&EPA
   United States
   Environmental Protection
   Agency
   The Benefits of Reducing Nitrate
   Contamination in Private Domestic Wells
   Under CAFO Regulatory Options
   December 2002

-------
U.S. Environmental Protection Agency
      Office of Water (43 03 T)
   1200 Pennsylvania Avenue, NW
      Washington, DC  20460
        EPA-821-R-03-008

-------
The Benefits of Reducing Nitrate Contamination in Private
     Domestic Wells Under CAFO Regulatory Options
                       Christine Todd Whitman
                           Administrator
                         G. Tracy Mehan III
                Assistant Administrator, Office of Water
                          Sheila E. Frace
               Director, Engineering and Analysis Division
                          Linda Chappell
                            Economist
                   Engineering and Analysis Division
                   Office of Science and Technology
                 U.S. Environmental Protection Agency
                       Washington, D.C. 20460
                          December 2002

-------
ACKNOWLEDGMENTS AND DISCLAIMER
This report was prepared by Stratus Consulting Inc. under the direction and review of
the Office of Science and Technology.

Neither the United States  government  nor any of its  employees, contractors,
subcontractors, or other employees makes any warranty, expressed or implied, or
assumes any legal liability or responsibility for any third party's use of, or the results
of such use of, any information, apparatus, product, or process discussed in this report,
or represents that its use by such a third party would not infringe on privately owned
rights.

-------
                                    CONTENTS
Exhibits	  vii
Acknowledgments and Disclaimer	ix
Acronyms  	xi

Executive Summary	S-l

Chapter 1    Introduction and Objectives

       1.1    Overview of Benefit Assessment Method	1-2
       1.2    Report Structure	1-4

Chapter 2    Loadings and Well Nitrate Concentrations

       2.1    Relationship between Nitrogen Loadings and Well Nitrate Concentrations ... 2-1
             2.1.1   Included Variables	2-1
             2.1.2   Omitted Variables 	2-3
       2.2    Data Sources  	2-4
             2.2.1   USGS Retrospective Database	2-4
             2.2.2   1990 U.S. Census	2-5
             2.2.3   National Pollutant Loading Analysis	2-5
       2.3    Regulatory Scenarios Used for Benefits Analysis	2-6

Chapter 3    Modeling Well Nitrate Concentrations

       3.1    Model Variables	3-1
       3.2    The Statistical Model	3-5
       3.3    Fitted Values  and Scenario Modeling 	3-6
       3.4    Discrete Changes from above the MCL to below the MCL  	3-8
       3.5    Incremental Changes below the MCL	3-8
       3.6    Timeline Following Scenario Implementation  	3-10

Chapter 4    Valuation: Benefits Transfer

       4.1    Benefits Transfer Methods	4-1
             4.1.1   Transfer an Average Value  	4-1
             4.1.2   Transfer a Function  	4-2
             4.1.3   Calculate a Metafunction	4-2
             4.1.4   Calibrate a Preference 	4-3
       4.2    Choice of Methods	4-4

-------
                                            IV
Chapter 5    Groundwater Valuation Studies

       5.1    Literature Search and Review  	5-1
       5.2    Overview of Groundwater Nitrate Valuation Studies 	5-1
              5.2.1  Crutchfield et al., 1997	5-1
              5.2.2  De Zoysa,  1995  	5-2
              5.2.3  Delavan, 1997  	5-3
              5.2.4  Edwards,  1988	5-4
              5.2.5  Giraldez and Fox, 1995  	5-5
              5.2.6  Hurley et al., 1999 	5-6
              5.2.7  Jordan and Elnagheeb, 1993  	5-7
              5.2.8  Poe and Bishop, 1992  	5-8
              5.2.9  Sparco, 1995  	5-10
              5.2.10 Walker and Hoehn, 1990	5-11
              5.2.11 Wattage, 1993  	5-12
       5.3    Evaluating Studies for Benefits Transfer	5-13
              5.3.1  Purpose of Rating Studies Based on Quality and Applicability	5-13
              5.3.2  Criteria for Ranking Based on Applicability	5-14
              5.3.3  Criteria for Ranking Based on Study Quality 	5-17
              5.3.4  Scoring Matrix	5-21
       5.4    Ranking of Nitrate Valuation Studies  	5-22
       5.5    Values for Benefits Transfer to CAFOs	5-23

Chapter 6    Benefit Calculations

       6.1    Total Annual Values  	6-1
       6.2    Discounting and Aggregating to Net Present Values	6-2
       6.3    Discounted Benefits	6-3
       6.4    Annualized Discounted Benefit Estimates	6-4
       6.5    Alternative Specification of Timepath: Discontinuation of New Regulations
              in 27th Year	6-5
       6.6    Sensitivity Analysis	6-6
              6.6.1  Ranges of Value Estimates 	6-6
              6.6.2  Discount Rate	6-7
              6.6.3  Time Line until Steady State is Achieved  	6-7
              6.6.4  Benefits for Changes under the 10 mg/L MCL	6-8
       6.7    Omissions, Biases, and Uncertainties  	6-10

References	R-l

-------
                                            V
Appendices

A     Nitrogen Sources and Well Data
B     Statistical Models
C     Summary of Groundwater Valuation of Nitrate Contamination Literature
D     Assessment of Data Used to Estimate Benefits

-------
                                      EXHIBITS


1-1    Analysis Plan and Data Sources	1-3

2-1    Characteristics of Benefits Analysis Scenarios 	2-8

3-1    Percentage of Wells Exceeding the MCL  	3-2
3-2    Summary Statistics	3-5
3-3    Gamma Regression Results	3-6
3-4    Characteristics of Benefits Analysis Scenarios 	3-7
3-5    Expected Reductions in Number of Households with Well Nitrate
       Concentrations above 10 mg/L  	3-9
3-6    Mean and Median Reductions in Nitrate Concentrations for Wells with
       Concentrations between 1 and 10 mg/L at Baseline  	3-10

5-1    Scoring Matrix for Groundwater Valuation Studies  	5-15
5-2    Ranking of Studies Based on Scoring Exercise	5-22
5-3    Groundwater Valuation Applicability and Quality Matrix  	5-23
5-4    Consumer Price  Index — All Urban Consumers — U.S. City Average —
       All Items  	5-23
5-5    Mean Annual WTP per Household 	5-24
5-6    Willingness-to-Pay Values Applied to Benefits Transfer  	5-26

6-1    Undiscounted Annual Values under CAFO Regulatory Scenarios	6-1
6-2    Timepath of Undiscounted Benefit Flows	6-2
6-3    Discounted Value of Annual Benefits Using 3%, 5%, and 7% Discount Rates
       Option 2/3 Scenario 7  	6-3
6-4    Total Present Value of Option/Scenarios Using Different Rates of Discount	6-4
6-5    Annualized Present Value of Option/Scenarios Using Different
       Rates of Discount	6-5
6-6    Benefits under Alternative Scenario of Regulatory Discontinuation
       in 27 Year	6-6
6-7    Annualized Benefits under Alternative Scenario of Regulatory Discontinuation
       in 27 Year	6-7
6-8    Change in Value for Crossing 10 mg/L	6-8
6-9    Sensitivity to Changes in Discount Rate  	6-9
6-10   Sensitivity to Changes in Time until Steady State	6-10
6-11   Sensitivity to Benefits from Changes below the MCL  	6-11
6-12   Omissions, Biases, and Uncertainties in the Nitrate Loadings Analysis	6-12

-------
                   ACKNOWLEDGMENTS AND DISCLAIMER
This report has been reviewed and approved for publication by the Engineering and Analysis
Division, Office of Science and Technology. This report was prepared with the support of Stratus
Consulting Inc. under the direction and review of the Office of Science and Technology.

Neither the United States government nor any of its employees, contractors, subcontractors, or
other employees makes any warranty, expressed or implied, or assumes any legal liability or
responsibility for any third party's use of, or the results of such use of, any information,
apparatus, product, or process discussed in this  report, or represents that its use by such a third
party would not infringe on privately owned rights.

The author thanks Stratus Consulting Inc. for their assistance and support in performing the
underlying analyses supporting the conclusions described in this report. Particular thanks are
given to Jeffrey K. Lazo, Robert S. Raucher, Tom Ottem, Marca Hagenstad, and Megan Harrod.
Additional analysis and support was provided by Maurice Hall of CH2M Hill of Redding,
California, and Don Waldman of the University of Colorado at Boulder.

-------
                                   ACRONYMS
AFOs        animal feeding operations
CAFOs      confined animal feeding operations
CARL       Colorado Association of Research Libraries
CES         constant elasticity of substitution
CPI          Consumer Price Index
CVM        contingent valuation method
CWS        community water supply
DOE        dichotomous choice question followed by an open-ended valuation question
ELGs        effluent limitations guidelines
EPA         U.S. Environmental Protection Agency
EVRI        Environmental Valuation Resource Inventory
GI           gastrointestional
IOE          information on current local government expenditures on public health and safety
             services followed by an open-ended valuation question
MCL        Maximum Contaminant Level
NAWQA     National Water Quality Assessment
NPDES      National Pollutant Discharge Elimination System
NPLA       National Pollutant Loading Analysis
pv           present value
VBSs        vegetated buffer strips
VSI          value of statistical illness
VSL         value of statistical life
WTP        willingness-to-pay
YN          yes/no
YNP         yes/no and protest votes

-------
                              EXECUTIVE SUMMARY
Confined animal feeding operations (CAFOs) can contaminate aquifers and thus impose health
risks and welfare losses on those who rely on groundwater for drinking water or other uses. Of
particular concern are nitrogen and other animal waste-related contaminants (which come from
manure and liquid wastes) that leach through soils and ultimately reach groundwater. Nitrogen
loadings convert to elevated nitrate concentrations at household and public water system wells,
and elevated nitrate levels in turn pose a risk to human health.

The federal health-based National Primary Drinking Water Standard for nitrate is 10 mg/L. This
Maximum Contaminant Level (MCL) applies to all  community water supply systems, but not to
households that rely on private wells. As a result, households served by private wells are at risk
of exposure to nitrate concentrations above 10 mg/L, which EPA considers unsafe for sensitive
subpopulations (e.g., infants). Nitrate above concentrations of 10 mg/L can cause
methemoglobinemia ("blue baby syndrome") in bottle-fed infants (National Research Council,
1997), which causes a blue-gray skin color, irritableness or lethargy, and potentially long-term
developmental or neurological effects. Generally, once nitrate intake levels are reduced,
symptoms abate. If the condition is not treated, however, methemoglobinemia can be fatal. No
other health impacts are consistently attributed to elevated nitrate concentrations in drinking
water;  however, other health effects are suspected.

U.S. Census data (1990) show that approximately 13.5 million households located in counties
with animal feeding operations (AFOs) are served by domestic wells. According to the
nationwide USGS Retrospective Database (1996), the concentrations of nitrate in 9.45% of
domestic wells in the United States exceed the 10 mg/L threshold. Thus, EPA estimates that
approximately 1.3 million households in counties with AFOs are served by domestic wells with
nitrate concentrations above 10 mg/L.

EPA's proposed revisions to the National Pollutant  Discharge Elimination System (NPDES)
regulation and effluent guidelines would affect the number and type of facilities subject to
regulation as CAFOs, and would also introduce new requirements governing the land application
of manure. As a result, EPA anticipates that its regulatory proposal will reduce nitrate levels in
household wells. In light of clear empirical evidence from the economics literature that
households are willing to pay to reduce nitrate concentrations in their water supplies —
especially to reduce concentrations from above the MCL to below the MCL — the anticipated
improvement in the quality of water drawn from private domestic wells represents a clear
economic benefit. This report estimates these benefits for each of the 12 regulatory scenarios
evaluated.

-------
                               EXECUTIVE SUMMARY > S-2
Exhibit S-l provides an overview of the approach used to estimate the benefits of well nitrate
reductions. The analysis begins by developing a statistical model of the relationship between
nitrate concentrations in private domestic wells and a number of variables found to affect nitrate
levels, including nitrogen loadings from CAFOs. It then applies this model, in combination with
the projected change in nitrogen loadings from CAFOs under each regulatory scenario, to
characterize the distribution of expected changes in well nitrate concentrations. Next, the analysis
applies this distribution to the number of households served by private domestic wells to
calculate (1) the increase in the number of households served by wells with nitrate concentrations
that are below the MCL and (2) the incremental change in nitrate concentrations for households
currently served by wells that are below the MCL. Finally, the analysis employs estimates of
households' values for reducing well nitrate concentrations to develop a profile of the economic
benefits of anticipated improvements in well water quality.

Regression Analysis: Baseline Model

The approach begins with the use of regression analysis to develop a model characterizing the
empirical relationship between well nitrate concentrations and a number of variables that may
affect nitrate levels, including nitrogen loadings from CAFOs. The primary purpose of the model
is to estimate the effects of nitrogen loadings from CAFOs on domestic well nitrate
concentrations.  The model also accounts for other sources of nitrogen and well characteristics
that could affect this relationship. Controlling for other sources  of nitrogen in the model ensures
that decreases in nitrogen loadings from CAFOs as a result of regulatory activities will not
overestimate impacts on well nitrate concentrations.

The variables included in the model are based  on a review of hydrogeological  studies that have
observed statistical relationships between groundwater nitrate concentrations and various other
hydrogeological and land use factors. Data for the dependent variable, domestic well nitrate
concentrations, were obtained from the USGS  Retrospective Database. Data were compiled for
2,985 observations in 374 counties.1 The regression model includes variables characterizing
nitrogen loadings from animal feeding operations [data obtained from the National Pollutant
Loading Analysis (NPLA)], agricultural fertilizers and atmospheric deposition (data obtained
from the USGS Retrospective Database), and septic systems (data obtained from the 1990
U.S. Census). The model also includes variables describing well depth, soil type, and land use
characteristics around the well (data obtained from the USGS Retrospective Database).
1. There are 678 counties with estimated nitrate loadings >0. Of these, 374 have one or more wells with
enough data available to be included in the analysis.

-------
                               EXECUTIVE SUMMARY > S-3
                                      Exhibit S-l
                             Overview of Analytic Approach
      Data Sources
               Analysis
  NPLA

  Retrospective Database

  U.S. Census

  Ag census
Baseline model: Statistical model estimation
Nitrates = p+px + ... + PX + e
      NPLA scenarios
      Calculation of changes in well
      nitrates under options/scenarios
        U.S. Census
     Change in number of households
     above lOmg/LMCL
     Change in nitrates 1 < N < 10 mg/L
      Benefits transfer
           Net present value of
           nitrate reductions
                                                   Annualized benefit estimates for
                                                   CAFO regulatory options
Calculation of Changes in Well Nitrates

After estimating the regression model using baseline loading information, the model was used to
estimate expected values for well nitrate concentrations, both for baseline and for each of the
12 alternative regulatory scenarios. The calculation of expected values under each scenario
employed data on AFO nitrogen loadings obtained from the NPLA; these loadings vary across
the regulatory scenarios, reflecting different manure application rates, manure management

-------
                               EXECUTIVE SUMMARY > S-4
practices, and other factors. To examine the impact of alternate regulatory scenarios on well
nitrate concentrations, the AFO loadings variable is the only independent variable that changes
value; the values for all other variables are held constant. Exhibit S-2 shows the reductions
nationally in total nitrogen loadings from CAFOs under the different regulatory options/scenarios
derived from the NPLA for the 2,637 counties in the NPLA indicated as having CAFOs.
Exhibit S-2
Nitrogen Loadings from CAFOs: Mean, Total, and Percent Reduction from Baseline
(2,637 counties)
Option/Scenario
Baseline
Option 1 — Scenario 6
Option 1 — Scenario 7
Option 1 — Scenario 8
Option 1 — Scenario 9
Option 2/3 — Scenario 6
Option 2/3 — Scenario 7
Option 2/3 — Scenario 8
Option 2/3 — Scenario 9
Option 5 — Scenario 6
Option 5 — Scenario 7
Option 5 — Scenario 8
Option 5 — Scenario 9
Mean (pounds
per county)
225,506
177,739
186,852
186,849
188,542
173,403
182,465
182,460
184,233
191,161
194,906
194,902
195,737
Total (pounds
nitrogen)
594,660,440
468,697,687
492,729,289
492,721,424
497,185,643
457,263,669
481,159,066
481,147,874
485,822,603
504,090,590
513,968,264
513,957,068
516,158,080
Percent
Reduction from
Baseline
0%
21%
17%
17%
16%
23%
19%
19%
18%
15%
14%
14%
13%
Source: Calculations based on NPLA (TetraTech, 2002).
       Discrete Changes from above the MCL to below the MCL

As noted above, under the baseline scenario, it is estimated that approximately 1.3 million
households in counties with AFOs are served by domestic wells with nitrate concentrations
above 10 mg/L. To estimate the impact of alternative CAFO standards on the number of wells
that would exceed the nitrate MCL, the mean percentage reduction in nitrate concentrations
predicted under each regulatory scenario was applied to the  observed nitrate concentration values
that the USGS Retrospective Database reports.

-------
                               EXECUTIVE SUMMARY > S-5
Based on the resulting values, the percentage reduction in the number of wells with nitrate
concentrations exceeding 10 mg/L was calculated. These values were then applied to the baseline
estimate of the number of households in counties with AFOs that are served by domestic wells
with nitrate concentrations above 10 mg/L. Based on this analysis, it is estimated that the
regulatory scenarios evaluated would bring between 106,000 and 149,000 households under the
10 mg/L nitrate threshold. Exhibit S-3  shows the number of households expected to have well
nitrate concentrations reduced from above the MCL to below the MCL for each of the options/
scenarios.
Exhibit S-3
Expected Reductions in Number of Households with Well
Nitrate Concentrations above 10 mg/L and in Total Nitrates under 10 mg/L
Scenario
Option 1 — Scenario 6
Option 1 — Scenario 7
Option 1 — Scenario 8
Option 1 — Scenario 9
Option 2/3 — Scenario 6
Option 2/3 — Scenario 7
Option 2/3 — Scenario 8
Option 2/3 — Scenario 9
Option 5 — Scenario 6
Option 5 — Scenario 7
Option 5 — Scenario 8
Option 5 — Scenario 9
Reduction in Number
of Households
above the MCL
148,705
120,823
120,823
120,823
148,705
120,823
120,823
111,529
144,058
106,882
106,882
106,882
Total Expected National
Nitrate Reduction
(mg/L)a
854,326
716,007
716,007
695,662
927,730
788,287
788,305
768,221
836,895
717,982
717,995
701,889
a. For wells at or below the MCL at baseline and above 1 mg/L.
       Incremental Changes below the MCL

Households currently served by wells with nitrate concentrations below the 10 mg/L level may
also benefit from incremental reductions in nitrate concentrations. For purposes of this analysis,
it is assumed that such incremental benefits would be realized only for wells with baseline nitrate
concentrations between 1 and 10 mg/L; presumably, an individual would not benefit if nitrate
concentrations were reduced to below background levels, which are assumed to be 1 mg/L.
Incremental reductions in nitrate concentrations for wells that remain above the MCL are not

-------
                               EXECUTIVE SUMMARY > S-6
calculated because we do not have reliable value estimates to apply to these changes. We also
have not calculated values for incremental changes below the MCL for households that are above
the MCL at baseline and below the MCL after new regulations. These values are potentially
already captured by benefit estimates used in the benefits transfer for wells achieving safe levels.
This analysis thus takes a conservative approach to benefits estimation.

For each regulatory scenario, the mean and median reduction in nitrate concentrations for wells
with baseline values between 1 and 10 mg/L was estimated. The  last column of Exhibit S-3
indicates the aggregate reduction in mg/L expected nationally for wells with nitrate levels below
the MCL before new regulations. Between 5.3 and 5.8 million  households would benefit from
these incremental reductions depending on the option and scenario.

Valuation of Predicted Reductions in Well Nitrate Concentrations

The benefit valuation analysis relies on a benefits transfer approach to value predicted reductions
in well nitrate concentrations. Three general steps were used to identify and apply values for
benefits transfer. First,  a literature search identified potentially applicable primary studies.
Second, we evaluated the validity and reliability of the studies  identified. Primary evaluation
criteria included the applicability and quality of the original study, each evaluated on multiple
criteria such as  sample size, response rates, significance of findings in statistical analysis, etc.
And, third, values for application to  CAFO impacts were selected and adjusted. Through the
review and evaluation of the relevant literature, three studies were selected to provide the
primary values used for the benefit transfer:

+      Poe and Bishop (1992): per household values for changes in well nitrate concentrations
       from above the  MCL to below the MCL.

>•      Crutchfield et al. (1997): values incremental changes in nitrate concentrations below the
       MCL.

>•      De Zoysa (1995): values incremental changes in nitrate concentrations below the MCL.

The Consumer Price Index (CPI) was used to convert the annual  mean household willingness-to-
pay values obtained from these studies to 2001  dollars. Exhibit S-4 shows the point value
estimates used for benefits transfer.

-------
                               EXECUTIVE SUMMARY > S-7
Exhibit S-4
Willingness-to-Pay Values Applied to Benefits Transfer
Study
Poe and
Average
Bishop
of Crutchfield et al. and De Zoysa
Value
Annual WTP per household for reducing nitrates
from above the MCL to the MCL
Annual WTP per mg/L between 10 mg/L and 1 mg/L
2001$
$583.00
$2.09
       Total Annual Benefits

Based on the benefit estimates from Exhibit S-4 and the changes in well nitrates under the
potential regulatory options/scenarios indicated in Exhibit S-3, Exhibit S-5 indicates the
estimated total annual (undiscounted) benefits. These values are then adjusted for the timing of
the reductions in well nitrates and discounted over the time frame of the analysis.
Exhibit S-5
Undiscounted Annual Values under CAFO Regulatory Scenarios
(millions 2001$)
Scenario
Option 1 — Scenario 6
Option 1 — Scenario 7
Option 1 — Scenario 8
Option 1 — Scenario 9
Option 2/3 — Scenario 6
Option 2/3 — Scenario 7
Option 2/3 — Scenario 8
Option 2/3 — Scenario 9
Option 5 — Scenario 6
Option 5 — Scenario 7
Option 5 — Scenario 8
Option 5 — Scenario 9
Total WTP for
Discrete Reduction
to MCL
$86.70
$70.44
$70.44
$70.44
$86.70
$70.44
$70.44
$65.02
$83.99
$62.31
$62.31
$62.31
Total WTP for
Incremental Changes
below 10 mg/L
$1.79
$1.50
$1.50
$1.45
$1.94
$1.65
$1.65
$1.61
$1.75
$1.50
$1.50
$1.47
Total
$88.48
$71.94
$71.94
$71.89
$88.63
$72.09
$72.09
$66.63
$85.74
$63.81
$63.81
$63.78

-------
                               EXECUTIVE SUMMARY > S-8
       Timing of Benefits

It is estimated that approximately 75% of affected wells would realize the new predicted nitrate
levels within 20 years (Hall, 1996). Assuming that the number of wells achieving these levels
increases linearly over time, this translates to approximately 3.7% of wells achieving new steady
state conditions each year. This analysis assumes this rate, so that all affected wells reach the new
levels in 27 years.

       Discounting

Three discount rates are used to calculate the net present value of the benefits from reductions in
domestic well nitrate levels: 3%, 5%, and 7%.

       Annualized Benefit Estimates

Because the benefit flows are uneven over time, the annualized values are presented. The
annualized present value represents the constant level of benefits that would yield the same
discounted present value, using the same rate of discount, as the uneven flow of benefits.
Exhibit S-6 presents the annualized benefit estimates for the total annual benefits shown in
Exhibit S-5. For instance, for Option 5 Scenario 7, using the 27 year timepath  and a 3% discount
rate, the present value of benefits would be $1,458.4 million. A constant benefit flow of
$43.75 million discounted at 3% shown in Exhibit S-6 for Option 5  Scenario 7 would generate
$1,458.4 million in total present value of benefits, also discounted at 3%.

-------
EXECUTIVE SUMMARY > S-9
Exhibit S-6
Annualized Present Value of Option/Scenarios Using Different Rates of Discount
(millions 2001$)
Scenario
Option 1 — Scenario 6
Option 1 — Scenario 7
Option 1 — Scenario 8
Option 1 — Scenario 9
Option 2/3 — Scenario 6
Option 2/3 — Scenario 7
Option 2/3 — Scenario 8
Option 2/3 — Scenario 9
Option 5 — Scenario 6
Option 5 — Scenario 7
Option 5 — Scenario 8
Option 5 — Scenario 9
3%
Annualized Value
$60.67
$49.32
$49.32
$49.29
$60.77
$49.43
$49.43
$45.68
$58.78
$43.75
$43.75
$43.73
5%
Annualized Value
$49.20
$40.00
$40.00
$39.98
$49.29
$40.08
$40.08
$37.05
$47.67
$35.48
$35.48
$35.46
7%
Annualized Value
$41.03
$33.36
$33.36
$33.34
$41.11
$33.43
$33.43
$30.90
$39.76
$29.59
$29.59
$29.58

-------
                                     CHAPTER 1
                       INTRODUCTION AND OBJECTIVES
The U.S. Environmental Protection Agency (EPA) is revising and updating the two primary
regulations that ensure that manure, wastewater, and other process waters generated by confined
animal feeding operations (CAFOs) do not impair water quality. The proposed regulatory
changes affect the existing National Pollutant Discharge Elimination System (NPDES)
provisions that define and establish permit requirements for CAFOs, and the existing effluent
limitations guidelines (ELGs) for feedlots, which establish the technology-based effluent
discharge standard that applies to regulated CAFOs. The existing regulations were promulgated
in the 1970s. EPA is revising the regulations to address changes in the animal industry sectors
over the last 25 years and to clarify and improve implementation of CAFO requirements.

CAFOs can contaminate groundwater and thus cause health risks and welfare losses to people
relying on groundwater for their potable supplies or for other uses. Of particular concern are
nitrogen and other animal waste-related contaminants (which  come from manure and liquid
wastes) that leach through the soils and the unsaturated zone and ultimately reach groundwater.
Nitrogen loadings convert to elevated nitrate concentrations at household and community system
wells, and elevated nitrate levels in turn pose a risk to human health. The proposed regulation
will generate benefits by reducing nitrate levels in household wells, and there is clear empirical
evidence from the economics literature indicating that households are willing to pay to reduce
nitrate concentrations in their water supplies.

The federal health-based National Primary Drinking Water Standard for nitrate is 10 mg/L, and
this Maximum Contaminant Level (MCL) applies to all  community water supply (CWS)
systems. Households relying on private wells are not subject to the federal MCL for nitrate;
however, levels above 10 mg/L are considered unsafe for sensitive subpopulations (e.g., infants).
Nitrate above concentrations of 10 mg/L can cause methemoglobinemia ("blue baby syndrome")
in bottle-fed infants (National Research Council, 1997), which causes a blue-gray skin color,
irritableness or lethargy, and potentially long-term developmental or neurological effects.
Generally, once nitrate intake levels are reduced, symptoms abate. If the condition is untreated,
however, methemoglobinemia can be fatal. No other health impacts are consistently attributed to
elevated nitrate concentrations in drinking water.

U.S. Census (1990) data show that there are currently approximately 13.5 million households
with domestic wells located in counties with animal feedlot operations. CAFOs present a
potential contaminant source to groundwater, particularly via nitrogen leached from manure.
Manure from these operations is generally managed either by  storing it in a waste lagoon, where
waste has the potential to leak through the lining or overflow onto the surrounding ground and

-------
                          INTRODUCTION AND OBJECTIVES > 1-2
leach nitrogen into the groundwater, or by spreading it on surrounding farm fields, where,
depending on the rate and timing of the applications, the soil hydrology, and precipitation, nitrate
may leach into the groundwater. Nitrate is of particular concern because it leaches easily into
groundwater, and is one of the most frequently found groundwater contaminants (Lichtenburg
and Shapiro, 1997).

CAFOs are currently covered by existing effluent guidelines at 40 CFR Part 412 and permit
regulations at 40 CFR Part 122. The effluent guidelines regulations, which require the largest
CAFOs to achieve zero discharge of waste to surface waters except under extreme storm events,
have not been sufficient to resolve water quality impairment from feedlot operations. Under the
current permit regulations, a CAFO is a facility in one of the following three categories:

+      more than 1,000 animal units  confined at the facility

+      301-1,000 animal units confined and the facility also meets one of the specific  criteria
       addressing the method of discharge [40 CFR Part 122 Appendix B]

>•      designated as a CAFO on a case-by-case basis if the NPDES-authorized permitting
       authority determines that it is  a significant contributor of pollution to waters of the
       United States [40 CFR part 122.23(c)].

This report estimates benefits for national reductions in nitrate concentrations in private domestic
wells achieved by changing regulations for effluents from CAFOs. Benefits achieved via this
regulation for public and surface water systems are considered elsewhere in this regulatory
analysis. The proposed regulatory options include different criteria for the definition of a CAFO,
therefore changing the number of operations that will have to comply with the proposed
regulations. They also include requirements for the quantity and rate of land application of
manure, as well as water quality reporting. The current regulations address only controls at the
feedlot; land application of manure is not addressed. This analysis evaluates the potential benefits
from eight regulatory scenarios.
1.1    OVERVIEW OF BENEFIT ASSESSMENT METHOD

The assessment of benefits of well nitrate reductions from CAFO regulations followed the
multistep process outlined in Exhibit 1-1.

