ECOSYSTEM SERVICES FROM LOW INPUT CROPPING SYSTEMS AND THE
PUBLIC’S WILLINGNESS TO PAY FOR THEM
By
Huilan Chen

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
MASTER OF SCIENCE
Agricultural, Food and Resource Economics
2010

ABSTRACT
ECOSYSTEM SERVICES FROM LOW INPUT CROPPING SYSTEMS AND THE
PUBLIC’S WILLINGNESS TO PAY FOR THEM
By
Huilan Chen
This thesis intends to value off farm ecosystem services from low input cropping systems by
estimating the public’s willingness to pay for them. Chapter one summaries what kinds of off
farm ecosystem services will increase by adopting low input cropping systems and how much
they will increase. It quantifies field experiment results on ecosystem services at the state
level under widespread adoption of low input cropping systems and identifies a set of
ecosystem service increases that are thought to significantly affect the general public. Five
ecosystem service changes including reductions in drain dredging,

flood damage, high

nitrate drinking water wells, eutrophic lakes, and greenhouse gas emission were identified as
having a significant impact on the general public. Chapter one lays the scientific base for the
services valued in chapter two.
Chapter two focuses on the economic valuation of ecosystem service changes. Two
ecosystem service changes, eutrophic lakes reduction and greenhouse gas emission reduction
are selected. The contingent valuation survey is applied. A mail survey was sent to 6000
randomly selected Michigan residents. Results show that Michigan residents are willing to
pay for a decrease in eutrophic lakes, but only the 40% who are concerned about global
warming are willing to pay for a decrease in greenhouse gas emission. Another finding is that
people’s probability of having zero willingness to pay is endogenous. It is sensitive to the
scope of ecosystem service changes and some demographic variables such as household
income and attitudes towards global warming. The implication is that studies that do not
allow the share of people with zero willingness to pay to adjust to the magnitude of
ecosystem service changes will undervalue increases in ecosystem services.

Copyright by
Huilan Chen
2010

ACKNOWLEDGEMENTS

The completion of this thesis is the sweetest fruit on my long road to my master’s
degree in the Department of Agricultural, Food and Resource Economics. I’d like to share it
with my family, my professors and my friends who have helped me and supported me along
the road.
First and foremost, I’d like to thank my major professor Dr. Frank Lupi for his guidance
in my study and research. I cannot imagine how I would have designed and managed a large
mail survey without his careful guidance. I would not have my work presented at academic
conference without countless meetings, discussions and working hours with him. I’m deeply
impressed by his academic expertise, intelligence and devotion to guiding students. I’m also
grateful to his support for my decision to pursue a career in other fields.
I’d like to also thank members of my thesis committee, Dr. Scott M. Swinton and Dr. G.
Phillip Robertson. They are top researchers with strong expertise in their field as well as nice,
considerate and responsible teachers.
I am also grateful to the support of Kellogg Biological Station’s Long-Term Ecological
Research and the members of the National Science Foundation’s Human and Social
Dynamics program Grant No. 0527587 for their support in my graduate study. I would
especially like to thank Sara P. Syswerda for providing me important field experiment data.
Many thanks are also given to the faculty and staff at the department who have provided
lots of help in arranging my survey and other academic issues.
Also thanks to my friends who gave me support in my study and life, especially Min
Chen, Shan Ma, Feng Song, Binlei Gong, Yuan Shao, Xiaobo Li, Christine Jolejole, Susan
Chen, Jake Richer-Gilbert and Tim Komarek for their help in my survey. Thanks to many
other friends and neighbors who shared the ups and downs of my life at MSU. The joy and

iv

memories you brought me brightened the gloomy sky of Michigan’s winter.
I owe thanks to undergraduate students Tim Nance, Derek Kuhn, Brendan Lammers and
Jeremy Rapp for helping me code the questionnaire and conduct interviews. I also appreciate
Dr. John Loomis who sent me his contingent valuation questionnaire and Dr. Sieglinde Snapp
who provided me a lot of information on low input cropping systems.
Last but not least, I’d like to thank my family. They are the backbones of my life. It is
all because of their love and support that I’m able to navigate every exciting journey of my
life.

v

TABLE OF CONTENTS

LIST OF TABLES................................................................................................................. ix
LIST OF FIGURES ................................................................................................................xi
LIST OF ABBREVIATIONS ................................................................................................xii
INTRODUCTION....................................................................................................................1

CHAPTER 1 OFF-FARM ECOSYSTEM SERVICES FROM SELECTED
LOW-INPUT CROPPING SYSTEMS: SCALING UP FROM FIELD EXPERIMENT
RESULTS TO STATE-WIDE IMPACTS IN MICHIGAN..................................................3
Abstract.....................................................................................................................................3
1. Introduction..........................................................................................................................4
2 Off-farm ecosystem services by changing farming practices............................................7
2.1 Soil erosion related ecosystem services ...........................................................................7
2.2 Surface freshwater related ecosystem services ................................................................7
2.3 Nitric oxide, nitrogen dioxide and nitrous oxide emissions .............................................8
2.4 Ground water related ecosystem services ........................................................................9
2.5 Global warming ................................................................................................................9
2.6 Off-farm pesticide pollution ...........................................................................................11
2.7 Human health .................................................................................................................11
2.8 Animal and insect population .........................................................................................12
3 Quantifying ecosystem services: possible mapping of functions from cropping systems
to off-farm ecosystem service changes .................................................................................12
3.1 Environmental quality functions and ecosystem service functions: a conceptual
framework for quantification................................................................................................12
3.2 Estimation methods for environmental quality functions and ecosystem service
functions ...............................................................................................................................15
3.3 Specification of the problem ..........................................................................................16
3.3.1 Selected low-input cropping systems in the LTER at Michigan State University...16
3.3.2 Linkages and ecosystem service changes we want to measure ...............................19
3.3.3 Specification of environmental quality functions and ecosystem functions............22
4 Function estimation and results.........................................................................................24
4.1 Soil erosion and related ecosystem services...................................................................24
4.1.1 Soil erosion estimates ..............................................................................................24
4.1.2 Drain dredging cost..................................................................................................26
4.1.3 Flood damage cost ...................................................................................................28
4.2 Phosphorus and eutrophic lakes .....................................................................................29
4.2.1 Phosphorus runoff function......................................................................................29
4.2.2 Reduction in eutrophic lakes....................................................................................29
4.3 Nitrate and drinking water quality..................................................................................33
4.3.1 Nitrate leaching reduction........................................................................................33
4.3.2 Drinking water quality .............................................................................................35
4.4 Greenhouse gas emission and global warming ..............................................................39
vi

