Abt Associates Inc. 2000).    
 Some researchers, such as Frost et al. (2002) and Burnett and Hahn (2001), disputed the 
finding of this cost-benefit analysis and found that the costs far outweighed the benefits, 
especially for small systems and their customers.  The average cost for arsenic treatment to those 
in the smallest system size class (systems serving less than 100 households) per year was $327 
(1999) per household, about 10 times the average cost per household requiring treatment per year 
nationwide ($33 (1999)) (Abt Associates Inc. 2000).   Despite this cost, the EPA found that this 
burden was affordable for households in small systems nationwide since small system costs did 
not exceed the expenditure margin of $500 (1997) (Abt Associates Inc. 2000).   
40 
 
 To the author?s knowledge, two ex post reviews of the affordability of arsenic have been 
conducted.  A review of arsenic treatment alternatives for fourteen tribal communities in Arizona 
found that the cheapest treatment alternative was within the expenditure margin but in excess of 
2.5% of the community?s MHI (Jones and Joy 2006).  In a study of thirty-six treatment sites 
throughout California, Hilkert Colby et al. (2010) found that 15% of the systems in the study 
exceeded the 2.5% MHI affordability criteria. 
 This chapter sets out to compare the current EPA method for affordability, the 
expenditure margin approach, and their proposed methodology, an incremental burden approach, 
in the context of the arsenic rule.  It also sets out to evaluate if the ability of these methods to 
find water treatment unaffordable and thereby enable states to grant systems permission to use 
variance technologies (i.e. point-of-entry and point-of-use systems that may produce water at 
lower quality but cheaper cost for consumers) has been compromised by the use of the MHI and 
whether a lower household income, the first quartile household income (Q1HI), is a better fit 
with both the purpose and function of affordability determinations. 
Data and Methods 
 An evaluation of the the EPA?s expenditure margin and incremental burden affordability 
criteria using a geographically diverse set of systems in the United States that had EPA installed 
treatment for arsenic removal between July 2003 and July 2011 is performed and its results 
discussed in this chapter.  The evaluations of affordability were also performed at two different 
time points.  First, using 2000 data to model how affordability determinations would have 
performed when the arsenic MCL was initially promulgated.  Second, this is also performed 
using 2010 data to see how it would perform under the most recently available census data.  
While national screens using the projected annual costs (Abt Associates Inc. 2000) with 
41 
 
adjustment for inflation using the Consumer Price Index (BLS 2012) were performed, system-
by-system tests using system-specific cost data (Wang and Chen 2011) were also done to 
determine whether a significant number of systems exceed the national screening criteria.   
 Making affordability determinations requires information on the costs of arsenic 
treatment, baseline drinking water treatment before the arsenic rule and socioeconomic data.  
Arsenic treatment cost data came from the EPA Office of Research and Development?s Arsenic 
Removal Technology Demonstration Project (ORD Project).  The ORD Project selected fifty 
sites nationwide that required treatment for arsenic in drinking water.  The EPA selected and 
funded the treatment regime, any associated engineering costs and installation and operation 
costs in exchange for the ability to collect information on performance and cost to help other 
systems make decisions about selecting arsenic removal technology.  The EPA then published 
this data in a series of forty-nine reports (one of the selected sites does not have a report prepared 
as of June 2013).  Of the forty-nine sites with reports, eight of these sites are non-transient, non-
community systems and one serves a seasonal resort. These nine were not considered in this 
study since those systems do not serve households.  Therefore, the remaining forty sites were 
used for this study (sites, as well as key system and community properties, can all be found in the 
Appendix). 
 Costs for arsenic treatment were recorded during the one-year study period and reported 
by the EPA (Wang and Chen 2011).  Capital and start-up costs were annualized over a 20-year 
period with an interest rate of 7% in keeping with standard government practice.  For some 
adsorption and ion exchange systems, breakthrough had yet to occur by the end of the ORD 
Project study, so costs for media replacement were projected as part of the ORD Project reports.  
Many of these reports list the number of households or connections served.  Reports that did not 
42 
 
provide the number of households did list the total population served and Census Bureau (CB) 
data on household size (2010a; 2000) was used to estimate the number of households served.  
The total cost for treatment was divided by the number of households to come up with average 
cost per household.  The average cost per household was adjusted for inflation using Bureau of 
Labor Statistics (BLS) data to convert into 2000 and 2010 values (BLS 2012).  The EPA used the 
2000 EPA CWS survey (CWSS) (EPA 2002b) to determine national costs for water treatment 
when they performed their original affordability analysis.  This same data was used for the 
national screens in this paper for both 2000 and 2010.  System-specific data used for more local 
analysis utilized regional estimates of water costs based on data collected as part of the 2000 
long form census (Rubin 2004).  Because this estimate includes both water and wastewater bills, 
the household cost for water treatment alone was isolated based on data collected as part of the 
Water and Wastewater Rates 2000 Survey performed by Raftelis Financial Consulting (2000).  
 Data from the CB and BLS was used for the socioeconomic variables needed in making 
affordability determinations under Method 2b.  For the year 2000, household size, median 
income, income distribution and poverty rate were all taken from the 2000 Decennial Census 
performed by the CB (2000).  For 2010, total population was taken from the CB?s 2010 
Decennial Census (2010a) and household size, median income, income distribution and poverty 
rate were taken from the CB?s American Community Survey 2010 5-Year Estimates (2010b).  
All unemployment statistics came from BLS (2010; 2000). 
 The EPA?s expenditure margin and incremental burden methods were both used in this 
study as national screens using ex ante regulatory cost estimates (from before implementation) 
and as ex post estimate of arsenic?s affordability using ORD Project cost data.   This analysis was 
started by conducting a national screen based on the EPA expenditure margin affordability 
43 
 
methodology (Method 1).  The sum of the Abt Associates Inc. (2000) predicted cost for arsenic 
treatment and the average baseline water cost from the 2000 CWSS was compared to 2.5% of the 
MHI in 2000 and 2010 for systems serving up to 10,000 people.  Following this calculation, it 
was determined how many systems in the sample ended up exceeding the affordability rule 
compared to the expected results from the nationwide screen.  The average cost per household 
per system for arsenic treatment as reported by the ORD Project reports was then added to the 
average annual water costs per household for the area.  This total cost was then divided by the 
2000 U.S. MHI at the national level.  This process was then repeated with 2010 MHI data and 
inflation adjusted cost data.  Any system producing an expenditure-to-income ratio greater than 
2.5% was flagged for having unaffordable water treatment for arsenic removal. 
 The 2006 incremental burden EPA affordability methodology (Method 2) was applied by 
performing an incremental burden national screen (Method 2a).  The expected cost for arsenic 
treatment alone for each size class was compared to 0.25%, 0.50% and 0.75% of the national 
MHI. Following the national screen the average annual cost for arsenic treatment per household 
at the 40 demonstration sites was divided by the 2000 national MHI to determine how many 
systems in the set exceeded the affordability threshold.  The same was done with the 2010 MHI.   
 The EPA?s guidance for the states on identifying EDCs (Method 2b) that could 
potentially qualify for variances based on socioeconomic conditions was then applied.  To do 
this the socioeconomic data collected from the CB and BLS from 2000 and 2010 (1999-2000 and 
2009-2010 for unemployment data) was utilized.  The MHI, poverty rate and two-year average 
unemployment rate for the municipalities and counties the demonstration sites were located in 
was divided by the national MHI, poverty rate and two-year average unemployment rate (for the 
unemployment rate only Reno, NV and the New England sites were analyzed using municipal 
44 
 
data as the BLS does not report unemployment for municipalities with a population of less than 
25,000 outside of New England).  The EPA proposed that if the ratios exceeded any of these 
limits, 0.65 for MHI, 2 for poverty level, or 2 for unemployment, the arsenic removal treatment 
for drinking water could be potentially unaffordable due to an inability to raise the funds needed 
to pay.   
 As previously discussed, some of the suggestions made by the EEC were incorporated 
into the EPA?s 2006 proposal of new affordability criteria, including using regional data and 
applying the national test as an initial screening procedure.  Many key proposals, most notably 
the need to use a below median household income, were unincorporated.  The EEC gave many 
reasons to examine a lower income level.  The most significant reason for this recommendation 
was within-system income differences.  Because all households within a system pay for the 
drinking water treatments, not just the median households, the affordability to houses below the 
median income level should be considered.  The EEC also noted significant income variation 
within systems of the same size-class.   Using the MHI for an entire size class ignores variations 
within the size class and so poorer systems are held to the same expenditure level as the richer 
systems (EPA-EEC 2002). 
 The EEC suggested a few alternative household income levels that could potentially be 
utilized in affordability determinations.  They suggested the 25th or 10th percentile and 1, 1.5 or 2 
standard deviations below the mean ? with the 25th percentile and 1.5 standard deviations below 
the mean being ?reasonable? (EPA-EEC 2002).    Household income is not normally distributed.  
Most household income distributions have a mode well below the mean and a small number of 
households with very high incomes.  Because of this non-normal distribution, 1.5 standard 
deviations below the mean is below zero and even with the top 5% of the distribution removed 
45 
 
(to reduce the effect of outliers) this level of income would only be a few hundred dollars for 
most systems.  The 25th percentile of household income was therefore selected to represent low 
income in developing the following three new potential methods for affordability determinations: 
 1.  Substitute 2.5% of the 25th percentile of household income for 2.5% of the MHI into   
      the national screen (Method 3 in Table 3-1). 
 2.  Substitute the 25th percentile of household income for MHI in the national screens that 
      look at the incremental burden (Method 4 in Table 3-1). 
 3.  Use of 2.5% of the 25th percentile of household income as a threshold limit for a    
     decentralized decision making processes about treatment affordability (Method 5 in    
     Table 3-1). 
 The 25th percentile of household income was identified from the 2000 U.S. Census (CB 
2000) and the 2006-2010 American Community Survey (CB 2010b) data.  These income levels 
were then applied in the above affordability criteria. 
RESULTS OF THE AFFORDABILITY METHODOLOGIES 
 There are 14 possible rules analyzed in this study.  Four possible MHI rules use the 
projected costs:  total cost for water treatment cannot exceed 2.5% of national MHI (Method 1) 
and arsenic cost for water treatment cannot exceed 0.25%, 0.50%, or 0.75% of national MHI 
(Method 2a and 2b).  These four rules can also be applied using site- specific costs from the 
ORD reports for a total of eight rules using the national MHI.  These same four rules that use 
MHI and predicted costs can also use Q1HI (Methods 3 and 4), for an additional four rules.  A 
final two rules compare site-specific cost data to 2.5% of the municipal and county Q1HI 
(Method 5).  These 14 rules are used in both 2000 and 2010, for a total of 28 different 
affordability determination scenarios.  
46 
 
 Based on the expenditure margin method, the results of this analysis here confirm the 
EPA?s original evaluation that arsenic removal would be affordable for all system size classes.  
In neither 2000 nor 2010, did total predicted water treatment costs exceed 2.5% of the MHI 
(Table 3-2).  In 2000, the national MHI was $41,994 with 2.5% of that being $1,050; for 2010, 
the national MHI was $51,914 with 2.5% equal to $1,298.  None of the individual sites exceed 
these levels either, with total water treatment costs for the 40 sites ranging from $174.94-$709.55 
and an average of $346.04 (2000) (in 2010, $221.53-$898.50 with an average of $438.19 
(2010)).   
Table 3-2 ? Arsenic and total water treatment costs in 2000 and 2010 
 2000 2010 
System 
Class 
Size 
(people 
served) 
Estimates 
for Arsenic 
Treatment 
per 
Household 
 
Average 
Cost of 
Water per 
Household 
 
Total Cost 
of Water 
(Including 
Arsenic) per 
Household  
Estimates 
for Arsenic 
Treatment 
per 
Household  
Average 
Cost of 
Water per 
Household 
 
 Total Cost of 
Water 
(Including 
Arsenic) per 
Household 
 
?100  345 235 580 437 300 737 
101-500 172 232 404 218 294 512 
501-
3000 
65 289 354 82 366 448 
3001-
10,000 
40 253 293 50 320 370 
 
 Arsenic treatment was not affordable for all systems when the EPA incremental burden 
criteria are applied.  Starting with the national screen, the relevant thresholds are $315, $210 and 
$105 in 2000 and $389, $260 and $130 in 2010 (for 0.75%, 0.50% and 0.25% MHI, 
respectively).  Under these thresholds, the ?100 size class qualifies for variances under all three 
thresholds and the 101-500 size class qualifies for variances under the 0.25% MHI threshold 
only.  Figure 3-1 shows the results when using system-specific cost data of the burden imposed 
Table 3-2 ? Arsenic and Total Water Treatment Costs in 2000 and 2010 
47 
 
by arsenic treatment.  Five systems have arsenic treatment costs in excess of 0.75% of the  
 
Figure 3-1 - Systems with arsenic treatment cost violating one of the incremental burden 
standards 
 
national MHI. Of those five, three have populations greater than 100 people and so would not 
qualify for a variance under the national screen.  The cost per household for arsenic removal 
imposes a burden exceeding 0.50% MHI for ten systems, seven of which do not qualify for a 
variance under the national screen.  Of the forty total systems, over half (22) have costs for 
arsenic removal in excess of 0.25% of MHI and six of these 22 systems serve a population of  
greater than 500 people and so would not qualify for variances under the national screen. 
 Following the national screen, the next step was to examine the ability of the three 
criteria for identifying disadvantaged communities in Method 2b that qualify for variances.  
Depending on the year and the geography used (municipal-level or county-level data), between 
1-10 systems could potentially qualify for a variance.  Of these three criteria, the MHI less than 
65% of the national MHI qualifies the most systems for a variance.  In total, 15 systems would 
48 
 
have qualified using 2000 municipal-level data, 14 using 2000 county-level data, 17 using 2010 
municipal-level data and 8 using 2010 county-level data (an additional four systems qualify 
when the national screening threshold is 0.25% of MHI).  Figure 3-2 shows the systems that 
qualify based on the incremental burden criteria. 
 
Figure 3-2 - Systems qualifying for variances based upon proposed EPA criteria in (a) 2000 and 
(b) 2010   
 
 The impact of using an income level below the MHI, specifically the 25th percentile of 
household income, has on affordability determinations was examined next.  Starting with the use 
of the 25th percentile of national household income in a screening procedure (Method 3), the 
49 
 
usage of a below median income level significantly altered the results of affordability 
determinations.  The 25th percentile income in 2000 was $22,197 (2000) and 2.5% of that is 
$555; in 2010 it was $26,428 and $661.  Compared to the total cost for water treatment (baseline 
plus arsenic treatment) found in Table 3-2, the treatment for arsenic removal is unaffordable in 
systems serving less than 100 people in both 2000 and 2010.   
 Limiting the incremental burden of arsenic treatment to 0.75%, 0.50% and 0.25% of the 
national 25th percentile of household income ($166, $111 and $55 in 2000 and $198, $132 and 
$66 in 2010) to the predicted costs in a national screen (Method 4) leads to the conclusion that 
arsenic is unaffordable for systems serving 500 people or less no matter the threshold selected 
and for systems serving between 501 and 3,000 people when the threshold used is 0.25% of the 
25th percentile of household income. 
 This lower income level could also be used in decentralized decision-making processes 
using system-specific cost data and local income data to determine if the regulation is affordable 
(Method 5).  Figure 3-3 shows the systems that have unaffordable arsenic treatment under this 
method.  In 2000, three systems qualify when municipal income is used and six systems when 
county income levels are used.  In 2010, seven systems qualified when either municipal or 
county level Q1HI is used.  Only one system qualified in both years using either municipal or 
county incomes.  
HOW DO THESE EVALUATIONS COMPARE? 
 Figure 3-4 shows the number of potential affordability scenarios (of the 28 listed in the 
beginning of the previous section) in which a system qualifies for a variance or extension.  Two 
systems never have unaffordable treatment under any of the scenarios.  A quarter of the systems 
50 
 
violate only the 0.25% of the 25th percentile household income national screen rule in 2000 and 
2010. One-fifth of the systems qualify for a variance under half of the scenarios included.   
 