-------
                        INTRODUCTION AND OBJECTIVES > 1-3
                                    Exhibit 1-1
                          Analysis Plan and Data Sources
    Data Sources
NPLA

Retrospective Database

U.S. Census

Ag census
   NPLA scenarios
     U.S. Census
   Benefits transfer
               Analysis
Baseline model: Statistical model estimation
Nitrates = P + P, x, + • • • + P X +e
         ~o   ~1 'M        ~nAn
      Calculation of changes in well
      nitrates under options/scenarios
     Change in number of households
     above lOmg/LMCL
     Change in nitrates 1 < N < 10 mg/L
           Net present value of
           nitrate reductions
                                                Annualized benefit estimates for
                                                CAFO regulatory options

-------
                          INTRODUCTION AND OBJECTIVES > 1-4
To estimate the benefits achieved by reducing nitrogen loadings from animal manure and thus
improving groundwater quality, we first established baseline water quality under current loadings
and current regulations using available data on nitrate concentrations in individual wells. These
data, described further in Chapter 2, were obtained from a national database of groundwater
quality. We then used these baseline data for nitrate concentrations and data on current nitrogen
loadings by county to model the relationship between nitrate concentrations and nitrogen
loadings. Our model also included significant explanatory variables such as well depth and soil
hydrological characteristics that were identified from a literature survey. We then applied the
parameter estimates generated from this model to projected loadings under each regulatory
scenario to estimate changes in nitrate concentrations in the wells for each regulatory option.

From these data we established the percentage of wells above the MCL (10 mg/L) under each
scenario, and the nitrate reduction for wells that were already below 10 mg/L at baseline. We
then extrapolated these values to the total number of household units on private wells in the
country to estimate the number of households that would have nitrate concentrations reduced
from above the MCL to below the MCL, and how many households that were  already below the
MCL at baseline and would have  further water quality improvements under the regulatory
scenarios.

After reviewing studies that estimated household-level monetary benefits  of improving water
quality through reduced nitrate concentrations, we established a range of values for both reducing
nitrate from above the MCL to below the MCL and reducing nitrate concentrations in wells that
were already below the MCL at baseline. Using benefits transfer methods, we  then estimated the
total monetary benefits that could be achieved under each scenario, based on the  number of
households brought from above the MCL to below the MCL and the number of households that
achieved water quality improvements below the MCL.

Monetary benefits were estimated annually over a 100 year time period to capture the time path
until well nitrates would achieve a steady state following implementation  of each regulatory
option. We assumed that it would take 27 years to achieve the steady state. Discounting was
applied to determine net present values,  and these were then annualized to derive a benefit
estimate to be used in comparison to annualized cost estimates. Sensitivity analysis was
performed to examine how annualized benefit estimates change using different discount rates,
years until clean, and per household benefit values.
1.2    REPORT STRUCTURE

Chapter 2 discusses the choice of variables to include in modeling the relationship between
loadings from CAFOs and well nitrate concentrations, and data sources used in the analysis. This
chapter also includes information on the methods used to calculate loadings for each scenario and
descriptions of each scenario.

-------
                           INTRODUCTION AND OBJECTIVES > 1-5
Chapter 3 discusses the model of the relationship between nitrogen loadings and well nitrate
concentrations. Statistical analyses and parameter estimates from analyses based on this model,
assuming a gamma distribution, are included. Chapter 3 also discusses the results from running
the parameter estimates through each of the regulatory scenarios with different loadings and the
subsequent changes in well nitrate concentrations.

Chapter 4 discusses the benefits transfer method in detail.

Chapter 5 discusses the groundwater valuation studies used in this analysis, including a ranking
of their relevance to this study, the various methods that each used to estimate benefits, and their
respective values for reducing groundwater contamination.

Chapter 6 provides a summary of benefit estimates using the different assumptions regarding
which approach to apply for extrapolating from the model to the population, the time until a new
steady state is achieved, and the discount rate used. Omissions, biases, and uncertainties in the
analysis are discussed here.

References are provided for both the nitrate modeling and benefits analysis portions of this
report.

The appendices include information on nitrogen loading data sets,  details of the statistical
analyses of the nitrogen-nitrate relationship, and tables summarizing the literature used in the
benefits transfer analysis.

-------
                                    CHAPTER 2
            LOADINGS AND WELL NITRATE CONCENTRATIONS
This chapter identifies variables affecting nitrate contamination in wells that can be used to
model the relationship between nitrogen loadings and nitrate concentrations in wells. We then
review the sources of data used to model this to the regulatory scenarios to be used for benefits
analysis.
2.1    RELATIONSHIP BETWEEN NITROGEN LOADINGS AND WELL
       NITRATE CONCENTRATIONS

We selected the variables to include in the model used to predict nitrate concentrations in
groundwater under different regulatory scenarios based on our review of hydrogeological studies
that have observed statistical relationships between groundwater nitrate concentrations and
various other hydrogeological and land use factors. Although the groundwater monitoring and
modeling studies reviewed for this report covered different geographic areas and focused on
varying nitrogen sources (septic systems, agricultural fertilizers, animal feedlots), certain
variables were significant across many of the studies. These studies were generally regional or
local in scope, and obtained their data by sampling the wells directly.
2.1.1   Included Variables

Nitrogen application rates, whether from agricultural fertilizers, animal wastes, or private septic
systems, were the most consistent and significant factor affecting nitrate levels in wells across the
studies reviewed for this analysis (Rausch, 1992; Spalding and Exner, 1993; Clawges and
Vowinkel, 1996; Richards et al., 1996; Lichtenberg and Shapiro, 1997; Lindsey, 1997; Burrow,
1998; CDC, 1998; Letson et al., 1998; Nolan et al., 1998; Kerr-Upal et al., 1999).

Nitrate is found in groundwater because of surface applications of two forms of the nutrient
nitrogen: nitrate and amine groups (of which nitrogen is a component). Generally nitrogen from
fertilizer is already in the nitrate form, which leaches more readily into the soil. Nitrogen from
manure and septic systems generally occurs as large organic molecules called amine groups.
Once in the soil, these large molecules convert to nitrate and ammonia as microbes break down
the organic matter. The ammonia then volatizes as a gas into the atmosphere, and the nitrate
leaches through the soil and potentially into groundwater. This process takes  a few hours to a few
weeks, depending on the soil conditions (M. Hall, CH2MH111, pers. comm, Sept. 15, 2000).

-------
                   LOADINGS AND WELL NITRATE CONCENTRATIONS > 2-2
Studies that investigated the effects of animal manure production on groundwater nitrate
concentrations found manure to be positively correlated with groundwater nitrate. Animal waste
lagoons were associated with elevated groundwater nitrate concentrations, particularly as the
distance to the water table decreased (Miller et al., 1976; Ritter and Chirnside, 1990;
North Carolina Division of Water Quality, 1998). Farms that applied manure as fertilizer tended
to have higher nitrate concentrations in groundwater as well (Rausch, 1992; Swistock et al.,
1993; Clawges and Vowinkel,  1996; Richards et al., 1996; Lindsey, 1997; Letson et al., 1998;
Kerr-Upal et al., 1999).

Several studies focused on agricultural practices such  as type of crop and crop rotations, and how
they may be correlated with nitrate concentrations in groundwater. Swistock et al. (1993), Stuart
et al. (1995), and Lichtenberg and Shapiro (1997) found corn production to be associated with
higher nitrate levels because corn demands higher fertilizer input and extensive irrigation, which
increases the rate at which nitrate leaches to the groundwater. Spalding and Exner (1993) found
that groundwater beneath any row-cropped, irrigated area tended to have higher nitrate levels.
Rausch (1992) found that tillage practices, which change the amount of organic matter in the root
zone, and planting nitrogen-fixing legumes as a part of the crop rotation cycle decreased the
quantity of nitrate available for leaching and were associated with lower levels of nitrate in
groundwater.

The proximity of septic systems to wells was found to be a small, but significant, contributing
factor to elevated nitrate concentrations in groundwater in several studies (Carleton, 1996;
Richards et al., 1996; CDC, 1998; Nolan et al., 1998).

Well depth was also frequently found to be a significant factor, inversely related to nitrate
concentrations in wells, regardless of nitrate source (Detroy et  al.,  1988; Ritter and Chirnside,
1990; Kross et al., 1993; Spalding and Exner,  1993; Swistock et al., 1993; Sparco, 1995;
Lichtenberg and Shapiro, 1997; Ham et al., 1998; North Carolina Division of Water Quality,
1998). Swistock et al. (1993) found that wells deeper than 100 ft tended to have significantly
lower nitrate concentrations, and Kross et al. (1993) found that wells deeper than 45 ft generally
had much lower nitrate concentrations.

A number of studies identified at least one geological  characteristic as a significant  factor
affecting nitrate  concentrations. Two studies found unconfined aquifers to be associated with
elevated nitrate in groundwater (Lichtenberg and  Shapiro, 1997; Lindsey, 1997). Other studies
found higher nitrate levels associated with more permeable, well-drained soils (Ritter and
Chirnside,  1990; Spalding and Exner, 1993; Sparco, 1995; Burrow, 1998; Chen, 1998; Ham
et al., 1998; Nolan et al., 1998; Kerr-Upal et al., 1999). Several studies explored the possibility of
using DRASTIC, an index intended to reflect the groundwater pollution potential of a region.
DRASTIC  incorporates several hydrogeological factors: drainage, aquifer recharge  rate, aquifer
media, soil media, topography, impact of the vadose zone, and hydraulic conductivity of the
aquifer. All found positive correlations between county-level DRASTIC scores and groundwater
nitrate concentrations, but none were statistically significant. All agreed that DRASTIC scores

-------
                   LOADINGS AND WELL NITRATE CONCENTRATIONS > 2-3
are not reliable predictors of groundwater nitrate levels (U.S. EPA, 1990; Rausch, 1992;
Richards et al., 1996). We included DRASTIC scores in some early regression analyses, but they
did not strengthen the results and were thus dropped from further analysis.

Different types of land use near wells are also associated with higher groundwater nitrate. Several
studies found agricultural land use in general to be associated with higher groundwater nitrate
than other land uses (Rausch,  1992; Spalding and Exner, 1993; Swistock et al., 1993; Mueller
et al., 1995;  Sparco, 1995; Carleton, 1996; Clawges and Vowinkel, 1996; Richards et al., 1996;
Nolan et al., 1998). Results from Carleton's study, for example, suggest that nitrate
concentrations in West Windsor Township in New Jersey have decreased as residential use has
replaced agriculture.
2.1.2   Omitted Variables

Because of incomplete or unreliable national data, we did not include all the significant variables
identified in these studies. First, well construction and age were cited as significant variables in
several studies (Spalding and Exner, 1993; Swistock et al., 1993; Richards et al., 1996; Burrow,
1998; CDC, 1998). In general, older wells were more vulnerable to nitrate contamination because
the casing could be cracked, allowing surface contaminants to enter the groundwater. Different
construction materials and methods also affected how easily nitrate or other pollutants could
reach the groundwater supply via direct contamination at the wellhead. This variable, however, is
often unreliable because it is generally obtained by surveying well owners and relying on their
subjective assessment of how and when the well was constructed. No reliable data on well
construction were available nationally.

Second, the distance from a pollutant source to well was significantly correlated with
groundwater nitrate in several studies (Rausch, 1992; Swistock et al., 1993; CDC,  1998; Ham
et al., 1998; North Carolina Division of Groundwater Quality, 1998). Although spatial data were
available for well locations, no spatial data on the location of animal feedlots, cropland, and
septic systems were available for our analysis.

Two studies in the literature surveyed (Sparco, 1995; Lichtenberg and Shapiro, 1997) developed
models to predict nitrate concentrations in groundwater, based on the variables described above.
These models were not used in the final analysis because they incorporated either spatial or
temporal data that are not available for a national  level assessment. In addition, as discussed
below, our analysis indicates that a gamma distribution more closely matches the distribution of
nitrate concentrations than the linear and lognormal distributions assumed in  the other models.
Aside from these differences, the final model used similar variables and assumptions regarding
land use and hydrogeology.

-------
                   LOADINGS AND WELL NITRATE CONCENTRATIONS > 2-4
2.2    DATA SOURCES

The independent variables for the analysis were chosen based on the preceding literature review
to identify variables that have significant impact on nitrate concentration in groundwater. Data
availability also dictated which variables were included in the model. The data for this analysis
were obtained primarily from three sources: the USGS Retrospective Database, the National
Pollutant Loading Analysis (TetraTech, 2002),1 and the 1990 U.S. Census. Appendix A provides
additional detail on how these data sets were combined and some additional summary statistics.
2.2.1   USGS Retrospective Database

The Retrospective Database contains water quality and land use data from 10,426 wells sampled
from 725 counties in 38 states. The data were gathered between 1969 and 1992. Data relevant to
this analysis were:

*•      nitrate concentrations in the well, in mg/L

+      water use of the well (e.g., irrigation, domestic)

*•      nitrogen inputs from manure and fertilizer loadings

+      atmospheric nitrogen deposition

+      depth to water in the well

+      soil hydrologic group,  a measure that includes runoff potential, soil permeability, depth to
       water table, depth to an impervious layer, water capacity, and shrink-swell potential

>•      number of septic systems per acre near the well

>•      land use near the well

>•      region of the United States the well is located in.

Within any given county, the reported nitrogen loadings data used in the data analysis are the
same (nitrogen loading data vary between counties but not within counties). These data were
obtained from other published data sources (U.S. Census, U.S. Census of Agriculture, and
U.S. EPA fertilizer sales data) that report at a county level. Water use, well depth, and nitrate
1. The National Pollutant Loading Analysis (NPLA) (TetraTech, 2002) comprises three Excel spreadsheets
provided by EPA. These are described in Section 2.2.3.

-------
                   LOADINGS AND WELL NITRATE CONCENTRATIONS > 2-5
concentrations are reported by well. The Retrospective Database was the limiting data source for
this analysis because it includes only 725 counties out of approximately 3,100 counties in the
United States. Implicit in our use of these data and in our analysis is the assumption that the
Retrospective Database is representative of private domestic wells nationwide. Potential biases
related to this assumption are discussed in Chapter 6.

In the Retrospective Database, approximately 18% of the reported nitrate concentrations in
domestic wells were at or below the detection limit (0.05 mg/L). Because this database is  a
compilation of several databases, these nondetects are reported in several ways: at the detection
limit, at half the detection limit, and at zero. To standardize our data we set all values reported at
or below the detection limit to the detection limit. In addition, because this analysis is concerned
only with the benefits gained from reducing nitrate contamination in domestic wells, we
eliminated wells with nondomestic uses (stock, irrigation, urban, and unknown).
2.2.2   1990 U.S. Census

We obtained the total number of household units on wells nationwide and the number of
household units using septic systems in each county in the United States from the
1990 U.S. Census.2 The number of households on septic systems in each county, divided by the
total acres in the county, provided an estimate of septic system density for the analysis.
2.2.3   National Pollutant Loadings Analysis

The National Pollutant Loading Analysis (NPLA; Tetra Tech, 2002) provided estimates of
leached nitrate from animal feedlot operations under different regulatory options. The NPLA
developed a national estimate of pollutant load reductions expected from meeting the
requirements of revised animal feeding operation effluent guidelines.

The estimate is based on loadings for the current effluent guidelines (preregulation baseline) and
after the implementation of revised effluent guidelines (postregulation modeling scenarios). The
national estimate of nutrient, pathogen, and metal loadings is based on conditions identified on a
broad range of sample farms. These farm conditions consisted of animal groupings of various
size classes, current management practices and animal waste management systems, and
regionally based physiographic information regarding the soil, rainfall, hydrology, crop rotation,
and other factors for a given region of the country. Model farms were developed from county,
regional, and national data sources, including the 1997 Census of Agriculture data.
2. County level source of water data did not appear to be available from the 2000 U.S. Census.

-------
                   LOADINGS AND WELL NITRATE CONCENTRATIONS > 2-6
Total nitrate leached to groundwater was based on the size and type of operations in the area and
subsequent manure produced, crop nutrient removal rates, and the GLEAMS model. GLEAMS
can be used to evaluate the effects of various agricultural management practices on the
movement of pollutants to water sources, using hydrology, erosion,  and biochemical processes to
evaluate pollutant transport.

Along with the NPLA, the U.S. EPA also provided the estimated number of facilities of each size
in each county and the percentage of facilities that would be subject to regulation in each state.
We assume this percentage to be constant for all counties within that state. In general, all "large"
operations will be subject to regulations, and varying percentages of "medium" operations will be
regulated. These data included loadings from beef,  dairy, veal, swine, layer, broiler, and turkey
operations.

Details on how these data were combined to estimate total nitrogen  loadings in each county are
provided in Section  2.3.
2.3    REGULATORY SCENARIOS USED FOR BENEFITS ANALYSIS

The regulatory scenarios evaluated in this analysis are based on different combinations of two
factors: limits for land application of manure (options), and variations on how many facilities
will be subject to the regulation (scenarios). EPA analyzed nitrate loadings under 12 option-
scenario combinations plus the baseline conditions, for a total of 12 regulatory scenarios. All
regulatory scenarios will entail common criteria, which include best management practices in the
feedlots (stormwater diversions, lagoon/pond depth markers, periodic inspections, record
keeping); mortality handling requirements; nutrient management planning and record keeping
(soil and manure sampling requirements); and prohibition of manure application within 100 ft of
surface water, tile drain inlets, and sinkholes.

The land application options are based on either total nitrogen applied (Option 1),  total phosphate
applied (Option 2), or total phosphate applied plus covered lagoons (Option 5).3 The nitrogen and
phosphate content of the manure and subsequent manure application rates under these options are
based on the type of animal operation. Under all three options, manure will be land-applied at
allowable manure application rates, providing adequate nutrients for crop uptake, runoff, and
leaching.

The percentage of affected facilities differs according to the size of the facility and state. The
scenarios for the number of affected facilities determine how many small, medium, and large
facilities will be defined as CAFOs under the regulation, and thus become subject  to the
3. Option 3 is similar to Option 2 but also requires liners for lagoons. Since the leached nitrate loadings are the
same for our analysis under Options 2 and 3, these are reported simply as Option 2/3 throughout this report.

-------
                   LOADINGS AND WELL NITRATE CONCENTRATIONS > 2-7
nitrogen-based or phosphate-based limits. The size categories are based on the number of
animals at the facility and vary by animal type (see Exhibit A-l in Appendix A). For example, a
large beef farm is defined as any farm with more than 1,000 head of cattle, and a medium beef
farm is defined as those farms with 300 to 1,000 head of cattle. In comparison, by definition, a
swine farm is large if it has more than 2,500 pigs, and a turkey farm is large if it has more than
55,000 turkeys. Small facilities are not subject to the regulation, and therefore are  not included in
the baseline analysis. All large facilities are considered CAFOs and therefore subject to nitrogen-
based or phosphate-base limits. At baseline, some  medium-sized operations are regulated and
therefore produce varying nitrogen loadings.

Similarly, most dry poultry operations were assumed to produce unregulated loadings at baseline.
Under the regulatory scenarios, however, some of these operations will be regulated and produce
reduced loadings.

Exhibit 2-1 summarizes the key nutrients, percentage of facilities that will be regulated, and how
a CAFO will be defined, based on animal type and size, for each scenario.

In the NPLA, animal operations are divided into two general categories: those currently with
controls at the feedlot and those currently without  controls at the feedlot. Those currently with
controls are assumed to be in complete compliance with existing regulations. Operations with
controls are modeled to have different loadings than operations without controls. Different
loadings data are provided in the NPLA for operations with and without controls.

Loadings for the scenarios, including baseline, are calculated based on the assumption that
facilities with  controls produce one amount of loadings  and facilities without controls produce
loadings equivalent to baseline. For all scenarios, including baseline, the regulated percentage of
operations will produce "regulated loadings," and the remaining percentage will produce
"baseline loadings." The equation for calculating total loadings for one category of facility
(e.g., medium  beef) in one county is:

       Total Loadings for Type of Operation (AnimalX, SizeY) in a county =
        (% of facilities regulated * Scenario loadings-regulated * Number of facilities) +           (2-1)
        [(1 - % of facilities regulated) * Baseline loadings-unregulated * Number of facilities].

This equation  generates the total loadings for operations of each animal type and size in each
county. The loadings are then summed across all operations (all animal types and relevant facility
sizes) to get total county loadings.

-------
LOADINGS AND WELL NITRATE CONCENTRATIONS > 2-8
Exhibit 2-1
Characteristics of Benefits Analysis Scenarios
Regulatory Scenario
Baseline
Option 1 — Scenario 6
Option 1 — Scenario 7
Option 1 — Scenario 8a
Option 1 — Scenario 9b
Option 2/3 — Scenario 6
Option 2/3 — Scenario 7
Option 2/3 — Scenario 8a
Option 2/3 — Scenario 9b
Option 5 — Scenario 6
Option 5 — Scenario 7
Option 5 — Scenario 8a
Option 5 — Scenario 9b
Key Nutrient
Manure
application not
regulated
Nitrogen
Nitrogen
Nitrogen
Nitrogen
Phosphate
Phosphate
Phosphate
Phosphate
Phosphate
Phosphate
Phosphate
Phosphate
Percentage of Facilities Regulated
100% of large AFOs, plus medium
AFOs that meet certain requirements
100% of large and medium AFOs
100% of large AFOs, plus medium
AFOs that meet certain requirements
New NPDES conditions for identifying
medium-sized CAFOs, plus qualifying
dry poultry and immature swine and
heifer operations
100% of large AFOs, all medium
AFOs regulated under current rules
100% of large and medium AFOs
100% of large AFOs, plus medium
AFOs that meet certain requirements
New NPDES conditions for identifying
medium-sized CAFOs, plus qualifying
dry poultry and immature swine and
heifer operations
100% of large AFOs, all medium
AFOs regulated under current rules
100% of large and medium AFOs
100% of large AFOs, plus medium
AFOs that meet certain requirements
New NPDES conditions for identifying
medium-sized CAFOs, plus qualifying
dry poultry and immature swine and
heifer operations
100% of large AFOs, all medium
AFOs regulated under current rules
Size of Facility
Subject to Regulation
All large, some
medium
All large, all medium
All large, some
medium
All large, some
medium
All large, no medium
All large, all medium
All large, some
medium
All large, some
medium
All large, no medium
All large, all medium
All large, some
medium
All large, some
medium
All large, no medium
a. The benefits reported in later chapters for Scenario 8 represent the estimated benefits of regulating all large
and some medium facilities that meet new NPDES conditions. The difference between Scenario 8 and
Scenario 9 represents the increase in estimated benefits attributable to new regulations on the identified
medium facilities, given that all large facilities are regulated.
b. The benefits reported in later chapters for Scenario 9 represent the benefits attributable to new regulations
on all large facilities while adding no new regulations to medium facilities.

-------
                                    CHAPTER 3
               MODELING WELL NITRATE CONCENTRATIONS
A statistical model of the relationship between nitrogen loadings and well nitrate concentrations
was developed to analyze the effect of different regulatory options. An alternative to a statistical
model would be representative hydrogeological models, to examine how changes in nitrogen
loadings would translate into well nitrate concentrations. This approach was considered
infeasible because of time and budgetary constraints as well as the likely limitation on data
needed to generalize such models to the national level. As described below, though, the statistical
model attempts to capture the effects of several variables that would also be used in a
hydrogeological model, such as well depth, soil type, and land use.

The statistical modeling approach uses existing data to estimate the relationship between sources
of nitrogen and well nitrate concentrations. This approach allows us to control for non-CAFO
sources of nitrogen, including septic systems, fertilizers, and natural (background) levels of
nitrate.
3.1    MODEL VARIABLES

Analysis of the relationship between loadings and well nitrate concentrations is based on the
following linear model:

        Nitrate (mg/L) = 130 + 13! ag dummy + B2 soil group + B3 well depth               (3-1)
    + B4 septic ratio + B5 atmospheric nitrogen + B6 loadings ratio + B7 regional
                             dummy variables.

Dependent Variable

Nitrate concentration  is the dependent variable in this model, expressed in mg/L.

The percentage of drinking water wells with nitrate concentrations greater than 10 mg/L varies
widely, depending on well, hydrologic, and pollutant characteristics. Exhibit 3-1 summarizes the
widely varying percentages found in different studies. Given this wide range of values, EPA
determined that the USGS Retrospective Database, which estimates that 9.45% of domestic wells
have nitrate levels above 10 mg/L, contains a reasonable representation of affected wells in the
United States.

-------
                    MODELING WELL NITRATE CONCENTRATIONS > 3-2
Exhibit 3-1
Percentage of Wells Exceeding the MCL
Study
Agriculture Canada,
1993 (as cited by
Giraldez and Fox,
1995)
Andres 1991 (as cited
in Sparco, 1995)
CDC, 1998
Chen, 1998
Kross et al., 1993
National Water Quality
Assessment (NAWQA)
Database, USGS, 1998
Poe and Bishop, 1999
Retrospective Database,
USGS, 1996
Richards et al., 1996
Spalding and Exner,
1993
Swistock et al., 1993
U.S. EPA, 1990
USGS, 1986
Vitosh, 1985 (cited in
Walker and Hoehn,
1990)
Location
Ontario, Canada
Sussex County, Delaware
Illinois, Iowa, Missouri, Kansas, Nebraska,
Wisconsin, Minnesota, S. Dakota, N. Dakota
Nemaha Natural Resources District, Nebraska
Iowa
National
Portage County, Wisconsin
National
Ohio, Indiana, W. Virginia, Kentucky
Iowa, Nebraska, Kansas, Texas, N. Carolina,
Ohio
Pennsylvania
National
Upper Conestoga River Basin
Southern Michigan
Type of Well
Domestic farm
Rural
Domestic
Rural
Rural
All
Rural
Domestic
Rural
Rural
Private
Rural domestic
Rural
Rural
% Exceeding
10 mg/L
13
23
13.4
10
18
16.2
16
8.9a
3.4
20, 20, 20, 8.2, 3.2,
2.7, respectively
9
2.4
40+
34
a. 8.9% of all domestic wells in the Retrospective Database exceed 10 mg/L. From all domestic wells in this
database, of the wells with enough data in order to be included in our analysis, 9.45% of the wells exceeded
the 10 mg/L. As discussed further in Section 3.4, we use this 9.45% as the baseline percent of wells above the
MCL for our analysis.
Actual nitrate concentrations in groundwater reported in the Retrospective Database, which were
used to scale predicted values, ranged from 0 mg/L to 84.3 mg/L. Nitrate concentrations below
the detection limit were reported in one of three ways: at the detection limit (0.05 mg/L), at half
the detection limit, or at zero. To account for this variability, EPA set any nitrate concentration

-------
                     MODELING WELL NITRATE CONCENTRATIONS > 3-3
below 0.05 mg/L to 0.05 mg/L. Approximately 18% of the observations were at or below the
detection limit.1

The intercept (/?0) will capture ambient nitrate levels in the absence of human influences from
septic systems, medium and large AFOs, and atmospheric nitrogen deposition. Loadings
estimates for small-sized AFOs were not available. Thus, they are implicitly included in the
intercept term.

Independent Variables

The independent variables used to explain nitrate concentrations in well water are classified into
two groups: well and land characteristics, and nitrogen inputs. All data are from the
Retrospective Database unless otherwise noted.

       Well and Land Characteristics

AgDummy: This is a dummy variable for agricultural land use. The ag dummy variable was set
to 1 when the land use in the vicinity of the well was agricultural. For all other land uses (the
remaining categories were woods, range, urban, and other), the dummy was set to zero.

Soil Group: Soil group is a classification system that integrates several hydrological variables,
including runoff potential, permeability, depth to water table, depth to an impervious layer, water
capacity, and shrink-swell potential. Lower numbers have the greatest permeability and water
transmission rates, and are therefore more susceptible to surface pollutants (Mueller et al., 1995).

Well Depth: Well depths in the Retrospective Database ranged from 1 ft to 1,996 ft.  For
observations used in the regression analysis, the maximum well depth was 1,996 ft and the mean
depth was 170ft.
1. Alternative treatment of observations below the detection limit were evaluated using the gamma model
described below. These alternatives included setting nondetects equal to 0.001 mg/L and setting all nitrate
levels below 1 mg/L equal to 1 mg/L. These alternative specifications had little impact on the model overall,
and almost no impact on the estimated loadings parameter, which is the key component of the model for
CAFO loadings analysis.

-------
                    MODELING WELL NITRATE CONCENTRATIONS > 3-4
Regional Dummies: EPA defined five regions for use in this analysis: Central, Mid-Atlantic,
Midwest, Pacific, and Southwest.2 A regional dummy was created for each region (equal to one if
the well is in the region, equal to zero otherwise), to help account for regional differences not
captured by the other independent variables included in the model. The Midwest dummy was
used as the basis variable and was not included in the model. Thus the estimated parameters for
each of the other dummies indicate how nitrate levels in that region compare to nitrate levels in
the Midwest.