4.4.1 Greenhouse gas emission.........................................................................................39
4.4.2 Scale up global warming effect................................................................................39
5 Discussions ...........................................................................................................................40
Appendix 1.1 Calculating the percentage of soil erosion from corn-soybean rotation land to
soil erosion from total cropland in Michigan .......................................................................46
Appendix 1.2 Calculation of soil erosion reduction rates of different cropping systems ....48
References ...............................................................................................................................51

CHAPTER 2 MICHIGAN PUBLIC’S WILLINGNESS TO PAY FOR REDUCTIONS
IN EUTROPHIC LAKES AND GREENHOUSE GAS DUE TO CHANGES IN LAND
USES: A CONTINGENT VALUATION STUDY ...............................................................59
Abstract...................................................................................................................................59
1 Introduction.........................................................................................................................60
1.1 Agriculture as managed ecosystems...............................................................................60
1.2 Rationale for the study ...................................................................................................61
1.2.1 Policy and research gaps and challenges of the topic ..............................................61
1.2.2 Objective of the study ..............................................................................................64
1.3 Layout of the chapter......................................................................................................64
2 Current research gaps ........................................................................................................65
2.1 Gaps in valuing ecosystem services from agriculture ....................................................65
2.2 Gaps in spike models in contingent valuation................................................................68
3 Contingent valuation methodologies in valuing ecosystem services...............................68
4 Theoretical model................................................................................................................73
4.1 WTP conditional on WTP>0 ..........................................................................................75
4.2 Spike probability ............................................................................................................77
4.3 Welfare calculation .........................................................................................................77
4.4 Econometric model: random effects probit ....................................................................79
5 Model specification: Description of explanatory variables and hypotheses..................81
6 Survey design and implementation ...................................................................................82
6.1 Survey design .................................................................................................................82
6.1.1 Survey attributes and values ....................................................................................82
6.1.2 Hypothetical WTP questions design ........................................................................88
6.1.3 Experimental design of attributes ............................................................................89
6.1.4 Validation questions design......................................................................................90
6.2 Survey pretest .................................................................................................................91
6.3 Survey implementation...................................................................................................92
7 Data analysis........................................................................................................................93
7.1 Protest votes, misunderstanding and other misleading votes .........................................93
7.2 Descriptive statistics of explanatory variables ...............................................................94
7.3 Conditional WTP model.................................................................................................99
7.4 Spike probability model ...............................................................................................101
7.5 Comparison of the fixed spike model and the endogenous spike model in the
probability of voting yes and WTP estimates.....................................................................106
8 Conclusions........................................................................................................................ 113
Appendix 2.1 Responses in survey period (by week) ........................................................116
Appendix 2.2 Protests, disbelief and misunderstanding votes ........................................... 117
Appendix 2.3 Content analysis process..............................................................................118
Appendix 2.4 Themes developed from comments .............................................................120
Appendix 2.5 Survey return statistics.................................................................................127
vii

Appendix 2.6 Questionnaire...............................................................................................128
Appendix 2.7 Prenotice letter.............................................................................................140
Appendix 2.8 First round cover letter ................................................................................141
Appendix 2.9 Second round cover letter ............................................................................142
Appendix 2.10 First round reminder postcard....................................................................143
Appendix 2.11 Second round reminder postcard ...............................................................144
Appendix 2.12 Steps to generate experimental design.......................................................145
Appendix 2.13 Experimental design plan ..........................................................................146
References .............................................................................................................................149

viii

LIST OF TABLES

Table 1.1 cropping systems from the 2008 farmer’s willingness to accept payment for the
adoption of low input cropping systems survey .....................................................................18
Table 1.2 Soil erosion reduction rate of different cropping systems .......................................25
Table 1.3 Soil erosion reduction rates of different cropping systems ......................................26
Table 1.4 Reduced dredging cost of different cropping systems .............................................27
Table 1.5 Reduced flooding damage cost of different cropping systems ................................29
Table 1.6 Percentage reductions in total phosphorus concentration of different cropping
systems....................................................................................................................................32
Table 1.7 Reduced eutrophic inland lakes of different cropping systems ...............................32
Table 1.8 Nitrate leaching reduction of low input cropping systems in the KBS LTER
experiments (Syswerda et al, 2010)........................................................................................34
Table 1.9 Nitrate leaching reduction rate in different cropping systems in KBS research ......35
Table 1.10 Percentage reduction in nitrate concentration in different cropping systems ........37
Table 1.11 Reduction in unsafe wells in different cropping systems.......................................38
Table 1.12 Global warming potential reduction of different cropping systems compared with
conventional practices ............................................................................................................39
Table 1.13 Global warming effect of different cropping systems............................................40
Table 1.14 Summary of estimates of five ecosystem services changes of low input cropping
systems....................................................................................................................................44
Table 1.15 Calculating weighted relative soil erosion rate of different crops .........................46
Table 2.1 Explanatory variables, definitions, unit of measure and levels of survey attributes,
survey of 6000 Michigan households, autumn 2009..............................................................85
Table 2.2 Definition, observation, mean, standard deviation, minimum and maximum values
of variables .............................................................................................................................96
Table 2.3 Estimation results of proposed models ..................................................................104
Table 2.4 Protests, disbelief and misunderstanding votes......................................................117
ix