Figure 3-3 - Systems with arsenic treatment in excess of 2.5% of the Q1HI in (a) 2000 and (b) 
2010 
  
 The above results lead to two general facts.  First, as expected, the lower the allowable 
burden water treatment is allowed to place on communities, the more likely it is that systems will 
qualify for variances.  A second observation is that the more local the data used in the decision, 
the more likely it is that more systems will qualify for a variance.  These should be remembered 
during the comparison of potential affordability methods. 
51 
 
 
Figure 3-4 - Number of different affordability rule scenarios under which a system qualifies for   
     variances 
 
 There is a significant difference between the outcomes of the two EPA methods 
(Methods 1 and 2).  Under the expenditure margin approach no systems qualified for variances 
between 7-24 systems in the 40 systems (depending on the share of MHI allowed and the 
geography of the data used) have unaffordable arsenic removal costs according to the 
incremental burden approach.  If the purpose of revising the affordability criteria is to make it 
easier to trigger permission to use variances, the incremental burden method proposed by EPA 
meets this goal.  If the EPA wants to make it easier to trigger affordability decisions, they should 
select the 0.25% MHI and recommend the use of municipal data for states and counties trying to 
make more decentralized variance decisions.  If the EPA wants the rule to qualify only those 
systems experiencing a severe burden, then the EPA should select 0.75% MHI as its threshold 
and make use of county data as it qualifies fewer systems. 
 The results of each of the three criteria used by the EPA to identify EDCs were also 
compared to the other criteria and to the results of the national screens.  It is possible that the 
52 
 
individual EDC criteria are all qualifying the same communities.  It is also possible that these 
criteria are all identifying systems that may be qualifying under a national screen.  In either of 
these cases, some of the EDC criteria could be unnecessary because they are not identifying 
different systems for permission to use variances.  Observing which rules qualify systems for 
variances in 2000 and 2010, only a few systems meet the conditions for a variance under 
multiple economically disadvantaged criteria or one of the EDC criteria and under the national 
screen.  As such, redundancy of the criteria does not seem to be a significant issue and they are 
capable of identifying different sites.   
 A potential criticism of the use of low income methods is that their use could reduce the 
regulatory burden necessary to receive permission to use a variance to a level that is 
unacceptably low.  Comparing the results of the methods that used Q1HI to the results from 
methods that used national MHI show this is not necessarily the case.  A national screen 
composed of expected treatment cost in excess of 2.5% of the 25th percentile household income 
is as stringent as the EPA incremental burden screening criteria of 0.75% of MHI.  A national 
screen using 0.75% or 0.50% of the 25th percentile household income is as stringent as the 
incremental burden screening criteria of 0.25% of MHI.   
DISCUSSION 
 In 2002 the EEC evaluated the EPA's affordability criteria based on two conditions, 
economic efficiency and equity.  Economic efficiency is achieved when marginal benefit (MB) is 
equal to marginal cost (MC), i.e. when the benefit of producing one more unit is equal to the cost 
of producing this additional unit.  Figure 3-5 illustrates this for water quality.  Because of the 
difference in the economies of scale, the marginal cost for producing improvements in water 
quality for large systems is less than the marginal cost for small systems.  This produces two  
53 
 
different efficient results, one for large systems (qL) and one for small systems (qS).  Regulations 
are designed to be feasible for large systems, but apply to all systems independent of their size.  
This introduces an economic inefficiency into the small systems water quality market (what 
economists call a deadweight welfare loss, representing the area between the marginal benefit 
curve, the marginal cost curve, and qL).  Allowing variances, and making them easier to receive, 
makes it possible for small systems to approach qs which would make the market more efficient. 
 Considerations of equity and environmental justice are normative arguments to allow for 
variances when the costs of a regulation become excessive.  The EPA defines environmental 
justice 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 
$ 
Regulated Water Quality 
MB 
MC-Large 
MC-Small 
qS 
 
qL 
 
Figure 3-5 - Marginal benefits and costs for increasing regulated water quality 
54 
 
environmental laws, regulations and policies? (EPA 2011). For drinking water and 
environmental justice, the limited literature draws on two arguments.  The first is that drinking 
water contamination may correspond well to the existing power structure with minority and less 
affluent groups more likely to use contaminated water sources.  The second approach to 
environmental justice for drinking water argues that forcing systems to meet drinking water rules 
can force them to forgo cheaper ways to undertake similar risk reduction measures.  This second 
model of environmental justice is more relevant to the question of affordability (Cory and 
Rahman 2009). 
 Environmental justice, as a whole, focuses on the elimination of environmental inequity, 
which includes a component of geographical inequity-the unequal distribution of costs, benefits 
and resources over different locations (Rosenbaum 2008).  Drinking water regulations can fall 
into this definition of geographical inequity if the costs of meeting the regulation for small 
systems in rural areas are significantly larger than the costs faced by consumers in large systems.  
The case that geographic inequity is at play in the determination of drinking water regulations is 
also demonstrated by the emphasis the EPA places on the economics of drinking water provision 
for large systems in establishing regulations (EPA 2006, 2002).  Providing a way for small water 
system consumers to obtain treated water without paying an excessive amount for it by granting 
systems the ability to use variance technologies can serve as a remedy to geographic inequity and 
to promote environmental justice. 
 Though geographic inequity makes a case for variances in general, the use of a lower 
income limit is not a part of geographic inequity.  The EPA defines environmental justice in part 
as consideration of interests of individuals irrespective of income (EPA 2011).   Using an income 
level below the MHI explicitly seeks to promote the fair treatment of people regardless of 
55 
 
income.  Using a lower income level therefore allows the EPA to promote environmental justice 
for all persons irrespective of income and socioeconomic class.     
 The goal of increasing the ease with which systems trigger national affordability criteria 
is contrasted by the need of water providers to supply their customers with safe water.  Variance 
technologies are required to produce water at a quality that is ?protective of public health? which 
is a different standard and can be of a lower quality than that produced under the MCL.  As such, 
granting a system the right to use variance technology could very well introduce increases in 
health risk to those in small systems.  Because of this added risk, EPA and the states have to be 
selective in the number of systems that are granted permission to use variances. 
 An ideal affordability metric therefore needs to balance the two conflicting objectives of 
promoting environmental justice and protecting health.  Therefore, any metric must be able to 
grant systems variances when necessary.  But the metric should not grant too many variances in 
order to minimize the increase in risk to consumers.  Where to strike the balance between these 
two forces is ultimately a subjective call.  The author suggests that the rules at the extremes, the 
total costs for water greater than 2.5% of MHI (which allowed for no variances) and the national 
screens for incremental treatment burden greater than 0.25% MHI (which found that all systems 
serving fewer than 500 people plus all small systems in 492 counties) is undesirable as they are 
too strict or too generous (respectively). 
 Another important factor to consider when deciding the right level of balance between 
the need to grant variances to systems and to protect public health is that the cost estimates made 
before regulations are implemented typically overestimate the cost (Harrington et al. 2000; 
NCEE 2012).  Taking this into account, affordability metrics should be stricter than desired.  
Making this adjustment allows for the affordability determination procedure to select systems 
56 
 
that deserve variances without selecting systems that have their costs over-predicted.  There are 
two methods that best meet this. Method 2, which compares the cost of a new regulation to 
0.75% of the national MHI with the disadvantaged community approach, grants variances to 
systems serving less than 100 people and all other small systems in 492 counties.  Method 5, 
setting affordability for water treatment at 2.5% of the county's 25th percentile household income 
paired with the criteria for identifying a disadvantaged community (in Method 2b),  grants 
variances to systems in the economically disadvantaged counties and no more than about 10,000 
other systems serving fewer than 100 people. 
 Both these methods end up qualifying approximately the same number of systems (about 
14,000 in all, which is an intermediate value between the extremes of no small systems and all 
30,000 small systems) but the 25th percentile household income allows for a more targeted 
approach opposed to the blanket determination for the 0.75% of national MHI.   
 Other problems undermine the use of an incremental burden approach.  The NDWAC 
(2003) recommended the incremental burden approach of Method 2 as a way of avoiding the 
?adding up? problem in which early regulations add up to consume the entire expenditure margin 
and preclude the implementation of new rules; the EPA is considering an incremental approach 
because of this.  However, Table 3-3 demonstrates that the adding up problem does not occur 
System Size 1998 Expenditure 
Margin1 (1998) 
1998 Expenditure 
Margin (2006) 
Expenditure Margin 
Based on 2006 CWSS2 
(2006) 
25-100 $537 $664 $899 
101-500 $920 
501-1,000 $463 $572 $907 
1,001-3,300 $871 
3,301-10,000 $474 $586 $942 
Table 3-3 ? Expenditure margin 
1From (63 FR 42032) 
2Calculated as 2006 MHI (USCB, 2006) less the Annual Residential Revenue per Connection from the 
2006 CWSS (EPA, 2009) 
 
57 
 
and that the expenditure margin has actually increased since its original proposal in 1998.   
 Furthermore, the incremental burden approach introduces another form of an adding-up 
problem; a series of regulations could all be found affordable individually whereas their 
cumulative burden (once added up) could be unaffordable.  The NDWAC (2003) did consider 
this possibility and recommended that cumulative burdens still be examined but the EPA did not 
mention this in their proposal of this new affordability methodology.  Because of this new 
version of the adding up problem, Method 5 should be further preferred over the incremental 
burden approach of Method 2 with a limit of 0.75% MHI.  An argument can be made that 
Methods 2 and 5 could both be applied with systems having to meet one of the two criteria in 
order to qualify for a variance as this would serve to balance out the weaknesses of the two 
methods.  However, it is very likely that over a series of new drinking water rules, affordability 
determinations would be controlled by just the results of Method 5 (or any cumulative method 
that is adopted). 
 Using county-level data runs the risk of granting variances to systems that do not need it 
based on their municipal conditions.  This leads to problems from the perspective of protecting 
public health by exposing communities to risks they could afford to reduce.  However, the actual 
choice between using the municipal and county data depends on administrative objectives.  
Whether it is the EPA or the various state primacy agencies in charge of administering 
determination of affordability, the data should be readily available to them. This lends itself to 
adopting county-level criteria as the data is more readily available.  This leads to the 
recommendation that a variance be granted to a system located in a county where the cost of 
treatment is more than 2.5% of the county's 25th percentile income, the county's MHI is less than 
65% of the national income, the poverty rate is twice the national average, or the average two-
58 
 
year unemployment is twice the national average.  Figure 3-6 shows the systems qualifying 
under this rule. 
 
Figure 3-6 - Systems qualifying for variances under proposed rule in (a) 2000 and (b) 2010 
CONCLUSION 
 This chapter sought to perform an affordability determination using the EPA's 
expenditure margin affordability method, the incremental burden methodology they proposed in 
March of 2006 and using a lower income level in the metric by using the 2001 arsenic MCL as a 
case study.  While the results presented here agree with the EPA's original affordability decision 
59 
 
that the arsenic rule was affordable for all systems, the 2006 incremental burden criteria would 
have found arsenic to be unaffordable for some systems.  The main outcome of this chapter was 
an analysis of the effect of using the 25th percentile income, instead of the MHI, for affordability 
methodologies in order to address concerns of environmental justice.   
 The criteria proposed by the EPA in 2006 addressed the biggest concern raised about the 
1998 method, namely that the bar had been set too high, by making significant changes to all 
parts of the method.  However, the incremental burden method still leaves much to be desired.  
For example due to its incremental approach, it fails to place a maximum cap on drinking water 
costs.  A far more important shortcoming is the EPA declining to take into account explicitly the 
impacts regulatory burden has on low income households.  The method put forward in this paper 
addresses this concern. 
 Ultimately selecting an affordability methodology is a subjective decision that requires 
balancing risk, economics and political considerations.  It is the author?s position that the usage 
of a 2.5% of the county's 25th percentile income threshold with the EPA's incremental burden 
criteria for identifying disadvantaged communities (Methods 5 and 2b) meets various standards 
that define an ideal affordability methodology (helps meet conditions of economic efficiency, 
grants some systems variances as needed by environmental justice, not overly generous with the 
number of variances to protect public health, accounts for overestimating of costs and 
affordability practicality).  
 
 
60 
 
  
 
 
CHAPTER 4 
ADJUSTING THE DEFINITION OF AN ECONOMICALLY DISADVANTAGED 
COMMUNTIY 
Introduction 
 During debate on the SDWA in 1974, Representative Del Latta stated that, ?nobody is, of 
course, against the purpose of this bill, but the real question is that?how can that (referring to a 
community that had used up its bonding authority) community come up with the money to meet 
the standards that some EPA directive might set forth??  As Rep. Latta pointed out no one had 
any answers for him; the SDWA did not contain any provisions relating to the affordability of 
water service (Congr. Rec. 1974).  This question remained unanswered for two decades. 
 In 1995 the Congressional Budget Office used SDWA regulations as a case study for 
unfunded mandates (U.S. Environmental Protection Agency 1998), and in the development of 
the 1996 SDWA amendments Congress became aware that the current regulatory framework 
posed a significant challenge for small water systems.  Congress incorporated the concept of 
affordability into the amendments and directed the EPA to consider affordability when 
establishing regulations.  For each regulation considered, the EPA would analyze the costs for 
implementing the possible technologies and determine which treatment technologies would be 
affordable for each size of system.  If they were unable to find a treatment technology that met 
the standard, systems would be allowed to apply for the use of variance technologies.  These 
variance technologies would provide a level of treatment that was protective of human health 
61 
 
(but maybe not to the same level as the MCL) that would still be affordable to water systems.  
The EPA was left to determine what criteria would be used to determine affordability (Tiemann 
2010).  Congress also created the Drinking Water State Revolving Fund, whose purpose was to 
provide states with funding for drinking water projects with special attention to disadvantaged 
communities (Beecher and Shanaghan 1998; Levine 2012). 
 The context of drinking water regulations was not the EPA?s first time considering 
affordability as part of the regulatory process.  The EPA had already considered and codified 
criteria for affordability of its regulations in nearly two dozen areas in the Offices of Water; 
Solid Waste; Compliance Monitoring; and Policy, Planning and Evaluation for regulatory 
frameworks under the Clean Water Act; the Toxic Substances Control Act; and the 
Comprehensive Environmental Response, Compensation and Liability Act (Beecher and 
Shanaghan 1998; U.S. Environmental Protection Agency 1998).  In the 1990s, the EPA 
addressed affordability agency-wide with a panel under the Office of Policy, Planning and 
Evaluation.  The panel established two different, two-step methods for measuring the 
affordability of EPA regulations.  In both methods the first step was a basic burden screen that 
examined if costs were a significant share of household income.  The second step of the first 
method then screened again the systems that determined if these systems would face a challenge 
in obtaining the necessary funding for improvements by assessing their access to financial 
resources if they had an unaffordable burden placed on them as determined in the first screen.  In 
the second method, systems would petition either the EPA or their state for regulatory relief if 
their costs to finance the project would be too high or if the costs would exceed their capability, 
which left the systems with the burden of proof (U.S. Environmental Protection Agency 1998). 
62 
 
 When the EPA made its affordability criteria for the 1996 Amendments, they included 
only the burden screen (Method 1 in Table 3-1) (63 FR 42032).  They defined that a rule would 
be unaffordable if it caused the total amount of drinking water costs to exceed 2.5% of MHI per 
year for the system?s size class (Cooke 2005; Hilkert Colby et al. 2010; Pontius 2008).  The 
arsenic rule quickly questioned the validity of this method and in 2001, Congress charged the 
EPA with reevaluating their affordability methodology both in general and in the case of the 
arsenic rule (EPA 2002a). As far as evaluating the affordability for the arsenic rule, the debate 
focused on identifying disadvantaged communities for relief.  The most significant contribution 
to this discussion came from Rubin (2002), who examined definitions for disadvantaged 
communities from state drinking water programs and federal agencies.  Rubin identified three 
criteria that could be used for this purpose:  (1) MHI less than 65% of the national MHI, (2) a 
poverty rate twice the national average and (3) two-year average unemployment twice the 
national average.  If a community met any of these criteria they would be considered 
disadvantaged.  These criteria represent the socioeconomic status of the community which 
represents the ability of water systems to fund improvements through their customers (EPA 
1998). 
 When the EPA published their new method in the Federal Register in 2006 they 
incorporated Scott Rubin?s work on identifying EDCs and brought drinking water affordability 
determinations into the 1990 Office of Policy, Planning and Evaluation format.  The first step 
was a national burden screen.  If a new rule would cost an average system in a particular size 
class greater than one of three (0.25%, 0.50%, or 0.75%) shares of MHI, then the systems in that 
size class would qualify for variances.  For systems that had affordable treatment based on this 
first screen, a second round would be used to further identify systems located in EDCs meeting 
63 
 
the criteria Rubin identified above.  These EDCs could then request variances from their states 
(71 FR 10671). 
 The purpose of this chapter is two-fold.  First, the author will determine how many 
counties meet the EPA criteria for an EDC using up-to-date socioeconomic data.  Second, the 
author will evaluate how adjusting the multiples (0.65 for MHI, 2 for poverty and 
unemployment) used in the criteria for defining an EDC affects the number of counties that can 
qualify for a variance.  In identifying their criteria, the EPA did not publish work that looked at 
other ratios than those that were under consideration.  Their criteria were adopted based on 
recommendations by Rubin (2002) who tried to develop numerical definitions of EDCs based 
upon review of criteria states had already adopted for their drinking water programs and 
qualifying conditions for communities to receive assistance grants from various state agencies.  
However, the states that had implemented affordability criteria tended to be more affluent and so 
even their most stringent criteria may not have been appropriate for national use.  For instance, 
New Jersey?s affordability criteria was 65% of the state?s MHI.  According to the 2000 census 
65% of New Jersey?s MHI was $35,845, a value greater than the MHI for the ten poorest states 
in the US at the time.  By performing a sensitivity analysis on these ratios, the author hopes to 
identify a more appropriate combination of ratio and criteria. 
 The remainder of this paper proceeds as follows.  Next is an evaluation the evolution of 
the concept of affordability at the EPA over time, followed by a description of the data and 
methods used in the sensitivity analyses of the criteria, as well as introducing and explaining the 
justifications for several alternative ratios for the criteria used.  The number of counties and an 
approximate number of small (less than 10,000 people served) CWSs (contained in the Safe 
Drinking Water Information System) that qualify at the EPA proposed criteria is calculated.  
64 
 