       Nitrogen Inputs

Septic Ratio: The septic ratio is equal to the number of housing units using septic systems per
acre in the county. The number of septic systems was obtained from the 1990 U.S.  Census.
County size (in acres) was taken from the 1992 Census of Agriculture.

Atmospheric Nitrogen: Estimated atmospheric nitrogen deposition in the area near  each well is
included in the Retrospective Database. The values used in the regression ranged from 0.54 to
8.92 pounds per acre.

Loadings Ratio: EPA calculated total nitrate loadings in each county as the total estimated
leached nitrogen from AFOs (both from manure application and from a variety or sources at or
near the AFO production areas), and from the application of fertilizers. EPA divided this total  by
total county acres to create a consistent unit across all counties. The assumption is that, in
general, once nitrate leaches into the groundwater it is dispersed in a volume of groundwater
proportional to the  county size. EPA obtained estimates of leached nitrogen from manure and
fertilizer loadings from  the NPLA (TetraTech, 2002).

Exhibit 3-2 lists summary statistics for the dependent and independent variables for the
2,985 observations used in the regressions described below.
2. The composition of the five regions is:

Midwest: ND, SD, MN, WI, IA, IL, MI, IN, MO, NE, KS
Mid-Atlantic: ME, NH, VT, NY, MA, RJ, CT, OH, NJ, PA, DE, MD, VA, WV, KY, TN, NC
Pacific: CA, OR, WA, AK, HA
Central: ID, MT, WY, NV, UT, CO, AZ, NM, TX, OK
South: AR, LA, MS, AL, GA, SC, FL.

-------
                     MODELING WELL NITRATE CONCENTRATIONS > 3-5
Exhibit 3-2
Summary Statistics
Variable
Nitrate Concentrations
Loadings Ratio
Atmospheric Nitrogen
Well Depth
Soil Group
Septic Ratio
Ag Dummy
Central Region Dummy
Mid-Atlantic Region Dummy
Pacific Region Dummy
South Region Dummy
N
2,985
2,985
2,985
2,985
2,985
2,985
2,985
2,985
2,985
2,985
2,985
Mean
3.57
2.02
5.07
170.07
2.42
0.03
0.78
0.06
0.39
0.12
0.07
Standard
Deviation
6.51
4.16
1.87
136.11
0.66
0.03
0.42
0.25
0.49
0.33
0.26
Minimum
0.05
0.00
0.54
1.00
1.00
0.00
0.00
0.00
0.00
0.00
0.00
Maximum
84.30
18.35
8.92
1,996.00
4.00
0.15
1.00
1.00
1.00
1.00
1.00
3.2    THE STATISTICAL MODEL

EPA used regression analysis to estimate the statistical model described in Equation 3-1 using
the data sources discussed in Section 2.2. EPA evaluated several different statistical models and
chose a "gamma model" because it best fit the data.3 The gamma model and the other statistical
models EPA tested are discussed in detail in Appendix B.

Exhibit 3-3 provides the output of the gamma regression model. Most of the explanatory
variables are significant. The exceptions are atmospheric nitrogen and the septic ratio. In
addition, all have the expected sign. This implies that the model produces intuitive results and
that the independent variables do help explain the variation in the nitrate levels. In particular the
regression results indicate that wells on agricultural land (ag dummy) have a higher well nitrate
concentrations. Wells located under less permeable soils (soil group) and deeper wells (well
depth) have lower well nitrate concentrations.  The positive parameter estimates for the three
sources of nitrogen (septic systems, atmospheric deposition, and animal feeding operations)
indicate that each source positively contributes to well nitrate concentrations. The model can thus
be used to help understand how changes in the independent variables (e.g., nitrogen loadings,
3. The term gamma model is used because the chosen regression is based on a gamma distribution, rather than
the normal distribution (as is used in ordinary least squares regression), or another type of distribution.

-------
                    MODELING WELL NITRATE CONCENTRATIONS > 3-6
Exhibit 3-3
Gamma Regression Results
Variable
Intercept
Loadings Ratio
Atmospheric Nitrogen
Well Depth3
Soil Group
Septic Ratio
Ag Dummy
Central Region Dummy
Mid-Atlantic Region Dummy
Pacific Region Dummy
South Region Dummy
Alpha
Parameter
Estimate
2.201
0.046
0.032
-0.171
-0.384
1.618
0.686
-0.076
-0.165
0.812
-0.907
0.497
Standard
Error
0.194
0.007
0.028
0.012
0.044
1.728
0.064
0.160
0.098
0.117
0.127
0.010
Asymptotic
T-Statistic
11.352
6.543
1.144
-13.782
-8.660
0.936
10.663
-0.475
-1.691
6.918
-7.170
50.639
Significance
0.000
0.000
0.253
0.000
0.000
0.349
0.000
0.635
0.091
0.000
0.000
0.000
Mean log-likelihood = -1.85646.
N = 2,985.
a. In the model, well depth is scaled to units of hundreds of feet.
well depth, land use) affect the expected level of nitrate at the well. Therefore EPA used this
model as the basis of its analysis of how reducing nitrogen loadings from CAFOs will affect
nitrate concentrations in domestic drinking water wells.
3.3    FITTED VALUES AND SCENARIO MODELING

After estimating the gamma model using the baseline loading information, expected values for
well nitrate concentrations were calculated using baseline loadings from 2,985 observations and
loadings from the 12 regulatory scenarios. As described above, the 12 regulatory scenarios are
based on different manure application rates, manure management practices, and monitoring
requirements. Loadings for the 12 regulatory scenarios were input into the model to estimate well
nitrate concentrations under these scenarios. In the analysis, the loadings ratio is the only variable
that changes across scenarios.

-------
                     MODELING WELL NITRATE CONCENTRATIONS > 3-7
Expected well nitrate concentrations under the 12 loadings scenarios were compared with the
expected well nitrate concentrations using the baseline loadings. EPA used the changes projected
from the model to calculate percentage differences in expected well nitrate concentrations under
the different regulatory options and scenarios. These were calculated by dividing the difference
from baseline for the expected values from the 12 different scenarios by the expected values from
the baseline loadings. These percentage differences were then applied to the actual nitrate
concentrations, the observed well nitrate concentrations from the Retrospective Database, to
calculate well nitrate concentrations under the various scenarios. The expected percentage
changes in nitrate concentration for each scenario are summarized in the last two columns of
Exhibit 3-4.
Exhibit 3-4
Characteristics of Benefits Analysis Scenarios
Regulatory Scenario
Option 1 — Scenario 6
Option 1 — Scenario 7
Option 1 — Scenario 8
Option 1 — Scenario 9
Option 2/3 — Scenario 6
Option 2/3 — Scenario T
Option 2/3 — Scenario 8a
Option 2/3 — Scenario 9
Option 5 — Scenario 6
Option 5 — Scenario 7
Option 5 — Scenario 8
Option 5 — Scenario 9
Change in Nitrate Loadings from
CAFOs (calculated across all
counties in the loadings dataset)b
Mean %
Reduction
21.05%
16.00%
16.02%
14.46%
22.46%
17.38%
17.40%
15.77%
14.12%
11.52%
11.54%
10.35%
Median %
Reduction
13.34%
8.42%
8.42%
5.71%
15.24%
9.74%
9.74%
7.32%
8.65%
5.18%
5.18%
3.38%
Nitrate (mg/L), Predicted by
Gamma Model
Mean %
Reduction
2.23%
1.89%
1.89%
1.83%
2.41%
2.07%
2.07%
2.02%
2.16%
1.88%
1.88%
1.84%
Median %
Reduction
0.40%
0.31%
0.31%
0.18%
0.43%
0.35%
0.35%
0.22%
0.32%
0.28%
0.28%
0.16%
a. Proposed scenarios.
b. Includes loadings from fertilizer application.
As indicated in the literature surveyed, although nitrogen loadings from CAFOs are significant
contributors to nitrate concentrations in wells, they are not the only important factor. Therefore
an analysis that does not incorporate these other factors, and assumes that the relationship
between nitrate concentrations and nitrogen loadings is directly proportional, will overestimate
the potential changes in nitrate concentrations due to decreased loadings. Exhibit 3-4 summarizes

-------
                    MODELING WELL NITRATE CONCENTRATIONS > 3-8
changes in nitrate concentrations as predicted by the gamma model, compared with percentage
changes that would be assumed if only changes in loadings (shown in the second and third
columns of Exhibit 3-4) were used to estimate nitrate concentrations.

EPA tested the ability of the gamma model to estimate small nitrate concentrations by comparing
the model's intercept with the natural, or ambient, level of nitrate in groundwater in the United
States.4 Using the mean values for soil group and well depth and setting all other variables to
zero (i.e., setting the ag dummy and  all human nitrogen sources to zero), the model predicts an
ambient nitrate concentration in the Midwest region of 1.32 mg/L on nonagricultural lands.
Using the same approach, the predicted value on agricultural land is 2.63 mg/L. Several studies
report natural nitrate levels ranging between 2 and 3 mg/L (Poe and Bishop, 1992;  Kross et al.,
1993; Poe, 1998), although one study suggests that 3 mg/L may be too high, given  the high
number of wells with nitrate levels below the detection limit in many groundwater monitoring
studies (Spalding and Exner,  1993).  Giraldez and Fox (1995) report that natural nitrate
concentration in groundwater is generally about 1.0 mg/L. Therefore the model's estimate of
1.32 mg/L on non-agricultural land seems to be a reasonable estimate of nitrate concentrations in
the absence of the pollution from septic systems, atmospheric deposition, and AFOs.
3.4    DISCRETE CHANGES FROM ABOVE THE MCL TO BELOW THE MCL

Census data show that approximately 13.5 million households in the United States use domestic
wells and are located in counties with animal feedlot operations.5 Based on the USGS
Retrospective data, 9.45% of these wells currently exceed 10 mg/L.6 This is roughly 1.3 million
domestic wells. Applying the percentage reductions, between 107,000 and 149,000 households
that are above the MCL at baseline are expected to be brought under 10 mg/L. Results are
displayed in Exhibit 3-5.
3.5    INCREMENTAL CHANGES BELOW THE MCL

Many households on wells with nitrate concentrations below the MCL at baseline may also gain
benefits from incremental changes in nitrate concentrations below the 10 mg/L level and above
the natural level, which is assumed to be 1 mg/L (see discussion in Section 3.3). Thus EPA
4. Technically, the intercept term includes ambient levels of nitrates as well as those induced by loadings from
AFOs with less than 300 AUs since these are not included in the loadings data.

5. The NPLA data indicates that 2,637 counties in the United States have AFOs.

6. Thus 9.45% of the wells in the Retrospective Database that had enough information to be included in the
gamma model (see discussion in Section 3.1) were found to have nitrate concentrations above the MCL.

-------
                    MODELING WELL NITRATE CONCENTRATIONS > 3-9
Exhibit 3-5
Expected Reductions in Number of Households with Well
Nitrate Concentrations above 10 mg/L
Regulatory Scenario
Option 1 —
Option 1 —
Option 1 —
Option 1 —
Option 2/3 -
Option 2/3 -
Option 2/3 -
Option 2/3 -
Option 5 —
Option 5 —
Option 5 —
Option 5 —
a. Proposed
Scenario 6
Scenario 7
Scenario 8
Scenario 9
— Scenario 6
— Scenario T
— Scenario 8a
— Scenario 9
Scenario 6
Scenario 7
Scenario 8
Scenario 9
Reduction Using Expected
Percentage Change
148,705
120,823
120,823
120,823
148,705
120,823
120,823
111,529
144,058
106,882
106,882
106,882
scenarios.
assumed that these incremental benefits are gained only for wells beginning with concentrations
between 1 and 10 mg/L. EPA did not calculate values for incremental changes where well
concentrations remain above the MCL because reliable value estimates do not exist for changes
in incremental nitrate concentrations above the MCL.

For households that start above the MCL preregulation and move below the MCL
post-regulation, EPA also did not calculate values for incremental changes below the MCL.
Based on the available valuation literature (see Chapter 5) there are no reliable estimates for
valuing incremental changes below the MCL in addition to valuing changes reductions to the
MCL; thus counting both values could double count some portion of the benefits for these
households. Exhibit 3-6 shows the average reduction in nitrate concentrations for wells between
1 and 10 mg/L at baseline, for each of the scenarios. Approximately 5.77 million households will
benefit from these incremental reductions.

-------
                   MODELING WELL NITRATE CONCENTRATIONS > 3-10
Exhibit 3-6
Mean and Median Reductions in Nitrate Concentrations for Wells
with Concentrations between 1 and 10 mg/L at Baseline
Scenario
Option 1 — Scenario 6
Option 1 — Scenario 7
Option 1 — Scenario 8
Option 1 — Scenario 9
Option 2/3 — Scenario 6
Option 2/3 — Scenario T
Option 2/3 — Scenario 8a
Option 2/3 — Scenario 9
Option 5 — Scenario 6
Option 5 — Scenario 7
Option 5 — Scenario 8
Option 5 — Scenario 9
Mean
Reduction in
[N] (mg/L)
0.14
0.12
0.12
0.11
0.15
0.13
0.13
0.13
0.14
0.12
0.12
0.11
Median
Reduction in
[N] (mg/L)
0.02
0.02
0.02
0.02
0.03
0.02
0.02
0.02
0.02
0.02
0.02
0.01
Households
Benefitting from
Incremental
Nitrate
Reductions
5,785,564
5,729,800
5,729,800
5,543,918
5,813,446
5,771,623
5,771,623
5,595,036
5,427,742
5,399,860
5,399,860
5,292,978
Total Expected
National Nitrate
Reduction
(mg/L)
854,326
716,007
716,007
695,662
927,730
788,287
788,305
768,221
836,895
717,982
717,995
701,889
a. Proposed scenarios.
3.6   TIMELINE FOLLOWING SCENARIO IMPLEMENTATION

Once new animal waste management practices are implemented, a time lag will exist between
implementation of these practices and realization of lower nitrate concentrations in wells and the
benefits from these reductions. The length of this time lag may be highly variable for any given
well and depends on a number of site-specific variables. The following is a brief description of
some of the more important variables affecting the time lag in response.

Depth to the saturated groundwater at the location where waste is applied affects the length of
time required for lower concentration (assuming improved waste management at the surface)
water to reach the groundwater. A considerable amount of water is stored in the unsaturated soil
zone beneath agricultural areas. When new "fresh" water leaches below the zone of plant rooting
(root zone), it replaces the uppermost water in this unsaturated storage, and "pushes" some of the
lower water into the saturated groundwater where it can move laterally toward surrounding wells.

-------
                    MODELING WELL NITRATE CONCENTRATIONS > 3-11
In many cases, relatively little change occurs in the nitrate concentration of the water between the
bottom of the root zone and the top of the saturated groundwater. While the progression of the
freshwater is not uniform because of faster flow along paths of preferential flow, generally the
fresh water must replace all the stored water in the unsaturated zone before an improvement is
seen in the groundwater immediately beneath the site of application.

In agricultural areas of the United States, depths to groundwater may vary from a few feet to over
100 ft. While some selected regions may have shallow or deep groundwater, these depths do not
vary clearly according to regional patterns, since they are determined as much by landscape
position and geology as by climate. Shallow groundwater is found in riparian areas and river
valleys of the arid West as well as on the Atlantic coastal plain.

The amount of excess water and properties of the soil or rock in this unsaturated zone also affect
the length of time required for the fresh water to reach the groundwater. A coarse-textured
material such as a sandy soil may only hold 1 inch of water for each foot of soil. In this case,  1 ft
of excess water infiltrating (a reasonable amount for a humid climate or a moderate irrigation in a
semi-arid climate) would move the "front" of cleaner water an average of 12 ft downward.
However, less coarse media such as a soil with moderate clay content may easily hold an average
of 3 inches  of water per foot of soil, so the same excess water infiltration will move the leading
edge of the  cleaner water only 4 ft downward.

Other factors  that influence how quickly the nitrate concentration at a well responds to improved
surface management are the amount of groundwater present, the distance between the well and
the point of waste application, and the velocity and direction of regional groundwater flow. In a
highly conductive aquifer with a steep groundwater gradient, the water may move a mile or more
in a year. In other cases, 10 ft or 20 ft in a year is more realistic.  In addition to how fast the
groundwater flows, the  amount of "older" water in the aquifer from which a well is drawing will
affect how quickly the response to improved management is reflected in a well. If the well is
drawing from 100 vertical ft of an aquifer, the upper levels of the aquifer may have nitrate
concentrations reflecting relatively recent management on nearby lands, and the lower levels of
the aquifer  still reflect poor management from prior years. Other local factors such as pumping of
other wells  and other sources of aquifer inflow such as leakage from nearby reservoirs or water
exchange with rivers combine to make the question of lag in well water response time highly
variable and site specific.

To estimate the value of improved groundwater quality from implementation of new CAFO
waste regulations, an estimate of the response time of an average well is needed. More
specifically, a realistic estimate is needed of how much time it will take after regulatory
implementation for the benefit of improved nitrate concentrations to be realized at the wellhead.

In sandy soils in central Kansas, Townsend et al. (1996) observed a response in the top layers of
the shallow groundwater, approximately 30 feet below the ground surface, in the first year after
implementation of improved surface management. The concentrations in this uppermost layer

-------
                    MODELING WELL NITRATE CONCENTRATIONS > 3-12
continued to improve and had dropped from near 25 mg/L to near 5 mg/L in six years. However,
nitrate concentrations approximately 20 ft lower in the aquifer continued to increase during the
same period.

Simulations by Hall (1996) of nitrate concentrations in the alluvial aquifer along the South Platte
River in northeastern Colorado suggest that significant improvements in nitrate concentrations in
the aquifer were realized as soon as a few years after implementation of improved management
practices. However, in these simulations, reductions in concentrations continued for more than
50 years, with relatively rapid improvements in the first 15 years and a decreasing rate of
improvement in later years as the simulated concentrations in the aquifer approached a new
steady state. The new steady state was somewhat reflective of the leaching concentrations under
the improved management scenario.

The South Platte alluvial setting is  a highly conductive aquifer with modest regional groundwater
gradients. The saturated groundwater at both the Kansas and Colorado sites is also somewhat
shallow. The response times in these cases are likely to be more rapid than for the United States
as a whole. Considering the range of aquifer depths and characteristics that might be expected,
we have assumed that 75% of the reduction in nitrate concentrations at the well heads will be
realized in 20 years. The drop in nitrate concentration is likely to be nonlinear, with more rapid
declines in early years. The shape of the concentration curve through time is unknown, however,
and the additional decline in concentration in later years becomes increasingly small. Without
better information, for this analysis EPA has made the conservative assumption that the
concentration curve is linear, resulting in an estimated period of 27 years for improved CAFO
waste management to improve an aquifer to its new equilibrium (i.e., "clean") status.

-------
                                     CHAPTER 4
                       VALUATION: BENEFITS TRANSFER
Several approaches could be used to estimate the benefits from changes in well nitrate
concentrations. The first issue to address is whether to obtain primary data on potential benefits
or use existing data. Given limited time and budget constraints, collecting primary data for a
nationwide sample is not feasible. We thus decided to apply a benefits transfer approach to
existing studies of household values for reduced well nitrate contamination.

"Benefits transfer" refers to the "application of existing valuation point estimates or valuation
function estimates and data that were developed in one context to value a similar resource and/or
service affected by the discharge of concern" [59 FR  1183]. In other words,  benefits transfer
entails applying empirical results obtained from a primary research effort conducted at one site
and set of circumstances to another (similar) site and  set of circumstances. In this manner,
existing research findings from a "study site" can be used as an expeditious  means of drawing
inferences regarding the magnitude of benefits or damages associated with a change in resource
conditions at a "policy site."
4.1    BENEFITS TRANSFER METHODS

There are four ways to transfer benefits: transfer an average value, transfer a function, calculate a
metafunction, or calibrate a preference. Crutchfield et al. (1997) discuss transferring an average
value and transferring a function, preferring transferring a function if data are available on the
sociodemographic characteristics of the original study and the policy site. Walsh et al. (1992)
develop what is essentially a meta-analysis of outdoor recreation demand studies for use in
benefits transfer analysis, and Boyle et al. (1994) present preliminary results of a meta-analysis of
groundwater valuation studies. Smith et al. (1999) discusses the preference calibration approach.
These four approaches are ordered in terms of increasing data requirements, increasing costs of
implementation, and increasing sophistication of the value estimates provided.
4.1.1   Transfer an Average Value

Transferring an average value has been the most common approach to benefits transfer. It entails
subjective evaluation on the part of the researcher to evaluate the validity and reliability of the
original studies and to make reasonable assumptions in transforming the original values into
those to be used in the new application. Transferring an average value can in a sense be a

-------
                           VALUATION: BENEFITS TRANSFER > 4-2
qualitative meta-analysis. Adjustments are often made based on the characteristics of the original
scenarios and the new scenario and on sociodemographic characteristics of the affected
population (e.g., income). Primary evaluation criteria would include:

+      the relevance of the commodity being valued in the original studies to the policy options
       being considered for CAFOs

>•      the quality (robustness) of the original study, evaluated on multiple criteria such  as
       sample size, response rates, and significance of findings in statistical analysis.

Much of the summary analysis of existing studies necessary for the average value method is also
necessary for the next three approaches. At a minimum, the initial work required for an average
value approach provides an initial assessment of the quality and availability of data that  could be
used in the other approaches.
4.1.2  Transfer a Function

Transferring a function from a specific study is generally more limited than using average values
from a number of different studies. Our evaluation of nitrate related groundwater valuation
studies does not reveal any one study that would be best suited for this approach. The primary
limitation in transferring a function is the fact that none of the studies involves a national sample
of values for reducing nitrate contamination. The applicability of a single local or regional study
to a national benefits assessment requires careful consideration of the likely representativeness of
the original study. Loomis (1992) further examined the benefit transfer function  approach and
empirically tested for the transferability of a function between states. Loomis' findings suggest
that benefit functions are not always directly transferable between states. This suggests that,
whatever method is adopted, spatially distinct benefit estimates should be examined for
consistency when transferring benefit estimates.
4.1.3  Calculate a Metafunction

Meta-analysis is a set of statistical procedures used to assess results across independent studies
that address a related set of research questions. It is a method for combining the effect sizes from
several studies; it is essentially an analysis of analyses (Wolf, 1986). A metafunction is the end
product of a meta-analysis in which the marginal effects of study or scenario characteristics on
willingness to pay are estimated. Such a function could potentially be used in a new policy
situation by inputting the relevant scenario characteristics for the policy analysis to derive the
relevant value estimate.

As discussed in Chapter 5, we identified 11 studies that derive values for reducing nitrates in
groundwater. Our examination of the 11 nitrate valuation studies suggests that a meta-analysis of

-------
                          VALUATION: BENEFITS TRANSFER > 4-3
these was not reasonable for the current benefits transfer. There is considerable difference in the
basic nature of many of the studies, which limits the number that would be usable in a meta-
analysis. There are significant differences in the commodities being valued (e.g., certain current
cleanup versus potential future cleanup of a portion of contaminated waters) and the types of
values being elicited (e.g., use values versus total values versus option values).
4.1.4   Calibrate a Preference

Preference calibration is a relatively new approach to benefits transfer analysis that builds on
existing methods and attempts to develop a utility-theoretic approach to benefits transfer (Smith
et al., 1999). Rather than deriving a transfer function, this approach attempts to derive a model of
preferences based on results from prior studies. This method may prevent errors in the other
approaches that may bias value estimates either up or down. Preference calibration requires
several steps:

1.      Specification of a preference ordering that dictates how a "representative" individual
       makes decisions (such as a constant elasticity of substitution, CES).

2.      Identification of relationships, axioms, and assumptions (such as utility maximizing
       behavior, demand is obtainable using Roy's identity, or a choke price exists) necessary so
       that the preference parameters are identified.

3.      Derivation of a closed-form solution for a willingness-to-pay (WTP) function
       (e.g., compensating  variation) and addition of supplemental data to identify the unknown
       parameters. Using data on consumer surplus values associated with marginal and/or
       incremental change  in environmental quality to be valued by the benefits transfer and
       other information on variables such as income,  rent, or travel costs for the representative
       individual, the implied values of the parameters are backed out of the WTP function.

4.      With the identified and estimated parameters, the WTP function is  now estimated and any
       set of environmental variables can be input to generate other Hicksian consumer surplus
       estimates.

Smith et al. (1999) do not claim the new approach necessarily results in smaller error. In fact, the
authors state, ". . . the measure from preference calibration is simply a more complex set of
numerical calculations." The advantage of preference calibration is that it is based on utility-
theoretic behavioral theory. Preference calibration is expected to rely on a much larger set of
assumption, axioms, economic relationships, and possible supplemental data than either the unit
value approach or meta-analysis. The data  requirements for preference calibration and the
additional assumptions required to choose  any one particular functional form may outweigh the
benefits of using a more theory-based approach.

-------
                          VALUATION: BENEFITS TRANSFER > 4-4
4.2    CHOICE OF METHODS

The average value approach is the most feasible one for analysis of potential benefits under the
proposed regulatory options. In part this choice is made because of the difference between the
benefits transfer approach used here and those generally discussed in the literature. Most
literature discusses the transfer of benefits from a specific study situation to another specific
policy situation. Adjustments are then made based on differences between the "study site" and
the "policy site." In the case of benefits of CAFO regulations, the "policy site" is all counties in
which potentially regulated CAFOs are located. Given limited resources, it is not feasible to
identify individual county characteristics in a manner that would allow the use of a transfer
function. In particular, we do not have information on income or other sociodemographic
characteristics of those individuals living in any given county who obtain their water from a
private well, as opposed to sociodemographic characteristics of the general population of the
county. In part to control for this, we use benefit estimates from studies that focus on private well
users in situations likely to be similar to that around CAFO locations. In this manner, the  original
studies are more likely to already have captured sociodemographic characteristics of the "policy
situation" population.

As noted above, we do not believe that there is sufficient information in the studies considered
below to use a transfer function or to develop a meta-analysis that would provide information
significantly better than that gained from the average price approach because  of the limited
number of studies and the significant methodological differences between them.  The same
scarcity of information and limited resources  preclude the use of the preference calibration
approach.

-------
                                    CHAPTER 5
                    GROUNDWATER VALUATION STUDIES
5.1    LITERATURE SEARCH AND REVIEW

The objective of the literature search was to identify studies that had developed or elicited values
for changes in groundwater quality. A number of studies deal with groundwater contamination
not related to nitrates. We limit the discussion here to those that focused on values for reductions
in or prevention of increases in nitrate contamination for drinking water wells. Our evaluation of
the literature led us to eliminate some studies that were of poorer overall quality or for which
only limited information was available.

We identified 11 relevant studies through an extensive search of literature using databases,
listservers, and the bibliographies of similar studies that addressed groundwater valuation. The
databases searched for this study were the Colorado Association of Research Libraries (CARL),
which includes the holdings of several university libraries in Colorado and the West, and the
Environmental  Valuation Resource Inventory (EVRI), a database compiled by Environment
Canada that includes empirical studies on the economic value of environmental benefits and
human health effects. Messages were sent to the ResEcon listserver, which  includes
approximately 700 individuals in the field of natural resource and environmental economics,
soliciting suggestions for articles pertaining to groundwater valuation and nitrate contamination.
Finally, several references cited in the studies that we identified using the databases and listserver
were used as well.
5.2.   OVERVIEW OF GROUNDWATER NITRATE VALUATION STUDIES

The following is a brief overview of the 11 studies we evaluated for inclusion in the benefits
transfer. Some of the information about these studies came from more than one report or paper
based on the study. Where relevant, we identified the most recent information about each study
from available literature. Summary information on these studies is presented in Appendix C.
5.2.1   Crutchfield et al., 1997

Crutchfield et al. (1997), Crutchfield et al. (1995), and Crutchfield and Cooper (1997) evaluated
the potential benefits of reducing or eliminating nitrates in drinking water by estimating average
WTP for safer drinking water. They received survey responses from 819 people in rural and

-------
                        GROUND WATER VALUATION STUDIES > 5-2
nonrural areas in four regions of the United States (Indiana, Nebraska, Pennsylvania,
Washington). Using the contingent valuation method (CVM) with valuation questions in a
dichotomous choice format, respondents were asked what their willingness to pay would be to
have the nitrate levels in their drinking water a) reduced to "safe levels," and b) completely
eliminated. Respondents were told that this would be accomplished using a filter installed at their
tap, and the cost for the installation and maintenance of the filter would be paid to a local water
agency. Respondents were also asked sociodemographic characteristics such as income, age,
education, and whether they currently use treated or bottled water. Crutchfield et al. used a
bivariate probit estimation for responses to the dichotomous choice questions. Across all regions,
the calculated willingness to pay per household to reduce nitrates to safe levels ranged from
$45.42/month to $60.76/month, with a mean of $52.89. The willingness to pay to remove nitrates
from drinking water ranged from $48.26/month to $65.1 I/month, with a mean of $54.50. Besides
income and program  cost, Crutchfield et al. found two variables to be significantly related to a
respondent's willingness to pay: "years lived in ZIP code" was positively correlated and "age of
respondent" was negatively correlated.