Table 2.5 Themes names, descriptions and example quotes..................................................120
Table 2.6 Survey return categories and quantities .................................................................127
Table 2.7 Attributes and levels of the contingent valuation design .......................................146
Table 2.8 Experiment design plans for each version of the questionnaire.............................146

x

LIST OF FIGURES

Figure 1.1 Mapping linkages from agricultural management factor to off-farm ecosystem
service changes .......................................................................................................................20
Figure 1.2 Distribution change of eutrophic lakes after a uniform phosphorus reduction ......30
Figure 1.3 Distribution changes of nitrate wells after a uniform nitrate reduction..................36
Figure 2.1 Frequencies of votes by program costs ..................................................................96
Figure 2.2 Probability of voting yes as a function of environmental change variables.........109
Figure 2.3 Unconditional mean WTP estimates as a function of environmental change
variables................................................................................................................................ 111
Figure 2.4 Responses in survey period (by week) .................................................................116
Figure 2.5 Protests, disbelief and misunderstanding votes as a percentage of total vote ......117

xi

LIST OF ABBREVIATIONS
KBS

Kellogg Biological Station

CSW

Corn-Soybean-Wheat Rotation

CS

Corn-Soybean Rotation

LTER

Long Term Ecological Research

WTP

Willingness to Pay

NASS

National Agricultural Statistics Service

USDA

United States Department of Agriculture

PES

Payment for Ecosystem Services

CV

Contingent Valuation

ABM

Attribute Based Method

NO

Nitric oxide

NO2

Nitrogen dioxide

N2O

Nitrous oxide

NOx

Nitrogen oxides

CH4

Methane

CO2

Carbon dioxide

GHG

Greenhouse gas

RUSLE

Revised Universal Soil Loss Equation

PSNT

Pre-sidedressing nitrate test

i.i.d.

Identically and independently distributed

EPA

Environmental Protection Agency

MCL

Maximum Contaminant level

TP

Total phosphorus

MDEQ

Michigan Department of Environmental Quality

xii

INTRODUCTION
Agriculture is one of the most important managed ecosystems. Management decisions
affect agricultural production outputs as well as off-farm ecosystem services. These off-farm
ecosystem services are often public goods that affect the public’s well-being. The public has
certain demands for off-farm ecosystem services but there are not generally markets to
provide them. “Payments for Environmental Services” (PES) programs are often used to
provide incentive to farmers to supply off-farm ecosystem services. Current PES research
often focuses on the supply-side, measuring costs to land owners for changing land
management practices, but less attention has been paid to the demand side.
This thesis intends to investigate the demand side of these ecosystem services. Three
research questions are investigated. First, what are these ecosystem services? Second, are
they significant enough to have much impact on the general public? Third, if some of them
significantly benefit the general public, what are people willing to pay to get these ecosystem
services? The three research questions construct the basic structure of the thesis. By
answering these three questions, a demand side picture of ecosystem service from agriculture
is drawn.
In Chapter one, I try to answer the first and second question. It quantifies field
experiment results on ecosystem services from low input cropping systems at the state level.
By doing this, it identifies a set of ecosystem service changes that are thought to significantly
affect the general public. Further, it measures how much these ecosystem services might
change at the state level under widespread adoption of low input cropping systems. Chapter
one lays the scientific base for the services valued in chapter two. First, it describes
quantitatively the goods to be measured in chapter two. Second, it establishes the possible
range of the goods to be valued in the contingent valuation survey in chapter two.
Chapter two focuses on the third research question. From chapter one, two ecosystem
1

services are selected to be valued in this chapter. One is reduction in eutrophic lakes and the
other is reduction in greenhouse gas emission. The contingent valuation method is applied to
value the two ecosystem service changes. People’s willingness to pay is estimated as a
function of ecosystem service changes. The demand curves for the ecosystem services can be
derived from this function.

2

CHAPTER 1 OFF-FARM ECOSYSTEM SERVICES FROM SELECTED
LOW-INPUT CROPPING SYSTEMS: SCALING UP FROM FIELD EXPERIMENT
RESULTS TO STATE-WIDE IMPACTS IN MICHIGAN

Abstract
Chapter 1 attempts

to answer the research questions of which ecosystem services are

generated from low input cropping systems, how large they would possibly be if the systems
were adopted statewide, and whether they have a major effect on the general public in
Michigan. This chapter lays the scientific basis of the types and quantities of ecosystem
services to be valued in the next chapter.
This chapter first reviews scientific literature on the ecosystem services related to
agricultural management practices and sorts out how agricultural management decisions
would affect intermediate environmental changes such as soil erosion, nitrogen and
phosphorus loss, greenhouse gas emission etc. These intermediate environmental changes
will cause a set of off-farm ecosystem service changes. A mapping of linkages that goes from
agricultural management practices to off-farm ecosystem service changes is set up. Based on
the mapping, five final ecosystem service changes are identified and quantified at the state
level. They include some of the supporting services and regulating services such as reduced
drain dredging, reduced flood damage, reduction in eutrophic lakes, reduction in high nitrate
drinking water wells, and reduced greenhouse gas emission.
The estimated maximum cost reduction in dredging from wide-spread adoption of low
input cropping systems is about 1 million dollars per year, and the estimated cost reduction in
flood damage is about 1.4 million dollars per year. By adopting low input cropping systems,
it is estimated that Michigan may see a reduction of about two hundred eutrophic lakes and
one thousand high nitrate wells that are above EPA’s Maximum Contaminant level at the
maximum. The maximum reduction of greenhouse gas from low input cropping systems is

3

estimated to be about 800 million kilograms each year. Finally, after thoroughly considering
all of these services, their scale, and the other available valuation methods, we need to decide
which ecosystem services to value in chapter 2.

1. Introduction
Agriculture has important ecosystem services functions including providing food, fiber
and fuel. These are provisioning ecosystem services, as categorized by the Millennium
Ecosystem Assessment(Millennium Ecosystem Assessment 2005). These ecosystem services
have explicit prices because agricultural products are traded in the market. Farmers are able
to capture the full economic value of them. These ecosystem services can be considered
private goods. On the other hand, agriculture provides other off-farm supporting and
regulating ecosystem services or disservices to soil, water and the atmosphere. For example,
soil erosion rate on cultivated lands in major agricultural regions usually ranges from 10 to
100 t/ha/year, which is at least 10 times of the soil formation rate (Pimentel et al 1987). Land
degradation has harmed soil productivity and increased offsite environmental effects such as
drains and reservoir damage, flooding etc, (Pimentel et al 1987) Excessive nitrogen runoff
from agriculture has possible negative environmental impacts such as coastal hypoxia, nitrous
oxide and nitrogen dioxide emissions into the atmosphere, and increased nitrate concentration
in groundwater (Robertson and Vitousek 2009). These ecosystem services or disservices from
agriculture are affecting both the general public and the farming community. People are not
able to prevent others from getting them, and the consumption of these goods from some
people does not affect the consumption from others. Therefore, most of the off-farm
ecosystem services can be considered as public goods.
Certain kinds of farming practice changes are able to increase off-farm ecosystem
services. The long term ecological research (LTER) of row crop agriculture at the Kellogg