This discussion is followed by a presentation of the results of how many counties and CWSs 
qualify under other ratios for these three criteria.  The sixth section analyzes two different 
methods of measuring EDCs.  Finally, the last section presents conclusions and 
recommendations for the affordability criteria. 
Data and Methods 
  The socioeconomic data for this work included the MHI, poverty rate for individuals and 
unemployment rates for the 3,143 counties and county equivalents in the 50 states.  MHI and 
poverty rate for all 3,143 counties and county equivalents were taken from the U.S. CB?s 2010 
American Community Survey (ACS) 5-year estimates (CB 2010).  Local, seasonally-unadjusted, 
monthly unemployment rates from November 2011-December 2012 for all but one of the 
counties and county equivalents were taken from the BLS (BLS 2013).   
 The number of small (10,000 or fewer customers) systems and the population served by 
these systems in these counties was taken from the EPA?s Safe Drinking Water Information 
System-Federal Version (SDWIS/FED) (EPA 2013).  The SDWIS/FED serves as a 
clearinghouse for the general public and provides information on systems, their characteristics 
and violations.  It is not a complete inventory of all drinking water systems in the country 
because it relies on states and systems to submit the required information, but it is the most 
exhaustive catalog for the entire U.S. that is available to the general public.  The number of 
systems and population are accurate as of October 2011. 
 This socioeconomic data and drinking water system counts were combined into a GIS 
database for identifying and displaying counties that met the various affordability criteria.  
Initially, the current number of counties, systems and population served by these systems that 
could qualify for variances using the proposed economically disadvantaged criteria were 
65 
 
identified.  For the entire United States, the MHI and poverty rate from the 2010 American 
Community Survey 5-year estimates were $52,762 and 14.3% and average unemployment at 
8.7% from November 2011-December 2012.  The thresholds therefore were $34,295 (0.65 x 
52,762) for MHI, 28.6% (2 x 14.3%) for poverty, 17.4% (2 x 8.7%) for unemployment.   
 To determine the impact that altering the ratios for the various factors has on selecting 
counties that qualify for variances, the author adjusted the various ratios over a wide range.  For 
income, the author selected a strict, yet reasonable, condition of 50% of national MHI to a 
generous 100% MHI ($26,381-$52,762).  The author also examined the impact of using the 
national nonmetropolitan area median household income (non-MA MHI), since many federal 
agencies use this as their income level for determining program eligibility (for 2010, this value 
was $41,440 (CB 2010b)).  The range for the national non-MA MHI is between 50-100% 
($20,720-$41,440).  For the poverty criteria, the range over which it will be considered was 
between the national average and three times the national average (14.3%-42.9%).  
Unemployment was studied over the range from the national average to three times the national 
average (8.7%-26.1%). 
 There are several ratios deserving of special attention because they are used by state 
drinking water agencies or federal programs that provide assistance to counties in need.  Income 
values include:  incomes below the poverty level ($22,113 (CB 2010b)), Department of 
Agriculture?s Rural Utilities Service grants for water and wastewater treatment; MHI less than 
$17,000 in 1989 (corresponding to $29,895 in 2010 dollars), EPA?s ?Road to Financing?; and 
MHI below 80% of nonmetropolitan area MHI ($33,152), Department of Agriculture?s Rural 
Utilities Service grants for water and wastewater treatment, grant eligibility for the Rural 
Development Administration and the Small Business Administration?s Historically Underutilized 
66 
 
Business Zones program. For poverty rate, the only other definition that uses poverty rate is the 
Department of the Treasury?s community development program, which defines EDCs as having 
a poverty rate of at least 30%. For unemployment these ratios are:  the average (8.7%), based on 
the EPA?s ?The Road to Financing?; 1 percentage point above the average (9.7%), based on the  
Table 4-1 - Definitions of EDCs considered 
Category Source Level 
Median Household Income Thresholds 
65% MHI EPA?s Drinking Water Affordability $34,295 
 Sensitivity Analysis Range (MHI) $26,381 - $52,762 
 Sensitivity Analysis Range (non-MA MHI) $20,720 - $41,440 
Poverty Level Agriculture Dept.?s Rural Utilities Service $22,113 
$17,000 (in 1989) EPA?s ?Road to Financing? $29,985 
80% non-MA MHI Agriculture Dept.?s Rural Utilities Service $33,152 
 Rural Development Administration 
 Small Business Administration?s HUBZones 
Poverty Rate Thresholds 
200% Poverty Rate EPA?s Drinking Water Affordability 28.6% 
 Sensitivity Analysis Range 14.3% - 42.9% 
30% Treasury Department 30% 
Unemployment Rate Thresholds 
200% Unemployment EPA?s Drinking Water Affordability 17.4% 
 Sensitivity Analysis Range 8.7% - 26.1% 
Avg. Unemployment EPA?s ?Road to Financing? 8.7% 
Avg. + 1% Commerce Dept.?s Economic Dev. Admin. 9.7% 
120% Unemployment State of Washington 10.4% 
125% Unemployment EPA?s CSO Financial Capacity Review 10.9% 
140% Unemployment Small Business Administration?s HUBZones 12.2% 
150% Unemployment Treasury Department 13.1% 
 
67 
 
Department of Commerce?s Economic Development Administration; 20% greater than average 
(10.4%), based on Washington state?s definition of a distressed county; 25% greater than average 
(10.9%), based on the EPA?s Combined Sewer Overflow Financial Capacity Review; 140% of 
the average (12.2%), the Small Business Administration?s Historically Underutilized Business 
Zones program; and 1.5 times the average (13.1%), based on the Treasury Department?s 
community development programs (Rubin 2002; EPA 1998; U.S. Small Business Administration 
2013).  All these definitions of EDCs, including the EPA drinking water criteria and the range 
over which the sensitivity analysis was performed over can be found in Table 4-1. 
Results and Discussion Using Proposed EPA Criteria 
 Table 4-2 shows the number of counties that could qualify under the 2006 EPA proposed 
criteria, the number of small systems in those counties and customers served by these systems.  
Of the 3,143 counties and county equivalents in the United States, 511 meet at least one of the 
three socioeconomic criteria the EPA has proposed to identify counties that qualify for variances.  
According to the SDWIS/FED database, in these 511 counties there are 3,812 small water 
systems serving a population of 5,629,431 customers.  Most of these counties qualify because of 
the MHI < 65% MHI criteria; a total of 383 counties meet only this criteria.  Nine counties meet 
only the poverty rate criteria and six counties meet only the unemployment criteria.   
 
Criterion 
Number of 
Counties 
Number of 
Small Systems 
Population 
Served 
MHI ? 0.65 National MHI 495 3,630 5,527,726 
Poverty Rate ? Twice National Average 122 832 1,480,165 
Unemployment Rate ? Twice National Average 10 111 94,247 
One or More of the Above 511 3,812 5,629,431 
  
Table 4-2 - Counties and systems qualifying for variances under proposed EPA criteria 
68 
 
 Figure 4-1 shows the counties that qualify for a variance based upon the EPA?s criteria.  
Many states do not have any counties qualifying while several states have a multitude.  Regions 
that have many counties qualifying for variances include the Sun Belt stretching from North 
Carolina to southern California and the southern portions of the Appalachian Mountains 
(primarily in Western Virginia and Kentucky).   
 
 While the number of systems that qualified when the EPA originally proposed these 
variance criteria and the application of these criteria to today?s socioeconomics situation has 
decreased, both the number of counties that qualify and the number of small system customers 
affected increased.  For counties, an additional 101 total counties (up from 410 in the original 
estimate) qualified using the most recent data.  The EPA estimated that the total number of 
systems in these counties was 4,249 and the counties identified in this present analysis contain 
Figure 4-1 - Counties potentially qualifying for variances at different EDC ratios 
69 
 
437 fewer systems.  The total number of customers potentially affected by these criteria 
increased in this latest analysis by 144,273 people after 5,485,158 customers were served by 
systems that qualified in the 2006 EPA analysis.  There is some indication that system 
consolidation has occurred with 1,275 fewer small systems in the October 2011 SDWIS/FED 
database.  In the decade between 2001 and 2011, a net 480 systems (in SDWIS/FED databases in 
both years) moved from serving fewer 10,000 customers to serving greater than 10,000 
customers (Rubin 2013b). 
 The individual criteria also showed a mixed response.  The income criteria showed the 
most dramatic increase between the original analysis and the results here, an additional 139 
counties, 145 systems and 1,155,049 customers qualified because of income.  The poverty 
criteria also increased the number of counties, systems and customers potentially qualifying for 
variances, 41, 300 and 529,960, respectively.  Finally, the unemployment criteria showed a 
significant decrease in the numbers qualifying: 70 fewer counties, 809 fewer systems and 
1,296,979 fewer people.  
 The unemployment criteria qualified fewer systems is most likely because that while the 
MHI and the national poverty rate increased in between the EPA?s analysis and this, they did not 
increase as significantly as unemployment did.  The MHI increased just a little more than $1000 
(in real dollars) and the poverty rate increased by a percentage point.  Unemployment showed a 
much more significant increase of nearly two points in the period between the EPA?s proposal of 
these criteria and the period used in this analysis.  This increase set the bar much higher for 
systems trying to qualify for variances based on their unemployment than was set for qualifying 
based on MHI and poverty rate. 
70 
 
 This demonstrates a potential problem with the EPA?s method.  In a recession, 
unemployment tends to increase and lost income due to this unemployment drives poverty rates 
up.  This has been true for the Great Recession (starting in December 2007 and ending in June 
2009 from which the recovery is still ongoing) and recessions in general (Pilkauskas et al. 2012).  
This means that if a regulation passes the first step of an EPA affordability determination, the 
incremental burden, whether a system passes the second step, the identification of disadvantaged 
communities, could very well depend on when in the business cycle affordability is being 
evaluated.  If the economy is expanding (and with it MHI increasing and poverty and 
unemployment rates decreasing), the levels of need that will have to be met will be less stringent.  
In a recession, however, the levels of poverty and unemployment will be higher making it harder 
for communities to qualify for variances.  This is unfortunate, because when poverty and 
unemployment rates are high is exactly when systems will have trouble affording treatment.  A 
few alternative arrangements that can alleviate these problems exist and will be considered in a 
following section. 
Sensitivity Analysis 
 Figures 4-2, 4-3 and 4-4 show the number of counties, the number of systems in those 
counties that could qualify and customers of these small systems that could be affected under 
various ratios for the MHI, the non-MA MHI, the poverty rate and the unemployment rate.  Half 
of the MHI qualifies 63 counties and 373 systems while the most generous 100% of MHI 
potentially affects 2,602 counties and 32,685 small systems.  For half of non-MA MHI, five 
counties and ten systems are potentially affected with the entire non-MA MHI qualifying 1,422 
counties and 13,597 small systems.  The toughest poverty rate criteria, three times the national 
rate, qualifies 51 systems in ten counties and the most generous, the national average, finds  
71 
 
1,642 counties with 21,210 small systems potentially qualifying for variances.  Finally, three 
times the unemployment rate qualifies two counties and 55 small systems and 1,044 counties and 
15,751 small systems have unemployment greater than the national average. 
 
Figure 4-2 ? Counties qualifying at various ratios of (a) MHI and non-MA MHI and (b) poverty 
and unemployment rates for the EDC criteria during the sensitivity analysis 
  
0 
500 
1000 
1500 
2000 
2500 
0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 
Co
un
ties 
Pot
en
tial
ly 
Qu
ali
fyi
ng
 
Ratio of MHI to National MHI To Qualify as an EDC 
MHI non-MA MHI 
 -    
 500  
 1,000  
 1,500  
1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3 
Co
un
ties 
Pot
en
tial
ly 
Qu
ali
fyi
ng
 
Ratio of Poverty/Unemployment to National Average to Qualify as an EDC 
Poverty Unemployment 
(b) 
(a) 
72 
 
 
 
Figure 4-3 ? Small systems qualifying at various ratios of (a) MHI and non-MA MHI and (b) 
poverty and unemployment rates for the EDC criteria during the sensitivity analysis 
  
 -    
 5  
 10  
 15  
 20  
 25  
 30  
0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 
Sm
all
 Sy
ste
ms 
Pot
en
tial
ly 
Qu
ali
fyi
ng
 
(Th
ou
san
ds)
 
Ratio of MHI to National MHI to Qualify as an EDC  
MHI non-MA MHI 
(a) 
 -    
 5  
 10  
 15  
 20  
1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3 
Sm
all
 Sy
ste
ms 
Pot
en
tial
ly 
Qu
ali
fyi
ng
 
(Th
ou
san
ds)
 
Ratio of Poverty/Unemployment Rate to National Average to Qualify as an EDC 
Poverty Unemployment 
(b) 
73 
 
 
 
Figure 4-4 ? Customers potentially affected at various ratios of (a) MHI and non-MA MHI and 
(b) poverty and unemployment rates for the EDC criteria during the sensitivity analysis 
  
 -    
 5  
 10  
 15  
 20  
 25  
 30  
 35  
 40  
0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 
Cus
tom
ers
 Po
tent
ially
 Af
fec
ted 
(M
illi
ons
) 
Ratio of MHI to National MHI to Qualify as an EDC 
MHI non-MA MHI 
(a) 
 -    
 5  
 10  
 15  
 20  
 25  
 30  
1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3 
Cu
sto
me
rs 
Pot
en
tial
ly 
Aff
ec
ted
 (M
illi
on
s) 
Ratio of Poverty/Unemployment Rate to National Average to Qualify as an EDC  
Poverty Unemployment 
(b) 
74 
 
The criteria used by other systems showed significant variability in the amount of 
systems they would qualify for permission to use variances as can be seen from Table 4-3.  From 
the toughest standard, MHI below the poverty line (granting permission to 14 counties and 68 
small systems), to the most generous, unemployment above the national average (1,044 counties 
and 15,751 systems), there is a large variability in other definitions that could be applied to the 
EPA?s affordability determination.  In general, the EPA?s proposed criteria are more stringent 
than these proposed criteria:  unemployment is the strictest of all of them and poverty only less 
strict than the other standard.  The MHI is of middle difficulty compared to the other standards. 
 