Evaluation: An important advantage of the Crutchfield et al. valuation approach is that they
surveyed individuals  in four different areas of the country, thus providing value estimates more
representative of national values. The annual WTP to reduce nitrates to the safe level
($52.89/month x 12 months) is $634.68. Crutchfield et al. compared  annual per household WTP
estimates from their study to three others (including Jordan and Elnagheeb, 1993, described
below). Values for reducing nitrates to either safe levels or to zero are higher in Crutchfield et al.
than the other three studies. Crutchfield  et al.'s estimate of $634.68/household/yr is not
unreasonably higher than the $412-$484/household/yr values discussed in Poe and Bishop (1992)
below. The difference in values between the two programs is likely to be representative of values
for incremental reductions in nitrates in drinking water. The difference between reducing nitrates
to zero and reducing nitrates to safe levels is $1.61  per month. For a change between the  MCL of
10 mg/L and 0 mg/L, this represents a per mg/L monthly WTP of $0.16, which is $1.92 per mg/L
annually (in 1997$).
5.2.2   De Zoysa, 1995

De Zoysa (1995) and Randall and De Zoysa (1996) discuss a contingent valuation study designed
to estimate the benefits from three environmental services in the Maumee River basin in
northwestern Ohio, including stabilization and reduction of nitrate levels. Rural and urban areas
in the river basin were sampled and one out-of-basin urban area was sampled, with 427 returned
questionnaires. Using a dichotomous choice format, a portion of the respondents were asked
whether they would pay different amounts, via a one time special tax, to reduce nitrate
contamination from fertilizer applied to fields. Under the hypothetical scenarios, nitrate
concentrations would be reduced from the current range of 0.5-3.0 mg/L to a range of
0.5-1.0 mg/L. Individuals were also asked questions regarding sociodemographic characteristics,

-------
                        GROUND WATER VALUATION STUDIES > 5-3
preferences for priorities for public spending, and how they used the resource in question
(e.g., how many trips they had taken to the area). From these responses, Randall and De Zoysa
formed two datasets: one that included only yes/no (YN) responses, and another that included
yes/no and protest votes (YNP). The multivariate analysis was conducted using a probit model;
income, the level of priority placed on groundwater protection, and the interest in increasing
government spending on education, healthcare, and vocational training all were positively and
significantly correlated with willingness to pay to improve groundwater quality.

Randall and De Zoysa reported various  WTP estimates using median and lower bound mean
estimates for groundwater, surface water, and wetlands programs or combinations of these
programs. For this analysis, we examine "stand-alone" WTP estimates for groundwater programs
that would reduce nitrates in groundwater. Median WTP for groundwater ranged from $71.03 for
the YN responses to $20.80 for YNP responses. Lower bound mean WTP for groundwater
ranged from $88.49 for the YN responses to $52.78 for YNP responses. Randall and De Zoysa
expressed a preference for the YNP models because they felt there was no strong reason to
assume that the protest responders had nonzero values. They also stated that for policy purposes
the mean values are the appropriate measure for which the "lower bound mean" provides a lower
bound estimate.

Evaluation: The reduction in groundwater nitrate levels is from a range of 0.5 to 3.0 mg/L to a
range of 0.5 to 1.0 mg/L. Taking range means, the reduction in nitrates is  from 1.75 mg/L to
0.75 mg/L, or a reduction of 1.0 mg/L. Using the lower bound mean values from the YNP model,
this represents a WTP of $52.78 per mg/L change in nitrate concentrations for incremental
changes below the 10 mg/L MCL. Since the valuation question was posed as a one-time special
tax, we can annualize the $52.78 per mg/L, which represents a net present value (since the
program would continue indefinitely). Using a 3% discount rate, this translates into an annual
WTP  of $1.61 per mg/L ($2.69 using a  5% discount rate and $3.76 using  a 7% discount rate).
5.2.3   Delavan, 1997

Using a CVM survey of 1,000 residents in two counties in southeastern Pennsylvania (with a
68% response rate), Delavan (1997) estimated willingness to pay to improve groundwater quality
(in 10 years, 75% of wells would meet the MCL). Delavan used CVM with two survey formats:
one presented a dichotomous choice question followed by an open-ended valuation question
(DOE), and the other presented information on current local government expenditures on public
health and safety services followed by an open-ended valuation question (IOE). Subjects were
also asked questions on their duration of residence, the current quality/safety of their water, and
their prior knowledge of water quality issues. Respondents were told that they would be assessed
a special tax annually for 10 years to increase the percentage of wells satisfying the MCL from
50% to 75% in their area. Tobit analysis was used to model the relationship between explanatory
variables and open-ended WTP, and a logit model was used to model protest bidders. Mean

-------
                        GROUND WATER VALUATION STUDIES > 5-4
annual WTP was $44.78 for the DOE surveys and $29.26 for the IOE surveys with protest
bidders, and $67.85 and $47.16, respectively, without protest bids. Delavan found that at
household incomes above $50,000, respondents' concern for their own safety as it relates to
drinking water, the priority respondents feel that government should place on protecting
groundwater, and respondents' perception of safety with and without the program were all
significantly and positively correlated with respondents' willingness to pay. He also found that
males were more likely to pay more for groundwater protection.

Evaluation: Delavan designed and thoroughly pretested the survey instrument and received a
reasonably strong response rates (68%) from a reasonably large sample (889). He tested and
controlled for protest bids and examined numerous hypotheses regarding respondents' attitudes
and values with respect to groundwater nitrate pollution. Although 40% of the respondents are on
private wells, regression analysis does not indicate a significant difference in WTP between
private well users and other water users.

Delavan elicited annual WTP for 10 years for a program to reduce the percentage of wells not
meeting the MCL from 50% to 25% (increase safe wells from 50% to 75%). Assuming
individuals perceive this as their own chance of having a well  above the MCL and assuming a
"linear in probabilities" utility function, the value for going from unsafe to safe for an individual
household with  certainty will be four times that of going from 50% to 75% certainty. Based on
these assumptions, annual WTP each year for 10 years from the IOE group without protests will
be $188.64. Annualizing this from a 10 year payment to a payment in perpetuity yields annual
WTP per household for reducing nitrates from unsafe to safe of $48.89, $74.22, and $94.96,
respectively, for 3%, 5%, and 7% discount rates. Given the assumptions made to translate the
Delavan values  into annual WTP estimates, we do not consider these estimates as reliable as
others that value WTP in a manner more consistent with those needed for benefits transfer to
CAFOs.
5.2.4   Edwards, 1988

Edwards (1988) conducted a contingent valuation study of household willingness to pay to
prevent uncertain future nitrate contamination of groundwater on Cape Cod, Massachusetts. The
785 respondents (585 provided useable responses), 89% of whom used a public water system,
were renters and both resident and nonresident property owners. The groundwater supply was
currently assumed to be safe, but fertilizer and sewage posed a potential problem because Cape
Cod relies on a sole source aquifer and measured nitrate levels had been increasing. Edwards
used dichotomous choice questions to estimate how much people would pay, using four payment
vehicles: (1) an annual bond to be paid in perpetuity, (2) a voluntary contribution, (3) water bills,
and (4) an unspecified payment mechanism. No significant difference was found between the
different payment vehicles. Edwards used a logit model to generate parameter estimates.
Edwards reported a WTP of $1,623 per household per year, for a management plan that would

-------
                         GROUND WATER VALUATION STUDIES > 5-5
increase the probability of supply from 0.0 to 1.0. Respondents' income, interest in ensuring safe
groundwater for future generations, and probability of how long they will live on Cape Cod were
all significantly and positively correlated with their willingness to pay for groundwater
protection.

Evaluation: Using the logit model with mean sociodemographic characteristic values, an annual
WTP for a certain water supply is calculated as $1,623  per year (1987$). This value is higher
than those found in other studies reviewed here, for several possible reasons. Edwards
specifically valued option price and option values, which may include risk premiums that some
of the other studies may not include. The unique characteristics of Cape Cod involving a sole
source aquifer suggest that WTP values will be higher there than in other locations with
alternative water resources. If nonuse values are a large component of Edwards' value estimate
because of the uniqueness of Cape Cod, then his value  estimate will be higher than those for less
unique locations more typical of counties with CAFOs. The high mean income of the sample
($55,000 in 1987$) is likely to lead to higher WTP estimates compared to other (lower mean
income) rural water users nationwide. Thus value estimates from Edwards probably represent an
upper bound if they are to be used in benefits transfer.
5.2.5   Giraldez and Fox, 1995

Giraldez and Fox (1995) conducted a cost-benefit analysis of controlling groundwater pollution
from agricultural use of nitrogen fertilizer in the village of Hensall (population 1,155 in 1986), in
southwestern Ontario. Nitrate concentrations in two wells in the village had recently exceeded
10 mg/L. These wells are sources for a public water distribution system that apparently does not
treat the water prior to delivery. Based on willingness-to-pay values from other studies, Giraldez
and Fox used three approaches to estimating values for reducing nitrates: (1) value of human life
as present value  of lifetime average earning, (2) value of statistical life (VSL) based on wage-risk
premiums, and (3) CVM. Based on values from CVM studies by Hanley (1989) and Edwards
(1988), Giraldez and Fox aggregated a cost of nitrate contamination for the entire village of
Hensall to range between about $30,0001 and $700,000 per year, depending whether bequest and
option values are included in the calculation. Based on a lifetime earnings approach, annual costs
ranged from $693 to $6,289 for the entire village. Using VSL estimates, Giraldez and Fox
estimated an annual benefit range of $984 to $111,639 for the village for reducing mortality
related to nitrate contamination. Potential mortality from nitrates is in infants only. The authors
concluded that because substantial uncertainty in both the benefits and costs calculations, they
could not decisively indicate whether the health benefits of reduced nitrate concentrations
justified the cost of changing local agricultural practices.
1. All dollars from Giraldez and Fox as reported in Canadian dollars. It is unclear what year Giraldez and Fox
are reporting dollar values for.

-------
                        GROUND WATER VALUATION STUDIES > 5-6
Evaluation: This study is primarily a benefits transfer study, which limits its use for the current
valuation exercise because we can simply use the primary studies if they are relevant. Giraldez
and Fox attempted to use two non-CVM approaches for deriving value estimates. It is generally
believed that the use of value of lifetime earnings is not an appropriate measure of welfare
impacts involving mortality risks (Freeman, 1993). It also seems unlikely that VSL estimates
from wage-risk studies can be directly applied to infant mortality risks. The value estimates
providing secondary value information from Hanley (1989) and Edwards (1988) imply values
between $72.73/year and $1,696.97/year (presumably in 1995$ Canadian), although as discussed
above Edwards provides a mean WTP of $l,623/year (1987$).
5.2.6   Hurley et al., 1999

Hurley et al. (1999) used data from a contingent valuation study in Clark and Adams counties in
Iowa to determine rural residents' willingness to pay to delay, by 10, 15, and 20 years, nitrate
contamination of their water supply from large animal confinement facilities. Baseline water
quality was not specified, although several highly publicized spills from these types of facilities
had occurred recently, and both counties rely heavily on surface water supplies for drinking
water. The authors mailed 1,000 surveys to a random sample of residents, of which 332 were
completed thoroughly and returned. Apparently 26% of respondents (about 85 total) were on
private groundwater wells (not municipal or rural water supply). It also appears that there could
be significant scenario rejection in this survey because less than 50% of respondents stated any
WTP for any delay in nitrate contamination and less than 10% stated WTP for  10 or 20 year
delays in nitrate contamination.

An ordered probit specification, with thresholds adjusted for possible anchoring, was used to
analyze the results. The results showed that higher education, income, and expected length of
time to remain in the community were positively and significantly correlated with willingness-to-
pay values. Male respondents were significantly less inclined to pay for water protection than
females. Based on analysis of these referendum questions, the willingness to pay ranged  from
$118.13 (for a 10 year delay) to $190.75 (for a 20 year delay) per year for a household with
sample mean characteristics.

Evaluation: A low overall response rate (33%), a small sample of private well users (85), and
potentially high scenario rejection bring results from this study into question for use in benefits
transfer. Some aspects of the scenario are unclear, such as what payment mechanism is used in
the valuation scenario. WTP in this study was elicited for delays in nitrate contamination, and
this does not translate directly into WTP for reducing current nitrates in private wells.
Furthermore, this study does not distinguish clearly between groundwater and surface water
nitrate contamination. We thus feel we cannot reliably translate values from this  study to
groundwater contamination from CAFOs without making significant assumptions to derive per
household annual WTP estimates for current benefits.

-------
                        GROUND WATER VALUATION STUDIES > 5-7
5.2.7   Jordan and Elnagheeb, 1993

Jordan and Elnagheeb (1993) conducted a contingent valuation study of residents' willingness to
pay for improvements in drinking water quality, using data from a statewide survey of a random
sample of 567 Georgia residents. Of the 199 complete responses received, 78% of respondents
were on public water systems and 22% (40 subjects) used private water systems. Water quality
was rated as "poor" by 27% of public users and 13% of private users. Respondents on private
wells were told to imagine that nitrate levels currently exceeded safety standards and those on
public  supply were told to imagine that nitrate levels were increasing (from an unspecified
baseline to an unspecified endpoint). Nitrate impacts were indicated as being due to nearby
agricultural activities. Respondents were asked how much they would be willing to pay (circling
one of seven values between $0 and $100) to "avoid the risk of increasing nitrate in [their]
drinking water." Public and private water users were given two separate scenarios to value:
private wells users were told they would be provided installation and maintenance of filtering
equipment and public system users were told that the water supplier would guarantee safe
drinking water. The cost for these services would be paid monthly, in perpetuity, through the
water bill for public users and a fee for private users.

Jordan and Elnagheeb used both OLS and maximum likelihood functions to generate parameter
estimates for their WTP model. The mean WTP for public and private water users, respectively,
was $128.20/household per year and $157.6I/household per year (1993$). The median was
$69.89/household per year for public users and $93.95/household per year for private users.
Respondents' income, years of education, and degree of uncertainty regarding their water quality
were positively and significantly correlated with the amount they were willing to pay. Females
and respondents who lived on farms were willing to pay more to avoid increases in nitrate in
their drinking water.

Evaluation: Jordan and Elnagheeb had a low overall response rate (35%) and a small sample of
individuals on private wells (38 after rejecting outliers). The scenario is unclear because  it
specifies nitrate levels currently somewhere above safe levels. The survey appears to be vague on
actual health impacts and specifies nitrate reduction to safe levels with little clarification of what
this means. Nitrate control is at the point of use for private wells and thus values are primarily
use values (no action is indicated to prevent aquifer contamination). Jordan and Elnagheeb did
not report the number or percentage of zero bids, and thus it is difficult to evaluate potential
scenario rejection. The best point estimate for private well owners' WTP for reducing nitrate
contamination to safe levels is $157.6I/household per year (1994$), which is primarily a use
value.

-------
                        GROUND WATER VALUATION STUDIES > 5-8
5.2.8   Poe and Bishop, 1992

Poe and Bishop (1992, 1999), and Poe (1993, 1998) conducted a contingent valuation study in
rural Portage County, Wisconsin, to estimate conditional incremental benefits of reducing nitrate
levels in household wells. The area had extensive nitrate problems, and previous research
suggested that 18% of private wells in the area exceeded the MCL. Two WTP valuation
scenarios are discussed in the various Poe and Bishop papers: WTP for a program to keep all
wells in Portage County at or below the MCL and WTP for a program to reduce well  nitrates in
all wells by 25%. Sources of nitrates identified in the information materials included "septic;
tanks, farm,  lawn, and garden fertilizers; livestock holding areas; and abandoned wells." In
particular, Poe and Bishop were interested in how providing respondents with information on
their own well nitrate concentrations was related to willingness to pay for nitrate reductions.

The survey thus comprised two stages. In the first stage, individuals were asked to submit water
samples from their tap and to complete an initial questionnaire. In the second stage, the
individuals were provided with their nitrate test results, general information about nitrates, and a
graphical depiction of their exposure levels relative to natural levels and the MCL, and they were
asked to complete contingent valuation questions. Poe and Bishop found no sample selection
bias between the first and second survey stages. Poe addressed potential nonlinearities by
allowing for a nonlinear WTP function where the degree of convexity or concavity is estimated
based on the data.

A total of 271 completed Stage 2 responses were received. In general, Poe and Bishop found that
respondents' knowledge of their water quality and awareness of the health effects of nitrates to be
positively and significantly correlated with willingness to pay. The various Poe and Bishop
papers report different WTP values for different types of analysis and for different portions of the
data set.

In their 1992 working paper,  Poe and Bishop (1992) report a mean ex post WTP of $257.10 per
household per year for a program to keep all wells in Portage County at or below the MCL. Poe
(1993) reports per household per year mean WTP values for a program to keep all wells in
Portage County at or below the MCL for different information levels and depending on whether
the individual had a prior test of actual well nitrate levels. These mean reported WTP values are
$199.73/household/yrNINT, $961.16/household/yr WINT, $244.32/household/yrNIWT, and
$526.63/household/yr WIWT (where MNT, WINT, NIWT, and WIWT mean "no information-no
test," "with information-no test," "no information-with test," and "with information-with test,"
respectively).2 Poe then calculates a mean WTP for prevention of well nitrates above  the MCL of
2. Values reported here from Table VII.2.3.2 from Poe (1993) for the mean and median values based on
1,000 draws using a Duffield and Paterson Simulation method for estimating mean WTP values.

-------
                        GROUND WATER VALUATION STUDIES > 5-9
$484 per household per year for households with a 100% probability of future contamination.3 In
terms of policy uses, it could be argued that the $484 value estimate represents the best informed
and most relevant value statement from respondents and thus should be used for benefits transfer.

Poe (1993) also calculates an imputed WTP for a 1 mg/L reduction (or increase) in nitrates as a
function of initial nitrate levels. A maximum per mg WTP of ~$120 is seen when initial nitrate
levels are close to 10 mg/L.  Above 10 mg/L the per mg WTP falls off to zero at about 22 mg/L.
Below 10 mg/L the per mg WTP falls to about $90 per mg when the initial level is 4 mg/L.
While this is an order of magnitude greater than Crutchfield et al. (1997) or De Zoysa (1995), it
is more in line with WTP values derived by Sparco (1995) for incremental changes in nitrate
concentrations of $123.56 per mg/L.

Poe (1998) reports WTP for the program to keep all wells in Portage County at or below the
MCL as a function of the individuals' observed well nitrate concentrations. Estimated WTP
values varied, as expected, by the results of the respondent's nitrate test. Those with a nitrate
level of 2 mg/L would pay $84.07/year, whereas a respondent with 40 mg/L of nitrate would be
willing to pay $515.59/year  to keep nitrate levels below the MCL.

Poe and Bishop (1999) also  estimated a nonlinear WTP function, including both single-power
and cubic formulations. They report WTP for the program to reduce well nitrates in all wells by
25%. Using the cubic function,  Poe and Bishop show that incremental benefits increase between
2 mg/L  and 14.5 mg/L and then fall to zero at about 22.5 mg/L. Since a 25% reduction from
14.5 mg/L would reduce nitrate levels to very near the MCL, this reduction could be considered  a
WTP to reduce nitrates to safe levels. The estimated WTP for a 25% reduction from 14.5 mg/L is
reported by Poe and Bishop  as $412 per year per household.

Evaluation: Overall, the high quality of the Poe and Bishop study suggests that benefit estimates
from this work are likely to  be reliable and valid. The Poe and Bishop work is based on a well
developed theoretical model of respondents' willingness to pay (e.g., Poe and Bishop is one of
the only studies to empirically assess potential nonlinearities in the WTP function). Survey
development, implementation, and analysis meet or exceed standards for CVM studies at the
time of the study. Poe and Bishop is also the only work we reviewed where respondents had
empirical information on the nitrate levels  in their own wells. Although the stage 2 sample size is
not large (271), the quality of the data is likely to be higher than for larger samples using less
well developed surveys.  The surveyed population (rural Wisconsin)  is most likely representative
of individuals facing potential well nitrate contamination from CAFOs.

Two value estimates from Poe and Bishop are the most applicable for benefits transfer. The first
is the mean WTP of $484/household/yr from the scenario of a program to keep all  wells in
Portage County at or below  the MCL for a household with a 100% probability of future
3. $347-$655 95% confidence interval (Table VII.2.4.2; Poe, 1993).

-------
                        GROUNDWATER VALUATION STUDIES > 5-10
contamination. The second is the $412/household/yr incremental value for a program to reduce
well nitrates by 25% for a well with a current nitrate concentration of 14.5 mg/L.
5.2.9   Sparco, 1995

Sparco (1995) used conjoint analysis to estimate the benefits of reduced groundwater
contaminant concentrations and subsequent risks of illness in Sussex County, Delaware. The
county is predominantly agricultural, and nitrate is a common pollutant in the groundwater. A
survey of private wells (Andres, 1991) found nitrate levels at or above 10 mg/L in 23% of the
county's wells, and 50% of households rely on their own drilled or dug wells for water.
Respondents were surveyed at public gatherings such as state fairs, and were asked to rate
preferences over four cards, including different attribute levels  of willingness to pay, nitrate
levels, atrazine levels, fecal coliform, and illness characteristics, as well as "attitudinal" questions
regarding the respondent's opinion on government intervention, agriculture, and the
environment.

Respondents were told that the contamination originated from agricultural activities. Sparco used
an ordered probit regression to analyze the responses. The total number of respondents was not
specified. The mean annual WTP (calculated from the ordered probit model) to reduce nitrate
contamination by 1 mg/L was $123.56. Calculated WTP values for 1 in 10,000 reductions in one-
week illness now or gastrointestinal (GI) cancer in 20 years of $129.58 and $370.72, respectively,
imply extremely large "value of statistical illness" (VSI) estimates. VSIs of nearly $13 million
for one week of illness now and of $37 million for GI cancer in 20 years seem implausible
compared to common value of statistical life estimates between $5 and $10 million (Chestnut
et al., 1997). A pro-environmental attitude was significant and negatively correlated with WTP
for nitrate reduction, and antigovernment intervention, and pro-farm viewpoints were significant
and correlated with WTP. While the signs of all three principal components appear to be
unexpected in the regression model, Sparco suggested that the signs of these three factors
indicate that survey respondents are supportive of farming in the county and believe that the
government should adopt a laissez-faire approach toward environmental regulation.

Evaluation: The methods  and analysis used in this study are good and predate current methods
in stated preference analysis using conjoint methods. Several issues, though, suggest limits to the
reliability and validity of value Sparco's estimates for use in benefits transfer. The sample is
nonrandom and the final sample size and response rates are unspecified. The apparently incorrect
signs  on attitudinal variables from the principal components analysis raise questions about the
model estimates. The experimental design had a significant effect on preference statements, and
it is unclear how this factors into value calculations. Sparco did not separate values between
private well users and municipal or community system users. And, as stated above, the value
estimates for illness characteristics seem implausibly high, casting some doubt on the reliability
of value estimates for incremental changes in nitrate concentrations of $123.56 per mg/L.

-------
                       GROUNDWATER VALUATION STUDIES > 5-11
5.2.10  Walker and Hoehn, 1990

Using information obtained primarily from an engineering model of the costs of water
purification technology, Walker and Hoehn (1990) developed a model of economic damages of
nitrate contamination in rural Michigan. The area has a history of elevated nitrate concentrations,
with a study reporting 34% of rural drinking water wells exceeding the MCL for nitrates (Vitosh,
1985). Over 95% of the rural residential water supply comes from groundwater. The authors
calculated net economic damages as the sum of producer and consumer surplus. The model
requires three components: a residential water demand function, a precontamination supply
function, and a post-contamination supply function. The demand function was assumed to be
linear, based on the quantity of water used per household, the average water price, household
income, rainfall, and the number of persons in the household. The precontamination supply
function is the incremental cost of providing water before contamination occurs, and is simply a
linear relationship with the initial price of water. The post-contamination function is the
incremental cost of providing water after nitrate contamination, and is the same as the
precontamination function plus the additional incremental cost of removing nitrates.

The incremental cost of nitrate removal was estimated from a sample of costs for nitrate removal
generated from the engineering model. The incremental costs are thus entirely determined by the
parameters of the engineering model. Based on these three functions, Walker and Hoehn
estimated that total damages from nitrate contamination range from $40 to $330/household per
year, depending on the treatment location, household water consumption, the price of water, the
damages and benefits per household, household income, the level of nitrate contamination, and
an estimate of annual costs for point-of-use nitrate removal.

Evaluation: This study deals with public water supply cost savings as a measure  of benefits from
reducing or avoiding nitrate contamination. Although it is not directly transferable to private
wells, WTP values to prevent nitrates in public water systems may indicate use values for
prevention of nitrates in private wells for comparable uses of drinking water. Based on
incremental value estimates from the damage model, an average household with a $15,000
income in a community of 500 households would be willing to pay $65/yr (1983$) for prevention
of nitrate contamination. Since the Walker and Hoehn model incorporates economies of scale to
estimate per household damages, the value per household in a 500 household community is lower
than that of a one household community (e.g., a private well). A $65/year estimate from Walker
and Hoehn thus could represent a lower bound estimate of use values. The estimate represents an
avoided cost measure of welfare  change based on the parameterization of the engineering model.
Because the validity of this model cannot be judged based on the information provided, it is not
possible to determine the validity of this avoided cost measure.

-------
                        GROUNDWATER VALUATION STUDIES > 5-12
5.2.11  Wattage, 1993

Wattage (1993) conducted a contingent valuation survey to elicit WTP for improved water
quality in the predominantly rural Bear Creek watershed in central Iowa. The purpose of the
survey was to determine values for vegetated buffer strips (VBSs) in terms of benefits for
groundwater protection. A single survey instrument was used to reach farmers and the general
public and asked different questions of each group. The survey involved multiple valuation
questions for several different "commodities" involving impacts to surface and groundwater from
agriculture. The valuation scenarios were not fully specified: there was no explanation of a
payment vehicle or of a program for achieving groundwater protection and cleanup. Based on the
discussion in the report, it is apparent that VBSs are the program that will provide improved
groundwater conditions. In the survey instrument, though, discussion and questions about VBSs
come after the valuation questions.

The 346 respondents were farmers, absentee owners, and town residents. Fifty percent of
respondents were on private wells; over 90% of respondents relied on groundwater for drinking
water supplies. Groundwater quality was of major concern to many of the respondents: only 16%
ranked water quality as suitable for human drinking purposes. Using both open-ended WTP
questions and dichotomous choice formats, respondents were asked how much they would be
willing to pay for programs to reduce contamination of groundwater and surface water supplies.
Wattage estimated a mean monthly WTP of $80, using both probit and logit models, finding that
the different models had little impact on the final estimation results. Wattage also used an
integration method to generate a conditional WTP estimate from the logit model of $49 per
month per household. The year of analysis is uncertain. Wattage found that income was
positively correlated with an individual's WTP, and the respondent's perception of current
groundwater quality and the distance from the respondent's land to the potentially polluted creek
were negatively correlated with an individual's WTP.

Evaluation: Given problems in scenario presentation, it seems likely that there is significant
misunderstanding of the scenario or potential scenario rejection. This position is supported by the
fact that only 32% of respondents strongly agreed that VBS could be effective in reducing
contamination from runoff. Given the information in the report and based on the survey
instrument, it is not possible to determine exactly what commodity is being valued or whether
this represents WTP for moving from unsafe to safe drinking water (since it is unclear what
initial conditions are). Since the endpoint is safe water and the baseline may also be safe water,
average value statements would be an underestimate for cleaning up unsafe water.

Using the value estimate from integration under the logit curve and the sample means for
sociodemographic characteristics yields a conditional WTP of $49 per month per household.  This
translates to an annual WTP of $588, which is larger than the cost of point-of-use controls of
$330. While the larger WTP may represent additional consideration of nonuse values such as

-------
                        GROUNDWATER VALUATION STUDIES > 5-13
protection of aquifers, these values most likely represent an upwardly biased estimate of values
for protection of groundwater from nitrate contamination.
5.3    EVALUATING STUDIES FOR BENEFITS TRANSFER

5.3.1   Purpose of Rating Studies Based on Quality and Applicability

The purpose of this work is to identify estimates of the benefits from changes in well nitrate
concentrations that are applicable for this benefit estimation for potential CAFO regulations.
Desvousges et al. (1992) developed five criteria that they used to guide the selection of studies
used in their application of the technique to a surface water quality issue. In essence, their five
criteria are that the studies to be transferred (1) be based on adequate data, sound economic
method, and correct empirical technique (i.e., "pass scientific muster"); (2) evaluate a change in
water quality similar to that expected at the policy site; (3) contain regression results that
describe willingness to pay as a function of socioeconomic characteristics; (4) have a study site
that is similar to the policy site (in terms of site characteristics and populations); and (5) have a
study site with a similar market as the policy site. NOAA condenses the five Desvousges et al.
criteria into three considerations: (1) comparability of the users and of the resources and/or
services being valued and the changes resulting from the discharge of concern, (2) comparability
of the change in  quality or quantity of resources and/or services, and (3) the quality of the studies
being used for transfer [59 FR 1183].