4

Biological Station has shown that several low input cropping systems are able to provide
offsite ecosystem services such as better carbon sequestration, less nitrate leaching to waters,
less phosphorus runoff from soil (Robertson and Vitousek 2009, Daroub et al 2001, Syswerda
et al 2010). Including cover crop in a rotation and organic farming have better water retention
in soil, reduce soil erosion and preserve soil organic matter that sustains crop growth
(Reganold et al 1987). Other scientific literature recognizes the positive environmental effect
of specific cropping systems from field experiments (Borin et al 1997, Brye et al 2002).
On the other hand, field experiment results on the ecosystem increases from low input
cropping systems are not sufficiently informative in the policy context. In the perspective of
decision makers, if the changes to low input cropping systems can bring are significant at the
state or national scale, then it is meaningful to provide economic incentives to adopt them.
Otherwise any policies that encourage them will not make significant changes so they are not
worth implementing. Field experiments do not give decision makers such information
because ecosystem service changes are measured at the field scale instead of at a state or
national scope. In order to give more insights on the ecosystem service changes at a wider
scope, it is necessary to link field experiment results to biophysical simulation models that
can simulate effects over large areas and estimate this scale up effects of those field
experiments at the regional scale.
The scale-up effect study is insightful in that it estimates a portfolio of multiple
ecosystem service changes from various cropping systems. Decision makers are able to
decide which ecosystem service changes they value most and then allocate more public
money on their supply. Current public payment programs are generally compensating farmers
for their direct cost of adopting environmentally friendly farming practices. They pay farmers
according to their cost and the number of acres they participate in the program. These
payment programs fail to induce farmers to supply what the public want most. (Smith 2006)

5

Moreover, these public payment programs do not consider the marginal benefits of
different ecosystem services in the supply portfolio so they do not maximize the supply of the
most valuable ecosystem services under certain budget constraints. As a result, they are not
efficient (Smith, 2006). To design an economically efficient incentive to increase the supply
of ecosystem services from agriculture, the marginal benefit of increasing ecosystem services
needs to be estimated (Costanza et al 1998, Howarth and Farber, 2002).
In order to get an accurate and consistent estimate of the economic value of the
ecosystem service bundle, a holistic approach is necessary in valuing multi-dimensional
environmental goods (Randall 2002). Adding up piece-wise value estimates of each
dimension of the good will overestimate the holistic value because the method ignores the
budget constraint and substitution effect in consumer’s valuation process (Hoehn and Randall
1989). Therefore, studies on the aggregate effect on ecosystem service changes are very
important in deciding what the multi-dimensional environmental good is, which is a
fundamental step in the holistic valuation of the good (Loomis et al 2000).
This chapter will focus on estimating the scale-up effect of ecosystem service changes
from low input cropping systems. By imposing reasonable assumptions in the calculation, the
paper will expand field study results to state wide environmental changes in Michigan and
translate them into ecosystem service changes that can be perceived by the public. The
second part of the paper will give an overview of how farming practices contribute to several
ecosystem service changes. The third part will set up a map of linkages from cropping
systems to perceivable ecosystem service changes and identify the target cropping systems
and ecosystem service changes we want to measure. The fourth part will describe the general
method to quantify ecosystem service changes. The fifth part will present detail calculations
and discuss the concluding results and the limitation of the estimates.

6

2 Off-farm ecosystem services by changing farming practices

2.1 Soil erosion related ecosystem services
Tillage method plays a key role in reducing soil erosion. No tillage, ridge tillage and
other conservation tillage methods can reduce soil erosion significantly (FAO 2003).
Conservation tillage is usually better at holding back moisture in soil and retains more
organic matter in soil by keeping plant and crop residuals in soil longer. The added organic
matter increases the soil’s ability to retain water and help reduce soil erosion (Reganold et al
1987, NRCS 2009).
Another farming practice that helps preserve soil is the use of cover crops. Cover crops
such as clover, alfalfa and legumes improve the water-holding capacity of the root zone in the
field, making it better at retaining water in soil. They also increase soil water infiltration
(Joyce et al 2002).
Reduced soil erosion has on-site ecosystem services such as improving soil fertility and
increasing crop production (Pimentel et al 1987, Pimentel and Kounang, 1998). These onsite
ecosystem services will not be discussed in the paper because they are benefiting farmers
instead of the general public. They can be considered as private goods. Offsite ecosystem
services from reduced soil erosion come from less sediment damage, which include reduced
earth-dam failures, less siltation of harbors and channels, increased reservoir storage,
preservation of wildlife habitat, less disruption of stream ecology, decreased flooding and
cost of water treatment. (Gray and Leiser, 1989, Clark 1985).

2.2 Surface freshwater related ecosystem services
Most phosphorus compounds are not volatile and basically stored in solid forms such as
earth, dust and human bodies. Most of them are neither soluble nor stable in water. Therefore,

7

phosphorus loss from agricultural land to the waters is mainly in the form of soil erosion
(Correll 1998, Hutciiinson 1957). As a result, soil erosion is closely related to phosphorus
loading from soil to the surface freshwater ecosystems and any farming practices that
influence soil erosion on land, such as tillage method and cover crops, also have an impact on
phosphorus runoff (Poudel et al 2001, Borin et al 1997).
Phosphorus is the limiting factor of the trophic status of freshwater ecosystems in the
US (Correll 1998). As phosphorus fertilizer input to the soil exceeds agricultural production
outputs, the excess phosphorus fertilizer finally goes to aquatic ecosystems and causes
eutrophication in some of the aquatic environment (Carpenter et al 1998). Now phosphorus
from fertilizer runoff is the key factor of the trophic status of lakes (Carpenter et al 1998,
Robertson 1996, Sharpley et al 2001). Eutrophication of waters affects habitat for fish and
other aquatic lives and recreational activities in those freshwater ecosystems.