Table 4-3 ? Other regulatory definitions of EDCs 
Category Source Level Counties Systems 
MHI Thresholds 
Poverty Level Ag. Dept.?s Rural Utilities Service $22,113 14 68 
$17,000 ($1989) EPA?s ?Road to Financing? $29,985 172 1017 
80% non-MA MHI Ag. Dept?s Rural Utilities Service, 
Rural Development Admin.,  
Small Business Admin.?s HUBZones 
$33,152 397 2,779 
Poverty Rate Thresholds 
Poverty Rate<30% Treasury Department 30% 92 606 
Unemployment Thresholds 
Avg. Unemployment EPA?s ?Road to Financing? 8.7% 1,044 15,751 
Avg. + 1% Commerce Dept.?s Econ. Dev. Admin. 9.7% 680 9,544 
120% Unemployment State of Washington 10.4% 474 6,244 
125% Unemployment EPA?s CSO Financial Capacity Rev. 10.9% 365 5,004 
140% Unemployment Small Business Admin.?s HUBZones 12.2% 189 2,491 
150% Unemployment Treasury Department 13.1% 114 1,648 
75 
 
 The results of the sensitivity analysis can be used to develop a set of alternative ratios that 
produce similar results to the original determinations by the EPA, assuming that the number of 
systems that would be granted variances in 2006 was acceptable.  To do this, the ratios will be 
adjusted until they identify a percentage of all small systems that is similar to the ones identified 
by the EPA in 2006.  The ratios that meet these conditions are 0.646 for MHI (3,445 small 
systems), 2.16 for poverty (518 small systems) and 1.71 for unemployment (920 small systems).  
These three criteria identify 4,190 systems as potentially qualifying for variances or 
approximately 8.9% of all small systems (the EPA originally found that 8.8% of all small 
systems in 2006 would qualify).  As should have been expected MHI multiple (which qualified 
about the same number of systems in both 2006 and today) remained relatively unchanged, the 
poverty multiple was made significantly higher and the unemployment factor significantly 
decreased. 
 An alternative way of assessing the ratios to determine appropriate levels would be to 
base them on the criteria least likely to grant variances used by other agencies and comparing the 
difficulty in the EPA?s affordability criteria.  Table 4-2 shows the number of counties qualifying 
and systems affected by the other measures used by various programs and states in determining 
EDCs.  The strictest standards are the Department of Agriculture?s for household income (county 
MHI below the poverty level) and the Treasury Department?s for poverty (above 30%) and 
unemployment (above 150% the national average which is 13.1%) rates.  Under these conditions 
2,077 small systems in 178 counties could potentially qualify for permission to use variance 
technology.  This is significantly fewer systems than under the EPA?s proposed method, mostly 
because of the severe tightening of the threshold required for systems to meet and qualify based 
76 
 
on MHI.  The Department of Agriculture?s standard is a significant drop as the poverty level is 
2/3rds of 65% of the MHI.   
Alternatives to the Current Ratios and Measures 
 The current methods run into problems when used during a recession or before the 
economy fully recovers from a recession.  As highlighted above, when the economy is in 
recession, poverty and unemployment rates are high, which makes the thresholds that must be 
met in order to qualify for variances even higher at a time when flexibility is most needed in 
granting permission to use variance technology.  There are two other alternative ways of 
handling these ratios, setting fixed thresholds that do not change to reflect the general health of 
the economy or setting the threshold as a certain number of percentage points above the current 
level of a socioeconomic indicator.  It is important to note that finding an appropriate level for 
these thresholds is fundamentally a political question.  In the methods outlined below, the author 
makes two assumptions to address potential political concerns.  First, the author assumes systems 
can demonstrate the need for variances if it is located in a county that has a poverty or 
unemployment rate that is twice a recent national average rate.  Second, the author assumed that 
the amount that the poverty or unemployment rate had to be in excess of in 2006 in order to 
qualify for variances (i.e. the poverty and unemployment rates in 2005) were acceptable amounts 
to set the threshold by. 
 For the first method, thresholds could be identified that would stay the same at all times 
and would therefore be independent of the fluctuations of the business cycle.  There are two 
possible ways of doing that.  First, the EPA could adopt the thresholds used in the original 
analysis as permanent thresholds.  For the first method, the thresholds are $30,935 (65% of MHI 
in 2005 adjusted only for inflation to 2010), average poverty rate of 26.6% (13.3% of the 
77 
 
population was in poverty in 2005) and average unemployment of 10.6% (for the period from 
Jan 2004-Dec 2005, average unemployment was 5.3%).  Using the 2010 American Community 
Survey data and the fixed ratio based on EPA?s 2006 criteria, 6,797 systems in 573 counties 
potentially qualify for variances (1,491 systems in 235 counties for income, 1,341 systems in 178 
counties for poverty and 5,687 systems in 428 counties for unemployment).  Table 4-4 shows the 
results for this method and the methods considered in the remainder of this section. 
Table 4-4 ? Alternative criteria for EDC definitions 
Method Description Criteria Applied Counties Affected 
(Systems in Parentheses) 
Thresholds Are Fixed Over Time 
Original poverty and unemployment 
thresholds from 2006 proposal are 
adopted permanently. 
Poverty Rate (26.6%) 178 (1,341) 
Unemployment (10.6%) 428 (5,687) 
Total 573 (6,797) 
Twice the long term average poverty 
and unemployment rates are adopted 
permanently. 
Poverty Rate (28.1%) 133 (1,023) 
Unemployment (14.0%) 91 (1,486) 
Total 481 (4,517) 
Thresholds Vary Over Time 
2006 poverty and unemployment rates 
added to current rates to develop 
thresholds. 
Poverty Rate (27.6%) 147 (1,130) 
Unemployment (14.0%) 79 (1,195) 
Total 545 (4,787) 
Average poverty and unemployment 
rates added to current rates to develop 
thresholds. 
Poverty Rate (28.3%) 128 (999) 
Unemployment (15.5%) 34 (450) 
Total 521 (4,173) 
Consumption Measures 
Twice the SNAP participation rate SNAP Rate (18.5%) 316 (2,448) 
15% + Households with rent greater 
than 35% of income 
Rent Expenditure (44.4%) 132 (2,967) 
   
78 
 
 Alternatively, longer term averages for MHI, poverty rate and unemployment rate could 
be used to account for variability in these metrics over the business cycle.  Using the 1-year 
estimates of the American Community Survey from 2005-2011 and BLS unemployment data 
over the same period, 65% of the MHI (adjusted for inflation to 2010) was $33,441, twice the 
average poverty rate of 28.1% (the average poverty rate over the period was 14.0%) and twice 
the average unemployment of 13.7% (average unemployment was 6.8% over the period).  Using 
the 2010 census data and the average ratio method, 4,517 systems in 481 counties potentially 
qualify for variances (3,060 systems in 420 counties for income, 1,023 systems in 133 counties 
for poverty and 1,486 systems in 91 counties for unemployment).  
 Under the second method the threshold will be set as a certain amount of percentage 
points above the current poverty and unemployment measures.  There are two different ways to 
do this.  First, we can use the poverty and unemployment rates the EPA used in proposing this 
EDC definition (13.3% and 5.3%, respectively) and add this to the current poverty and 
unemployment rates to develop the thresholds.  This would lead to the adoption of an EDC 
definition of counties that have poverty rates greater than 27.6% or unemployment rates greater 
than 14.0%.  In 2010, 147 counties with 1,130 small systems exceeded the poverty threshold and 
1,195 small systems in 79 counties met the unemployment threshold.  With the addition of the 
65% of MHI criteria, 4,787 small systems in 545 counties could potentially qualify for variances.   
 The second technique uses long-term average poverty and unemployment rates (14.0% 
and 6.8%) instead of the rates from the original EPA analysis.  Doing so sets a poverty threshold 
of 28.3% and an unemployment threshold of 15.5%.  Using the most recent data, 999 systems in 
128 counties exceed the poverty rate threshold and 450 small systems in 34 counties meet the 
79 
 
unemployment threshold.  If these two criteria are used in conjunction with the MHI criteria, 
4,173 small systems in 521 counties could potentially qualify for variances. 
 There exists another source of measures for whether a system could potentially need 
permission to use variance technologies.  The purpose of variance technologies is to prevent 
environmental injustices to consumers by allowing those whose incomes could be strained by 
mandating centralized treatment technologies to have cheaper point-of-use and point-of-entry 
technologies (Cory and Rahman 2009).  The money that is saved can then be spent on other risk-
reducing technologies or on other household necessities (Balazs et al. 2012; Beecher and 
Shanaghan 1998; Cory and Rahman 2009; Jones and Joy 2006; EPA 1998).  A new drinking 
water regulation that forces households to make a trade-off between paying for water and paying 
for another necessary expenditure should be found unaffordable.  This is a question not of 
income resources, but of consumption decisions faced by the consumers and the economics 
literature identifies several metrics that capture these decisions better. 
 There is evidence in the literature that suggests that poverty is better measured by looking 
for concrete examples of economic hardship (Citro and Michael 1995; Pilauskus et al., 2012).  
Consumption measures offer several advantages over using income as a measure of poverty and 
affordability.  The biggest such advantage is that consumption measures for the same household 
are more stable over time than income measures.  Families consume based not entirely on current 
income, but on expected streams of income.  In essence, households that experience temporary 
household income shocks (often from unemployment or changes in the family structure) do not 
necessarily adjust their consumption patterns, often by tapping into savings, borrowing from 
friends or family, or using credit to make up the difference between income and expenses and 
thereby maintain a constant standard of living.  Consumption taps into (predictions of) a 
80 
 
household?s longer term income and therefore the ability of households to afford their basic 
needs (Citro and Michael 1995; Meyer and Sullivan 2003). 
 Consumption measures and (more importantly) ways of identifying households that are 
poor based on consumption measures can be found in the literature.  Identifying households 
based on being consumption poor is typically done by analyzing material hardship in four 
domains: food security, housing security, access to medical treatment and ability to pay bills 
(Pilkauskas et al. 2012).  While many studies and surveys by academics and government 
agencies (most notably the Survey of Income and Program Participation sponsored by the CB) 
capture various measures of material hardships, none of them have significant data that is 
organized on a county-by-county basis and so these studies are of little use in affordability 
determinations.  Fortunately there exist three potential characteristics measured by the CB as part 
of the American Community Survey that are presented on a county level and could potentially 
allow for operationalization of three of the four categories of material hardship.  The lack of 
health insurance can serve as a proxy measure of ability to obtain medical care; however, 
because not everyone decides to purchase insurance, it is a poor measure of material hardship on 
the medical care dimension.  The second, rent as a share of income above 35% of income, can 
serve as an indicator that a household is spending a large amount of income on housing and 
expenses on other items or rises in utility costs could pose a potential problem for the household 
budget.  Finally, participation rates in the Supplemental Nutrition Assistance Program (SNAP, 
also known more colloquially as the Food Stamps program) can serve as measures of a lack of 
food security on the justification that SNAP participants have income that is insufficient to cover 
their food expenses.  These last two measures of hardship populations will serve as indicators 
that a community could potentially not afford water treatment. 
81 
 
 The criteria studied for these two measures will be if food stamps participation in a 
county is greater than twice the national average or if the share of population with rent in excess 
of 35% of their income is 15% higher than the national average.  For this first measure, the 
national SNAP participation rate is 9.3% and the county threshold is 18.5%.  Using 2010 ACS 5-
year estimates, 316 counties with 2,448 small systems could qualify for variances with this 
threshold.  In the rent as a share of income measure, 38.6% of the U.S. population has rent that 
exceeds 35% of their income and so if a county has 44.4% of its population paying more than 
35% of their income in rent it can qualify for a variance.  Nationwide, 132 counties and 2,967 
systems could potentially qualify for permission to use variances under this criteria.  These two 
measures identify approximately the same number of counties and systems as the definitions 
considered currently under consideration by the EPA, but since they more accurately represent 
the consumption choices faced by consumers that are the heart of affordability determinations, 
that these consumption measures are desirable alternatives to the EPA measures.   
Conclusions 
 This study demonstrates that, using recent data from the CB and BLS, at least 3,812 
systems serving less than 10,000 customers located in 511 counties could potentially qualify 
according to EPA?s 2006 proposed affordability criteria for permission to use variance 
technologies to treat drinking water to the levels mandated by the EPA under the SDWA.  This is 
about five hundred less systems than when this methodology was initially proposed in 2006, 
despite more counties qualifying for variances.  While some of this reduction may be due to 
consolidation of small systems, fewer counties qualified because of their unemployment level 
than did in 2006.  The high national level of unemployment in the study period demonstrate a 
potentially significant weakness for the EPA methodology, namely that it is likely to grant fewer 
82 
 
variances when the health of the economy is poor and systems and customers most in need of 
them.   
 In order to address this problem two different methods of establishing threshold criteria 
were considered, setting permanent threshold levels or using material hardship measures that are 
less likely to be affected by short term income shocks like unemployment.  The results of these 
methods are reported in Table 4-4.  These alternative methods either do not change during the 
business cycle or fluctuate significantly less.  These criteria therefore can help alleviate the risk 
that systems qualifying for variances depend on the health of the whole United States? economy 
(instead of just the health of the local economy).  Furthermore, with the consumption indicators 
of household levels of income, EDC definitions can consider a potentially more 
methodologically sound consideration of poverty (Citro and Michael 1995).  These proposed 
revisions should be taken into consideration to the EPA and state drinking water primacy 
agencies in developing affordability criteria for their drinking water programs.    
  
83 
 
 
 
 
CHAPTER 5 
 
CONCLUSIONS AND RECOMMENDATIONS 
 Statistical data was used to study the implications of various methods for determining 
whether treatment to meet drinking water standards set by the EPA is affordable.  There were 
three hypotheses of this study:  
1) The EPA?s current and proposed methodologies are insufficient because the current 
methodology is too difficult to trigger and the proposed methodology is not strict enough 
when it comes to granting variances;  
2) Using the EEC proposed income level of the first quartile household income (Q1HI) as 
the income used for affordability determinations can successfully identify systems that 
deserve variances, without being too lenient; and  
3) That other definitions of EDCs will successfully identify these communities.   
These three hypotheses gave rise to three objectives: 
1) Application of the EPA?s current and proposed criteria to make a determination about the 
affordability of the arsenic MCL for small water systems to study performance of EPA 
affordability methodologies; 
2) Use of the EEC?s recommendation for a lower income level, the first quartile, as the 
income for the burden measure for an affordability determination of the arsenic MCL for 
small water systems to demonstrate the ability of the first quartile household income ; and 
84 
 
3) Performance of a sensitivity analysis on EPA socioeconomic status variance criteria for 
EDCs and testing of other definitions of EDCs. 
These three objectives were completed by the use of statistical analysis of treatment costs and 
socioeconomic data.  The conclusions of these studies are outlined below followed by 
recommendations for future research and policy makers. 
CONCLUSIONS 
 The results of a study of the costs and burdens associated with the removal of arsenic in 
drinking water in dozens of sites located across the United States were detailed in Chapter 3.  In 
doing so it addressed Hypotheses 1 and 2 and Objectives 1 and 2.  Arsenic removal was selected 
as the regulatory burden studied for two reasons.  First, the arsenic rule was the first rule under 
the SDWA to have its affordability significantly challenged.  This challenge was used to critique 
the expenditure margin approach to drinking water affordability.  Furthermore, the discussion 
over affordability of the arsenic rule is still ongoing, with the rule being questioned on the 
grounds of affordability in the published literature more than a decade after its finalization in 
2001.  Second, a large set of readily available cost data has been compiled and released by the 
EPA that represents a wide array of arsenic treatment costs and socioeconomic conditions.  This 
cost data was used to compute the annual costs for arsenic removal in forty CWSs located 
throughout the United States.  Socioeconomic data from these forty sites (MHI, Q1HI, poverty 
rate and unemployment rate) was compiled from the CB and BLS.  This data was used to run the 
EPA?s expenditure margin and incremental burden/disadvantaged community approaches to 
affordability determinations.  The Q1HI was then substituted for MHI in the affordability 
determination methodologies. 
85 
 
 The results of this study confirm the EPA?s original affordability determination, that the 
arsenic regulation was affordable for all systems based on the predicted average costs.  Even 
when affordability determinations were made on a system-by-system basis using the actual costs, 
the arsenic rule was affordable and none of the costs exceeded 2.5% of the system?s customer-
base?s MHI.  The incremental burden/disadvantaged community approach is more generous with 
granting variances.  Systems serving 500 or fewer people qualify for variances based on the 
incremental burden criteria when an acceptable burden for treatment is set at 0.25% of national 
MHI.  Under the most likely scenario for identifying EDCs (analysis done at the county level), 
only one additional system qualifies for variances.  In total 18 of the 40 systems in the sample 
would qualify for variances based on this approach.  When the arsenic removal costs are 
compared to the Q1HI, seven systems have treatment costs in excess of 2.5% of their Q1HI. 
 An analysis of the metrics used by the EPA to identify the EDCs that qualify for 
variances was carried out in Chapter 4.  This analysis addressed Hypothesis and Objective 3.  
The MHI, poverty rate and unemployment rate were collected for the 3,143 counties and county 
equivalents in the U.S. states and Washington, D.C. and analyzed using a geographic information 
system.  This data was first used to update the total number of counties, small systems and 
population potentially affected by the EPA?s definition of an EDC using 2010 data.  Following 
the updated count of affected systems, a sensitivity analysis was performed by adjusting the 
ratios and several specific definitions of EDC used by other agencies and programs.  Lastly, a 
few other measures, the population with rent in excess of 35% of their income and participation 
in the Supplemental Nutrition Assistance Program (food stamps) were used to capture 
consumption choices faced by consumers. 
86 
 
    In this analysis, 3,812 small systems located in 511 counties qualified using the EPA?s 
definition of an EDC with the most recent socioeconomic data.  While fewer systems qualified 
with this data compared to the EPA?s original analysis, more counties and more people were in 
these potentially affected systems.  The results of the sensitivity analysis were used to identify a 
new set of ratios (0.646 for MHI, 2.16 for poverty and 1.71 for unemployment) that found the 
same proportion of systems potentially qualifying for variances as in the original analysis based 
on the assumption that the percentage of systems qualifying was politically acceptable.  Finally, 
alternative indicators that households and therefore systems may not be able to afford water 
treatment, SNAP participation and share of households paying more than 35% of income in rent, 
were examined and identified 316 counties with 2,448 systems and 132 counties with 2,967 
systems respectively potentially qualifying for variances under those two criteria. 
 Together these two chapters examined the EPA?s affordability methodology and found 
both the method the EPA has implemented and has considered and found them lacking.  The 
methods themselves ignore those who are most impacted by a higher cost of drinking water 
treatment regulatory burdens, the poor.  This thesis sought to better develop the environmental 
justice literature on drinking water regulations by studying different ways to bring the situations 
of the poor directly into the affordability determination process.  Chapter 3 introduced the usage 
of a below MHI, the 25th percentile household income, to explicitly address the incomes of the 
poor in using expenditure-to-income ratios as part of the determination process.  Chapter 4 
addressed the inadequacy of the ratios during economic downturns and proposed two alternative 
criteria to be used in classifying a community as economically disadvantaged that better identify 
houses that have a true need for relief from drinking water regulations. 
 