In a general sense, items (2), (4), and (5) of Desvousges et al. and items (1) and (2) of NOAA are
concerned with the applicability of an original study to a policy site. Items (1) and (3) of
Desvousges et al. and item (3) of NOAA are concerned with the quality of the original study. To
assess original studies for use in the benefits transfer for benefits assessments from CAFO
regulations, we assess the applicability and the quality of the original studies on several criteria.

The 11 studies summarized in Appendix C represent a diverse range of valuation exercises. To
the extent feasible, information was obtained or derived from each report or paper for
28 categories of information used to characterize the studies. While this is largely a qualitative
assessment, the purpose of the following discussion is to make this assessment as transparent as
possible. Because applicability to CAFOs and quality of the value estimates are distinct concepts,
we want to rate these characteristics of the studies separately. Overall,  the goal of the rating
process is to identify studies that elicit high-quality values (reliable and valid) and which are
most applicable to the benefits assessment. There are three steps in undertaking the rating
process:

1.      identify study characteristics upon which to judge applicability and quality
2.      assign scores to the studies based on these characteristics

-------
                        GROUNDWATER VALUATION STUDIES > 5-14
3.     assign weights to these scores for aggregating scores into unidimensional measures of
       applicability and quality.

We assigned scores according to the criteria discussed below and identified in Exhibit 5-1. For
this rating schema, the weighting on the various characteristics related to quality or applicability
is simple so that the effect of changing the weighting scheme will be transparent.
5.3.2   Criteria for Ranking Based on Applicability

The first criterion for ranking the groundwater valuation studies is applicability. Applicability
refers to the relationship between values elicited in the groundwater valuation studies and benefit
estimates necessary for application to the analysis of CAFO regulatory options. Values necessary
for benefit analysis of CAFO regulatory options primarily involve potential health risks related to
elevated nitrate levels in drinking water. While CAFOs may introduce other contaminants into
drinking water, nitrate contamination is a primary focus of regulatory options. Criteria for
evaluation of study applicability include characteristics of the original studies such as:

+      location (urban, rural, etc.)
>•      water supply/groundwater use (percent on wells)
+      contaminants (scenario involves nitrate contamination of groundwater)
>•      source of contaminants (scenario involves conditions similar to those relevant for
       CAFOs)
>•      value estimates are for the correct theoretical construct (e.g., total WTP for reducing
       groundwater contamination from nitrates).

       Location

In general, urban residents are not on private groundwater wells and thus have less experience
with potential groundwater contamination. A higher applicability rating was given to studies that
are primarily rural than to those with urban/rural or purely urban samples. Concentrating on rural
populations is also more likely to be similar to the population of individuals on private wells to
which we apply benefit estimates.  Since we do not have national sociodemographic information
specific to the population on private wells, focusing the transfer on studies conducted with more
rural populations helps account for potential  income differences between rural and urban
populations.

       Water Supply

Studies received a higher score if more than 50% of the respondents indicated that they were
currently using groundwater for their primary water supply. Again, the policy population is

-------
GROUNDWATER VALUATION STUDIES *• 5-15
Exhibit 5-1
Scoring Matrix for Groundwater Valuation Studies
Scoring Criteria
Location (urban, rural,
etc.)
HH H2O Supply/GW
Use
Contaminants
Source of
Contaminants
Values Estimated
Published/Peer
Reviewed?
Lype of Study
Survey Implement
Respondents
Response Rate
Groundwater Baseline
Change in
Groundwater Scenario
Credibility of Scenario
Change
Scoring
Rural = 2;
Rural/urban = 1;
Urban/other = 0
> = 50% on wells = 1;
<50% = 0
Nitrates = 2; nitrates +
other = 1; Not nitrates = 0
CAFOs/Agr = 2; Mixed
sources w/ag = 1;
Not specified = 0
WTP =1; Other = 0
Peer rvw. = 2;
Dissert. = 1 ; Other = 0
Primary data = 1;
Other = 0
Mail/in person = 1;
Other = 0
>1000 = 2;500-1000 = 1;
<500 = 0
>70% = 2;40%-70%=1;
<40% = 0
Specified = 1;
Not specified = 0
Defined change = 1;
Undefined or vague = 0
Assessed credibility = 1;
Didn't asses = 0
Applic-
able
/
/
/
/
/








Quality





/
/
/
/
/
/
/
/
Crutchfleld
etal.
1
0
2
0
1
T
1
0
2d
1
0
1
1
Delavan
1
0
2
1
1
0
1
1
1
0
1
1
1
De
Zoysa
1
0
2
2
1
1
1
1
2
0
1
1
1
Edwards
1
0
2
0
1
2
1
1
1
2
1
1
1
Giraldez
and Fox'
2
1
2
2
0
2
0
0
0
0
1
1
0
Hurley
etal.
1
0
2
2
1
2
1
1
Oe
of
0
1
0
Jordan
and
Elnagheeb
1
lb
2
2
1
2
1
1
0
0
1
0
0
Poe and
Bishop
2
1
2
2
1
2
1
1
0
2
1
1
1
Sparco
1
1
1
1
1
1
1
0
0
0
0
1
0
Walker
and
Hoehn
2
0
2
2
0
2
0
0
0
0
1
1
0
Wattage
2
0
0
1
1
1
1
1
0
0
0
0
0

-------
GROUNDWATER VALUATION STUDIES *• 5-16
Exhibit 5-1 (cont.)
Scoring Matrix for Groundwater Valuation Studies
Scoring Criteria
Valuation
Methodology
Payment Vehicle
Duration of Payment
Vehicle
Analysis
Significant
Explanatory Variables
Scoring
Valid = 1;
Questionable = 0
Specified = 1;
Not specified = 0
Continuous = 2;
One time = 1; Other = 0
Advanced = 1; Other = 0
Validity indicated = 1;
Other = 0
Applic-
able





Quality
/
/
/
/
/

Total Applicability
Total Quality
Crutchfield
etal.
1
1
2
1
1
Crutchfield
etal.
4
14
Delavan
1
1
2
1
1
Delavan
5
12
De
Zoysa
1
1
1
1
1
De
Zoysa
6
13
Edwards
1
1
2
1
1
Edwards
4
16
Giraldez"
I8
0
0
0
1
Giraldez
7
6
Hurley
etal.
0
0
2
1
1
Hurley
etal.
6
9
Jordan
and
Elnagheeb
1
1
2
1
1
Jordan
and
Elnagheeb
7
11
Poe and
Bishop
1
1
2
1
1
Poe and
Bishop
8
15
Sparco
1
0
2
1
1
Sparco
5
8
Walker
and
Hoehn
0
0
0
1
0"
Walker
and
Hoehn
6
5
Wattage
0
0
2
1
1
Wattage
4
7
a. Benefits transfer study and thus many categories are not applicable.
b. Using analysis for private wells only.
c. Crutchfield and Cooper, 1997.
d. Based on indication of 819 usable responses and -50% response rate.
e. Only 85 private well users in the analysis.
f. 44.7% returned: 33. 2% usable.
g. Valid for benefits transfer.
li. Significant explanatory variables in Walker and Hoehn are entirely the result of generating data using an engineering model of incremental costs of water production.

-------
                        GROUNDWATER VALUATION STUDIES > 5-17
individuals on private wells, and thus studies of this population are more applicable for benefits
transfer.

       Nitrate Contamination of Groundwater

We considered primarily valuation studies that present a scenario of nitrates as a source of
contamination in groundwater. Nitrate contamination scenarios are more likely to present
individuals with impacts and risks that are similar to those necessary for the valuation of CAFO
control benefits. While some studies indicated other contaminants in addition to nitrates, we
placed higher weight on values identified as specifically associated with nitrates. While other
scenarios will also elicit values for reducing risks of drinking contaminated groundwater, they
may involve health risks different from those from nitrate contamination.

       Relationship of Valuation Scenario to CAFOs

Some of these studies consider sources other than CAFOs or agricultural sources. While values
for reduced health risks from groundwater contamination may be elicited in other studies, it
seems likely that studies specifically considering scenarios similar to CAFOs or agricultural
contamination will be more amenable to benefits transfer. In addition, CAFO-type contamination
sources and their regulation may involve decisions and impacts that are different from other
contamination sources such as air deposition or contamination from septic systems.

       Valuation Scenario

While most of the studies elicit total values for reduced contamination, some are designed to
elicit option values. While these are theoretically valid values, we need to further consider their
applicability to the regulatory options under consideration. In particular we rated studies as to
whether they elicited willingness to pay as the appropriate theoretical construct applicable for
policy analysis. In addition, studies directly eliciting values for reducing nitrate contamination in
individuals' own wells are more directly transferred to the current policy scenario than studies
valuing prevention of future possible contamination (e.g., Edwards, 1988) or the probability of
contamination in a group of wells (e.g., Delavan, 1998).
5.3.3  Criteria for Ranking Based on Study Quality

Analysis of study quality is based on evaluation of the validity and reliability of the value
estimates derived in the groundwater valuation studies. This is primarily a qualitative exercise
examining multiple facets of the studies under consideration. Based on suggested criteria as to
what contributes to a valid and reliable stated preference valuation study, we identified
characteristics of these studies that indicate reliability and validity (Bishop et al.,1997). Criteria
for evaluation of study quality include:

-------
                        GROUNDWATER VALUATION STUDIES > 5-18
       >      published/peer reviewed
       >•      type of study (design/method)
       *•      survey implementation
       >•      respondents: number and well usage
       +      response rate
       >•      groundwater baseline
       +      change in groundwater scenario
       >•      credibility of scenario change
       +      valuation method
       >•      payment vehicle
       +      duration of payment vehicle
       >•      analysis (method of empirical estimation)
       +      significant explanatory variables.

       Peer Reviewed

Peer reviewed publications may provide more reliable and defensible value estimates than
nonreviewed reports. To this end we also considered PhD dissertations to be more reliable than
master's theses because they have generally undergone more rigorous review and meet a higher
standard than master's theses or general  staff publications. While we do not mean to say that
master's work or staff publications cannot be of as high or higher quality than peer reviewed
work, there is more evidence that peer reviewed work has met an accepted professional standard.

       Type of Study

We placed a higher rating on studies that elicit empirical values from actual households as
opposed to being theoretical modeling exercises. Some of the studies are primarily theoretical
exercises that do not elicit primary data from households (e.g., Walker and Hoehn, 1990). As
such, these studies may not provide information on values directly transferable for the benefits
assessment.

       Survey Implementation

Survey implementation is defined here as the method of conducting the survey. In general
telephone surveys are less likely to generate reliable data in CVM surveys because of the
abbreviated nature of telephone surveys. While some researchers favor in-person surveys, mail
surveys have been shown to generate reliable responses (Dillman, 2000). In our evaluation of
study quality, we also noted studies that  did not involve a random sample (e.g., Sparco, 1995) to
minimize potential sample selection bias (see below on response rates).

-------
                        GROUNDWATER VALUATION STUDIES > 5-19
       Respondents

For contingent valuation surveys, it is important that a sufficient sample size has been used to
ensure representativeness of the value estimates. While there is no clear-cut rule for assessing
adequate sample size in CVM studies, statistical methods used in sampling design can indicate
sample sizes necessary to obtain estimates of population parameters. For instance, with a
population size of 1 million, a sample size of 1,066 is needed to estimate a 95% confidence
interval with a ±3% sampling error (Dillman, 2000, see also Kalton, 1983). When evaluating the
number of respondents, we also attempted to identify those respondents on private wells because
many studies elicit values from other water users (e.g., municipal).

       Response Rate

Higher response rates are used as an indication of the representativeness of the value estimates
and as an indication of overall study quality. Because of potential sample selection and
nonresponse biases (Mitchell and Carson,  1989), response rates above 70%  are considered good
for CVM surveys, while those below 40% are rated as poor for evaluating these studies.

       Groundwater Baseline

A full definition of the commodity being valued includes identifying baseline conditions. The
survey instrument must either specify baseline conditions or elicit individuals' perceptions of
baseline conditions (Fischhoff and Furby,  1988). In our evaluation of study quality, we identified
studies where baseline is actually defined or elicited in the survey instrument as opposed to only
mentioned in the study report. Not specifying baseline in the survey leaves the commodity
inadequately defined.

       Change in  Groundwater Scenario

Scenario development is essential in CVM studies to ensure that individuals understand the
valuation exercise and that the values elicited are for the commodity being studied (Fischhoff and
Furby, 1988). Several aspects of the study design fall under the concept of scenario development,
including identifying baseline groundwater conditions, identifying changes in groundwater
conditions as discussed above, specifying the source of contamination,  assessing the credibility
of the scenario, and using a  realistic payment vehicle. This study quality criterion evaluates
whether the change in groundwater quality is specified, because if it is not, we cannot determine
exactly what commodity is being valued.

       Credibility of Scenario Change

To elicit a valid value statement from individuals, the proposed program or commodity change
must be credible to respondents. Credibility depends on how the program is described to

-------
                        GROUNDWATER VALUATION STUDIES > 5-20
individuals and the perceived likelihood of whether or not such a program would ever be
provided or would even be possible to provide. A not credible scenario is likely to induce
scenario rejection and misstatements of actual values. Studies were scored depending on whether
or not they had assessed the credibility of the scenario to respondents (e.g., attempted to identify
scenario reject!on).

       Valuation Method

The method for estimating the value of a commodity has to be appropriate for the value being
estimated. As part of implementing correct valuation methods, the appropriate population needs
to be sampled, the correct type of value (e.g. WTA or WTP) elicited from that population, and
the appropriate method applied for deriving the value. For instance, if values are elicited from
non-groundwater users for cleaning up drinking water that comes from groundwater, these values
likely are to be different than values that groundwater users would have. Additionally,
engineering cost  models cannot be used to derive individuals' WTP values because such models
are based on different theoretical values (i.e., costs, not welfare values).

       Payment Vehicle

Numerous types of payment vehicle can be proposed in a CVM survey. CVM researchers
generally feel that the payment vehicle should be well defined and plausibly related to the
commodity being valued (Morrison et al., 2000).  The payment vehicle should be assessed for
adequacy in pretests or in quantitative analysis (Carson, 1997) as in Edwards (1988). We ranked
studies lower if they do not  specify a payment vehicle.

       Duration of Payment Vehicle

Similar to the requirement that the payment vehicle be commensurate with the commodity, the
duration of the payment should be reasonably related to the duration of the commodity or
program providing the commodity being valued. Since most groundwater nitrate control
programs and benefits are continuous, we rated studies with continuous (e.g., Poe and Bishop,
1992) or multiyear (e.g., Delavan, 1998) payment vehicles, e.g., monthly water bills, higher than
those with one-time payments (e.g., De Zoysa, 1995). Likewise, we rated lower those studies that
do not appear to specify the payment vehicle duration, because this indicates inadequate
commodity definition.

       Methods of Analysis

Statistical analysis includes appropriate econometric methods (e.g., probit or logit models rather
than ordinary least squares for qualitative choice surveys or tobit for truncated at zero, open-
ended WTP questions) and adequate reporting on the results of statistical analysis. In general, all
of the studies present reasonably high quality analysis where applicable.

-------
                         GROUNDWATER VALUATION STUDIES > 5-21
       Significant Explanatory Variables

Economic theory suggests that willingness to pay is related to certain sociodemographic
characteristics; for example, it is generally positively related to income. Other relationships are
expected, although not based on microeconomic theory. For instance, rural residents are expected
to be willing to pay more for clean groundwater from private wells than urban dwellers who rely
on public water supplies. Ceterisparibus, individuals who use private wells are expected to be
willing to pay more than those on public supplies, even in rural areas. Perceptions of water
quality also can be expected to be related to WTP for reducing nitrates in drinking water. For
several studies the likelihood that an individual would live in an area in the future was positively
correlated with WTP for safe drinking water.
5.3.4  Scoring Matrix

Most of the screening information items presented in Appendix C were used for these
assessments. Characteristics summarized in Appendix C but not used for the assessment were
year of analysis, place, who was asked, actual groundwater baseline condition, number of survey
versions, and the values actually estimated.4 Based on these characteristics and scoring criteria,
Exhibit 5-1 presents the scoring matrix for the 11  nitrate valuation studies evaluated. The
"scoring" column indicates the scoring method for evaluating the various studies using the
criteria discussed above for applicability and quality. Several of the criteria apply only to primary
data collection (e.g., contingent valuation surveys) such as survey implementation, respondents,
response rate,  credibility of scenario change, valuation methodology, and payment vehicle.
Studies that are not based on primary data collection thus score low on these criteria and are not
likely to be included in the benefits transfer assessment. Checkmarks in the applicability and
quality columns indicate which scores were summed to aggregate the study characteristics to the
unidimensional applicability and quality scores at the bottom of the exhibit.

The scoring was undertaken without weighting the various characteristics for importance in
determining applicability or quality  of study. A weighting scheme was derived to provide  more
reliable assessment.5
4. "Place" does play a role in that the Edwards study is not weighted highly in the benefits transfer in part
because of the unique location of the study. It involved a sole source aquifer in a unique location (Cape Cod)
where mean income of respondents is most likely higher than would be expected at typical rural sites where
CAFO impacts are expected.

5. The weighting scheme was based on collaborative professional judgment with EPA and consultant
economists.

-------
                       GROUNDWATER VALUATION STUDIES > 5-22
5.4    RANKING OF NITRATE VALUATION STUDIES

Using the scoring from Exhibit 5-1, we sorted the studies into high, medium, and low categories
based on their applicability and reliability for use in CAFO analysis. Our results are shown in
Exhibit 5-2. It must be emphasized that these scorings and rankings are not intended as
judgments of the studies except for purposes of their use in benefits assessments for CAFO
regulatory options. Many aspects of these studies that explore important theoretical or
methodological issues are not as applicable for the benefits assessment and thus may receive low
weights. Possible applicability scores range from 0 to 8. Studies scoring from 0 to 4 were rated as
low, 5 and 6 as medium, and 7 and 8 as high. Possible quality scores range from 0 to 17. Studies
scoring from 0 to 9 were rated as low,  10 to 13 as medium, and 14 and above as high. Exhibit 5-2
summarizes the scoring and rating according to this criterion.
Exhibit 5-2
Ranking of Studies Based on Scoring Exercise
Study
Crutchfield et al.
De Zoysa
Delavan
Edwards
Giraldez and Fox
Hurley et al.
Jordan and Elnagheeb
Poe and Bishop
Sparco
Walker and Hoehn
Wattage
Total
Applicability
4
6
5
4
7
6
7
8
5
6
4
Total Quality
14
12
12
16
6
9
11
15
8
5
7
Total
Applicability
low
medium
medium
low
high
medium
high
high
medium
medium
low
Total Quality
high
medium
medium
high
low
low
medium
high
low
low
low
Based on the scoring and qualitative rankings, Exhibit 5-3 indicates where these studies fall
across the two dimensions of applicability to CAFOs and quality of studies.

-------
                     GROUNDWATER VALUATION STUDIES > 5-23
Exhibit 5-3
Groundwater Valuation Applicability and Quality Matrix

Quality
of Study
High
Medium
Low
Applicability of Study to CAFOs
High
Poe and Bishop, 1992
Jordan and Elnagheeb, 1993
Giraldez and Fox, 1995
Medium

De Zoysa, 1995
Delavan, 1998
Hurley et al., 1999
Sparco, 1995
Walker and Hoehn, 1990
Low
Crutchfield et al., 1997
Edwards, 1988

Wattage, 1993
5.5   VALUES FOR BENEFITS TRANSFER TO CAFOs

We applied the CPI to convert the annual mean household willingness-to-pay values obtained
from these studies to 2001 dollars. Exhibit 5-4 shows the CPI values used for these conversions.
Exhibit 5-4
Consumer Price Index — All Urban Consumers — (CP - U)
U.S. City Average — All Items (1982-1984 = 100)
Year
1983
1984
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
Annual
99.6
103.9
109.6
113.6
118.6
124.0
130.7
136.2
140.3
144.5
148.2
152.4
156.9
160.5
163.0
166.6
172.2
177.1
Source: U.S. Bureau of Labor Statistics, 2000.

-------
                        GROUNDWATER VALUATION STUDIES > 5-24
Exhibit 5-5 shows summary mean per household annual WTP in 2001 dollars for several of the
studies discussed above. Not all values are shown for all reports.
Exhibit 5-5
Mean Annual WTP per Household
Study Reference
Crutchfield et al., 1997
De Zoysa, 1995
Delavan, 1997
Edwards, 1988
Jordan and Elnagheeb, 1993
Poe and Bishop, 1992
Sparco, 1995
Year of
Analysis
1994
1994
1996
1987
1991
1991
1993
Mean Annual Household WTP in 2001 dollars
$758.40 to reduce nitrates to safe level
$22.90 to reduce from 10 mg/L to 0 mg/L
($2.29 per mg/L)
$63.07 (lower bound mean)
$1.89 per mg/L (using 3% discount rate)
$212.92 IDE w/o protest bidders (see Section 5.2.3)
$2,530.22 to increase probability of supply from 0.0 to
1.0
$204.94 (private wells)
$535 (25% reduction in nitrates to safe level)
$629 (households with 100% probability of future
contamination)
$15 1.44 per mg/L
Based on this summary, WTP values for reducing nitrate contamination to safe levels fall into a
range between $60 and $2,500 a year. The exact interpretation of the commodity varies for these
studies, as discussed above in the study evaluations. For reasons outlined there, we feel Edwards'
$2,500/year represents a high estimate not directly applicable to the conditions of CAFO counties
nationwide. Also as discussed above, the Delavan and De Zoysa values represent either lower
bound estimates or value estimates that are not reliably translated into those necessary for CAFO
benefit transfer assessment. Jordan and Elnagheeb's small sample,  unclear scenario, and potential
scenario rejection make their value estimate less reliable than Poe and Bishop, but may provide a
lower bound value for nitrate reductions. Poe and Bishop's work represents the most rigorous
analysis and provides the only value estimates based on respondents knowing their actual well
nitrate levels.

For estimates of the per mg/L values for nitrate reductions, Sparco's value estimates appear to be
implausibly high,  especially relative to the values for potentially larger total mg/L reductions
from unsafe to safe levels. The Crutchfield et al. estimate for WTP per mg/L under the MCL
provides a lower bound estimate that we can conservatively use in the benefits transfer.

-------
                        GROUNDWATER VALUATION STUDIES > 5-25
The Crutchfield et al. value estimate for reducing nitrates to safe levels are derived from a more
diverse sample than Poe and Bishop. The Crutchfield et al. WTP estimate is
$758.40/household/yr (2001$). As indicated in Exhibit 5-2, though, we ranked the Crutchfield
et al. study as being of low applicability for benefits transfer to CAFOs primarily because they
did not specify the source of the nitrate contamination in their scenario and less than 50% of their
respondents were on private wells. We thus consider the Crutchfield et al. values as a possible
upper bound for application for this benefits transfer. We thus rely primarily on the average of
Poe and Bishop's two WTP estimates as reliable estimates of WTP for reducing nitrates to safe
levels (from above the MCL to below the MCL). The average of these two estimates is $583.00
per household per year.

We use the average of De Zoysa and Crutchfield et al. for changes in incremental nitrate
concentrations below the MCL. The values from Poe and Bishop are expressed as willingness to
pay per year as long as the individual lives in the county, and thus can  be directly translated to the
policy scenarios.

In De Zoysa's study, the reduction in groundwater nitrate levels is from a range of 0.5 to
3.0 mg/L to a range of 0.5 to 1.0 mg/L. Taking range means, the reduction in nitrates is thus from
1.75 mg/L to 0.75 mg/L, or a reduction of 1.0 mg/L. Using the annual  lower bound mean values,
this represents a WTP of $63.07 per mg/L (in 2001$)  change in nitrate concentrations for
incremental changes below the 10 mg/L MCL. Using  a 3% discount rate, this translates into an
annual WTP of $1.89.

Crutchfield et al. report monthly willingness-to-pay values for reducing nitrates, and thus we
adjust their values to an annual WTP per mg/L. They report values for reducing nitrates from
above the MCL to the MCL and from above the MCL to zero. The difference between these two
values is taken as the value of reducing nitrate concentrations from the MCL of 10 mg/L to
0 mg/L. Using the monthly willingness-to-pay values  reported in Crutchfield et al., we calculated
a per-year per-mg/L value for incremental changes in  nitrate concentrations below 10 mg/L. This
adjustment assumes a "linear" value per mg/L between 10 mg/L and 0 mg/L, indicating no
threshold effects. The resulting value, $2.29 per mg/L per household per year (in 2001$), is
applied to changes in well nitrate concentrations between  10 mg/L and 1 mg/L, assuming that
there is a natural, or ambient, background level of 1 mg/L of nitrates in groundwater.

For purposes of benefits transfer we use an average of the values from the De Zoysa and
Crutchfield et al. of $2.09 per household per year per mg/L (in 2001$). Exhibit 5-6 shows the
point value estimates used for benefits transfer.

-------
GROUNDWATER VALUATION STUDIES > 5-26
Exhibit 5-6
Willingness-to-Pay Values Applied to Benefits Transfer
Study
Poe and
Average
Bishop
of Crutchfield et al. and De Zoysa
Value
Annual WTP
Annual WTP per mg/L
between 10 mg/L and 1 mg/L
2001$
$583.00
$2.09

-------
                                 CHAPTER 6
                         BENEFIT CALCULATIONS
6.1   TOTAL ANNUAL VALUES

Exhibit 6-1 shows the undiscounted annual benefit estimates when all the effects of reduced
nitrogen loadings have been achieved at the well. The second column shows the benefits derived
from reductions in the number of households above the MCL, and the third column shows
benefits from incremental reductions between 1 mg/L and 10 mg/L for households that were
below the MCL before regulatory changes. The last column shows total annual national
undiscounted benefits.
Exhibit 6-1
Undiscounted Annual Values under CAFO Regulatory Scenarios (2001$)
Scenario
Option 1 — Scenario 6
Option 1 — Scenario 7
Option 1 — Scenario 8
Option 1 — Scenario 9
Option 2/3 — Scenario 6
Option 2/3 — Scenario 7
Option 2/3 — Scenario 8
Option 2/3 — Scenario 9
Option 5 — Scenario 6
Option 5 — Scenario 7
Option 5 — Scenario 8
Option 5 — Scenario 9
Total WTP for
Discrete Reduction to
MCL
$86,695,000
$70,440,000
$70,440,000
$70,440,000
$86,695,000
$70,440,000
$70,440,000
$65,021,000
$83,986,000
$62,312,000
$62,312,000
$62,312,000
Total WTP for
Incremental
Changes below
10 mg/L
$1,786,000
$1,496,000
$1,496,000
$1,454,000
$1,939,000
$1,648,000
$1,648,000
$1,606,000
$1,749,000
$1,501,000
$1,501,000
$1,467,000
Total
$88,481,000
$71,936,000
$71,936,000
$71,894,000
$88,634,000
$72,087,000
$72,087,000
$66,627,000
$85,735,000
$63,813,000
$63,813,000
$63,779,000

-------
                             BENEFIT CALCULATIONS > 6-2
6.2    DISCOUNTING AND AGGREGATING TO PRESENT VALUES

Exhibit 6-2 shows the timepath of undiscounted benefits under the primary assumptions used in
the benefits assessment. As discussed in Section 3.6, we assume that impacts from nitrogen
reductions will be translated into reduced well nitrate concentrations in a linear manner over
27 years. Benefits thus increase from the year of implementation until the 27th year when all the
effects of reduced nitrogen loadings have been achieved at the well. From the 27th year onward
the benefits are equal to the total benefits when all of the effects of reduced nitrogen loadings
have been achieved at the well, as shown in Exhibit 6-2. The top line in Exhibit 6-2 shows the
timepath of benefits for the Option 2/3 — Scenario 6, the lower line shows the timepath of
benefits for Option 5 — Scenario 9, which produces the lowest benefits, and Option 2/3 —
Scenario 8 falls within these bounds.

                                    Exhibit 6-2
                      Timepath of Undiscounted Benefit Flows
             o
             o
             O
                               20         40         60
                                     Years from Implementation
        80
100
                            •Option5 - Scenario 9
                            Option 2 - Scenario 8
Option 2 - Scenario 6
In calculating present values we use an infinite time horizon. Exhibits 6-2 and 6-3 show the
timepath for undiscounted or discounted present to 100 years for illustrative purposes only.
Benefits received in the distant future (e.g., 100 years plus) are only a small percentage of total
benefits even at the lowest discount rate used in this analysis (3%).

-------
                              BENEFIT CALCULATIONS > 6-3
                                      Exhibit 6-3
       Discounted Value of Annual Benefits Using 3%, 5%, and 7% Discount Rates
                                Option 2/3 — Scenario 8
                                20         40        60        80

                                     Years after Implementation
                          100
                              -X- 3 Percent
•5 Percent
- 7 Percent
6.3    DISCOUNTED BENEFITS

Exhibit 6-3 shows the timepath of discounted benefits for Option 2/3 — Scenario 8 using a 3%,
5%, and 7% rate of discount. As can be seen, the present value of benefits increases over time as
the number of wells achieving the steady state following regulation increases and then decreases
from the maximum toward zero benefits because of the discounting of the future benefits.