2.3 Nitric oxide, nitrogen dioxide and nitrous oxide emissions
Nitric oxide, nitrogen dioxide and nitrous oxide from soil come from biological
processes of nitrification, denitrification and chemodenitrification. Better nitrogen storage
helps reduce agricultural emission of NO, NO2 and N2O.
Nitrogen fertilizers contribute to NO, NO2 and N2O increase (Bouwman 1996) because
inorganic nitrogen fertilizer increases the process of nitrification and denitrification.
Agricultural management practices like fertilizer type, fertilization/irrigation method,
harvesting, burning and tillage, land use change and spatial patterns of agricultural
management have influences on NO, NO2 and N2O emission from soil (Hall et al 2002).
Tillage is considered an important factor in nitrogen retention. Tillage method significantly
influences nitrogen conservation by slowing the biological decomposition of nitrogen from
soil organic matter. The slow-down of the process can improve nitrogen mineralization and

8

immobilization (Bayer et al 2000, Angers et al 1997, Halvorson et al 2002) and reduce NOx
emissions (Venterea et al 2005). Cover crops are beneficial in nitrogen fixation (Drinkwater
et al 1998) but field experiments have shown mixed results on its effect on N2O emission.
Some have shown that cover crops increase N2O emission (Bouwman 1996) but others have
shown cover crops reduced N2O emission during thaw (Wagner-Riddle and Thurtell 1998).
Biogenic nitric oxide and nitrogen dioxide (NOx) from soil has proven to have
significant contribution to regional air pollution. Williams et al (1992) estimate that in
summer months, NOx emissions from soil can contribute to 14% of combustion emissions in
the US; 66% of these NOx emissions come from fertilized agriculture. Nitrous oxide is an
important kind of greenhouse gas that has 310 times of the global warming potential of
carbon dioxide (Forster et al 2007).

2.4 Ground water related ecosystem services
Tillage methods, fertilizer use and fertilization methods that reduce NOx emission also
reduce nitrate leaching into ground water. Cover crops help in fixing nitrogen in soil and
reduce nitrogen loss to groundwater (Poudel et al 2001, Borin et al 1997). The
pre-sidedressing nitrate test (PSNT) is a useful tool for farmers to apply moderate but not
excessive amount of nitrogen fertilizer and helps reduce excessive nitrogen loss.
Extensive use of nitrogen fertilizer in agriculture has contributed to the ground water
contamination. Nitrate pollution often occurs in agricultural irrigation areas where fertilizer
input is heavy and nitrate leaching rate is high (Power and Schepers 1989). Nitrogen from
crop fertilizer is one of the major sources of nitrate in groundwater (MDEQ 2003).

2.5 Global warming
Net global warming potential from agricultural systems is the sum effect of soil organic

9

carbon storage, methane emission and oxidation and nitrous oxide emissions (Mosier et al
2005). Carbon sequestration is the process of removing carbon from the atmosphere and
depositing it in a reservoir (UNFCCC 2010). Several farming practices have effects either on
greenhouse gas emissions, carbon sequestration or nitrogen retention. First, reduced nitrogen
fertilizer input is helpful in reducing nitrous oxide, nitrogen dioxide and carbon dioxide. It is
also an effective way to increase methane oxidation (Palm et al 2002). It is a farming practice
that mitigates global warming.
Conservation tillage is one of the important management practices to enhance carbon
sequestration and nitrogen retention. In conventional tillage, crop residuals are more easily
decomposed by microorganisms and release more carbon dioxide (Reicosky and Lindstrom
1993). Conservation tillage leaves 30% or more crop residues covering the soil (Soil Science
of America 2009), especially in no tillage case, crop residuals are decomposed more slowly
so that carbon and nitrogen are largely in the form of soil organic matter in soil (Bayer et al
2000, Dick et al 1998). Conventional tillage cropping system can emit eight times more of
net global warming potential than the no-till system (Robertson et al 2000). Data from global
field experiments show that on average, a switch from conventional tillage to no-till can
sequester 57 ± 14 grams of carbon per square meter per year (West and Post 2002).
Some field experiments show that cover crop helps increase carbon pool in soil (Lal et
al 2004, Reicosky et al 1995). However, some study indicates that the benefits of cover crops
on increased carbon sequestration can be offset by frequent tillage (Sainju et al 2002) while
another proves the other way around (Robertson et al 2000). The effect of combining cover
crop and frequent tillage is not clear.
Rotation complexity also enhances carbon sequestration. According to West and Post’s
estimate (2002), enhancing rotation complexity can sequester an average 8- 32 grams of
carbon per square meter per year, excluding a change from continuous corn to corn-soybean

10

rotation which may not result in a significant accumulation of soil organic matter.

2.6 Off-farm pesticide pollution
The rate and extent of the emission during pesticide application, predominantly as spray
particle drift, depends primarily on the application method such as equipment and technique,
the formulation and environmental conditions. The emission after application depends
primarily on the properties of the pesticide, soils, crops and environmental conditions such as
temperature, moisture, wind speed etc (Van den Berg et al 1999, Scholtz et al 2002).
Pesticide emission is a major source of onsite air pollution but it sometimes also has
offsite effects. Some pesticides have been shown to be transported through the air and
decomposed in aquatic environment and become a regional environmental problem (Glotfelty
et al 1987, Voldner and Schrueder 1990, Scholtz et al 2002, Majewski et al 1998, Baker et al
1996). That pesticide can get several thousand fold concentrated in fog droplets may cause
possible health problems (Glotfelty et al 1987).

2.7 Human health
High nitrate exposure in drinking water may cause Methemoglobiemia, commonly
known as blue baby syndrome, a blood disorder that impairs the ability of hemoglobin to
transport oxygen and carbon dioxide, leading to tissue hypoxemia and even death (Wright et
al 1999). It mainly affects infants under six months old (MDEQ 2004, Walton 1951).
Possible negative effects on human health of nitrate include reproductive and
developmental defects in offspring. Health studies have shown some signs of high risks but
the association is not sufficiently determined (Fan and Steinberg, 1996).
Health studies on pesticide have confirmed some negative health effects. Direct
exposure of organophosphates is associated with acute health problems such as nausea,

11

dizziness, vomiting, headaches, abdominal pain, skin and eye problems and chronic health
problems such as respiratory problems, memory disorders, dermatologic conditions,
childhood cancer, depression, neurologic deficits, miscarriages and birth defects (Ecobichon
1996, Zahm and Ward 1998, Daniels et al 1997). Direct exposure of farmers and indirect
exposure of their children are much more risky compared with the offsite public (Eskenazi et
al 1999, Coye 1986, McCauley et al 2006, Baker et al 1996, Fenske et al 2002, Fenske et al
2000).