87 
 
RECCOMEDATIONS 
 The impact of price and household income on the quantity of water demanded is well 
documented in the literature (Agthe and Billings 1987; Espey et al. 1997; Fenrick and Getachew 
2012).  However there is significantly less (i.e. to the knowledge of the author, none) to be found 
that discusses how households make the trade-off between water and other essential needs. This 
lack of research will hamper decision making and development of the tools necessary to analyze 
the question of affordability.  Fundamental research, probably in the form of detailed case studies 
or targeted survey research, on when households face the decisions to make these trade-offs, how 
they decide what to give up, what is essential for their needs and how these decisions impact 
their quality of life need to be completed.  By answering those questions, policy makers can 
better understand what the signs are that households are about to make these tough decisions and 
how to assist those who are already making them.   
 Only after research has been completed on households that face affordability problems 
can policy makers and those that seek to provide them with decision support tools decide what 
signs indicate that the a non-negligible number of households in a system?s customer base could 
be pushed over the edge into unaffordable water bills.  The current indicators (low MHI in a 
county, high poverty rate in a county, high unemployment rate in a county) and those proposed 
in this thesis (low Q1HI in a county, high participation rate in SNAP programs in a county, high 
proportion of a county?s population that spends more than 35% of household income on housing) 
have theoretical justifications and may serve as valid indicators.  However, there is little 
empirical evidence to demonstrate that if a county has a low median or Q1HI or high poverty, 
unemployment, SNAP participation or household expenditures on housing, that they have a need 
to grant variances to their water systems.  Once research has identified what the signs are that 
88 
 
water will be unaffordable to households, the above indicators can be verified or new ones with 
stronger empirical foundations can be developed.  However, any indicator that can be 
empirically verified will need to be further evaluated for suitability in use.  Is the indicator 
selective enough to highlight counties where there is a need for variances as opposed to being too 
generous?  Is the indicator easily available by use of data that is aggregated at a county-level or 
will the EPA, state agencies, counties or systems themselves (the worst option) be responsible 
for collecting and analyzing the data?  Those questions also will need to be answered if policy 
makers are to use the appropriate affordability criteria in their decision making. 
 Once indicators have been identified, it will be up to analysts to develop the tools that 
collect, compile and make the data useful to policy makers and regulators.  While the exact 
nature of the decision support tool and its content depends on the nature of the affordability 
methodology adopted, there are several features that will most certainly need to be included, and 
in several cases the information may need to be updated.  The decision tool will need to include 
any socioeconomic variables that are used in the methodology.  If a method that incorporates 
household expenditures on drinking water is used, this data will need to be included and updated 
since the last data set on cost of drinking water was done in 2006 (U.S. Environmental Protection 
Agency 2009) and ignores regional variations (the last study that took regional variations into 
account was performed using 2000 census data (Rubin 2004, 2005)).  Because the number of 
individuals and systems granted variances is a crucial factor in understanding their impact on 
consumers, any decision tool should incorporate the total number of systems and individuals 
affected by variances so that increases in health risks can be better predicted.  This data is readily 
available from the EPA in the form of the SDWIS/FED database.  The decision support tool 
89 
 
developed from this data would be useful to policy makers in informing them about the impacts 
on public health when making decisions about drinking water affordability. 
 A further complication that policy researchers will need to tackle and incorporate into 
affordability metrics is the rising cost of distribution system maintenance.  Maintenance and 
replacement costs are expected to cost the American public a trillion dollars in the coming 
decades (American Society of Civil Engineers 2013), and this will most certainly impact the cost 
and affordability of drinking water provision outside of the regulatory framework.  Finding a 
way to properly incorporate maintenance expenditures in addition to regulatory changes into an 
affordability framework will be crucial to account for the costs consumers face for water and will 
have major ramifications on policy outcomes.  A related issue is how external funding (e.g. the 
Drinking Water State Revolving Fund) for drinking water treatment and distribution systems 
affects the costs faced by consumers and the affordability of water treatment.  Finding ways to 
incorporate the availability of external funding sources into affordability metrics will also need 
to be done as well to ensure that systems that receive permission to use variance technologies 
truly have no way to fund the upgrades they need 
 In developing their affordability criteria, the EPA decided to focus on the ?typical or 
average situation, rather than a worst case scenario? (U.S. Environmental Protection Agency 
2002b).  The purpose of providing small systems with permission to use variances is to ensure 
that customers in those systems are protected from excessive treatment ? associated expenses so 
that the money can be spent on other household essentials and risk-reducing measures.  
Affordability criteria should be designed with that purpose in mind.   
 This research addresses how a concern for the poor and those unable to pay for water 
services can be inserted into the regulatory framework for drinking water standards.  
90 
 
Incorporating a below MHI allows the EPA to take explicitly address the needs of these groups.  
The results outlined in Chapter 3 showed that using a lower income, while most certainly 
incorporating a concern for those who would struggle to support increases in water charges, is 
not quite the ?worst-case scenario? it sounds like it has the potential to be.  Because using the 
first quartile household income is not a worst-case scenario, and given its advantage of being a 
more accurate representation of the burden on households that are most affected by new drinking 
water rules, the EPA and state primacy agencies should adopt the method outlined in Chapter 3.  
This method of granting variances to communities where drinking water expenses per household 
to 2.5% of the Q1HI (or located in EDCs), is therefore strongly recommended. 
 Chapter 4 also illustrates the effects of how the EPA?s definition of an EDC in finding in 
favor of granting systems permission to use variance technology.  It demonstrated how the 
measures used to indicate need could be altered to give them a better methodological justification 
without dramatically adjusting the number of systems that could potentially qualify.  It also 
pointed out a serious flaw in the current EPA definition of an EDC, that it depends on business 
cycles in a way that makes it harder to be called an EDC when the general economy is in a 
recession.  The recommended definitions of an EDC outlined in this chapter (fixed definitions 
that do not vary over time; definitions that fluctuate over the business cycle, but less than the 
current method; and definitions that tap into measures of longer-term, consumption measures of 
income resources and predictions ? food stamp participation and expenditures on housing) 
reduce the effects of the business cycle on drinking water affordability.  Because of this reduced 
effect, they should also be considered by the EPA and state primacy agencies for their definitions 
of EDCs.   
91 
 
 While the regulatory burden studied for this thesis was the arsenic rule, it?s not hard to 
imagine other regulations running into questions of affordability.  Under the 1996 SDWA 
Amendments every potential drinking water rule (except for microbial contaminants and 
indicators of microbial contamination rules) is evaluated by the EPA for affordability to small 
systems.  While most rules won?t be classified as unaffordable (unless the EPA fails to adopt any 
changes to its affordability measures), it is not unforeseeable that a few rules will.    
Affordability therefore needs to be something that can be determined reasonably easily and 
consistently going forward.  As a result of this, readily available data is crucial to any useful 
affordability methodology.  All the data used in the methods are already collected by the Census 
Bureau and the Bureau of Labor Statistics, and most are currently reported as summary statistics.  
This data reporting makes it easier to use them in an affordability methodology adopted by the 
EPA or state primacy agencies based on these results. 
 Affordability and EDCs are not unique concepts to EPA drinking water regulations.  
Affordability is a central concept in nearly two dozen environmental regulations at the EPA 
(Beecher and Shanaghan 1998; U.S. Environmental Protection Agency 1998).  State primacy 
agencies are also compelled to adopt affordability methods for making determinations for 
making decisions on which systems to provide money to as part of the Safe Drinking Water 
Revolving Fund.  Several federal agencies target grant money to EDCs, each with their own 
definition of an EDC.  Many of these definitions could be updated to reflect the research 
undertaken here. 
  
  
92 
 
 
 
 
References 
Abt Associates Inc. (2000). Arsenic in Drinking Water Rule Economic Analysis. Bethesda, MD. 
Agthe, D. E., and Billings, R. B. (1987). ?Equity, Price Elasticity, and Household Income under 
Increasing Block Rates for Water.? The American Journal of Economics and Sociology, 
46(3), 273?286. 
American Society of Civil Engineers. (2013). ?2013 Report Card for America?s Infrastructure.? 
<http://www.infrastructurereportcard.org/> (Mar. 19, 2013). 
Amini, M., Abbaspour, K. C., Berg, M., Winkel, L., Hug, S. J., Hoehn, E., Yang, H., and 
Johnson, C. A. (2008). ?Statistical modeling of global geogenic arsenic contamination in 
groundwater.? Environmental science & technology, 42(10), 3669?75. 
Announcement of Small System Compliance Technology Lists for Existing National Primary 
 Drinking Water Regulations and Findings Concerning Variance Technologies, Notice of 
 lists of technologies and upcoming release of guidance and support documents, 63 Fed. 
 Reg. 42032-42048 (August 6, 1998).  
Arrow, K. J., Cropper, M. L., Eads, G. C., Hahn, R. W., Lave, L. B., Noll, R. G., Portney, P. R., 
Russell, M., Schmalensee, R., Smith, V. K., and Stavins, R. N. (1996). ?Is There a Role for 
Benefit-Cost Analysis in Environmental, Health, and Safety Regulation?? Science, 272, 
221?222. 
Baird, G. M. (2010). ?Water Affordability?: Who?s Going to Pick Up the Check??? Journal of the 
American Water Works Association, 102(12), 16?23. 
Balazs, C. L., Morello-Frosch, R., Hubbard, A. E., and Ray, I. (2012). ?Environmental justice 
implications of arsenic contamination in California?s San Joaquin Valley: a cross-sectional, 
cluster-design examining exposure and compliance in community drinking water systems.? 
Environmental health?: a global access science source, Environmental Health, 11(1), 84. 
Bates, M.N., Smith, A. H., and Cantor, K. P. (1995). ?Case-control study of bladder cancer and 
arsenic in drinking water.? American Journal of Epidemiology, 141, 523?530. 
Beecher, J. A., and Shanaghan, P. E. (1998). ?Water affordability and the DWSRF.? Journal of 
the American Water Works Association, 90(5), 68?75. 
93 
 
Brown, J. P. (2008). ?Risk Assessment for Arsenic in Drinking Water.? Risk Assessment for 
Chemicals in Drinking Water, R. A. Howd and A. M. Fan, eds., John Wiley & Sons, Inc., 
Hoboken, NJ, pp.213?266. 
Burnett, J. K., and Hahn, R. W. (2001). ?A Costly Benefit.? Regulation, 24(3), 44?49. 
Chen, C.J., Chuang, Y. C., Lin, T. M., and Wu, H. Y. (1985). ?Malignant neoplasms among 
residents of a blackfoot disease-endemic area in Taiwan: High-arsenic artesian well water 
and cancers.? Cancer Research, 45, 5895?5899. 
Chen, C.J., and Wang, C. J. (1990). ?Ecological correlation between arsenic level in well water 
and age-adjusted mortality from malignant neoplasms.? Cancer Research, 50, 5470?5474. 
Chen, H., Frey, M. M., Clifford, D., McNeill, L. S., and Edwards, M. (1999). ?Arsenic treatment 
considerations.? Journal of the American Water Works Association, 91(3), 74?85. 
Chiou, H. Y., Hsueh, Y. M., Liaw, K. F., Horng, S. F., Chiang, M. H., Pu, Y. S., Lin, J. S., 
Huang, C. H., and Chen, C.J. (1995). ?Incidence of internal cancers and ingested inorganic 
arsenic: A seven-year follow-up study in Taiwan.? Cancer Research, 55, 1296?1300. 
Cho, Y., Easter, K. W., and Konishi, Y. (2010). ?Economic evaluation of the new U.S. arsenic 
standard for drinking water: A disaggregate approach.? Water Resources Research, 46(10), 
1?16. 
Cho, Y., Konishi, Y., and Easter, K. W. (2007). Can Rural Communities Comply with the New 
Arsenic Standard for Drinking Water? St. Paul, MN. 
Citro, C. F., and Michael, R. T. (1995). Measuring Poverty: A New Approach. National 
Academies Press, Washington, DC. 
Cooke, S. C. (2005). ?Determining the EPA ?affordability criteria? for Small Water Systems 
Under the 1996 Clean Drinking Water Act.? 
Cory, D. C., and Rahman, T. (2009). ?Environmental justice and enforcement of the safe 
drinking water act: The Arizona arsenic experience.? Ecological Economics, Elsevier B.V., 
68(6), 1825?1837. 
Cuzick, J., Sasieni, P., and Evans, S. (1992). ?Ingested arsenic, keratoses, and bladder cancer.? 
American Journal of Epidemiology, 136, 417?421. 
Departments of Veterans Affairs and Housing and Urban Development, and Independent 
 Agencies Appropriations Act, 2002, Pub. L. No.107-272, 115 Stat. 651 (2001). 
Espey, M., Espey, J., and Shaw, W. D. (1997). ?Price elasticity of residental demand for water: 
A meta-analysis.? Water Resources Research, 33(6), 1369?1374. 
94 
 
Fenrick, S. A., and Getachew, L. (2012). ?Estimation of the effects of price and billing frequency 
on household water demand using a panel of Wisconsin municipalities.? Applied Economics 
Letters, 19(14), 1373?1380. 
Frey, M. M., Owen, D. M., Chowdhury, Z. K., Raucher, R. S., and Edwards, M. A. (1998). ?Cost 
to utilities of a lower MCL for arsenic.? Journal of the American Water Works Association, 
90(3). 
Frost, F. J., Tollestrup, K., Craun, G. F., Raucher, R., Stomp, J., and Chwirka, J. (2002). 
?Evaluation of costs and benefits of a lower arsenic MCL.? Journal of the American Water 
Works Association, 94(3), 71?80. 
Guo, H. R., Chiang, H. S., Hu, S. R., Lipsitz, S. R., and Monson, R. R. (1997). ?Arsenic in 
drinking water and incidence of urinary cancers.? Epidemiology, 8, 545?550. 
Gurian, P. L., and Small, M. J. (2002). ?Point-of-use treatment and the revised arsenic MCL.? 
Journal of the American Water Works Association, 94(3), 101?108. 
Gurian, P. L., Small, M. J., Lockwood, J. R., and Schervish, M. J. (2001a). ?Benefit-cost 
estimation for alternative drinking water maximum contaminant levels.? Water Resources 
Research, 37(8), 2213?2226. 
Gurian, P. L., Small, M. J., Lockwood, J. R., and Schervish, M. J. (2001b). ?Addressing 
Uncertainty and Conflicting Cost Estimates in Revising the Arsenic MCL.? Environmental 
Science & Technology, 35(22), 4414?4420. 
Harrington, W., Morgenstern, R. D., and Nelson, P. (2000). ?On the accuracy of regulatory cost 
estimates.? Journal of Policy Analysis and Management, 19(2), 297?322. 
Heflin, C., Sandberg, J., and Rafail, P. (2009). ?The Structure of Material Hardship in U . S . 
Households?: An Examination of the Coherence behind Common Measures of Well-Being.? 
Social Problems, 56(4), 746?764. 
Heinzerling, L. (2002). ?Markets for Arsenic.? Georgetown Law Journal. 
Hilkert Colby, E. J., Young, T. M., Green, P. G., and Darby, J. L. (2010). ?Costs of Arsenic 
Treatment for Potable Water in California and Comparison to U.S. Environmental 
Protection Agency Affordability Metrics1.? Journal of the American Water Resources 
Association, 46(6), 1238?1254. 
Hogue, C. (2001). ?Arsenic Debate: It?s All About Cost.? Chemical & Engineering News, 
79(42), 20?21. 
Hopenhayn-Rich, C., Biggs, M. L., Fuchs, A., Bergoglio, R., Tello, E. E., Nicolli, H., and Smith, 
A. H. (1996). ?Lung and kidney cancer mortality associated with arsenic in drinking water 
in Cordoba, Argentina.? Epidemiology, 7, 117?124. 
95 
 
Hulchanski, J. D. (1995). "The Concept of Housing Affordability:  Six Contemporary Uses of 
the Housing Expenditure-to-Income Ratio" Housing Studies, 10(4), 471?492. 
Hutchinson, J. (1887). ?Arsenic Cancer.? British Medical Journal, 2, 1280?1281. 
Jones, S. A., and Joy, N. (2006). ?The inequity of the Revised Arsenic Rule for very small 
community drinking water systems.? Environmental Science & Policy, 9(6), 555?562. 
Kapaj, S., Peterson, H., Liber, K., and Bhattacharya, P. (2006). ?Human health effects from 
chronic arsenic poisoning--a review.? Journal of environmental science and health. Part A, 
Toxic/hazardous substances & environmental engineering, 41(10), 2399?428. 
Kolstad, C. D. (2000). Environmental Economics. Oxford University Press, New York, NY. 
Kumar, A., Adak, P., Gurian, P. L., and Lockwood, J. R. (2010). ?Arsenic exposure in US public 
and domestic drinking water supplies: a comparative risk assessment.? Journal of exposure 
science & environmental epidemiology, 20(3), 245?54. 
Levine, R. L. (2012). ?The Need for Congressional Action to Finance Arsenic Reductions in 
Drinking Water.? Journal of Environmental Health, 75(4), 20?25. 
Mayer, S. E., and Jencks, C. (1989). ?Poverty and the Distribution of Material Hardship.? 
Journal of Human Resources, 24(1), 88?114. 
McGavisk, E., Roberson, J. A., and Seidel, C. (2013). ?Using community economics to compare 
arsenic compliance and noncompliance.? Journal American Water Works Association, 
105(3), 115?126. 
Memorandum for the Heads and Acting Heads of Executive Departments and Agencies, 
 Memorandum, 66 Fed. Reg. 7702 (January 24, 2001). 
Meyer, B. D., and Sullivan, J. X. (2003). ?Measuring the Well-Being of the Poor Using Income 
and Consumption.? The Journal of Human Resources, 38(Supplement), 1180?1220. 
Morales, K. H., Ryan, L., Kuo, T L, Wu, M M, and Chen, C J. (2000). ?Risk of internal cancers 
from arsenic in drinking water.? Environmental Health Perspectives, 108(7), 655?61. 
Mushak, P., and Crocetti, A.F., (1995). ?Risk and Revisionism in Arsenic Cancer Risk 
Assessment.? Environmental Health Perspectives, 103(7/8), 684?689. 
National Academy of Science. (1999). Arsenic in Drinking Water. National Academies Press, 
Washington, DC, 330. 
National Academy of Science. (2001). Arsenic in Drinking Water: 2001 Update. National 
Academies Press, Washington, DC, 244. 
96 
 