The total present value of any given scenario/option will be the area under the curve using the
given rate  of discount. Exhibit 6-4 shows the total discounted present value for all scenarios
using three different rates of discount: 3%, 5%, and 7%. Note that these numbers are presented in
millions of 2001$, so the discounted present value for Option 2/3 — Scenario 8 using a 3% rate
of discount is roughly $1,648  million. Using a 7% rate of discount, this falls to $478 million.

-------
                             BENEFIT CALCULATIONS > 6-4
Exhibit 6-4
Total Present Value of Option/Scenarios Using Different Rates of Discount
(millions 2001$)
Scenario
Option 1 — Scenario 6
Option 1 — Scenario 7
Option 1 — Scenario 8
Option 1 — Scenario 9
Option 2/3 — Scenario 6
Option 2/3 — Scenario 7
Option 2/3 — Scenario 8
Option 2/3 — Scenario 9
Option 5 — Scenario 6
Option 5 — Scenario 7
Option 5 — Scenario 8
Option 5 — Scenario 9
3%
Present Value
$2,022.19
$1,644.07
$1,644.07
$1,643.10
$2,025.69
$1,647.52
$1,647.53
$1,522.73
$1,959.44
$1,458.41
$1,458.41
$1,457.64
5%
Present Value
$984.01
$800.02
$800.02
$799.54
$985.72
$801.70
$801.70
$740.97
$953.47
$709.67
$709.67
$709.30
7%
Present Value
$586.20
$476.59
$476.59
$476.31
$587.22
$477.59
$477.59
$441.42
$568.01
$422.77
$422.77
$422.55
6.4    ANNUALIZED DISCOUNTED BENEFIT ESTIMATES

In addition to calculating the present value of estimated benefits, EPA developed an estimate of
the annualized benefits attributable to the regulatory scenarios analyzed; these annualized values
reflect the constant flow of benefits over time that would generate the associated present value.

The constant annual benefit^ that, over a period of n years, equals the estimated present value
(PV) of benefits is determined by:

                             A=PV(r)l{\-[\/(\+rr\},

where r represents the annual discount rate. As n approaches infinity, this equation simplifies to:

                                     A=PV*r .

EPA uses this equation to calculate the annualized benefits reported in this analysis. Exhibit 6-5
presents the annualized benefit estimates for the total present value benefits shown in
Exhibit 6-4. For instance, for Option 2/3 — Scenario 8, a constant benefit flow of $49.43 million

-------
                            BENEFIT CALCULATIONS > 6-5
Exhibit 6-5
Annualized Present Value of Option/Scenarios Using Different Rates of Discount
(millions 2001$)
Scenario
Option 1 — Scenario 6
Option 1 — Scenario 7
Option 1 — Scenario 8
Option 1 — Scenario 9
Option 2/3 — Scenario 6
Option 2/3 — Scenario 7
Option 2/3 — Scenario 8
Option 2/3 — Scenario 9
Option 5 — Scenario 6
Option 5 — Scenario 7
Option 5 — Scenario 8
Option 5 — Scenario 9
3%
Annualized Value
$60.67
$49.32
$49.32
$49.29
$60.77
$49.43
$49.43
$45.68
$58.78
$43.75
$43.75
$43.73
5%
Annualized Value
$49.20
$40.00
$40.00
$39.98
$49.29
$40.08
$40.08
$37.05
$47.67
$35.48
$35.48
$35.46
7%
Annualized Value
$41.03
$33.36
$33.36
$33.34
$41.11
$33.43
$33.43
$30.90
$39.76
$29.59
$29.59
$29.58
discounted at 3% would generate $1,648 million in total present value of benefits, also
discounted at 3%.
6.5   ALTERNATIVE SPECIFICATION OF TIMEPATH: DISCONTINUATION OF
      NEW REGULATIONS IN 27TH YEAR

A potential alternative timepath specification involves the analysis of a regulatory regime where
the proposed regulatory scenario would be in place for 27 years (until all reductions in nitrates
had been realized at the well) and then the regulations would revert to current (2002) regulations.
Under this scenario there would be an increase in benefits from the year of implementation until
the 27th year, and then a decrease in benefits until the 54th year when conditions are assumed to
have returned to current (2002). Exhibit 6-6 shows the maximum undiscounted annual value, the
present value, and the annualized value for these scenarios using a 3% rate of discount.

Under this alternative specification of the timepath for regulations, Exhibit 6-7 shows the
annualized benefits for the various options/scenarios using the three discount rates (3%, 5%,
and

-------
                             BENEFIT CALCULATIONS > 6-6
Exhibit 6-6
Benefits under Alternative Scenario of Regulatory Discontinuation
in 27 Year (3% rate of discount)
(millions 2001$)
Scenario
Option 1 — Scenario 6
Option 1 — Scenario 7
Option 1 — Scenario 8
Option 1 — Scenario 9
Option 2/3 — Scenario 6
Option 2/3 — Scenario 7
Option 2/3 — Scenario 8
Option 2/3 — Scenario 9
Option 5 — Scenario 6
Option 5 — Scenario 7
Option 5 — Scenario 8
Option 5 — Scenario 9
Maximum Undiscounted
Annual Value
$88.48
$71.94
$71.94
$71.89
$88.63
$72.09
$72.09
$66.63
$85.74
$63.81
$63.81
$63.78
Present Value
$1,133.72
$921.73
$921.73
$921.19
$1,135.69
$923.67
$923.67
$853.70
$1,098.54
$817.65
$817.65
$817.21
Annualized Value
$41.11
$33.42
$33.42
$33.40
$41.18
$33.49
$33.49
$30.96
$39.83
$29.65
$29.65
$29.63
6.6    SENSITIVITY ANALYSIS

6.6.1   Ranges of Value Estimates

As shown in Exhibit 5-5, Delavan (1997) reported a willingness to pay of $212.92 (see
Section 5.2.3) and Jordan and Elnagheeb (1991) reported a willingness to pay of $204.94 per
household per year (2001$). Using an approximation of $209 per household per year, Exhibit 6-£
shows how the annualized benefit estimates would change using this lower value for benefits to
households achieving the MCL. Alternatively, Exhibit 6-8 also uses Edwards (1988) reported
WTP of $2,530.22 (2001$) as an upper bound value for household benefits for achieving
the MCL.

-------
                             BENEFIT CALCULATIONS > 6-7
Exhibit 6-7
Annualized Benefits under Alternative Scenario of Regulatory Discontinuation
in 27 Year (3%, 5%, and 7% rate of discount)
(millions 2001$)
Scenario
Option 1 — Scenario 6
Option 1 — Scenario 7
Option 1 — Scenario 8
Option 1 — Scenario 9
Option 2/3 — Scenario 6
Option 2/3 — Scenario 7
Option 2/3 — Scenario 8
Option 2/3 — Scenario 9
Option 5 — Scenario 6
Option 5 — Scenario 7
Option 5 — Scenario 8
Option 5 — Scenario 9
3%
$41.11
$33.42
$33.42
$33.40
$41.18
$33.49
$33.49
$30.96
$39.83
$29.65
$29.65
$29.63
5%
$37.71
$30.66
$30.66
$30.64
$37.78
$30.72
$30.72
$28.40
$36.54
$27.20
$27.20
$27.18
7%
$33.78
$27.46
$27.46
$27.45
$33.84
$27.52
$27.52
$25.43
$32.73
$24.36
$24.36
$24.35
6.6.2   Discount Rate

As shown in Exhibit 6-9, compared to the basic parameters used in the analysis, increasing the
discount rate from 3% to 5% and 7% leads to a 18.9% and 32.4% reduction in estimated
annualized benefits, respectively.
6.6.3   Time Line until Steady State is Achieved

As shown in Exhibit 6-10, comparing 27 years to 20 years until steady state is achieved increases
the present annualized value by 9.8%. Spreading out time until steady state is achieved to
50 years decreases the present annualized value by 15.8%.

-------
                             BENEFIT CALCULATIONS > 6-8
Exhibit 6-8
Change in Value for Crossing 10 mg/L
Discount Rate
Years to Steady State
Value for Crossing
10 mg/L
Value for Changes below
10 mg/L
Scenario
Option 1 — Scenario 6
Option 1 — Scenario 7
Option 1 — Scenario 8
Option 1 — Scenario 9
Option 2/3 — Scenario 6
Option 2/3 — Scenario 7
Option 2/3 — Scenario 8
Option 2/3 — Scenario 9
Option 5 — Scenario 6
Option 5 — Scenario 7
Option 5 — Scenario 8
Option 5 — Scenario 9
3%
27
$583.00
$2.09
Annualized
Value
(2001$)
$60.67
$49.32
$49.32
$49.29
$60.77
$49.43
$49.43
$45.68
$58.78
$43.75
$43.75
$43.73
3%
27
$209.00
$2.09
Annualized
Value
(2001$)
$22.53
$18.34
$18.34
$18.31
$22.64
$18.44
$18.44
$17.08
$21.84
$16.34
$16.34
$16.32
Percent Change
in Annualized
Value
-62.9%
-62.8%
-62.8%
-62.9%
-62.7%
-62.7%
-62.7%
-62.6%
-62.8%
-62.6%
-62.6%
-62.7%
3%
27
$2,530.22
$2.09
Annualized
Value
(2001$)
$259.20
$210.63
$210.63
$210.60
$259.31
$210.73
$210.73
$194.58
$251.11
$186.45
$186.45
$186.43
Percent Change
in Annualized
Value
327.3%
327.1%
327.1%
327.2%
326.7%
326.4%
326.4%
326.0%
327.2%
326.1%
326.1%
326.3%
6.6.4   Benefits for Changes under the 10 mg/L MCL

Counting only the value for reductions from above the MCL to below the MCL does not have a
significant impact on the total annualized benefit estimate. As shown in Exhibit 6-11, reductions
of nitrate concentrations below the 10 mg/L MCL and above the 1 mg/L "background" level add
less than 5% to the estimated benefits.

-------
                             BENEFIT CALCULATIONS > 6-9
Exhibit 6-9
Sensitivity to Changes in Discount Rate
Discount Rate
Years to Steady State
Value for Crossing
lOmg/L
Value for Changes below
lOmg/L
Scenario
Option 1 — Scenario 6
Option 1 — Scenario 7
Option 1 — Scenario 8
Option 1 — Scenario 9
Option 2/3 — Scenario 6
Option 2/3 - Scenario 7
Option 2/3 — Scenario 8
Option 2/3 — Scenario 9
Option 5 — Scenario 6
Option 5 — Scenario 7
Option 5 — Scenario 8
Option 5 — Scenario 9
3%
27
$583.00
$2.09
Annualized
Value
(millions
2001$)
$60.67
$49.32
$49.32
$49.29
$60.77
$49.43
$49.43
$45.68
$58.78
$43.75
$43.75
$43.73
5%
27
$583.00
$2.09
Annualized
Value (millions
2001$)
$49.20
$40.00
$40.00
$39.98
$49.29
$40.08
$40.08
$37.05
$47.67
$35.48
$35.48
$35.46
Percent Change
in Annualized
Value
-18.9%
-18.9%
-18.9%
-18.9%
-18.9%
-18.9%
-18.9%
-18.9%
-18.9%
-18.9%
-18.9%
-18.9%
7%
27
$583.00
$2.09
Annualized
Value
(millions
2001$)
$41.03
$33.36
$33.36
$33.34
$41.11
$33.43
$33.43
$30.90
$39.76
$29.59
$29.59
$29.58
Percent
Change in
Annualized
Value
-32.4%
-32.4%
-32.4%
-32.4%
-32.4%
-32.4%
-32.4%
-32.4%
-32.4%
-32.4%
-32.4%
-32.4%
The per mg/L value used for changes below the MCL came from the Crutchfield et al. and
De Zoysa reports. As discussed in Chapter 5, Poe (1993) calculates an imputed WTP for a
1 mg/L reduction (or increase) in nitrates as a function of initial nitrate levels. A maximum per
mg WTP of ~$147 (2001$) is seen when initial nitrate levels are close to 10 mg/L. Below
10 mg/L the per mg WTP falls to about $100 (2001$) per mg when the initial level is 4 mg/L.
Sparco (1995) also estimated WTP for incremental changes in nitrate concentrations of
$142.46 per mg/L (2001$). Using a conservative lower bound for these estimates of $100 per
mg/L WTP value, the right-hand side of Exhibit 6-11 shows how much benefit estimate would
increase using these value  per mg/L estimates for incremental changes below the MCL.

-------
                             BENEFIT CALCULATIONS > 6-10
Exhibit 6-10
Sensitivity to Changes in Time until Steady State (20 and 50 years)
(all in 2001$)
Discount Rate
Years to Steady State
Value for Crossing
lOmg/L
Value for Changes below
lOmg/L
Scenario
Option 1 — Scenario 6
Option 1 — Scenario 7
Option 1 — Scenario 8
Option 1 — Scenario 9
Option 2/3 — Scenario 6
Option 2/3 — Scenario 7
Option 2/3 — Scenario 8
Option 2/3 — Scenario 9
Option 5 — Scenario 6
Option 5 — Scenario 7
Option 5 — Scenario 8
Option 5 — Scenario 9
3%
27
$583.00
$2.09
Annualized
Value
(2001$)
$60.67
$49.32
$49.32
$49.29
$60.77
$49.43
$49.43
$45.68
$58.78
$43.75
$43.75
$43.73
3%
20
$583.00
$2.09
Annualized
Value
(2001$)
$66.60
$54.15
$54.15
$54.11
$66.71
$54.26
$54.26
$50.15
$64.53
$48.03
$48.03
$48.01
Percent Change
in Annualized
Value
9.8%
9.8%
9.8%
9.8%
9.8%
9.8%
9.8%
9.8%
9.8%
9.8%
9.8%
9.8%
3%
50
$583.00
$2.09
Annualized
Value (2001$)
$51.08
$41.53
$41.53
$41.51
$51.17
$41.62
$41.62
$38.47
$49.50
$36.84
$36.84
$36.82
Percent Change
in Annualized
Value
-15.8%
-15.8%
-15.8%
-15.8%
-15.8%
-15.8%
-15.8%
-15.8%
-15.8%
-15.8%
-15.8%
-15.8%
6.7    OMISSIONS, BIASES, AND UNCERTAINTIES

Omissions, biases, and uncertainties are inherent in any analysis relying on several different data
sources, particularly those that were not created specifically for that analysis. Exhibit 6-12
summarizes the omissions, biases, and uncertainties for this analysis. The column labeled "likely
impact on benefit" indicates how the benefit estimate is influenced by the omission, bias, or
uncertainty indicated for that row. For instance, in the row on "well location selection," the
benefit estimates discussed above may be positively biased (higher than true value) if the
sampled wells in the Retrospective Database are mainly in areas with nitrate problems.

-------
                              BENEFIT CALCULATIONS > 6-11
Exhibit 6-11
Sensitivity to Benefits from Changes below the MCL
Discount Rate
Years to Steady State
Value for Crossing 10 mg/L
Value for Changes below
10 mg/L
Scenario
Option 1 — Scenario 6
Option 1 — Scenario 7
Option 1 — Scenario 8
Option 1 — Scenario 9
Option 2/3 — Scenario 6
Option 2/3 — Scenario 7
Option 2/3 — Scenario 8
Option 2/3 — Scenario 9
Option 5 — Scenario 6
Option 5 — Scenario 7
Option 5 — Scenario 8
Option 5 — Scenario 9
3%
27
$583.00
$2.09
Annualized
Value
(2001$)
$60.67
$49.32
$49.32
$49.29
$60.77
$49.43
$49.43
$45.68
$58.78
$43.75
$43.75
$43.73
3%
27
$583.00
$0.00
Annualized
Value
(2001$)
$59.44
$48.30
$48.30
$48.30
$59.44
$48.30
$48.30
$44.58
$57.58
$42.72
$42.72
$42.72
Percent
Change in
Annualized
Value
-2.0%
-2.1%
-2.1%
-2.0%
-2.2%
-2.3%
-2.3%
-2.4%
-2.0%
-2.4%
-2.4%
-2.3%
3%
27
$583.00
$100.00
Annualized
Value
(2001$)
$118.02
$97.39
$97.39
$95.99
$123.05
$102.34
$102.35
$97.25
$114.96
$91.95
$91.95
$90.85
Percent
Change in
Annualized
Value
94.5%
97.5%
97.5%
94.7%
102.5%
107.1%
107.1%
112.9%
95.6%
110.2%
110.2%
107.7%
Alternatively, the benefit estimates discussed above may understate true values if, as indicated in
the row on "per household value for reducing well nitrates to the MCL," the benefit estimates
from Poe and Bishop are lower bound estimate of true values.

Data availability limited the variables included in this statistical analysis for the gamma model.
Several variables, such as well construction and well age, proximity of wells to a pollutant
source, and aquifer volume, composition and flow direction, were not included in this analysis
even though they were significant factors in other studies.

-------
BENEFIT CALCULATIONS > 6-12
Exhibit 6-12
Omissions, Biases, and Uncertainties in the Nitrate Loadings Analysis
Variable
Likely Impact
on Benefit"
Comment
Well, land, and nitrate data
Geographic coverage
Well location selection
Year of sample
Nitrate loadings from
AFOs with 0-300 AU
Loadings estimates across
counties in the NPLA
loadings dataset
Percent of wells above 10
mg/L
Sampling methods
Unknown
Positive
Unknown
Positive
Positive
Unknown
Unknown
Date availability limited the well samples used in the statistical
modeling to those from approximately 374 counties nationwide.
Wells sampled in the Retrospective Database may not be random.
Samples may come from areas with problems with nitrate.
Samples taken over 23 years. Land use and other factors
influencing nitrate concentrations in the vicinity of the well may
have changed over time.
Data for the smallest AFOs were not included in this analysis
because they will not be affected by the proposed regulations.
This may subsequently underestimate total loadings, resulting in
an overestimate of the impact of nitrogen loadings on well nitrate
concentrations.
Average loadings estimates for counties included in the
Retrospective Database are greater than in non-USGS counties.
Estimated nitrate reductions in non-USGS counties may thus be
overstated.
Based on the Retrospective Database, EPA assumes that 9.45% of
wells currently exceed the MCL. If the true national percent is
lower (higher) our analysis overstates (understates) benefits.
Data set compiled from data collected by independent state
programs, whose individual methods for measuring nitrate may
differ.
Model variables
Well construction and age
Spatial data
Unknown
Unknown
No reliable data available nationally.
No national data available on the distance from well to pollutant
source.

-------
                              BENEFIT CALCULATIONS > 6-13
Exhibit 6-12 (cont.)
Omissions, Biases, and Uncertainties in the Nitrate Loadings Analysis
Variable
Likely Impact
on Benefit"
Comment
Benefit calculations
Per household value for
reducing well nitrates to
the MCL
Years until wells achieve
steady state.
Exclusion of values for
changes for wells still
above the MCL after new
regulations
Exclusion of values for
incremental changes for
wells above the MCL
before new regulations but
below the MCL after new
regulations
Negative
Negative
Negative
Negative
The Poe and Bishop values generally appear to be a lower bound
estimate of households' WTP for reducing nitrates to the MCL.
The analysis assumes a linear path over 27 years until reduced
nitrogen loadings would result in most wells achieving reduced
nitrate concentrations. A large portion of wells (especially
shallower wells) may achieve this on a much faster time path.
Changes in nitrate concentrations for wells that are still above the
MCL after new regulations are not valued because EPA does not
have reliable value estimates for changes incremental changes
above the MCL.
Changes in nitrate concentrations for wells that were above the
MCL before new regulations, but below after new regulations, are
not calculated since such values may be captured in benefit
estimates used to value changes from above the MCL to below
the MCL nitrate concentrations.
a. "Positive" impact implies that estimated benefits may be overstated; "negative" means that estimated
benefits may be understated if the bias, omission, or uncertainty is not corrected for in the benefit estimate
calculation.
This analysis assumes constant nitrate concentrations and loadings over time (including the past
when data in the Retrospective Database were collected), omitting the potentially significant time
lag associated with nitrate transport through soil and into the aquifer. This may be a significant
source of error, considering that the loadings data are based on current conditions, the nitrate
concentrations were sampled over a 20 year period, and nitrates may take decades to reach the
groundwater.

With respect to the issue of loadings estimates across counties in the NPLA loadings dataset,
there may be a potential bias due to selection of wells sampled for nitrate testing. Counties that
had wells included in the gamma model dataset have different characteristics than counties not
included. This may be because wells that are more likely to have higher nitrates because of
conditions in their surrounding area are more likely to be tested. We attempted to explore this
issue with the sample selection model discussed in Appendix B. Overall, our results suggest little
impact due to potential sample selection.

-------
                                   REFERENCES
Agriculture Canada. 1993. Ontario Farm Groundwater Quality Survey. Agriculture Canada,
Ottawa.

Andres, A.S. 1991. Results of the Coastal Sussex County, Delaware Groundwater-Water Quality
Survey. Report of Investigations no. 49. Delaware Geological Survey, Newark.

Bishop, R., P.A. Champ, T.C. Brown, and D.W. McCollum. 1997. "Measuring the Non-Use
Values: Theory and Empirical Application." In Determining the Value of Non-Marketed Goods:
Economic, Psychological, and Policy Relevant Aspects of Contingent Valuation Methods.
RJ. Kopp, W.W. Pommerehne, and N. Schwarz (eds.), Kluwer Academic Publishers, Dordrecht,
The Netherlands, pp. 59-81.

Boyle, K., G. Poe, and J. Bergstrom.  1994. "What do we know about groundwater values?
Preliminary Implications from a Meta-Analysis of Contingent-Valuation Studies." American
Journal of Agricultural Economics December, 1055-1064.

Burrow, K.R.  1998. Occurrence of Nitrate and Pesticides in Groundwater beneath Three
Agricultural Land-Use Settings in the Eastern San Joaquin Valley, California, 1993-1995.
U.S. Geological Survey, Denver, CO.

Carleton, G.B. 1996. Nitrate in Groundwater and Surface Water in a Residential Subdivision,
West Mercer Township, New Jersey. U.S. Geological Survey, Denver, CO.

Carson, R.T. 1997. "Contingent Valuation Surveys and Tests of Insensitivity to Scope." In
Determining the Value of Non-Marketed Goods: Economic, Psychological, and Policy Relevant
Aspects of Contingent Valuation Methods. RJ. Kopp, W.W. Pommerehne, andN. Schwarz
(eds.), Kluwer Academic Publishers, Dordrecht, The Netherlands, pp. 127-164.

CDC. 1998. A Survey of the Quality of Water Drawn from Domestic Wells in Nine Midwest
States. NCEH 97-0265. National Center for Environmental Health, Centers for Disease Control
and Analysis, Atlanta, GA.

Chen, A.H. 1998. Nitrate Concentrations in Groundwater in the Paleovalley Alluvial Aquifers of
the Nemaha Natural Resources District, Nebraska. U.S. Geological Survey, Lincoln, NE.

Chestnut, L.G., D. Mills, and A. Alberini. 1997. Monetary Valuation of Human Mortality Risks
in Cost-Benefit Analyses of Environmental Programs: Background Paper and Bibliography.

-------
                                  REFERENCES > R-2
Prepared for U.S. Environmental Protection Agency, Office of Policy Planning and Evaluation
by Hagler Bailly Services, Boulder, CO.

Clawges, R.M. and E.F. Vowinkel. 1996. "Variables Indicating Nitrate Contamination in
Bedrock Aquifers, Newark Basin, New Jersey." Water Resources Bulletin 32:1055-1066.

Crutchfield,  S.R. and J. Cooper. 1997. "Valuing Risk Reduction: The Example of Nitrates in
Drinking Water." Food Review January-April, 38-41.

Crutchfield,  S.R., J.C. Cooper, andD. Hellerstein. 1997. Benefits of Safer Drinking Water: The
Value of Nitrate Reduction. Agricultural Economic Report 752. U.S. Department of Agriculture,
Economic Research Service, Food and Consumer Economics Division.

Crutchfield,  S.R., P.M. Feather, and D.R. Hellerstein. 1995. The Benefits of Protecting Rural
Water Quality, An Empirical Analysis. U.S. Department of Agriculture, Economic Research
Service, Agricultural Economic Report Number 701.

De Zoysa, A.D.N.  1995. A Benefit Evaluation of Programs to Enhance Groundwater Quality,
Surface Water Quality and Wetland Habitat in Northwest Ohio.  Dissertation, The Ohio State
University, Columbus.

Delavan, W. 1997. Valuing the Benefits of Groundwater Protection from Nitrate Contamination
in Southeastern Pennsylvania. Master's Thesis, Department of Agricultural Economics and Rural
Sociology, Penn State University, State College.

Desvousges, W.H., M.C. Naughton, and G.P. Parsons. 1992. "Benefit Transfer:  Conceptual
Problems in Estimating Water Quality Benefits Using Existing Studies." Water Resources
Research 28(3):675-683.

Detroy, M.G., P.K.B. Hunt, and M.A. Holub. 1988. Groundwater Quality Monitoring Program in
Iowa: Nitrate and Pesticides in Shallow Aquifers. U.S. Geological Survey, Denver,  CO.

Dillman, D. 2000. "Reduction of Coverage and Sampling Errors." In Mail and Internet Surveys.
John Wiley and Sons, New York, pp.  194-213.

Edwards, S.F. 1988. "Option Prices for Groundwater Protection." Journal of Environmental
Economics and Management 15:475-487.

Fischhoff, B. and L. Furby. 1988. "Measuring Values: A Conceptual Framework for Interpreting
Transactions with Special Reference to Contingent-Valuation of Visibility." Journal of Risk and
Uncertainty  1:147-184.

-------
                                  REFERENCES > R-3
Freeman, A.M. 1993. Valuing the Longevity of Health. Chapter 10 in The Measurement of
Environmental and Resource Values. Resources for the Future, Washington, DC.

Giraldez, C. and G. Fox. 1995. "An Economic Analysis of Groundwater Contamination from
Agricultural Nitrate Emissions in Southern Ontario." Canadian Journal of Agricultural
Economics 43:387-402.

Hall, M. 1996. Simulation of Nitrates in a Regional Subsurface System: Linking Surface
Management with Ground Water Quality. PhD Thesis, Colorado State University, Fort Collins.

Ham, J.M., L.N. Reddi, C.W. Rice, and J.P. Murphy. 1998. Earthen Lagoons for Containing
Animal Waste: Review of Factors Affecting Seepage Losses and Groundwater Quality.
Department of Agronomy, Kansas State University, Manhattan.

Hanley, N. 1989. "Problems in Valuing Environmental Improvements Resulting from
Agricultural Change: The Case of Nitrate Pollution." In Economic Aspects of Environmental
Regulation in Agriculture, A. Dubgaard and A. Hjortshoj Nielsen, eds. Vauk Kiel KG,
Wissenschaftsverlag, Germany.

Hurley, T.M., D.  Otto, and J. Holtkamp.  1999. "Valuation of Water Quality in Livestock
Regions: An Application to Rural Watersheds in Iowa." Journal of Agricultural and Applied
Economics 31:177-184.

Jordan, J.L. and A.H. Elnagheeb. 1993. "Willingness to Pay for Improvements in Drinking Water
Quality." Water Resources Research 29:237-245.

Kalton, G. 1983.  Introduction to Survey Sampling. Sage University Paper 35. Newbury Park,
CA.

Kerr-Upal, M., T. Van Seters, and M. Stone. 1999. "Assessing the Risk of Groundwater Nitrate
Contamination in the Region of Waterloo, Ontario." Canadian Water Resources Journal
24:225-233.

Kross, B.C., G.R. Hallberg, and D.R. Bruner. 1993. "The Nitrate Contamination of Private Well
Water in Iowa." American Journal of Public Health 83:270-272.

Letson, D., N. Gollehon, and C. Mose. 1998. "Confined Animal Production and Groundwater
Protection." Review of Agricultural Economics 20:348-364.

Lichtenberg, E. and L.K. Shapiro. 1997. "Agriculture and Nitrate Concentrations in Maryland
Community Water System Wells." Journal of Environmental Quality 26:145-153.

-------
                                  REFERENCES > R-4
Lindsey, B.D. 1997. Nitrate in Groundwater and Streambase Flow in the Lower Susquehanna
River Basin, Pennsylvania and Maryland. U.S. Geological Survey, Denver, CO.

Loomis, J. 1992. "The Evolution of a More Rigorous Approach to Benefit Transfer." Water
Resources Research 28(3):701-705.

Miller, M.H., J.B. Robinson, and D.W. Gallagher. 1976. "Accumulation of Nutrients on Soil
Beneath Hog Manure Lagoons." Journal of Environmental Quality 5: 279-288.

Mitchell, R. and R. Carson, 1989. Using Surveys to Value Public Goods: The Contingent-
Valuation Method. Resources for the Future, Washington, DC.