2.8 Animal and insect population
Animal experiment results have shown that the reproduction and offspring development
are seriously hampered at very high nitrate level while no significant effect has been observed
at nitrate-nitrite concentration below 500 ppm (Fan et al 1987)
Acute exposure to pesticide may cause reduced activity, weight loss and physical
abnormalities (Hecnar 1995). Repeated exposure to low concentration of nitrate has negative
effect on the offspring survivorship, animal growth and neurodevelopment (Ekenazi et al
1999, Hecnar 1995).

3 Quantifying ecosystem services: possible mapping of functions from cropping systems to
off-farm ecosystem service changes
3.1 Environmental quality functions and ecosystem service functions: a conceptual
framework for quantification
Scientific literature measure the exact quantity of environmental changes caused by
farming practice changes. For example, they measure how many tons of nitrogen would leach
into ground water by adopting different cropping systems (Poudel et al 2001, Borin et al
1997), how many tons of carbon and nitrogen would be fixed in soil or emit into the
12

atmosphere if conservation tillage method was adopted (Robertson et al 2000, Reicosky and
Lindstrom 1993, Bayer et al 2000, Dick et al 1998). However, one environmental change
may provide one or more ecosystem services as a joint product (Fisher et al 2009). For
example, soil erosion reduction influences both sediment and phosphorus loading to waters.
Sediment reduction improves drainage siltation, local flooding, reservoir damage etc (Clark
1985). Less phosphorus loading reduces nutrients in lakes. One environmental quality change
may lead to several ecosystem service changes. Therefore, we need to understand how one
environmental quality change would be mapped into several ecosystem service changes then
we should estimate each ecosystem service change separately.
To map from agricultural practice changes to ecosystem service changes, we should go
through two steps: first, we need to know the quantitative linkages from cropping systems to
environmental quality changes, Second, we need to find out linkages from each
environmental quality change to relevant ecosystem services changes. The two sets of
quantitative linkages will be combined together and give a full picture of how much cropping
systems would change the “intermediates” (environmental qualities) and then how much the
“intermediates” would change the “end products” (ecosystem services).
In the first step, let us consider environmental quality as a function of a set of variables
(this set might include infinite number of variables). Environmental quality is a vector of n
dimensions which is denoted by

arbitrary functions

(q1 , q 2 , ..., q n )

q1 (X) , q 2 (X) ,

…,

. Then we can use a set of

q n (X)

to denote the general

environmental function where X is the vector of independent variables that influence
environmental quality. Agricultural practice factor vector C can be considered a subset of X
that is independent from other subset Y such as climate, industry, policy etc. In most cases,
environmental quality variables will influence one another. For example, soil erosion will

13

affect phosphorus loading from land to waters because phosphorus is usually in the form of
soil solid matters. Therefore, other environmental quality should also be included into the
functions. Then if C changes from

C0

then environmental quality change

(Δq1 , Δq 2 , ..., Δq n )

to

C1

while other variables are held as constant,
can be expressed as

Δq1= q1 (C1, q21, q31, …, qn1| Y)- q1(C0, q20, q30, …, qn0| Y) (1)
Δq2= q2 (C1, q11, q31, …, qn1| Y)- q2(C0, q10, q30, …, qn0| Y)

(2)

…
Δqn= qn (C1, q11, q21, …, qn-11|Y)- qn(C0, q10, q20, …, qn-10| Y) (3)
In the second step, we consider the multidimensional ecosystem services as a vector of
m dimensions (es1,
vector (q1,

es2,…, esm).

Then it is a set of functions of environmental quality

q2, …, qn) and other variable vector Z. Moreover, it is possible that ecosystem

services influence one another. So the ecosystem service functions are expressed as

es1= es1(q1, q2, …, qn, es2, es3 ,…, esm , Z)

(4)

es2= es2(q1, q2, …, qn, es1, es3 ,…, esm , Z)

(5)

…
esm= esm(q1, q2, …, qn, es1, es2 ,…, esm-1 , Z)

(6)

Vector Z contains all other exogenous variables that influence a particular ecosystem
service. Z can be climate, related ecological, geological, physical processes, economic and
policy factors etc. Suppose Z is holding constant, then a change in environmental quality
from (q1

0

, q20, …, qn0) to (q11, q21, …, qn1) by (Δq1, Δq2, …, Δqn) will cause

ecosystem service changes by
14

Δes1= es1(q11, q21, …, qn1, es21, es31 ,…, esm1 |Z)- es1(q10, q20, …,
qn0, es20, es30 ,…, esm0 |Z)

(7)

Δes2= es2(q11, q21, …, qn1, es11, es31 ,…, esm1|Z)- es2(q10, q20, …,
qn0, es10, es30 ,…, esm0 |Z)

(8)

…
Δesm= esm(q11, q21, …, qn1, es11, es21 ,…, esm1|Z)- esm(q10, q20, …,
qn0, es10, es20 ,…, esm-10 |Z)

(9)

3.2 Estimation methods for environmental quality functions and ecosystem service functions
If the functional forms of all q1

(C, q2, q3, …, qn| Y), q2 (C, q1, q3, …, qn|

Y), …, qn (C, q1, q2, …, qn-1| Y) and es1(q1, q2, …, qn, es2, es3 , …, esm
|Z), es2(q1, q2, …, qn, es1, es3 ,…, esm |Z), …, esm(q1, q2, …, qn, es1,
es2, …, esm-1 |Z) were estimated, we would be able to estimate the ecosystem change
vector

(Δes1, Δes2, …, Δesm). However, it is a very difficult task to estimate all the

environmental quality functions and ecosystem service functions. Some functions are
relatively easy to estimate and the model has been widely accepted. For example, the Revised
Universal Soil Loss Equation has been widely used to calculate how land management
factors would change soil erosion. For the first kind of functions, we can use these models
directly to solve our problems. Some functions have been estimated but the parameters of the
functions are only valid in the particular research area. For example, the hydrological models
that estimate the nitrate, phosphorus concentration in water. For the second kind of functions,

15

we should look for the closest estimates of the function parameters that best fit our
background conditions. If we cannot find an appropriate function to rely on, then we’ll have
to impose several assumptions to set up a special functional form. Then our task is to estimate
the parameters for the special functional form. For many other functions, they have not been
estimated so far. Examples are the health effect function, morbidity risk function of nitrate
and pesticide and the biodiversity function. For the last kind of functions, we’re not going to
estimate them in this paper.