National Center for Environmental Economics. (2012). Retrospective Study of the Costs of EPA 
Regulations?: An Interim Report of Five Case Studies. 
National Drinking Water Advisory Council (NDWAC). (2001). Report of the Arsenic Cost 
Working Group to the National Drinking Water Advisory Council. Washington, DC. 
National Drinking Water Advisory Council (NDWAC). (2003). Recommendations of the 
National Drinking Water Advisory Council to U.S. EPA on Its National Small Systems 
Affordability Criteria July 2003. Washington, DC. 
National Primary Drinking Water Regulations; Arsenic and Clarifications to Compliance and 
 New Source Contaminants Monitoring:  Delay of Effective Date, Final rule; delay of 
 effective date, 66 Fed. Reg. 28342-28350 (May 22, 2001). 
National Primary Drinking Water Regulations; Arsenic and Clarifications to Compliance and 
 New Source Contaminants Monitoring, Proposed rule; correction, 66 Fed. Reg. 42974-
 42975 (August 16, 2001). 
Oates, W. E. (2002). ?The Arsenic Rule: A Case for Decentralized Standard Setting?? 
Resources, 2(147), 16?18. 
Pilkauskas, N. V, Currie, J. M., and Garfinkel, I. (2012). ?The Great Recession , Public Transfers 
, and Material Hardship.? Social Service Review, 86(3), 401?427. 
Pontius, F. (2008). ?Are Drinking Water Regulations Affordable??? Journal of the American 
Water Works Association, 100(10), 24?34. 
Raftelis Financial Consulting. (2000). Year 2000 Water and Wastewater Rate Survey. Raftelis 
Financial Consulting, Charlotte, NC. 
Rector, R. E., Johnson, K. A., and Youssef, S. E. (1999). ?The Extent of Material Hardship and 
Poverty in the United States.? Review of Social Economy, 57(3), 351?387. 
Reimann, C., and Banks, D. (2004). ?Setting action levels for drinking water: are we protecting 
our health or our economy (or our backs!)?? The Science of the total environment, 332(1-3), 
13?21. 
Roberson, J. A., Lueders, M., Adams, C., and Rosen, J. S. (2009). ?Risk indexes for draft CCL3 
chemicals.? Journal of the American Water Works Association, 101(9), 64?74. 
Rosenbaum, W. A. (2008). Environmental Politics and Policy. CQ Press, Washington, DC, 132. 
Rubin, S. J. (2001). ?Criteria to Assess the Affordability of Water Service.?, National Rural 
Water Association, Duncan, OK. 
97 
 
Rubin, S. J. (2002). ?Criteria to Assess Affordability Concerns in Conference Report for H.R. 
2620.? National Rural Water Association, Duncan, OK. 
Rubin, S. J. (2004). ?The Cost of Water and Wastewater Service in the United States.? National 
Rural Water Association, Duncan, OK. 
Rubin, S. J. (2005). ?Census data shed light on US water and wastewater costs.? Journal of the 
American Water Works Association, 97(4), 99?110. 
Rubin, S. J. (2013a). ?Evaluating violations of drinking water regulations.? Journal American 
Water Works Association, 105(3), 137?147. 
Rubin, S.J. (2013b).  "Structural changes in the water industry from 2001 to 2011."  Journal 
American Water Works Association, 105(3), 148-156. 
"Safe Drinking Water Act" (1974) Congressional Record, 120, p. 36367. 
Scheberle, D. (2004). Federalsim and Environmental Policy. Georgetown University Press, 
Washington, DC, 124?153. 
Seidel, C. J., Najm, N., Blute, N. K., Corwin, C. J., and Angeles, L. (2013). ?National and 
California treatment costs to comply with potential hexavalent chromium MCLs.? 320?336. 
Simpson, R. D. (2011). Do Regulators Overestimate the Costs of Regulation?? Washington, DC. 
Small Drinking Water Systems Variances-Revision of Existing National-Level Affordability 
 Methodology and Methodology to Identify Variance Technologies That Are Protective of 
 Public Health, Notice 71 Fed. Reg. 10671-10685 (March 2, 2006). 
Smedley, P., and Kinniburgh, D. (2002). ?A review of the source, behaviour and distribution of 
arsenic in natural waters.? Applied Geochemistry, 17(5), 517?568. 
Smith, A. H., Hopenhayn-Rich, C., Bates, M N, Goeden, H. M., Hertz-Picciotto, I., Duggan, H. 
M., Wood, R., Kosnett, M. J., and Smith, M. T. (1992). ?Cancer risks from arsenic in 
drinking water.? Environmental health perspectives, 97(6), 259?67. 
Sunstein, C. R. (2002). ?The Arithmetic of Arsenic.? Georgetown Law Journal, (c), 1?53. 
Tiemann, M. (2007). Arsenic in Drinking Water?: Regulatory Developments and Issues. 1?6. 
Tiemann, M. (2010). Safe Drinking Water Act ( SDWA ): Selected Regulatory and Legislative 
Issues. Washington, DC. 
Tseng, W. P. (1977). ?Effects and dose-response relationships of skin cancer and blackfoot 
disease with arsenic.? Environmental Health Perspectives, 19, 109?119. 
98 
 
Tseng, W. P., Chu, H. M., How, S. W., Fong, J. M., Lin, C. S., and Yeh, S. (1968). ?Prevalence 
of skin cancer in an endemic area of chronic arsenicism in Taiwan.? Journal of the National 
Cancer Institute, 40, 453?463. 
Tsuda, T., Babazono, A., Yamamoto, E., Kurumatani, N., Mino, Y., Ogawa, T., Kishi, Y., and 
Aoyama, H. (1995). ?Ingested arsenic and internal cancer: A historical cohort study 
followed for 33 years.? American Journal of Epidemiology, 141, 198?209. 
United States Bureau of Labor Statistics (BLS), 2000.  Labor Force Statistics including the 
 National Unemployment Rate, 1999-2000. 
 
United States Bureau of Labor Statistics (BLS), 2010.  Labor Force Statistics including the 
 National Unemployment Rate, 2008-2010. 
 
United States Bureau of Labor Statistics (BLS), 2012.  CPI Inflation Calculator.  
 http://www.bls.gov/data/inflation_calculator.htm/   (accessed August 2012). 
 
United States Bureau of Labor Statistics (BLS), 2013.  Labor Force Statistics including the 
 National Unemployment Rate, 2011-2012. 
 
United States Census Bureau (CB), 2000.  2000 Decennial Census SF3 Sample Data. 
 
United States Census Bureau (CB), 2006.  2006 American Community Survey. 
 
United States Census Bureau (CB), 2010a.  2010 Decennial Census 2010 Demographic Profile 
 Data. 
 
United States Census Bureau (CB), 2010b.  2010 American Community Survey 5-Year 
 Estimates. 
United States Environmental Protection Agency (EPA). (1998). ?Information for States on 
Developing Affordability Criteria for Drinking Water.? U.S. Environmental Protection 
Agency, Washington. 
United States Environmental Protection Agency (EPA). (2000). The Drinking Water State 
Revolving Fund Program: Case Studies in Implementation III . Disadvantaged 
Communities. Washington, DC. 
United States Environmental Protection Agency (EPA). (2002a). System Survey Volume I?: 
Overview. Washington, DC, 9. 
United Staes Environmental Protection Agency (EPA). (2002b). Report to Congress?: Small 
Systems Arsenic Implementation Issues. Washington, DC. 
United States Environmental Protection Agency (EPA). (2009). 2006 Community Water System 
Survey Volume I?: Overview. Washington, DC, 10. 
99 
 
United States Environmental Protection Agency (EPA). (2011). ?How Does EPA Define 
Environmental Justice?? <http://compliance.supportportal.com/link/portal/23002/ 
23009/Article/32790/How-Does-EPA-define-Environmental-Justice> (Sep. 9, 2012). 
United States Environmental Protection Agency (EPA). (2013).  "SDWISFED Drinking Water 
Data" <http://water.epa.gov/scitech/datait/databases/drink/sdwisfed/pivottables.cfm> (Feb. 
15, 2013). 
United States Environmental Protection Agency-Science Advisory Board (EPA-SAB). (2001). 
Arsenic Rule Benefits Analysis: An SAB Review. Washington, DC. 
United States Environmental Protection Agency-Environmental Economics Committee (EPA-
EEC). (2002). Affordability Criteria for Small Drinking Water Systems: An EPA Science 
Advisory Board Report. Washington, DC. 
United States Small Business Administration. (2013). ?Understanding the HUBZone Program.? 
HUBZone Program, <http://www.sba.gov/content/understanding-hubzone-program> (Feb. 
19, 2013). 
Viscusi, W. K., Magat, W. A., and Huber, J. (1991). ?Pricing Environmental Health Risks: 
Survey Assessments of Risk-Risk and Risk-Dollar Trade-Offs for Chronic Bronchitis.? 
Journal of Environmental Economics and Management, 21(1), 32?51. 
Viscusi, W. K., Vernon, J. M., and Harrington Jr., J. E. (2000). Economics of Regulation and 
Antitrust. The MIT Press, Cambridge, MA. 
Wang, J. S., and Wai, C. M. (2004). ?Arsenic in Drinking Water ? A Global Environmental 
Problem.? Journal of Chemical Education, 81(2), 207?213. 
Wang, L., and Chen, A. S. C. (2011). Costs of Arsenic Removal Technologies for Small Water 
Systems: U.S. EPA Arsenic Removal Technology Demosntration Program. Powell, OH. 
Wilson, R. (2001). ?Underestimating Arsenic?s Risk: The latest science supports tougher 
standards.? Regulation, 24(3), 50?53. 
Woods, R. (2001). ?E.P.A. Announces Arsenic Standard for Drinking Water of 10 Parts per 
Billion.? U.S. E.P.A. News Releases, 
<http://yosemite.epa.gov/opa/admpress.nsf/e51aa292bac25b0b85257359003d925f/6d26c01
5b807156e85256af6007b9bed!OpenDocument> (Nov. 12, 2012). 
Wu, M.M., Kuo, T.L., Hwang, Y. H., and Chen, C.J. (1989). ?Dose-response relation between 
arsenic concentration in well water and mortality from cancers and vascular diseases.? 
American Journal of Epidemiology, 130, 1123?1132. 
 
  
100 
 
 
 
 
APPENDIX A   
 
LIST OF SITES 
 
 
Figure A-1 ? ORD Sites included in Chapter 3 study 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
101 
 
AN Desert Sands Mutual Domestic Water Consumers Association 
Anthony Dona Ana NM 
In a Metropolitan Area? Yes 
 
What Technology is Used? AM Annual Production (kgal): 40,395 
Arsenic Treatment Costs  
( per household): 
$155.55 Baseline Water Costs 
(per household): 
$260.84 
Total Water Costs  
(per household): 
$416.39 Number of Households  
in Systems: 
437 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$22,547 Median Household Income 
(2000): 
$29,808 
 (2010): $22,258  (2010): $36, 657 
25th Percentile Income (2000): $12,540 25th Percentile Income (2000): $14,859 
(2010): $14,927 (2010): $18,333 
Unemployment (2000):  Unemployment (2000): 7.00% 
(2010):  (2010): 7.20% 
Poverty (2000): 38.0% Poverty (2000): 25.4% 
  
  
AR United Water Systems 
Arnaudville St. Landry LA 
In a Metropolitan Area? No 
 
What Technology is Used? IR Annual Production (kgal): 101,152 
Arsenic Treatment Costs  
(per household): 
$42.34 Baseline Water Costs 
(per household): 
$185.21 
Total Water Costs  
(per household): 
$227.55 Number of Households  
in Systems: 
1200 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$21,600 Median Household Income 
(2000): 
$22,855 
 (2010): $23,083  (2010): $31,813 
25th Percentile Income (2000): $11,895 25th Percentile Income (2000): $9,766 
(2010): $11,190 (2010): $13,601 
Unemployment (2000):  Unemployment (2000): 6.45% 
(2010):  (2010): 7.55% 
Poverty (2000): 25.6% Poverty (2000): 29.3% 
(2010): 27.3% (2010): 28.3% 
 
 
 
 
102 
 
AV Oak Manor Municipal Utility District 
Alvin Brazoria TX 
In a Metropolitan Area? Yes 
 
What Technology is Used? AM Annual Production (kgal): 18,928 
Arsenic Treatment Costs  
(per household): 
$161.35 Baseline Water Costs 
(per household): 
$354.94 
Total Water Costs  
(per household): 
$516.29 Number of Households  
in Systems: 
189 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$38,576 Median Household Income 
(2000): 
$48,632 
 (2010): $46,110  (2010): $65,607 
25th Percentile Income (2000): $21,323 25th Percentile Income (2000): $26,120 
(2010): $22,443 (2010): $33,953 
Unemployment (2000):  Unemployment (2000): 5.95% 
(2010):  (2010): 8.45% 
Poverty (2000): 13.3% Poverty (2000): 10.2% 
(2010): 14.9% (2010): 10.6% 
 
 
BC City of Brown City 
Brown City Sanilac MI 
In a Metropolitan Area? No 
 
What Technology is Used? AM Annual Production (kgal): 51,334 
Arsenic Treatment Costs  
(per household): 
$85.44 Baseline Water Costs 
(per household): 
$264.42 
Total Water Costs  
(per household): 
$349.86 Number of Households  
in Systems: 
450 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$33,906 Median Household Income 
(2000): 
$36,870 
 (2010): $28,889  (2010): $40,818 
25th Percentile Income (2000): $21,214 25th Percentile Income (2000): $29,581 
(2010): $11,443 (2010): $21,979 
Unemployment (2000):  Unemployment (2000): 5.20% 
(2010):  (2010): 16.15% 
Poverty (2000): 11.7% Poverty (2000): 10.4% 
(2010): 25.6% (2010): 14.8% 
 
 
 
103 
 
BW White Rock Water Company 
Bow Merrimack NH 
In a Metropolitan Area? No 
 
What Technology is Used? AM Annual Production (kgal): 8,530 
Arsenic Treatment Costs  
(per household): 
$661.70 Baseline Water Costs 
(per household): 
$244.76 
Total Water Costs  
(per household): 
$906.46 Number of Households  
in Systems: 
96 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$79,329 Median Household Income 
(2000): 
$56,842 
 (2010): $101,731  (2010): $63,012 
25th Percentile Income (2000): $54,750 25th Percentile Income (2000): $28,017 
(2010): $58,048 (2010): $34,120 
Unemployment (2000): 1.85% Unemployment (2000): 2.25% 
(2010): 4.10% (2010): 5.60% 
Poverty (2000): 1.8% Poverty (2000): 5.9% 
(2010): 13.8% (2010): 8.1% 
 
 
CM City of Climax 
Climax Polk MN 
In a Metropolitan Area? Yes 
 
What Technology is Used? IR/IA Annual Production (kgal): 13,800 
Arsenic Treatment Costs  
(per household): 
$305.82 Baseline Water Costs 
(per household): 
$225.11 
Total Water Costs  
(per household): 
$530.93 Number of Households  
in Systems: 
111 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$24,688 Median Household Income 
(2000): 
$35,105 
 (2010): $45,938  (2010): $47,233 
25th Percentile Income (2000): $17,083 25th Percentile Income (2000): $18,881 
(2010): $31,560 (2010): $23,559 
Unemployment (2000):  Unemployment (2000): 4.25% 
(2010):  (2010): 5.90% 
Poverty (2000): 11.8% Poverty (2000): 10.9% 
(2010): 3.7% (2010): 13.0% 
 
 
 