Morrison, M.D., R.K. Blarney, and J.W. Bennett. 2000. Minimising Payment Vehicle Bias in
Contingent Valuation Studies. Environmental & Resource Economics 16(4):407-422.

Mueller, O.K., P.A. Hamilton, D.R. Helsel, KJ. Hitt, and B.C. Ruddy.  1995. Nutrients in Ground
Water and Surface Water of the United States — An Analysis of Data through  1992. USGS
Water-Resources Investigations Report 95-4031, U.S. Geological Survey.

National Research Council. 1997. Valuing Ground Water: Economic Concepts and Approaches.
National Academy Press, Washington, DC.

Nolan, B.T., B.C. Ruddy, KJ. Hitt, and D.R. Helsel. 1998. "A National Look at Nitrate
Contamination of Groundwater s." Water Conditioning and Purification 39:76-79.

North Carolina Division of Water Quality, Groundwater Section. 1998. Impact of Animal Waste
Lagoons on Ground Water Quality. Draft Report. North Carolina Department of Environment
and Natural Resources, Raleigh.

Poe, G.L. 1993.  Information, Risk Perceptions, and Contingent Value: The Case of Nitrates in
Groundwater. PhD Dissertation, University of Wisconsin-Madison.

Poe, G.L. 1998.  "Valuation of Groundwater Quality Using a Contingent Valuation-Damage
Function Approach."  Water Resources Research 34(12):3627-3633.

Poe, G. and R.C. Bishop. 1992. Measuring the Benefits of Groundwater Protection from
Agricultural Contamination: Results from a Two-Stage Contingent Valuation Study. Staff Paper
Series, University of Wisconsin-Madison.

Poe, G.L. and R.C. Bishop. 1999. "Valuing the Incremental Benefits of Groundwater Protection
When Exposure Levels Are Known." Environmental and Resource Economics 13:341-367.

-------
                                   REFERENCES > R-5
Randall, A. and D. DeZoysa. 1996. Groundwater, Surface Water, and Wetlands Valuation for
Benefits Transfer: A Progress Report. W-133 Benefits and Costs Transfer in Natural Resource
Planning, Ninth Interim Report. Department of Economics, Iowa State University, Ames.

Rausch, J.N. 1992. Sources of Ohio Farm Water Well Nitrate-Nitrogen Contamination and the
Willingness to Pay for Remediation. Masters Thesis, available from Ohio State University,
Columbus.

Richards, R.P., D.B. Baker, and L.K. Wallrabenstein. 1996. "Well Water Quality, Well
Vulnerability, and Agricultural Contamination in the Midwestern United States." Journal of
Environmental Quality 25:389-402.

Ritter, W.F. and A.E.M. Chirnside. 1990. "Impact of Animal and Waste Lagoons on
Groundwater Quality." Biological Wastes 34:39-54.

Smith, V.K., G. Van Houtven, and S. Pattanayak. 1999. Benefit Transfer as Preference
Calibration, Resources for the Future discussion paper 99-36.
http://www.rfforg/disc_papers/PDF_files/9936.pdf. Accessed 9/1/00.

Spalding, R.F. and M.E. Exner. 1993. "Occurrence of Nitrate  in Groundwater-A Review."
Journal of Environmental Quality 22:392-402.

Sparco, J.  1995. A Benefit-Cost Methodology for Analysis of Nitrate Abatement in Sussex
County, Delaware, Groundwater. PhD Dissertation, University of Delaware, Newark.

Stuart, M.A., FJ. Rich, and G.A. Bishop. 1995. "Survey of Nitrate Contamination in Shallow
Domestic Drinking Water Wells of the Inner Coastal Plain of Georgia." Journal of the National
Water Well Association 33:284.

Swistock, B.R., W.E. Sharpe, and P.O. Robillard. 1993. "A Survey of Lead, Nitrate, and Radon
Contamination of Individual Water Systems in Pennsylvania." Journal of Environmental Health
55:6-12.

TetraTech. 2002.  "Development of Pollutant Loading Reductions from Revised Effluent
Limitation Guidelines for Concentrated Animal Feeding Operations." Draft report prepared for
U.S. Environmental Protection Agency, Office of Science and Technology.

Townsend, M.A., R.O. Sleezer, and S.A. Macko. 1996. "Effects of Agricultural Practices and
Vadose Zone Stratigraphy on Nitrate Concentration in Ground Water in Kansas, USA." Water
Science Technology 3 3:4:219-226.

U.S. Bureau of Labor Statistics. 2000. Consumer Price Index.
ftp://ftp.bls.gov/pub/special.requests/cpi/cpiai.txt. Accessed 10/03/00.

-------
                                   REFERENCES > R-6
U.S. Census. 1990. http://venus.census.gov/cdrotn/loookup. Accessed 11/02.

U.S. EPA. 1990. National Survey of Pesticides in Drinking Water Wells. U.S. Environmental
Protection Agency, Washington, DC.

USGS. 1986. Occurrence of Nitrate and Herbicides in Groundwater in the Upper Conestoga
River Basin, Pennsylvania. Water Resources Investigations Report 85-4202, U.S. Geological
Survey.

USGS. 1996. Retrospective Database for Nutrients in Groundwater and Surface Water.
U.S. Geological Survey, http://water.usgs.gov/nawqa/nutrients/datasets/retrodata.html. Accessed
6/1/00.

USGS. 1998. National Water Quality Assessment Database. U.S. Geological Survey.
http://water.usgs.gov/nawqa/nutrients/datasets/cycle91/. Accessed 6/1/00.

Vitosh, M.L. 1985. Nitrogen Management Strategies for Corn Producers. Extension Bulletin
WQ06, Cooperative Extension Service, Michigan State University.

Walker, D.R. and J.P. Hoehn.  1990. "The Economic Damages of Groundwater Contamination in
Small Rural Communities: An Application to Nitrates." NorthcentralJournal of Agricultural
Economics 12:47-56.

Walsh, R., D. Johnson, and J. McKean. 1992. "Benefit Transfer of Outdoor Recreation Demand
Studies, 1968-1988." Water Resources Research 28(3):707-713.

Wattage, P.M. 1993. Measuring the Benefits of Water Resource Protection from Agricultural
Contamination: Results from a Contingent Valuation Study. PhD Dissertation, Iowa State
University, Ames.

Wolf, P.M. 1986. Meta-Analysis: Quantitative Methods for Research Synthesis. Sage University
Paper 59. Sage Publications, Newbury Park, CA.

-------
                                    APPENDIX A
                    NITROGEN SOURCES AND WELL DATA
Several individual datasets were combined to create the county level loadings data used to model
the relationship between nitrogen loadings and nitrate concentrations in private wells. The final
loadings dataset includes estimates of the total nitrogen loadings for each county under each
scenario, and was created by combining information from three different datasets provided by
EPA. These separate datasets contained information on the number of facilities in each county,
the percentage of these facilities that would be regulated under various scenarios, and the
loadings for each type of facility in each region of the country. These individual data files were
the loadings, facility, and state percent data files.

Loadings: The loadings dataset was provided as an Excel spreadsheet with multiple worksheets.
The file contains information on modeled  surface and leached nitrogen and phosphorous loadings
from a variety of sources, including on-site and off-site manure application, fertilizer application,
and loadings generated at farm production areas. A total of 250 facility types are included in the
dataset. The farm types are defined as the  combination of 10 animal types, 5 facility size
categories, and  5 regions. Loadings were estimated for baseline conditions and for four
regulatory options. Option 1  regulates loadings by setting limits on nitrogen application amounts.
Option 2 regulates loadings by setting limits on phosphorous application amounts. Option 3 is
similar to Option 2, but also requires liners for lagoons.1 Finally, Option 5 is similar to Option 2
but also requires that lagoons be covered.

Facility:  The facility dataset was provided as an Excel spreadsheet. This dataset identifies the
average number of facilities by animal type and size for 2,637 counties (including some counties
that have no facilities). The dataset identifies animal types of beef, veal, broilers, dairy, two types
of swine, wet layers, dry layers, turkey, and heifers. Facilities are ranked as small, medium,  or
large based on the definitions in Exhibit A-l. The dataset has 2,637 observations (one for each
county) and 55 variables, including identifier columns for the counties and number of facilities
for the different animal type and facility size.

State Percent:  The state percent dataset was provided as an Excel spreadsheet. The dataset
identifies the percentage of each facility type that will be regulated under each scenario
(including baseline) for each state.
1. Option 3 is similar to Option 2 but also requires liners for lagoons. As the leached nitrate loadings are the
same for our analysis under Options 2 and 3 these are reported simply as Option 2/3 throughout this report.

-------
                                   APPENDIX A > A-2
Exhibit A-l
Summary of Size Category Definitions for All Animal Types
Sector
Mature Dairy Cattle
Veal Calves
Cattle or Cow/Calf Pairs
Heifer
Swine (weighing over 25 kilograms)
Swine (weighing less than 25 kilograms)
Horses0
Sheep or Lambs0
Turkeys
Chickens (wet manure systems)
Chickens Other than Laying Hens (dry
manure systems)
Laying Hens (dry manure systems)
Ducks (dry operations)0
Ducks (wet operations)0
Large
More than 700
More than 1,000
More than 1,000
More than 1,000
More than 2,500
More than 10,000
More than 500
More than 10,000
More than 55,000
More than 30,000
More than 125,000
More than 82,000
More than 40,000
More than 5,000
Medium"
200-700
300-1,000
300-1,000
300-1,000
750-2,500
3,000-10,000
150-500
3,000-10,000
16,500-55,000
9,000-30,000
30,000-125,000
25,000-82,000
12,000-40,000
1,500-5,000
Small"
Less than 200
Less than 300
Less than 300
Less than 300
Less than 750
Less than 3,000
Less than 150
Less than 3,000
Less than 16,500
Less than 9,000
Less than 30,000
Less than 25,000
Less than 12,000
Less than 1,500
a. Must also meet one of two criteria to be defined as a CAFO.
b. Must be designated by the permitting authority.
c. Not included in final analysis.
Output — County Level Total Nitrogen Loadings Dataset: The output of combining these
datasets is the nitrogen loadings for each county for each of the options/scenarios for the
2,637 counties with AFOs. Data from 678 of these 2,637 counties are combined with data from
the USGS Retrospective Database (described below) for estimation of the gamma model.2 An
issue is whether the counties used for the gamma modeling are different in some manner from
those (1,959) not used for estimating the nitrogen-nitrate relationship. Exhibit A-2 shows mean
values for the average loadings and various sociodemographic data from these two groups of
counties.  The "percent difference" column indicates how much larger (or smaller for negative
values) the mean values are for counties used in the gamma modeling compared to counties not
used in the gamma model. The Z score from a Wilcoxon rank test show whether the differences
are statistically significant. All of the Z scores are significant at the 1% level. In general the
average county nitrogen loadings in the gamma model counties are higher than the excluded
2. Of these 678 counties, 374 have at least one well with enough data to be included in the analysis.

-------
APPENDIX A > A-3
Exhibit A-2
Comparison of Mean Loadings and Sociodemographics for Counties in the Loadings
Database Used in the Gamma Modeling
(for counties in model, n = 374; for counties not in model, n = 2,263)
Variable
Baseline nitrogen loadings
Option 1 — Scenario 6
Option 1 — Scenario 7
Option 1 — Scenario 8
Option 1 — Scenario 9
Option 2/3 — Scenario 6
Option 2/3 — Scenario 7
Option 2/3 — Scenario 8a
Option 2/3 — Scenario 9
Option 5 — Scenario 6
Option 5 — Scenario 7
Option 5 — Scenario 8
Option 5 — Scenario 9
Loadings per Acre (baseline)
Acres
Population13
Population Densityb
Percent of County Land in Farms'3
Median Household Incomeb
Housing Units'3
Mean (counties
included in the
gamma model)
410,019
299,521
319,964
319,959
323,081
289,843
311,410
311,402
314,788
315,212
327,810
327,802
329,605
0.99
528,201
83,296
0.13
0.62
26,196
33,451
Mean (counties
not included in
the gamma
model)
195,013
157,612
164,853
164,851
166,307
154,159
161,154
161,150
162,657
170,659
172,942
172,938
173,613
0.51
732,337
74,900
0.11
0.53
23,761
30,766
Percent
Difference
(from counties
not included)
-52%
-47%
-48%
-48%
-49%
-47%
-48%
-48%
-48%
-46%
-47%
-47%
-47%
-48%
39%
-10%
-11%
-14%
-9%
-8%
Z
(Wilcoxon
rank test)
9.66
10.72
10.61
10.61
10.59
10.71
10.63
10.63
10.63
10.81
10.67
10.67
10.66
9.76
-2.68
3.67
4.36
5.05
7.71
3.17
a. Proposed scenario.
b. nobs = 2,257 for "not in the retrospective database" and 374 for "in the retrospective database."

-------
                                    APPENDIX A > A-4
counties. In addition the included counties are somewhat smaller (30% smaller) and have a
roughly 10-30% larger population, higher median income, and greater number of housing units.
The included counties also have a larger portion of their land area in farms.

Septic Ratio: EPA calculated the number of septic systems per acre in each county using data
from the 1990 U.S. Census and the 1997  Census of Agriculture. This provides a proxy measure
for the contribution of septic systems to well nitrate concentrations. The number of household
units on septic systems for each county was reported in the U.S. Census, and the total acres per
county was reported in the Census of Agriculture.

The USGS Retrospective Database: As discussed in Section 2.2.1, the USGS Retrospective
Database contains water quality and land use data from 10,426 well samples from 725 counties
in 38 states. The data were gathered between 1969 and 1992.3

The dataset provides information on well location, well characteristics, pollution inputs, and well
water sample. Each observation provides well location information, including FIPS code, town,
state FIPS code, county FIPS code, study unit, well identification number, and latitude and
longitude. Well characteristics include water use (e.g., domestic, stock, public, or irrigation),  well
depth in feet, depth to water in feet, geographic region, soil hydrologic group, lithological
description of the aquifer, land use category (e.g., agricultural, woods, or urban), population
density in people per square kilometer, the ratio of pasture to cropland, and the ratio of woodland
to cropland.

Pollution input information includes atmospheric nitrogen input, fertilizer nitrogen input in tons
sold per square mile,  fertilizer plus atmospheric nitrogen inputs in tons per square mile, fertilizer
plus atmospheric nitrogen inputs in pounds per acre, manure nitrogen input in tons per square
mile, and the sum of nitrogen inputs. Well water sample information includes ammonia as
nitrogen in mg/L, nitrate as nitrogen in mg/L, total phosphate in mg/L,  and orthophosphate as
phosphorous  in mg/L, and the year of the sample.

Exhibit A-3 provides summary statistics on the observations from the USGS Retrospective
Database for  all observations in the dataset. This includes all water use types. Only a subset of
these observations (2,985 observations) were usable for the analysis described in Chapter 3
because of missing data. The mean well nitrate concentration is 2.89 mg/L, ranging from no
nitrates to 125.64 mg/L. Of the 10,426 observations in the retrospective dataset, 19.8% are at or
below 1.0 mg/L and 7.4% exceed the MCL of 10 mg/L.
3. Of these 725 counties, 374 are also estimated to have nitrate loadings >0.

-------
                                  APPENDIX A > A-5
Exhibit A-3
USGS Retrospective Database Summary Data
(all water use types)
Variable
Well Depth (feet)
Soil Hydrologic Group
Pop. Density (people per km square)
Atmospheric Nitrogen Input
Fertilizer Nitrogen Input (tons/mi sq)
Fertilizer plus Atmospheric Nitrogen
Input (tons/mi sq)
Fertilizer plus Atmospheric Nitrogen
Input (Ibs/acre)
Manure Nitrogen Input (tons/mi sq)
Sum of Nitrogen Inputs
Ratio of Pasture to Cropland
Ratio of Woodland to Cropland
Year of Sample
Nitrate as Nitrogen (mg/L)
Total Phosphate (mg/L)
N
9141
10419
10426
10426
10426
10426
10426
10426
10426
9981
9772
9289
10426
3336
Missing
1285
7
0
0
0
0
0
0
0
445
654
1137
0
7090
Mean
282.728
2.549
131.958
1.355
5.958
7.313
22.853
4.086
11.400
5.502
0.500
1982.509
2.886
0.069
Std Dev
400.204
0.729
427.002
0.598
6.210
6.374
19.920
5.614
9.887
19.374
1.391
5.629
5.958
0.263
Min.
1.000
1.000
0.000
0.172
0.000
0.208
0.650
0.000
0.219
0.006
0.000
1969.000
0.000
0.000
Max.
5310.000
4.000
13516.670
2.910
30.010
31.882
99.631
34.502
50.048
147.991
14.880
1992.000
125.640
7.500
Exhibit A-4 shows the distribution of well water use for observations in the USGS dataset.
Because the benefits transfer exercise is focused on domestic water well use, we limited analysis
to wells listed as domestic, which make up roughly 31% of the observations from the
Retrospective Database.
Exhibit A-4
Distribution of Well Water Use in Retrospective Database
Water Use
Domestic
Irrigation
Public
Stock
Unknown
Frequency
3226
838
1088
209
5065
Percent
30.94
8.04
10.44
2.00
48.58

-------
                                    APPENDIX A > A-6
Exhibit A-5 presents the summary information for only those wells listed as being for domestic
use. Of particular interest for the modeling described in Chapter 3 is the observation that the
average total of fertilizer sales and atmospheric nitrogen inputs (8.85 tons/square mile) exceeds
that from manure of 6.03 tons/square mile. This suggests that in understanding the potential
benefits of controlling nitrogen inputs to groundwater from CAFOs it is important to control for
non-CAFO nitrogen sources. In the analysis for this rule, EPA estimated leached nitrogen from
the application of fertilizer under each regulatory scenario. As discussed above, EPA used a
proxy measure, density of septic systems in a county, to control for nitrogen loadings from septic
systems.
Exhibit A-5
USGS Retrospective Database Summary Data
(domestic water use only)
Variable
Well Depth (feet)
Soil Hydrologic Group
Pop. Density (people per km square)
Atmospheric Nitrogen Input
Fertilizer Nitrogen Input (tons/mi sq)
Fertilizer plus Atmospheric Nitrogen
Input (tons/mi sq)
Fertilizer plus Atmospheric Nitrogen
Input (Ibs/acre)
Manure Nitrogen Input (tons/mi sq)
Sum of Nitrogen Inputs
Ratio of Pasture to Cropland
Ratio of Woodland to Cropland
Year of Sample
Nitrate as Nitrogen (mg/L)
Total Phosphate (mg/L)
N
3068
3225
3226
3226
3226
3226
3226
3226
3226
3143
3117
2789
3226
1006
Missing
158
1
0
0
0
0
0
0
0
83
109
437
0
2220
Mean
169.320
2.425
47.071
1.627
7.224
8.851
27.658
6.033
14.884
0.945
0.234
1983.068
3.548
0.068
Std Dev
135.569
0.654
136.469
0.595
5.992
5.999
18.746
7.271
10.343
2.432
0.437
5.542
6.406
0.291
Min.
1.000
1.000
0.045
0.172
0.000
0.215
0.672
0.000
0.219
0.012
0.000
1969.000
0.000
0.000
Max.
1996.000
4.000
2321.628
2.855
30.010
31.882
99.631
34.502
44.114
24.597
6.227
1991.000
84.300
6.400

-------
                                    APPENDIX B
                              STATISTICAL MODELS
As described in Section 3.2, the statistical analysis of the relationship between loadings and well
nitrate concentrations is based on the following linear model:

              Nitrate (mg/L) = 30 + 3! ag dummy + B2 soil group + B3 well depth
                   + B4 septic ratio + B5 atmospheric N + B6 loadings ratio

Well nitrate concentrations are the dependent variable in the analysis. Summary statistics on the
distribution of observed values for well nitrates indicate a nonnegative distribution with a
rightward skew (skew = 4.85) and a thick tail (kurtosis = 37.15) (see Exhibit B-l).
                                      Exhibit B-l
                         Nitrate Distribution: Observed Values
            35 ^

            30

            25

          c 20
          
-------
                                     APPENDIX B > B-2
The gamma and exponential distributions both allow for fitting of nonnegative, right skewed
distributions (no observations are assumed to be censored in the exponential or gamma models).
The gamma distribution has the density function:
                                        r(cc)

We used the gamma distribution instead of the more commonly used exponential distribution
since it is more general that the exponential model (includes the exponential specification as a
special case).1 The gamma distribution allows for the density function to be more flexible and
allows for more curvature in the distribution. To model the relationship between the nitrate levels
(y) and the independent variables, let Q. = exp(-/?x.). For this distribution,
E(yt) = aldi = aexp(/?x.) . Maximum likelihood methods are used to estimate the parameters.
The log likelihood function is:
This log likelihood was maximized using GAUSS software.2 Estimation results are displayed in
Exhibit B-2. All of the parameter estimates are significant at the 1% level and are of the expected
sign. From the gamma model, expected values can be calculated using:


                                 E(yI)=a/Q, =ccexp(j3x/.).
EPA tested the ability of the gamma model to estimate small nitrate concentrations by comparing
the model's intercept with the natural, or ambient, level of nitrate in groundwater in the United
States.3 Using the mean values for soil group and well depth and setting all other variables to
zero (i.e., setting the ag dummy and all human nitrogen sources to zero), the model predicts an
1. A likelihood ratio test of the difference between the exponential model (where CC is restricted to equal 1) and
the gamma model (where alpha is estimated) yielded a %2 statistic of 659.98, so that the null hypothesis that
CC = 1 is rejected at any level of significance (the 1% tail of the %2(1) distribution is at 6.63).

2. A range of starting values were used in the GAUSS program to examine the sensitivity of results to starting
values.  For all starting values for which the program converged, virtually the identical parameter estimates
were obtained.

3. Technically, the intercept term includes ambient levels of nitrates as well as those induced by loadings from
AFOs with less than 300 AUs since these are not included in the loadings data.

-------
                                   APPENDIX B > B-3
Exhibit B-2
Gamma Regression Results
Variable
Intercept
Loadings Ratio
Atmospheric Nitrogen
Well Depth3
Soil Group
Septic Ratio
Ag Dummy
Central Region Dummy
Mid-Atlantic Region Dummy
Pacific Region Dummy
South Region Dummy
Alpha
Parameter
Estimate
2.201
0.046
0.032
-0.171
-0.384
1.618
0.686
-0.076
-0.165
0.812
-0.907
0.497
Standard
Error
0.194
0.007
0.028
0.012
0.044
1.728
0.064
0.160
0.098
0.117
0.127
0.010
Asymptotic
T-Statistic
11.352
6.543
1.144
-13.782
-8.660
0.936
10.663
-0.475
-1.691
6.918
-7.170
50.639
Significance
0.000
0.000
0.253
0.000
0.000
0.349
0.000
0.635
0.091
0.000
0.000
0.000
Mean log-likelihood = -1.85646.
N = 2,985.
a. In the model, well depth is scaled to units of hundreds of feet.
ambient nitrate concentration in the Midwest region of 1.32 mg/L on nonagricultural lands.
Using the same approach, the predicted value on agricultural land is 2.63 mg/L. Several studies
report natural nitrate levels ranging between 2 and 3 mg/L (Poe and Bishop, 1992; Kross et al.,
1993; Poe, 1998), although one study suggests that 3 mg/L may be too high, given the high
number of wells with nitrate levels below the detection limit in many groundwater monitoring
studies (Spalding and Exner, 1993). Giraldez and Fox (1995) report that natural nitrate
concentration in groundwater is generally about 1.0 mg/L. Therefore the model's estimate of
1.32 mg/L on non-agricultural land seems to be a reasonable estimate of nitrate concentrations in
the absence of the pollution from septic systems, atmospheric deposition, and AFOs.

Other Models

In addition to the gamma model described above, several other model types were explored for
this analysis. Given the nature of nitrate contaminations, a nonnegative distribution is preferred.
The OLS and Tobit models discussed here were estimated to allow us to explore whether these
simpler models would suffice for purposes of modeling the nitrate-nitrogen relationship. The
OLS, Tobit, and Selection-Trunctation models were estimated using GAUSS Version 4.0.

-------
                                    APPENDIX B > B-4
Ordinary Least Squares (OLS)

OLS was used initially to model the loadings-well nitrate relationship to explore how well the
data could explain this relationship. Estimation results are displayed in Exhibit B-3.
Exhibit B-3
OLS Regression Results
Variable
Intercept
Loadings Ratio
Atmospheric Nitrogen
Well Depthb
Soil Group
Septic Ratio
Ag Dummy
Central Region Dummy
Mid-Atlantic Region
Dummy
Pacific Region Dummy
South Region Dummy
Parameter Estimate
4.907
0.197
0.176
-0.625
-1.234
-2.768
1.709
0.048
-0.262
3.292
-1.865
Standard Error
0.768
0.031
0.117
0.086
0.184
6.786
0.289
0.688
0.394
0.496
0.535
T Value
6.391
6.374a
1.511
-7.278a
-6.722a
-0.408
5.910a
0.070
-0.666
6.637a
-3.485a
F value = 31.946; Adjusted R2= 0.094.
N = 2,985.
a. Indicates significant at the 1% level.
b. In the model, well depth is scaled to units of hundreds of feet.
The results indicate that there are significant relationships between the dependent and most
independent variables. The signs are all of the expected direction. The coefficient on Loadings
Ratio is significant at the 1% level. It must be emphasized that there are a priori reasons to prefer
a distribution that does not allow for negative values in the dependent variable (well nitrate
concentrations), and thus the OLS and Tobit models were purely exploratory models.

Tobit

Since well nitrates at or below the detection limit were reported in a number of ways, nondetects
were set to 0.05 mg/L; 522 of the 2,985 observations had nitrate values reported at the detection
limit. Treating this as a censoring of the distribution, we used a  Tobit model to estimate the
parameter coefficients. Exhibit B-4 reports the Tobit model estimates.

-------
                                    APPENDIX B > B-5
Exhibit B-4
Tobit Regression Results
Variable
Intercept
Loadings Ratio
Atmospheric Nitrogen
Well Depth3
Soil Group
Septic Ratio
Ag Dummy
Central Region Dummy
Mid-Atlantic Region Dummy
Pacific Region Dummy
South Region Dummy
Sigma
Estimate
5.200
0.216
0.222
-0.731
-1.618
4.184
2.227
0.382
-0.543
3.447
-4.149
1.914
Standard Error
0.941
0.036
0.140
0.102
0.233
8.959
0.371
0.803
0.548
0.563
0.735
0.014
Chi-Square
5.529
5.981
1.588
-7.186
-6.950
0.467
6.011
0.476
-0.992
6.129
-5.646
132.804
Pr > ChiSq
0.000
0.000
0.112
0.000
0.000
0.641
0.000
0.634
0.321
0.000
0.000
0.000
Log likelihood: -1.90865.
N = 2,985.
a. In the model, well depth is scaled to units of hundreds of feet.
As seen in Exhibit B-4, the Tobit model produced generally strong results with significant
coefficient estimates of the correct sign. While the Tobit model is used for modeling
observations on non-negative values, in this case with observations truncated at nitrate
concentrations below the detection limit, using the model to fit expected values could still predict
negative nitrate concentrations. We thus used the  Tobit model purely to explore the data and the
relationships between dependent and independent variables as well as potential misspecifications
of the error term.

Exponential

As with the OLS model, the Tobit model may not be appropriate to use to explore the physical
relationship between nitrogen loadings and well nitrate concentrations because the Tobit model
assumes a censoring of true values at zero, and true nitrate concentrations are non-negative. We
thus explored the use of the exponential and gamma models as nonnegative distributions.
Assuming the_y, follow the exponential distribution, the density function is:
                                  f(y) = 6 exp(-6y)

-------
                                    APPENDIX B > B-6
Letting 0. = exp (- /&,.), the expected value ofy, is£(_y.) = 1/6,=
Maximum likelihood methods are used to estimate the parameters. The log-likelihood function
 s:
The only difference between the exponential and gamma models is that (X is set to 1 for the
exponential model. In the more general gamma model, a is estimated. As discussed above, a
was found to be significantly different from 1 and thus we felt the gamma model represented a
better model to use for scenario analysis for CAFOs. Exhibit B-5 presents the results of
estimating the exponential model using GAUSS.
Exhibit B-5
Exponential Regression Results
Variable
Intercept
Loadings Ratio
Atmospheric Nitrogen
Well Depth3
Soil Group
Septic Ratio
Ag Dummy
Central Region Dummy
Mid-Atlantic Region Dummy
Pacific Region Dummy
South Region Dummy
Parameter Estimate
1.502
0.046
0.032
-0.171
-0.384
1.616
0.686
-0.076
-0.165
0.812
-0.907
Standard Error
0.141
0.005
0.020
0.009
0.030
1.219
0.049
0.116
0.072
0.087
0.093
Asymptotic
T Value
10.641"
9.1923
1.571
-18.4153
-12.7233
1.326
13.9953
-0.652
-2.293a
9.3043
-9.8123
Mean log-likelihood = -2.07756.
N = 2,985.
a. In the model, well depth is scaled to units of hundreds of feet.
b. Indicates significant at the 0.01% level.
Most of the coefficients are significant at the 1% level. The exceptions are atmospheric nitrogen
deposition, the septic ratio, and the Central Region dummy. All the variables have the expected
sign. The coefficients are nearly identical to those  estimated in the gamma model except for the
alpha coefficient, which is implicitly restricted to 1 in the exponential model (Exhibit B-2), and
the septic ratio coefficient, which differs by 0.001. Note that the data for well depth was scaled in
order for GAUSS to converge to a solution.