3.3 Specification of the problem
3.3.1 Selected low-input cropping systems in the LTER at Michigan State University
The long term ecosystem research (LTER) of Kellogg Biological Station (KBS),
Michigan State University has conducted over 15 years of research on the ecosystem services
from row crop agriculture (Robertson 1999, Paul 1997). KBS LTER has identified that
compared with the conventional corn-soybean-wheat rotation, several low input cropping
systems can provide improved levels of one or more of the ecosystem services. Those low
input cropping systems have one or more of the following features:
Less or no tillage
Increased rotation complexity by putting cover crops in rotation
2/3 of conventional fertilizer input
2/3 of conventional pesticide
Effective fertilizer management by using Pre-Sidedress Nitrogen Test (PSNT) to test
nitrogen available in soil before applying nitrogen fertilizer.
The LTER study showed that low input cropping systems increased some offsite
ecosystem service changes (Farber et al 2009, Robertson et al 2000), including:
Lower global warming potential by better carbon sequestration and lower N2O, CO2

16

and CH4 emissions
Less nitrate leaching by lower nitrogen fertilizer use
Habitat for birds, insects and other animals
Other field experiment results showed other offsite environmental benefits from low
input cropping systems such as reduced phosphorus runoff (Daroub et al 2001) and reduced
soil erosion (Reganold et al 1987)
To investigate farmers’ willingness to accept payments to adopt low input cropping
systems, a farm-level survey was implemented by the Department of Food, Agricultural and
resource Economics, Michigan State University in 2008 (Jolejole 2009). The study estimates
the farm level supply curves of low input cropping systems. Table 1.1 showed the profiles of
some low input cropping systems. System 4 were consistent was modeled on one of the
KBS-LTER treatments.

17

Table 1.1 cropping systems from the 2008 farmer’s willingness to accept payment for the adoption of low input cropping systems survey
(Jolejole 2009)
Farming practices

Low-input cropping systems
1

2

3

Rotation

Corn-soybean rotation

Corn-soybean rotation

Corn-soybean-wheat
rotation

Corn-soybean-wheat
rotation

Cover Crops

No

Yes

Yes

Yes

Tillage

Chisel plow with
cultivation as needed

Chisel plow with
cultivation as needed

Chisel plow with
cultivation as needed

Chisel plow with
cultivation as needed

Soil Test

PSNT always

PSNT always

PSNT always

PSNT always

Fertilizer

Full rate

Full rate

Full rate

Reduced rate (1/3)

Pesticide

Full rate

Full rate

Full rate

Reduced rate (1/3)

18

4

3.3.2 Linkages and ecosystem service changes we want to measure
Figure 1.1 describes the linkages to quantify the ecosystem service (the “mapping”).
Cropping systems are decomposed into combinations of several agricultural management
factors: rotation complexity, tillage method, fertilizer input and pesticide use. These elements
have impacts on six dimensions of intermediate environmental quality: soil erosion,
phosphorus loading to waters, greenhouse gas emission, nitrogen loss as nitrate into waters,
nitrogen loss as NOx and N2O into the atmosphere and pesticide pollution. Furthermore,
there are interactions among intermediate environmental changes. Soil erosion is responsible
for the majority of phosphorus runoff into waters (Correll 1998). Nitrogen loss from farming
has various forms: nitrogen as nitrate that affects ground water quality, nitrogen as NO2 and
NO (NOx) that affects regional air quality and nitrogen as N2O that is a kind of greenhouse
gas.
Intermediate environmental quality will cause eight major categories of off-farm
ecosystem services. Less soil erosion from land will reduce sediment loading to drains and
rivers. The drain dredging cost will decrease. Moreover, less soil erosion will lower the
probability of local flooding by causing drainage siltation, raising river beds and water level
etc. Less phosphorus loading to fresh waters will improve the eutrophic status of lakes,
reducing excess plant growth and algal blooms. It will improve the joy of recreational
activities like fishing, boating and swimming. Nitrogen loss in the form of nitrate will go to
ground water. Reduced nitrogen loss from land will reduce the nitrate concentration in ground
water and improve drinking water quality. Nitrogen loss as NOx will improve the air quality.
Nitrogen as nitrous oxide emission contributes to the global warming. Less pesticide use can
reduce relevant health risk and increase the population of insects and animals.

19

Figure 1.1 Mapping linkages from agricultural management factor to off-farm ecosystem service change
For interpretation of the references to color in this and all other figures, the reader is referred to the electronic version of this thesis.

Agricultural
management factor

Intermediate
environmental changes
Less soil erosion

Off farm ES changes
we want to measure
Reduced drain
dredging
Lower incidents of
flooding

Complex
rotation
Less tillage

Less greenhouse
gas emissions

Less fertilizer
input
Less pesticide
input

N2O

Low-input
cropping
systems

Less phosphorus
loading to waters

As
nitrate

Less
nitrogen
loss

As
NOx

Less pesticide
into the air

20

Better recreation
use of waters
Mitigated global
warming
Improved drinking
water quality

Air quality

Reduced health risk
Higher population of
insects and animals

Not
going to
quantify

Finally, there are eight categories of off-farm ecosystem services we want to measure:
soil erosion would reduce drain dredging and incidents of local flooding; phosphorus loading
to waters would affect recreation use like fishing, boating and swimming; greenhouse gas
(GHG) emissions would affect global warming; nitrogen loss as nitrate would affect drinking
water quality; nitrogen loss as NOx would affect perceivable air quality; pesticide would
affect human health risk and higher population of insects and animals. These eight categories
of ecosystem service changes have direct economic implications to humans. Dredging cost
and flood damage are direct economic costs from soil erosion. The other six categories have
important economic value to humans, although they don’t have explicit prices.
In order to quantify the last three categories, the impacts must be significant, and
quantitative relationship between environmental input and health reaction is necessary.
Existing literature shows that although the exposure of nitrate/nitrate and pesticide in the
general public is detectable but is generally much lower than among agricultural field
workers (Majewski et al 1998, Fenske et al 2000, Baker et al 1996, Eskenazi et al 1999). The
reduction in nitrate/nitrite and pesticide exposures in the general public has an unknown
marginal health impact. Furthermore, quantitative results in literature are not sufficient to
estimate the reduced risk in human health or animal populations. Epidemiologic studies have
shown correlations between nitrate from drinking water and infant developmental effects but
the cause and effect relationship cannot be drawn (Fan and Steinberg 1996). Experimental
results with animals might not be applied to human health problems because the probability
that the general public would encounter such high concentration of nitrate/nitrate as in lab
experiment is very low (Fan and Steinberg 1996). Health impacts of pesticides are better
acknowledged but assessments of health risks under certain doses are very rare. There is one
of the few studies on the morbidity probability distribution analysis of ambient pesticide