104 
 
DM Charette Mobile Home Park 
Dummerston Windham CT 
In a Metropolitan Area? No 
 
What Technology is Used? AM Annual Production (kgal): 571 
Arsenic Treatment Costs  
(per household): 
$584.48 Baseline Water Costs 
(per household): 
$213.20 
Total Water Costs  
(per household): 
$797.68 Number of Households  
in Systems: 
14 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$46,121 Median Household Income 
(2000): 
$38,204 
 (2010): $62,591  (2010): $46,714 
25th Percentile Income (2000): $27,019 25th Percentile Income (2000): $20,800 
(2010): $41,250 (2010): $24,473 
Unemployment (2000): 1.80% Unemployment (2000): 2.75% 
(2010): 4.15% (2010): 6.50% 
Poverty (2000): 2.4% Poverty (2000): 9.4% 
(2010): 5.0% (2010): 11.1% 
 
 
DV Vintage on the Ponds 
Delavan Walworth WI 
In a Metropolitan Area? No 
 
What Technology is Used? IR Annual Production (kgal): 2,200 
Arsenic Treatment Costs  
(per household): 
$134.90 Baseline Water Costs 
(per household): 
$244.76 
Total Water Costs  
(per household): 
$379.66 Number of Households  
in Systems: 
52 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$42,551 Median Household Income 
(2000): 
$46,274 
 (2010): $45,218  (2010): $54,487 
25th Percentile Income (2000): $24,296 25th Percentile Income (2000): $26,393 
(2010): $22,678 (2010): $29,411 
Unemployment (2000):  Unemployment (2000): 2.65% 
(2010):  (2010): 9.05% 
Poverty (2000): 7.6% Poverty (2000): 8.4% 
(2010): 20.7% (2010): 11.7% 
 
 
 
105 
 
FE Town of Felton 
Felton Kent DE 
In a Metropolitan Area? Yes 
 
What Technology is Used? CF Annual Production (kgal): 38,200 
Arsenic Treatment Costs  
(per household): 
$107.90 Baseline Water Costs 
(per household): 
$271.56 
Total Water Costs  
(per household): 
$379.46 Number of Households  
in Systems: 
428 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$42,589 Median Household Income 
(2000): 
$40,950 
 (2010): $44,533  (2010): $53,183 
25th Percentile Income (2000): $27,642 25th Percentile Income (2000): $27,573 
(2010): $22,439 (2010): $28,725 
Unemployment (2000):  Unemployment (2000): 3.55% 
(2010):  (2010): 7.85% 
Poverty (2000): 10.4% Poverty (2000): 10.7% 
(2010): 11.8% (2010): 12.5% 
 
 
FL City of Fruitland 
Fruitland Payette ID 
In a Metropolitan Area? Yes 
 
What Technology is Used? IX Annual Production (kgal): 65,400 
Arsenic Treatment Costs  
(per household): 
$53.77 Baseline Water Costs 
(per household): 
$252.51 
Total Water Costs  
(per household): 
$306.28 Number of Households  
in Systems: 
1377 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$32,469 Median Household Income 
(2000): 
$33,046 
 (2010): $32.855  (2010): $43,559 
25th Percentile Income (2000): $18,345 25th Percentile Income (2000): $18,352 
(2010): $24.758 (2010): $22,793 
Unemployment (2000):  Unemployment (2000): 6.80% 
(2010):  (2010): 8.80% 
Poverty (2000): 11.9% Poverty (2000): 13.2% 
(2010): 18.2% (2010): 15.7% 
 
 
 
106 
 
GE Geneseo Hills Subdivision 
Geneseo Henry IL 
In a Metropolitan Area? Yes 
 
What Technology is Used? AM Annual Production (kgal): 14,868 
Arsenic Treatment Costs  
(per household): 
$329.56 Baseline Water Costs 
(per household): 
$226.30 
Total Water Costs  
(per household): 
$555.86 Number of Households  
in Systems: 
155 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$40,760 Median Household Income 
(2000): 
$39,854 
 (2010): $58,830  (2010): $49,164 
25th Percentile Income (2000): $25,208 25th Percentile Income (2000): $22,466 
(2010): $30,377 (2010): $26,935 
Unemployment (2000):  Unemployment (2000): 4.95% 
(2010):  (2010): 9.05% 
Poverty (2000): 5.3% Poverty (2000): 8.0% 
(2010): 5.0% (2010): 10.4% 
 
 
GF Orchard Highlands Subdivision 
Goffstown Hilsborough  NH 
In a Metropolitan Area? Yes 
 
What Technology is Used? AM Annual Production (kgal): 1,509 
Arsenic Treatment Costs  
(per household): 
$176.78 Baseline Water Costs 
(per household): 
$244.76 
Total Water Costs  
(per household): 
$421.54 Number of Households  
in Systems: 
42 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$55,833 Median Household Income 
(2000): 
$53,384 
 (2010): $76,171  (2010): $69,321 
25th Percentile Income (2000): $34,210 25th Percentile Income (2000): $30,377 
(2010): $39,513 (2010): $37,459 
Unemployment (2000): 2.45% Unemployment (2000): 2.70% 
(2010): 5.15% (2010): 6.40% 
Poverty (2000): 4.2% Poverty (2000): 6.3% 
(2010): 3.4% (2010): 7.2% 
 
 
 
107 
 
GV Town of Greenville 
Greenville Outagamie WI 
In a Metropolitan Area? Yes 
 
What Technology is Used? IR Annual Production (kgal): 24,051 
Arsenic Treatment Costs  
(per household): 
$31.07 Baseline Water Costs 
(per household): 
$245.95 
Total Water Costs  
(per household): 
$277.02 Number of Households  
in Systems: 
1502 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$61,381 Median Household Income 
(2000): 
$49,613 
 (2010): $81,405  (2010): $55,914 
25th Percentile Income (2000): $44,545 25th Percentile Income (2000): $29,824 
(2010): $54,418 (2010): $30,980 
Unemployment (2000):  Unemployment (2000): 4.7% 
(2010):  (2010): 8.30% 
Poverty (2000): 2.0% Poverty (2000): 2.70% 
(2010): 4.5% (2010): 8.5% 
 
 
HD Sunset Ranch Development 
Homedale Owyhee ID 
In a Metropolitan Area? No 
 
What Technology is Used? POU-RO Annual Production (kgal): n/a 
Arsenic Treatment Costs  
(per household): 
$554.45 Baseline Water Costs 
(per household): 
$252.51 
Total Water Costs  
(per household): 
$806.96 Number of Households  
in Systems: 
10 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$24,196 Median Household Income 
(2000): 
$28,339 
 (2010): $36,370  (2010): $33,441 
25th Percentile Income (2000): $13,840 25th Percentile Income (2000): $16,404 
(2010): $17,526 (2010): $20,544 
Unemployment (2000):  Unemployment (2000): 16.9% 
(2010):  (2010): 4.25% 
Poverty (2000): 20.3% Poverty (2000): 3.95% 
(2010): 18.5% (2010): 22.2% 
 
 
 
108 
 
LD Terry Trojan Water District 
Lead Lawrence  SD 
In a Metropolitan Area? No 
 
What Technology is Used? AM Annual Production (kgal): 18,790 
Arsenic Treatment Costs  
(per household): 
$145.02 Baseline Water Costs 
(per household): 
$248.93 
Total Water Costs  
(per household): 
$393.95 Number of Households  
in Systems: 
187 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$29,485 Median Household Income 
(2000): 
$31,755 
 (2010): $38,845  (2010): $42,356 
25th Percentile Income (2000): $17,241 25th Percentile Income (2000): $16,453 
(2010): $22,380 (2010): $21,323 
Unemployment (2000):  Unemployment (2000): 14.8% 
(2010):  (2010): 4.60% 
Poverty (2000): 12.9% Poverty (2000): 3.40% 
(2010): 12.4% (2010): 15.3% 
 
 
LI Upper Bodfish Well CH2-A 
Lake Isabella Kern CA 
In a Metropolitan Area? Yes 
 
What Technology is Used? AM Annual Production (kgal): 9,318 
Arsenic Treatment Costs  
(per household): 
$213.81 Baseline Water Costs 
(per household): 
$242.38 
Total Water Costs  
(per household): 
$456.19 Number of Households  
in Systems: 
200 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$19,813 Median Household Income 
(2000): 
$35,446 
 (2010): $19,627  (2010): $47,089 
25th Percentile Income (2000): $10,478 25th Percentile Income (2000): $18,204 
(2010): $12,863 (2010): $23,934 
Unemployment (2000):  Unemployment (2000): 9.85% 
(2010):  (2010): 15.15% 
Poverty (2000): 20.5% Poverty (2000): 20.8% 
(2010): 21.3% (2010): 20.6% 
 
 
 
109 
 
LW City of Lidgerwood 
Lidgerwood Richaland ND 
In a Metropolitan Area? No 
 
What Technology is Used? SM Annual Production (kgal): 65,7000 
Arsenic Treatment Costs  
(per household): 
$114.69 Baseline Water Costs 
(per household): 
$268.58 
Total Water Costs  
(per household): 
$383.27 Number of Households  
in Systems: 
385 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$25,887 Median Household Income 
(2000): 
$36,098 
 (2010): $31,731  (2010): $47,131 
25th Percentile Income (2000): $14,403 25th Percentile Income (2000): $19,461 
(2010): $21,891 (2010): $26,406 
Unemployment (2000):  Unemployment (2000): 2.70% 
(2010):  (2010): 4.65% 
Poverty (2000): 13.7% Poverty (2000): 10.4% 
(2010): 11.0% (2010): 10.7% 
 
 
NP Nambe Pueblo Tribe 
Nambe Pueblo Santa Fe NM 
In a Metropolitan Area? Yes 
 
What Technology is Used? AM Annual Production (kgal): 30,709 
Arsenic Treatment Costs  
(per household): 
$561.24 Baseline Water Costs 
(per household): 
$294.19 
Total Water Costs  
(per household): 
$855.43 Number of Households  
in Systems: 
150 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$30,452 Median Household Income 
(2000): 
$42,207 
 (2010): $41,343  (2010): $52,696 
25th Percentile Income (2000):  25th Percentile Income (2000): $23,071 
(2010): $19,565 (2010): $25,808 
Unemployment (2000):  Unemployment (2000): 3.20% 
(2010):  (2010): 6.25% 
Poverty (2000): 14.2% Poverty (2000): 12.0% 
(2010): 14.3% (2010): 14.4% 
 
 
 
110 
 
OK City of Okanogan 
Okanogan Okanogan WA 
In a Metropolitan Area? No 
 
What Technology is Used? CF Annual Production (kgal): 139,400 
Arsenic Treatment Costs  
(per household): 
$63.15 Baseline Water Costs 
(per household): 
$344.22 
Total Water Costs  
(per household): 
$407.37 Number of Households  
in Systems: 
1047 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$26,994 Median Household Income 
(2000): 
$29,726 
 (2010): $36,394  (2010): $38,551 
25th Percentile Income (2000): $13,079 25th Percentile Income (2000): $15,393 
(2010): $16,493 (2010): $18,404 
Unemployment (2000):  Unemployment (2000): 8.65% 
(2010):  (2010): 10.15% 
Poverty (2000): 24.3% Poverty (2000): 21.3% 
(2010): 22.7% (2010): 19.5% 
 
 
PF Seely-Brown Village 
Pomfret Windham CT 
In a Metropolitan Area? No 
 
What Technology is Used? AM Annual Production (kgal): 706 
Arsenic Treatment Costs  
(per household): 
$223.73 Baseline Water Costs 
(per household): 
$242.38 
Total Water Costs  
(per household): 
$466.11 Number of Households  
in Systems: 
32 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$57,938 Median Household Income 
(2000): 
$45,115 
 (2010): $68,278  (2010): $59,370 
25th Percentile Income (2000): $34,244 25th Percentile Income (2000): $23,905 
(2010): $41,128 (2010): $30,340 
Unemployment (2000): 2.05% Unemployment (2000): 2.80% 
(2010): 4.60% (2010): 10.9% 
Poverty (2000): 4.1% Poverty (2000): 8.5% 
(2010): 6.0% (2010): 11.4% 
 
 
 
111 
 
PW Village of Pentwaer 
Pentwater Oceana MI 
In a Metropolitan Area? No 
 
What Technology is Used? IR/IA Annual Production (kgal): 38,300 
Arsenic Treatment Costs  
(per household): 
$94.01 Baseline Water Costs 
(per household): 
$209.03 
Total Water Costs  
(per household): 
$303.04 Number of Households  
in Systems: 
450 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$38,542 Median Household Income 
(2000): 
$35,307 
 (2010): $44,524  (2010): $39,543 
25th Percentile Income (2000): $21,311 25th Percentile Income (2000): $19,582 
(2010): $27,766 (2010): $22,098 
Unemployment (2000):  Unemployment (2000): 6.25% 
(2010):  (2010): 15.35% 
Poverty (2000): 8.4% Poverty (2000): 14.7% 
(2010): 5.4% (2010): 19.2% 
 
 
RF Rollinsford Water/Sewer District 
Rollinsford Strafford NH 
In a Metropolitan Area? Yes 
 
What Technology is Used? AM Annual Production (kgal): 21,243 
Arsenic Treatment Costs  
(per household): 
$152.14 Baseline Water Costs 
(per household): 
$243.57 
Total Water Costs  
(per household): 
$395.71 Number of Households  
in Systems: 
654 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$48,588 Median Household Income 
(2000): 
$44,803 
 (2010): $66,161  (2010): $57,809 
25th Percentile Income (2000): $30,106 25th Percentile Income (2000): $25,873 
(2010): $33,268 (2010): $30,463 
Unemployment (2000): 1.05% Unemployment (2000): 2.60% 
(2010): 4.60% (2010): 6.05% 
Poverty (2000): 3.7% Poverty (2000): 9.2% 
(2010): 7.9% (2010): 11.3% 
 
 
 
112 
 
RN South Truckee Meadow General Improvement District 
Reno Washoe NV 
In a Metropolitan Area? Yes 
 
What Technology is Used? AM Annual Production (kgal): 22,885 
Arsenic Treatment Costs  
(per household): 
$47.33 Baseline Water Costs 
(per household): 
$276.33 
Total Water Costs  
(per household): 
$323.66 Number of Households  
in Systems: 
3410 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$40,530 Median Household Income 
(2000): 
$45,815 
 (2010): $48,895  (2010): $55,658 
25th Percentile Income (2000): $22,368 25th Percentile Income (2000): $25,692 
(2010): $25,259 (2010): $29,571 
Unemployment (2000): 4.05% Unemployment (2000): 3.65% 
(2010): 12.05% (2010): 12.25% 
Poverty (2000): 12.6% Poverty (2000): 10.0% 
(2010): 16.3% (2010): 12.6% 
 
 
RR Arizona Water Company 
Rimrock Yavapai Arizona 
In a Metropolitan Area? Yes 
 
What Technology is Used? AM Annual Production (kgal): 8,508 
Arsenic Treatment Costs  
(per household): 
$16.98 Baseline Water Costs 
(per household): 
$287.05 
Total Water Costs  
(per household): 
$304.03 Number of Households  
in Systems: 
1048 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$33,750 Median Household Income 
(2000): 
$34,901 
 (2010): $40,476  (2010): $43,290 
25th Percentile Income (2000): $20,286 25th Percentile Income (2000): $24,080 
(2010): $24,894 (2010): $24,031 
Unemployment (2000):  Unemployment (2000): 3.80% 
(2010):  (2010): 10.70% 
Poverty (2000): 9.1% Poverty (2000): 11.9% 
(2010): 11.4% (2010): 13.7% 
 
 
 
113 
 
SA City of Sabin 
Sabin Clay MN 
In a Metropolitan Area? Yes 
 
What Technology is Used? IR Annual Production (kgal): 12,200 
Arsenic Treatment Costs  
(per household): 
$199.96 Baseline Water Costs 
(per household): 
$225.11 
Total Water Costs  
(per household): 
425.07 Number of Households  
in Systems: 
175 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$43,523 Median Household Income 
(2000): 
$37,889 
 (2010): $62,292  (2010): $50,057 
25th Percentile Income (2000): $28,146 25th Percentile Income (2000): $19,999 
(2010): $39,481 (2010): $25,345 
Unemployment (2000):  Unemployment (2000): 2.85% 
(2010):  (2010): 4.90% 
Poverty (2000): 6.4% Poverty (2000): 13.2% 
(2010): 4.1% (2010): 12.0% 
 
 
SC Big Sauk Lake Mobile Home Park 
Sauk Centre Stearns MN 
In a Metropolitan Area? Yes 
 
What Technology is Used? IR Annual Production (kgal): 1,650 
Arsenic Treatment Costs  
(per household): 
$146.80 Baseline Water Costs 
(per household): 
$200.10 
Total Water Costs  
(per household): 
$346.90 Number of Households  
in Systems: 
50 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$37,644 Median Household Income 
(2000): 
$42,426 
 (2010): $47,061  (2010): $51,779 
25th Percentile Income (2000): $20,215 25th Percentile Income (2000): $19,029 
(2010): $25,649 (2010): $27,752 
Unemployment (2000):  Unemployment (2000): 3.00% 
(2010):  (2010): 7.60% 
Poverty (2000): 5.2% Poverty (2000): 8.7% 
(2010): 13.8% (2010): 12.9% 
 