-------
                                    APPENDIX B > B-7
Using the parameter estimates from the exponential model, we can calculate expected ambient
nitrate levels. Using mean values for well depth and soil type, and setting all anthropocentric
nitrogen sources equal to zero, the expected ambient well nitrate concentration in the Midwest is
1.32 mg/L for non-agricultural land and 2.62 mg/L for wells on agricultural land. These values
are within the range of natural or ambient nitrate concentrations as reported in Section 3.3.

Selection Model

Some of the models above perform relatively well in predicting nitrogen concentration.  Some are
statistically stronger than others. They all share a common weakness, though, in that the data
used in the model may not be an unbiased sampling of wells in the United States. Of the more
than 3,000 counties in the United States, only 374 were represented in the final database created
for this analysis. Geographically, these counties tend to be concentrated in the Midwest and
Middle Atlantic states (with a very large percentage of the wells sampled located in Nebraska and
Pennsylvania). The results presented Exhibit A-2 in Appendix A indicate that not only are the
counties used in the model geographically different than counties not used included in the model,
but also the characteristics of these counties appear to be quite different. In particular, it appears
that the wells included in the Retrospective Database tend to be located in counties with higher
than average levels of nitrogen loadings from manure and fertilizer.

To explore correcting for this potential problem EPA developed a selection model that includes
components which aim to capture the effects of sample selection bias. Ultimately, this model
generates estimates of nitrate concentration that are quite similar to the estimates from the
gamma model. Thus, EPA used the gamma model, which was used as the primary model for the
proposed rule, as the primary model for the final benefit analysis as well. The results of the
sample selection model are included here for comparison. The details of the sample selection
model follow.

Let y'j be the theoretical nitrate concentration of the /-th well in county j, for  j = l,...,J(~ 300)
and i = \,...,Hj (2-inj = 2985J. Concentration is modeled as a function of well characteristics and
county characteristics, both observed and unobserved:

                                  y'^pxi+z'jy+uj+e,,,                             (B-l)

where the Zj  are the observed covariates common to all wells in county7, the xtj  are the observed
characteristics of the particular well, and the random unobserved factors are u. ~ N(0, (f) and
ev ~^(0'°^), assumed mutually independent and independent of one another. That is,

                                    E(u,eu\=0 V/,./
and                                   v j  v)       >J
                                                                                    (B-2)
                                     E(UjUr)=0
                                     E(eueu,)=0  V/*/'.

-------
                                   APPENDIX B > B-8
Suppose nitrate concentration data for a well exists only if the county-specific component,
z'y+u , is sufficiently high. Furthermore, we can measure concentration only down to 0.05 per
mg/L. The selection model then is
                         {*     *                  /       \
                        yp.  if yp. >.05 and u}. >—z'^(regime Ij

                        05  if y"v <.05 and uj > —z'y(regime II)^
                                                              (B-3)
and there is information on y.. only if uj >—z'y. This is a censored/truncated regression model
(censored by the detection limit; truncated by the county rule). Let w.. =uj +ev. The likelihood is

                                                        >.05,M >-z'r)           (B-4)
which can be written
In Equation B-5 /(-,-) , is the bivariate normal density, and the bivariate probability is from the
normal distribution:
                       V.^    lYcAfV+a2  a2
                             N
                                .0
                    a2    a2
                                                                                 (B-6)
The contribution to the likelihood in the first data regime can be written
J_
>* <.05 and uj >—z^y\ we have

-------
                                   APPENDIX B > B-9
                       P|wff<05-

                       P[Wy<.05-

                             w..
                       P[-    1J
  +zjy)'uj>~z'y]=
  +z'Jy),-uJcr;
                                             •',-p
where  O is the cdf of the cumulative distribution function for the standardized bivariate normal
random vector with correlation
The results from the estimation of the sample selection model are presented in Exhibit B-6. The
results are very similar to the Tobit model reported in Exhibit B-4. All the significant parameters
have intuitive signs. The only variable in the model with a counterintuitive sign is atmospheric
nitrogen deposition, which has a negative but not statistically significant sign. Most of the other
parameters are significant at the 1% level of significance; the septic ratio and two of the regional
dummies are the exceptions.

Estimated Benefits

Exhibit B-7 presents the benefits of four scenarios (Scenarios 6 and 8, combined with Options 2
and 2), as estimated using the gamma, Tobit, and Selection models. The estimates in the table
range from a low of $66.88 million (Optionl-ScenarioS in the Selection model), to a high of
$94.07 million (Option2/3-Scenario6 in the Tobit model). The largest spread is found for
Optionl-Scenario6, where the difference between the minimum and maximum estimates from
the three models is $8.13 million.

It is interesting to note that the relative magnitudes of the total benefits vary by model, option,
and scenario. For Option 1, the  gamma model produces the largest benefits estimates. For
proposed option, Option 2/3, the gamma model produces estimated benefits that are smaller than
the selection model and only slightly larger than the Tobit model. It is not entirely clear why the
Selection model should produce benefits estimates that are larger than the gamma model. Since
the gamma model produces the most conservative estimate of benefits for the proposed rule, and
because it does not allow the prediction of negative concentrations, it is the preferred model for
the groundwater analysis.

-------
APPENDIX B
Exhibit B-6
Sample Selection Model Regression Results
Variable
Intercept
Loadings Ratio
Atmospheric Nitrogen
Well Depth3
Soil Group
Septic Ratio
Ag Dummy
Central Region Dummy
Mid-Atlantic Region Dummy
Pacific Region Dummy
South Region Dummy
02e
Parameter
Estimate
7.540
0.218
-0.116
-0.716
-2.317
1.100
2.220
0.052
0.385
3.580
-4.465
3.863
Standard
Error
1.056
0.040
0.141
0.105
0.221
10.653
0.495
1.108
0.593
0.609
0.895
0.012
Asymptotic T-
Statistic
7.143
5.472
-0.819
-6.811
-10.474
0.103
4.488
0.047
0.650
5.877
-4.991
320.014
Significance
0.000
0.000
0.413
0.000
0.000
0.918
0.000
0.963
0.516
0.000
0.000
0.000
Mean log-likelihood = -2.89556.
N = 2,985.
a. In the model, well depth is scaled to units of hundreds of feet.

-------
APPENDIX B »-B-ll
Exhibit B-7
Estimated Benefits from Various Models
Regulatory Scenario
Option 1 — Scenario 6
Option 1 — Scenario 6
Option 1 — Scenario 6
Model
Gamma
Tobit
Selection
Expected Reductions
in Number of
Households with Well
Nitrate Concentrations
above 10 mg/L
148,705
134,764
144,058
Total Expected
National Nitrate
Reduction
(mg/L) for Wells
with
Concentrations
between 1 and 10
mg/L at Baseline
854,326
851,575
1,085,613
Undiscounted
Annual Benefits
under CAFO
Regulatory
Scenarios
(Millions 2001$)
88.48
80.35
86.25

Option 1 — Scenario 8
Option 1 — Scenario 8
Option 1 — Scenario 8
Gamma
Tobit
Selection
120,823
116,176
111,529
716,007
695,699
891,660
71.94
69.18
66.88

Option 2/3 — Scenario 6
Option 2/3 — Scenario 6
Option 2/3 — Scenario 6
Gamma
Tobit
Selection
148,705
157,999
148,705
927,730
935,138
1,190,218
88.63
94.07
89.18

Option 2/3 — Scenario 8a
Option 2/3 — Scenario 8a
Option 2/3 — Scenario 8a
Gamma
Tobit
Selection
120,823
120,823
130,117
788,305
775,856
992,566
72.09
72.06
77.93
a. Proposed scenario.

-------
           APPENDIX C
SUMMARY OF GROUNDWATER VALUATION
OF NITRATE CONTAMINATION LITERATURE

-------
APPENDIX C > C-l
Study Reference
1. Published/Peer
Reviewed?
2. Year of Analysis
3. Place
4. Type of Study
5. Survey Implement
6. Respondents
7. Response Rate
8. Location (urban, rural,
etc.)d
9. Who Was Asked?
Crutch field et al.
USDA ERS Report
1994
IN, Central NE, PA, WA
Survey eliciting WTP for
improved water quality
Telephone
1600?b
50% (8 19 usable
responses)
Unspecified
Residents
Delavan
Master's thesis
1996
Southeastern PA (parts of
Lebanon and Lancaster counties)
Survey eliciting WTP for
improved water quality
Mail
1000 mailed
68.6%
75% of respondents live in
borough or city; 6.3% involved
with farming
Residents
de Zoysa
PhD dissertation
1994
Maumee River
Basin, northwest
Ohio
Survey eliciting
WTP for
improved water
quality
Mail
1050°
51% overall
Urban, suburban,
and rural
Urban and rural
residents in the
drainage, urban
residents outside
the drainage
Edwards
Journ. of Environmental
Economics and Management
1987
Cape Cod, MA
Survey eliciting WTP to
prevent contamination of
aquifer
Mail
1000 mailed
78.5%
(58.5% analyzable)
Primarily rural
Households listed in phone
book (renters, resident and
nonresident property owners)
Giraldez and Fox
Canadian Journ. of
Agricultural Economics
1995
Hensall, southwestern
Ontario
Metadata (lifetime
earnings, wage risk
studies, & CVM)
n.a.a
(n.a. = not applicable)
n.a.
n.a.
n.a.
n.a.
a. "n.a." indicates that either the information was not available or was not relevant to this study.
b. Crutch field et al. indicate that there were 819 usable responses and about a 50% response rate.
c. Of the 147 versions, 84 included the groundwater valuation scenario. These were randomly distributed proportionally to the 1,000 person sample.
d. Poe and Bishop (1992) define rural as census tracts that do not have municipally provided water. Although the definition of rural in most other studies is not
clarified, we interpret rural, as used in these studies, to mean areas with nonmunicipal water supply for domestic use.

-------
APPENDIX C > C-2
Study Reference
10. Household Water
Supply /Ground-water Use
1 1 . Actual Groundwater
Baseline Condition
12. Groundwater Baseline
Scenarios
1 3 . Change in
Groundwater Scenario
14. Credibility of Scenario
Change
15. Contaminants
16. Source of
Contaminants
17. Types of Values
Estimated
Crutch field et al.
Municipal?: IN 73%; NE
69%; PA 53%;
WA 74%
Unknown — between 17%
and 53% for the four
regions had heard about N
contamination
None given
If tap water has 50%
greater N levels than
EPA's MCL, how much to
reduce to min. safety
standards; how much to
completely eliminate
Not reported — several
questions were asked that
could be used to identify
scenario rejecters
Nitrates
Not specified
Primarily use values
(commodity is a point-of-
use filter)
Delavan
40% private wells, 60% public
sources (incl bottled water)
Perceived GW quality is 7 1 on a
scale of 0-100, w/0 as not safe
and 100 as definitely safe
50% of private wells meet
lOmg/L MCL
In 10 years, 75% of private wells
will meet MCL
Checked for scenario rejection
and also the scenario was very
specific
Nitrates
Fertilizer, septic, manure
Total
de Zoysa
Not specified
0.5to3.0mg/L
with some higher
Typical N
concentrations
range from
0.5-3 mg/L,
although some
are much higher
Reduce levels to
0.5-1 mg/L
Reduce N
contamination
from fertilizer
applied to
farm fields
Nitrates
Agricultural
fertilizer
Total
Edwards
89% public
11% private wells
Assumed that current water
quality is safe
Safe (state and county
systematically monitor
nitrate levels) — respondents
were told to assume no
health risks
Prevent uncertain nitrate
contamination of Cape Cod's
sole source aquifer
Although vague, respondents
were told to suppose the
program was possible
Nitrates
Fertilizer and sewage
(primarily sewage)
Option price
(use value)
Giraldez and Fox
n.a.e
King St Well > lOmg/L
York St Well high also
n.a.
n.a.
n.a.
Nitrates
n.a.
Total value benefits
transfer from CVM
e. 100% groundwater apparently from a public water supply distributing untreated well water.

-------
APPENDIX C > C-3
Study Reference
18. Valuation
Methodology
19. Payment Vehicle
20. Duration of Payment
Vehicle
2 1 . # of Survey Versions
22. Analysis
23. Mean Annual
Household WTP in Study
Year Dollars
24. Mean Annual
Household WTP in 2001
Dollars
Crutch field et al.
Dichotomous choice
Payment for local water
agency for filter
installation and
maintenance
Monthly, in perpetuity
Not specified, but multiple
Bivariate probit
$52.89/month (reduced N
to MCL);
$54. 50 (no N): $1.61
difference.
$63.20/month (reduced N
to MCL);
$65. 13 (no N): $1.92
difference.
Delavan
Dichotomous choice open ended
(DOE); informed open ended
(IOE)
Special tax
Annually for 10 years
2f
Tobit model8
DOE: $44.78 w/protest bidders
IOE: $29.26 w/protest bidders
DOE: $67.85 w/o protest bidders
IOE: $47. 16 w/o protest bidders
DOE: $50.55 w/protest bidders
IOE: $33.03 w/protest bidders
DOE: $76.59 w/o protest bidders
IOE: $53.23 w/o protest bidders
de Zoysa
Dichotomous
choice followed
by open-ended
Special tax
One time
147
Probit model
$52.78 lower
bound mean
(1994$ from
YNP model)
$63. 07 lower
bound mean
Edwards
Dichotomous choice
Fur versions: (1) bond,
(2) water bill,
(3) voluntary contribution,
and (4) unspecified
Annually
(in perpetuity)
10
logit
$1623 for a management
plan to increase the
probability of supply from
0.0 to 1.0
$2,530.22 for a management
plan to increase the
probability of supply from
0.0 to 1.0
Giraldez and Fox
(1) Loss of lifetime
earnings; (2) value of
statistical life; (3) total
value benefits transfer
from CVM
n.a.
n.a.
n.a.
n.a.
Based on disaggregating
value community value
estimate: $412. 50 per HH
($72.73/yrto$1696.97/yr)
Based on disaggregating
value community value
estimate:$479.36 per HH
($84.52/yrto
$l,972.00/yr)
f. Two "types" of survey (DOE and IOE). The DOE had eight versions differing only in the bid amount.
g. Also used a logit model to examine protest bids.

-------
APPENDIX C > C-4
Study Reference
25. Median Annual
Household WTP in Study
year dollars
26. Median Annual
Household WTP in 2001
Dollars
27. Range
28. Significant
Explanatory Variables
Crutch field et al.
n.a.
n.a.
$45.42-$60.76/month
Bid value (-)
income (+)
years lived in
ZIP code (+)
age (-)
Delavan
DOE: $5 w/protest bidders
IDE: $0 w/protest bidders
DOE: $50 w/o protest bidders
IOE: $25 w/o protest bidders
DOE: $5.98 w/protest bidders
IOE: $0 w/protest bidders
DOE: $59.75 w/o protest bidders
IOE: $29.88 w/o protest bidders
$29.26-$67.85
-Income (+)
-perceptions of increased safety
(+)
-age (-)
-concern for drinking water safety
(+)
-high priority placed on spending
for drinking water protection (+)
de Zoysa
$20.80 median
(1994$ from
YNP model)
$23. 48 median
(from YNP
model)
Not specified
-Income (+)
-high priority for
groundwater (+)
-increase gov't
spending on
education,
healthcare, and
vocational
training (+)
Edwards
n.a.
n.a.
n.a.
Bequest motivation (+)
income effect (+)
probability of future supply
(+)
probability of future demand
(+)
Giraldez and Fox
n.a.
n.a.
CVM: $29,938 -$669,487
per year for entire village
(higher estimate includes
option prices as well)
lifetime earnings/wage
risk: $693-$30,855
n.a.

-------
APPENDIX C > C-5
Study Reference
1. Published/Peer
Reviewed?
2. Year of Analysis
3. Place
4. Type of Study
5. Survey Implement
6. Respondents
7. Response Rate
8. Location (urban, rural,
etc.)
9. Who Was Asked?
10. Household Water
Supply /Ground-water Use
1 1 . Actual Groundwater
Baseline Condition
Hurley et al.
Journal of Agricultural and Applied
Economics
Apparently 1993
Clarke and Adams counties, IA
Survey eliciting WTP for delaying
water quality deterioration
Mail
1000 (500 to each county)
33.2%h
Rural — possibly some urban/rural
municipalities
Residents
75% use municipal or rural water
Not specified
Jordan and Elnagheeb
Water Resources Research
1991
Georgia (statewide sample)
Survey eliciting WTP for
improved water quality
Mail
567 mailed
35%
Unspecified mix of
community sizes
Residents
78% public
22% private wells
50% of wells contain
nitrates — did not specify %
exceeding the MCL — 27%
of public users rated water
quality poor, 13% of private
well users rated water
quality poor
Poe and Bishop
Environmental and Resource
Economics
1991
Portage County, WI
Survey eliciting WTP for
improved water quality
Mail
480 mailed
77.9% (ex-ante)
83% (ex post)
64.4% (2nd stage)
Rural
Residents not hooked up to
municipal water supply
100% on private wells
1 8% of wells had nitrate
contamination exceeding EPA
safety level -
16% of water tested > MCL
Sparco
PhD dissertation
1993
Sussex County, DE
Survey eliciting WTP for improved
water quality
booth at public gathering
3 occasions (# of respondents not
specified) (not a random sample)
Not specified
Predominantly rural
Passersby
61.9% of respondents use
individual wells; remainder use
municipal or community water
systems
N concentrations >10 mg/L in 23%
of samples(cited Andres 1991)
h. Doesn't indicate bad addresses; 44.7% returned of which 332 had usable data.

-------
APPENDIX C > C-6
Study Reference
12. Groundwater Baseline
Scenarios
1 3 . Change in Groundwater
Scenario
14. Credibility of Scenario
Change
15. Contaminants
16. Source of Contaminants
17. Types of Values
Estimated
18. Valuation Methodology
19. Payment Vehicle
Hurley et al.
Presumable currently safe
Delay N contamination in drinking
water for 10, 15, and 20 years,
assuming existing facilities would
result in contamination beyond legal
limits w/in 5 years.
Not assessed? No significant
difference in WTP over 10 to 20
years. High percent of zero WTPs.
Nitrates (from AFOs)
CAFOs (mostly hog)
Total value
Referendum (dichotomous choice)
Not specified?
Jordan and Elnagheeb
Not specified as individuals'
own water conditions —
baseline indicated as average
conditions over all of GA
(no individual probability of
>MCL specified)
Private wells: water supplier
provides new equipment, fee
includes installation and
maintenance public: water
supplier guarantees safe
drinking water for private
wells — specified N >MCL,
for public water specified N
increasing (not indicated
whether or not safe)
Examination of zero bidders
did not indicate any
significant scenario rejection
Nitrates
Agricultural activities
(fertilizers)
Total value (primarily use as
nitrate controls are at well
head not reductions in N in
the aquifer)
Close-ended payment card
("checklist")
Water bill for public users
costs for equipment to clean
nitrates from water for
private wells
Poe and Bishop
An increase in the number of
wells in Portage County with
nitrate contamination
Groundwater protection
program to reduce nitrates by
25% or to keep nitrate levels
below the MCL for all wells
in Portage County
Although vague, respondents
were told to suppose the
program was possible- the
survey was thoroughly
pretested
Nitrates
Agricultural activities and
other sources discussed in the
survey
Total value — option price
(use value)
Dichotomous choice,
referendum format
Higher taxes, lower profits,
increased costs and prices
Sparco
Not specified
WTP for a 1 part per million deer.
in N contamination
Not assessed
Nitrates, fecal coliform, atrazine
Agricultural activities (primarily
poultry manure from AFOs)
Marginal value
Conjoint analysis (contingent
rating)
Not specified

-------
APPENDIX C > C-7
Study Reference
20. Duration of Payment
Vehicle
2 1 . # of Survey Versions
22. Analysis
23. Mean Annual
Household WTP in Study
Year Dollars
24. Mean Annual
Household WTP in 2001
Dollars
25. Median Annual
Household WTP in Study
Year Dollars
26. Median Annual
Household WTP in 2001
Dollars
27. Range
28. Significant Explanatory
Variables
Hurley et al.
Annually
Not specified
Ordered probit
Not specified
Not specified
$118.13 (10 year delay) to $190. 75
(20 year delay) for household with
mean socio-economic characteristics
$118.13 (10 year delay) to $190. 75
(20 year delay) for household with
mean socio-economic characteristics
n.a.
Education (+)
likelihood that respondent will remain
in area longer than 5 yrs (+)
income (+)
Jordan and Elnagheeb
Monthly (in perpetuity)
One
Ordered probit
Public: $128.20/hh/yr
private: $157.61/hh/yr1
(primarily use)
Public:$166.70/hh/yr
private: $204.94/hh/yrJ
(primarily use)
Public: $69.89/hh/yr
private: $93.95/hh/yr
Public: $90.88/hh/yr
private: $122.1 6/hh/yr
$128.20 — $157.61/hh/yr
Income (+)1
gender (F+)
black (+)
education (+)
uncertainty (+)
live on farm (+)
Poe and Bishop
Annually, for as long as
respondent lives in the county
2
Logit
$199.73/hh/yrNINTk
$96 1.1 6/hh/yr WINT
$244.32/hh/yrNIWT
$526.63/hh/yr WIWT
$259.71 NINT
$1,249.79 WINT
$3 17.69 NIWT
$684.77 WIWT
$194.45/hh/yrNINT
$853. 46/hh/yr WINT
$242.58/hh/yrNIWT
$507.94/hh/yrWIWT
$252.84 NINT
$1,109.75 WINT
$3 15.43 NIWT
$660.47 WIWT
$199.73-$961.16/hh/yr
Knowledge (+)
quiz score (+)
Sparco
Annually, in perpetuity
8
Ordered probit
$123.56 per mg/L reduction in
nitrates
$ 1 5 1 .44 per mg/L reduction in
nitrates
n.a.
n.a.
n.a.
Pro-environment attitude (-)
cost (-)
health risks (-)
anti-government intervention (+)
pro-farm viewpoints (+)
i. Using unconditional mean values from maximum likelihood estimates after rejecting outliers.
j. Using unconditional mean values from maximum likelihood estimates after rejecting outliers.
k. NINT, WINT, NIWT, WIWT = No information-no test; with information-no test; no information-with test; with information-with test respectively.
1. Significant variables from maximum likelihood on private wells excluding outliers.

-------
APPENDIX C > C-8
Study Reference
1 . Published/ Peer Reviewed?
2. Year of Analysis
3. Place
4. Type of Study
5. Survey Implement
6. Respondents
7. Response Rate
8. Location (urban, rural, etc.)
9. Who Was Asked?
10. Household Water Supply/ Groundwater Use
1 1 . Actual Groundwater Baseline Condition
12. Groundwater Baseline Scenarios
1 3 . Change in Groundwater Scenario
14. Credibility of Scenario Change
15. Contaminants
16. Source of Contaminants
17. Types of Values Estimated
18. Valuation Methodology
19. Payment Vehicle
20. Duration of Payment Vehicle
2 1 . # of Survey Versions
22. Analysis
Walker and Hoehn
Northcentral Journal of Agricultural Economics
1983
Rural MI
Model for estimating N values based on
marginal cost of public treatment
n.a.
n.a.
n.a.
Rural
n.a.
>95% rural supply from GW
34% of 191 wells >10mg/L
Modeled specific scenarios
Modeled specific scenarios
n.a.
Nitrates
Agricultural activities
Damages (producer + consumer surplus) (use
values only)
n.a.
n.a.
n.a.
n.a.
Welfare theory
Wattage
PhD dissertation
1992?
Bear Creek watershed, central IA
Survey eliciting WTP for improved water quality
mail
345
40%
Predominantly rural
Farmers, absentee owners, town residents
50% private wells
43% municipal (also GW)
93% GW
Perceived: 16% ranked water quality as suitable for
human drinking purposes
Individuals' perceived water quality
Installing vegetative buffer strips (VBSs) to reduce
overland flow of contaminated water into GW & S W
supplies
32% of respondents strongly agree that VBS could
control N in the root zone- possibly significant
scenario rejection
Nitrates, pesticides; sediments
All runoff sources including: fertilizers, manure,
illegal wastes, gasoline
Total value
Dichotomous choice and open-ended. WTP and
WTA for various scenarios
Not specified
Monthly, as long as live in watershed
not specified
4 analyses: OLS, linear probability model, probit,
logit

-------
APPENDIX C > C-9
Study Reference
23. Mean Annual Household WTP in Study
Year Dollars
24. Mean Annual Household WTP in 2001
Dollars
25. Median Annual Household WTP in Study
Year Dollars
26. Median Annual Household WTP in 2001
Dollars
27. Range
28. Significant Explanatory Variables
Walker and Hoehn
n.a.
n.a.
n.a.
n.a.
$40-330/household/yrm
-Treatment location (point of use vs.
centralized)
-water consumption
-price of water
-damages and benefits per household
-household income
-nitrate contamination
Wattage
$80/month
$100.98/month
Not specified
Not specified
Not specified
Income (+)
distance from creek to land (+)
present GW quality (-)
m. $330/yr is based on annual cost of point-of-use nitrate removal.

-------
                                    APPENDIX D
           ASSESSMENT OF DATA USED TO ESTIMATE BENEFITS
The majority of the data EPA used to estimate the environmental and economic benefits
associated with the effluent guideline limitations for CAFOs are from existing sources. As
defined in the Office of Water 2002 Quality Management Plan (USEPA, 2002), existing (or
secondary) data are data that were not directly generated by EPA to support the decision at hand.
Existing data were used to model:

1.      reductions in leached nitrogen loadings resulting from new phosphorus-based and
       nitrogen-based manure application regulations which would apply to all large AFOs as
       well as any medium AFOs identified under new NPDES conditions,

2.      the reduction in nitrate concentrations in private drinking water wells, as a result of the
       new regulations, and

3.      the value of the reductions.

In keeping with the graded approach to quality management embodied in the quality management
plan, EPA must assess the quality of existing data relative to their intended use. The procedures
EPA used to assess existing data for use in estimating the benefits associated with effluent
guideline limitations for CAFOs varied with the specific type of data. In general, EPA's
assessment included:

>•      reviewing a description of the existing data that explains how the data were collected or
       produced (e.g., who collected the data, what data were collected; why were the data
       originally collected; when were the data collected; how were they collected; are the data
       part of a long-term collection  effort, or was this a one-time effort; who else uses the data;
       what level  of review by others have the data undergone?)

>•      specifying the intended use of the existing data relative to the CAFO final rule

>•      developing a rationale for accepting data from this source, either as a set of acceptance
       criteria, or as a narrative discussion

+      describing any known limitations with the data and their impact on EPA's use of the data.

Brief descriptions  of the data and their limitations are presented in Chapters 3 and 5 and
Appendices A and C, as each data source is introduced. In addition, Section 6.7 presents a

-------
                                    APPENDIX D > D-2
detailed accounting of known omissions, biases, and uncertainties in the analysis of the benefits
of reduced nitrate in private drinking water wells attributable to new CAFO regulations.

In searching for existing data sources and determining their acceptability, EPA generally used a
hierarchical approach designed to identify and utilize data with the broadest representation of the
industry sector of interest. EPA began by searching for national-level data from surveys and
studies by USDA and other federal agencies. When  survey or study data did not exist, EPA
considered other types of data from federal agencies.

Where national data did not exist, as the second tier, EPA searched for data from land grant
universities. Such data are often local or regional in  nature. EPA assessed the representativeness
of the data relative to a national scale before deciding to use the data. When such data came from
published sources, EPA gave greater consideration to publications in peer-reviewed professional
journals compared to trade publications that do not have a formal review process.

The third tier was data supplied by industry. Prior to proposal, EPA requested data from a variety
of industry sources, including trade associations and large producers. The level of review applied
to data supplied by industry depended on the level of supporting detail  that was provided. For
example, if the industry supplied background information regarding how the data were collected,
such as the number of respondents and the total number of potential respondents, EPA reviewed
the results, compared them to data from other potential sources to determine their suitably for use
in this rulemaking. If the data provided by industry originated from an identifiable non-industry
source (e.g., a state government agency), EPA reviewed the original  source before determining
the acceptability of the  data. In a limited number of instances, EPA conducted site visits to
substantiate information supplied by industry. In contrast, data supplied by industry without any
background information were given much less weight and generally were not used by EPA.
Further, some data that were supplied by industry  prior to the proposal  were included in the
proposal for comment. In the absence of any negative comments, such  data were relied on to  a
greater extent than data submitted by industry during the comment period itself.
REFERENCES

U.S. EPA. 2002. Office of Water Quality Management Plan. April 2002. EPA 821-X-02-001.

-------