 
 
114 
 
SD City of Sandusky 
Sandusky Sanilac MI 
In a Metropolitan Area? No 
 
What Technology is Used? IR Annual Production (kgal): 60,300 
Arsenic Treatment Costs  
(per household): 
$48.38 Baseline Water Costs 
(per household): 
$264.42 
Total Water Costs  
(per household): 
$312.80 Number of Households  
in Systems: 
1124 
 
Municipality Characteristics County Characteristics 
Median Household Income (2000): $33,667 Median Household Income 
(2000): 
$36,870 
 (2010): $29,250  (2010): $40,818 
25th Percentile Income (2000): $14,564 25th Percentile Income (2000): $20,581 
(2010): $13,350 (2010): $21,979 
Unemployment (2000):  Unemployment (2000): 5.20% 
(2010):  (2010): 16.15% 
Poverty (2000): 11.0% Poverty (2000): 10.4% 
(2010): 28.6% (2010): 14.8% 
 
 
SF Chateau Estates Mobile Home Park  
Springfield Clark OH 
In a Metropolitan Area? Yes 
 
What Technology is Used? IR & AM Annual Production (kgal): 16,700 
Arsenic Treatment Costs  
(per household): 
$162.66 Baseline Water Costs 
(per household): 
$260.84 
Total Water Costs  
(per household): 
$423.50 Number of Households  
in Systems: 
226 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$32,193 Median Household Income 
(2000): 
$40,340 
 (2010): $34,045  (2010): $44,141 
25th Percentile Income (2000): $17,102 25th Percentile Income (2000): $22,191 
(2010): $16,274 (2010): $22,968 
Unemployment (2000):  Unemployment (2000): 4.40% 
(2010):  (2010): 10.35% 
Poverty (2000): 16.9% Poverty (2000): 10.7% 
(2010): 25.8% (2010): 15.9% 
 
 
 
115 
 
ST City of Stewart 
Stewart McLeod MN 
In a Metropolitan Area? No 
 
What Technology is Used? IR & AM Annual Production (kgal): 19,133 
Arsenic Treatment Costs  
(per household): 
$165.45 Baseline Water Costs 
(per household): 
$240.59 
Total Water Costs  
(per household): 
$406.04 Number of Households  
in Systems: 
247 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$38,542 Median Household Income 
(2000): 
$45,953 
 (2010): $48,646  (2010): $58,544 
25th Percentile Income (2000): $21,781 25th Percentile Income (2000): $22,778 
(2010): $24,507 (2010): $32,326 
Unemployment (2000):  Unemployment (2000): 3.25% 
(2010):  (2010): 9.35% 
Poverty (2000): 7.1% Poverty (2000): 4.8% 
(2010): 18.3% (2010): 6.6% 
 
 
SV Queen Anne?s County 
Stevensville Queen Anne?s MD 
In a Metropolitan Area? Yes 
 
What Technology is Used? AM Annual Production (kgal): 28,106 
Arsenic Treatment Costs  
(per household): 
$140.44 Baseline Water Costs 
(per household): 
$283.47 
Total Water Costs  
(per household): 
$423.91 Number of Households  
in Systems: 
300 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$63,962 Median Household Income 
(2000): 
$57,037 
 (2010): $86,932  (2010): $81,096 
25th Percentile Income (2000): $46,299 25th Percentile Income (2000): $31,915 
(2010): $53,293 (2010): $67,289 
Unemployment (2000):  Unemployment (2000): 3.00% 
(2010):  (2010): 6.90% 
Poverty (2000): 2.5% Poverty (2000): 6.3% 
(2010): 3.5% (2010): 5.5% 
 
 
 
116 
 
TA Town of Taos 
Taos Taos NM 
In a Metropolitan Area? No 
 
What Technology is Used? AM Annual Production (kgal): 42,961 
Arsenic Treatment Costs  
(per household): 
$27.92 Baseline Water Costs 
(per household): 
$244.76 
Total Water Costs  
(per household): 
$272.68 Number of Households  
in Systems: 
2416 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$25,016 Median Household Income 
(2000): 
$26,762 
 (2010): $36,389  (2010): $35,441 
25th Percentile Income (2000): $11,556 25th Percentile Income (2000): $12,981 
(2010): $13,161 (2010): $18,167 
Unemployment (2000):  Unemployment (2000): 9.15% 
(2010):  (2010): 8.70% 
Poverty (2000): 23.1% Poverty (2000): 20.9% 
(2010): 30.8% (2010): 17.0% 
 
 
TE Golden Hills Community Service District 
Tehachapi Kern CA 
In a Metropolitan Area? Yes 
 
What Technology is Used? AM Annual Production (kgal): 34,039 
Arsenic Treatment Costs  
(per household): 
$17.42 Baseline Water Costs 
(per household): 
$242.38 
Total Water Costs  
(per household): 
$259.80 Number of Households  
in Systems: 
2917 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$29,208 Median Household Income 
(2000): 
$35,446 
 (2010): $46,067  (2010): $47,089 
25th Percentile Income (2000): $13,737 25th Percentile Income (2000): $18,204 
(2010): $20,374 (2010): $23,394 
Unemployment (2000):  Unemployment (2000): 9.85% 
(2010):  (2010): 15.15% 
Poverty (2000): 20.4% Poverty (2000): 20.8% 
(2010): 14.3% (2010): 20.6% 
 
 
 
117 
 
TF City of Three Forks 
Three Forks Gallatin MT 
In a Metropolitan Area? No 
 
What Technology is Used? IX Annual Production (kgal): 111,100 
Arsenic Treatment Costs  
(per household): 
$49.77 Baseline Water Costs 
(per household): 
$238.21 
Total Water Costs  
(per household): 
$287.98 Number of Households  
in Systems: 
730 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$34,212 Median Household Income 
(2000): 
$38,120 
 (2010): $43,058  (2010): $50,136 
25th Percentile Income (2000): $21,235 25th Percentile Income (2000): $21,739 
(2010): $27,133 (2010): $28,361 
Unemployment (2000):  Unemployment (2000): 3.25% 
(2010):  (2010): 6.45% 
Poverty (2000): 22.2% Poverty (2000): 12.8% 
(2010): 8.3% (2010): 13.5% 
 
 
TN Tohono O?Odlham Utility Authority 
Tohono O?Odlham Nation Pima AZ 
In a Metropolitan Area? Yes 
 
What Technology is Used? AM Annual Production (kgal): 5,755 
Arsenic Treatment Costs  
(per household): 
$359.60 Baseline Water Costs 
(per household): 
$238.21 
Total Water Costs  
(per household): 
$597.81 Number of Households  
in Systems: 
64 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
 Median Household Income 
(2000): 
$36,758 
 (2010): $24,496  (2010): $45,521 
25th Percentile Income (2000):  25th Percentile Income (2000): $20,062 
(2010): $7,961 (2010): $23,842 
Unemployment (2000):  Unemployment (2000): 3.45% 
(2010):  (2010): 9.20% 
Poverty (2000):  Poverty (2000): 14.7% 
(2010): 45.2% (2010): 16.4% 
 
 
 
118 
 
WA Springbrook Mobile Home Park 
Wales Androscoggin ME 
In a Metropolitan Area? Yes 
 
What Technology is Used? AM Annual Production (kgal): 955 
Arsenic Treatment Costs  
(per household): 
$344.13 Baseline Water Costs 
(per household): 
$287.64 
Total Water Costs  
(per household): 
$631.77 Number of Households  
in Systems: 
14 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$44,444 Median Household Income 
(2000): 
$35,793 
 (2010): $51,111  (2010): $44,470 
25th Percentile Income (2000): $28,252 25th Percentile Income (2000): $19,332 
(2010): $36,966 (2010): $23,243 
Unemployment (2000): 2.20% Unemployment (2000): 3.60% 
(2010): 7.50% (2010): 8.45% 
Poverty (2000): 8.0% Poverty (2000): 10.0% 
(2010): 4.1% (2010): 14.3% 
 
 
WL Hot Springs Mobile Home Park 
Willard Box Elder UT 
In a Metropolitan Area? No 
 
What Technology is Used? IR & AM Annual Production (kgal): 3,049 
Arsenic Treatment Costs  
(per household): 
$267.94 Baseline Water Costs 
(per household): 
$269.78 
Total Water Costs  
(per household): 
$537.72 Number of Households  
in Systems: 
46 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$52,150 Median Household Income 
(2000): 
$44,630 
 (2010): $57,202  (2010): $55,135 
25th Percentile Income (2000): $31,901 25th Percentile Income (2000): $27,583 
(2010): $37,514 (2010): $23,103 
Unemployment (2000):  Unemployment (2000): 4.15% 
(2010):  (2010): 8.85% 
Poverty (2000): 7.2% Poverty (2000): 7.1% 
(2010): 6.2% (2010): 8.6% 
 
 
 
119 
 
WM City of Wellman 
Wellman Terry TX 
In a Metropolitan Area? No 
 
What Technology is Used? AM Annual Production (kgal): 11,758 
Arsenic Treatment Costs  
(per household): 
$498.80 Baseline Water Costs 
(per household): 
$301.93 
Total Water Costs  
(per household): 
$800.73 Number of Households  
in Systems: 
95 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$50,833 Median Household Income 
(2000): 
$28,090 
 (2010): $51,250  (2010): $39,498 
25th Percentile Income (2000): $21,250 25th Percentile Income (2000): $14,291 
(2010): $32,484 (2010): $17,932 
Unemployment (2000):  Unemployment (2000): 6.05% 
(2010):  (2010): 7.10% 
Poverty (2000): 22.0% Poverty (2000): 23.3% 
(2010): 7.1% (2010): 16.6% 
 
 
WV Village of Waynesville 
Waynesville De Witt IL 
In a Metropolitan Area? No 
 
What Technology is Used? IR Annual Production (kgal): 10,731 
Arsenic Treatment Costs  
(per household): 
$106.53 Baseline Water Costs 
(per household): 
$202.48 
Total Water Costs  
(per household): 
$309.01 Number of Households  
in Systems: 
214 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$40,588 Median Household Income 
(2000): 
$41,256 
 (2010): $48,438  (2010): $45,347 
25th Percentile Income (2000): $22,727 25th Percentile Income (2000): $22,835 
(2010): $27,782 (2010): $26,021 
Unemployment (2000):  Unemployment (2000): 5.65% 
(2010):  (2010): 9.5% 
Poverty (2000): 7.8% Poverty (2000): 8.2% 
(2010): 4.5% (2010): 8.6% 
 
 
 
120 
 
VA City of Vale 
Vale Malheur OR 
In a Metropolitan Area? No 
 
What Technology is Used? IX Annual Production (kgal): 111,100 
Arsenic Treatment Costs  
(per household): 
$106.48 Baseline Water Costs 
(per household): 
$267.39 
Total Water Costs  
(per household): 
$373.87 Number of Households  
in Systems: 
749 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$27,065 Median Household Income 
(2000): 
$30,241 
 (2010): $31,307  (2010): $39,114 
25th Percentile Income (2000): $15,714 25th Percentile Income (2000): $16,739 
(2010): $16,183 (2010): $19,037 
Unemployment (2000):  Unemployment (2000): 8.40% 
(2010):  (2010): 10.60% 
Poverty (2000): 20.0% Poverty (2000): 18.6% 
(2010): 24.3% (2010): 22.7% 
 
 
VV Arizona Water Company 
Valley Vista Yavapai AZ 
In a Metropolitan Area? Yes 
 
What Technology is Used? AM Annual Production (kgal): 18,750 
Arsenic Treatment Costs  
(per household): 
$82.36 Baseline Water Costs 
(per household): 
$287.05 
Total Water Costs  
(per household): 
$369.41 Number of Households  
in Systems: 
757 
 
Municipality Characteristics County Characteristics 
Median Household Income 
(2000): 
$44,042 Median Household Income 
(2000): 
$34,901 
 (2010): $49,749  (2010): $43,290 
25th Percentile Income (2000): $23,602 25th Percentile Income (2000): $24,080 
(2010): $28,797 (2010): $24,031 
Unemployment (2000):  Unemployment (2000): 3.80% 
(2010):  (2010): 10.70% 
Poverty (2000): 9.7% Poverty (2000): 11.9% 
(2010): 10.8% (2010): 13.7% 
 
Notes:  All costs per household in 2010 prices.  All incomes in 2000 and 2010 are reported in 
nominal values. 
121 
 
Technology Key:  AM-Adsorptive Media 
        CF-Coagulation/Filtration 
        IR-Iron Removal 
                IR/IA-Iron Removal with Iron Addition 
        IX-Ion Exchange 
        POU-RO-Point of Use-Reverse Osmosis System 
        SM-System Modification 
 
  
122 
 
 
 
 
 
 
APPENDIX B 
 
COST EFFECTIVENESS OF ARSENIC TREATMENT FOR EACH SYSTEM SIZE 
CLASS 
 
 Tables B-1 and B-2 show the cost effectiveness of the 10 ?g/L rule for each of the nine 
size classes plus overall effectiveness based on the Abt Assoicates, Inc. (2000) estimates of cost 
and avoided cancer cases.  Table B-1 shows the effectiveness for the low estimate of avoided 
cancer cases (37.4 cancers each year) and B-2 shows the effectiveness for the high estimates 
(55.7 cancers each year).  The number of cancer cases each year was assigned to system size 
classes on the basis of the number of households in each system size class. The cost for the entire 
system size class was divided by the number of avoided cancer cases to find cost effectiveness 
for each system size class.  The cost for the average household in each system size class was then 
divided by cancer cases to determine how much each household was paying to avoid a case of 
cancer within its size class. Overall there is a difference of several orders of magnitude between 
the smallest systems and the largest systems or averages for all systems in the effectiveness of 
avoiding cancer cases. 
 
 
 
 
 
 
123 
 
Table B-1 - Cost effectiveness for arsenic treatment under low cancer occurrence estimates 
 System Size Class 
<100
 
101
-500
 
501
-1,0
00
 
1,001
-
3,300
 
3,301
-
10,000
 
10,001
-
50,000
 
50,001
-
100,000
 
100,001
-
1,000,000
 
>1,000,00
0 
Al
l 
Syst
em
s 
Households 
Served 
(thousands) 
26.4 104.4 101.9 289.0 475.6 997.9 469.2 936.6 1,185.0 4,585.8 
Share of 
Households 
(%) 
0.58 2.28 2.22 6.30 10.37 21.76 10.23 30.42 25.84 100.0 
Cancer 
Cases 
Avoided 
(cases/yr) 
0.215 0.851 0.831 2.357 3.879 8.138 3.826 7.638 9.664 37.4 
     How much does the entire system size class pay to avoid a cancer case in its size class?  
Costs for the 
Size Class 
($mil/yr) 
5.90 12.50 7.60 25.60 27.90 53.90 19.00 36.50 4.30 193.00 
Cost per 
Avoided 
Cancer Case 
($mil/cancer) 
27.44 14.68 9.15 10.86 7.19 6.62 4.97 4.78 0.44 5.16 
     How much does each household pay to avoid a cancer case in its size class?  
Costs for 
Each 
Household 
($/yr) 
326.82 162.50 70.72 58.24 37.31 32.37 24.81 20.52 0.86 31.85 
Costs per 
Avoided 
Cancer Case 
($/cancer) 
1,519.72 190.90 85.13 24.71 9.72 3.98 6.48 2.69 0.09 0.85 
 
 
 
 
 
 
 
 
 
 
 
 
124 
 
Table B-2 - Cost effectiveness for arsenic treatment under high cancer occurrence estimates 
 System Size Class 
<100
 
101
-500
 
501
-1,0
00
 
1,001
-
3,300
 
3,301
-
10,000
 
10,001
-
50,000
 
50,001
-
100,000
 
100,001
-
1,000,000
 
>1,000,00
0 
Al
l 
Syst
em
s 
Households 
Served 
(thousands) 
26.4 104.4 101.9 289.0 475.6 997.9 469.2 936.6 1,185.0 4,585.8 
Share of 
Households 
(%) 
0.58 2.28 2.22 6.30 10.37 21.76 10.23 30.42 25.84 100.0 
Cancer 
Cases 
Avoided 
(cases/yr) 
0.320 1.268 1.237 3.510 5.777 12.120 5.698 11.376 14.393 55.7 
     How much does the entire system size class pay to avoid a cancer case in its size class?  
Costs for the 
Size Class 
($mil/yr) 
5.90 12.50 7.60 25.60 27.90 53.90 19.00 36.50 4.30 193.00 
Cost per 
Avoided 
Cancer Case 
($mil/cancer) 
18.42 9.86 6.14 7.29 4.83 4.45 3.33 3.21 0.30 3.46 
     How much does each household pay to avoid a cancer case in its size class?  
Costs for 
Each 
Household 
($/yr) 
326.82 162.50 70.72 58.24 37.31 32.37 24.81 20.52 0.86 31.85 
Costs per 
Avoided 
Cancer Case 
($/cancer) 
1,020.42 128.18 57.16 16.59 6.53 2.67 4.35 1.80 0.06 0.57