21

concentration (Lee et al 2002). Nevertheless, results from one site specific study are not
conclusive of the generalization of the risk and dose relationship. Because existing literature
does not provide sufficient results to estimate the quantities of the last three categories of
ecosystem service changes, they are not quantified in this chapter.

3.3.3 Specification of environmental quality functions and ecosystem functions
We define the intermediate environmental quality vector from changing agricultural
management practices as (qs, qp, qGHG,

qn) where qs is the soil erosion, qp is the

phosphorus loading to waters, qGHG is the greenhouse gas emission and qn is the nitrate
leaching into groundwater. Then the changes in qs,

qp, qGHG, qn as Δqs, Δqp, ΔqGHG,

Δqn can be specified as:
Δqs= qs (C1, qp1, qGHG1, qn1| Y)- qs(C0, qp0, qGHG0, qn0| Y)

(10)

Δqp= qp (C1, qs1, qGHG1, qn1| Y)- qp(C0, qs0, qGHG0, qn0| Y)

(11)

Δqn= qn (C1, qp1, qs1, qGHG1|Y)- qn(C0, qp0, qs0, qGHG0| Y)

(12)

ΔqGHG= qGHG (C1, qp1, qs1, qn1| Y)- qGHG(C0, qp0, qs0, qn0| Y)
(13)
Where C is the agricultural management factor and Y is other exogenous variables that
hold constant.
Among the four variables,

qp is affected by qs and the rest are mutually unaffected.

Therefore, equation (1) to (4) can be simplified as

Δqs= qs (C1| Y)- qs(C0 | Y)

(14)

22

Δqp= qp (C1, qs1 | Y)- qp(C0, qs0 | Y)

(15)

Δqn= qn (C1 |Y) - qn(C0 | Y)

(16)

ΔqGHG= qGHG (C1 | Y)- qGHG(C0 | Y)
Then we define the ecosystem service vector as

esdw)

where

esdredge

(esdredge esflood esrecreation esgw

is the reduced dredging cost in drains and rivers,

reduced flood damage cost,
global warming and

(17)

esdw

esrecreation

esflood

is the

esgw

is the

is the recreational aspects of lakes,

is the drinking water quality. The set of ecosystem service

functions are

esdredge= esdredge(qs, qp, qGHG, qn, esflood, esrecreation , esgw, esdw , Z) (18)
esflood= esflood(qs, qp, qGHG, qn, esdredge, esrecreation , esgw, esdw, Z)
(19)
esrecreation= esrecreation(qs, qp, qGHG, qn, esdredge, esflood , esgw, esdw , Z)
(20)
esdw= esdw(qs, qp, qGHG, qn, esdredge, esflood, esrecreation , esgw , Z)
(21)
esgw= esgw(qs, qp, qGHG, qn, esdredge ,esflood, esrecreation , esdw, Z)
(22)
where Z is the vector of other exogenous variables that are held constant
The five categories of ecosystem services do not affect each other so the interaction
among ecosystem services can be ignored. Dredging cost and flood damage are affected by
23

qs. Recreation aspects of lakes are affected by qp. Global warming is affected by qGHG and
drinking water quality is affected by qn. Then equations (18)- (22) are rewritten as

esdredge= esdredge(qs, Z)

(23)

esflood= esflood(qs, Z)

(24)

esrecreation= esrecreation(qp, Z)

(25)

esdw= esdw(qn, Z)

(26)

esgw= esgw(qGHG, Z)

(27)

A change in environmental quality from

(qs0, qp0, qGHG0, qn0)

to

(qs1, qp1,

qGHG1, qn1) will cause ecosystem service changes by
Δesdredge= esdredge(qs1 |Z)- esdredge(qs0 |Z)

(28)

Δesflood= esflood(qs1 |Z)- esflood(qs0 |Z)

(29)

Δesrecreation= esrecreation(qp1 |Z)- esrecreation (qp0 |Z)

(30)

Δesdw= esdw(qn1 |Z)- esdw(qn0 |Z)

(31)

Δesgw= esgw(qGHG1 |Z)- esgw(qGHG0 |Z)

(32)

Then our task is to estimate equations (14)- (17) and (28)-(32)

4 Function estimation and results
4.1 Soil erosion and related ecosystem services
4.1.1 Soil erosion estimates
The Revised Universal Soil Loss Equation (RUSLE) is a popular model for calculating
soil erosion (Institute of Water Research Michigan State University 2008). It calculates the
24

soil erosion rate by the following formula:

qs = R * K * LS * C * P
Where R is the rainfall runoff erosivity factor; K is the soil erodibility factor which
represents the susceptibility of soil to erosion and the rate of runoff; L is slope length factor
and S is slope steepness; C is the cover management factor, which is the rotation each year; P
is the support practice factor.
Variables R, K, LS, P remain constant when changing land management practices. Then
the percentage change in soil erosion is

qs1/ qs0- 1= (R * K * LS * P * C1) / (R * K * LS * P * C0 ) -1=
C1/ C0 -1
qs1/ qs0- 1 is the percentage reduction in soil erosion for different crop rotations.
Denote this percentage reduction as w, then

Δqs= qs1- qs0= qs0 (qs1/ qs0- 1) = w qs0
Where

qs0

(33)

denotes soil erosion amount in the conventional cropping system,

qs1

denotes soil erosion in the target low input cropping system and w denotes the soil erosion
reduction rate. Plugging in

C0 as the corn-soybean rotation and C1 as corn-soybean-wheat

rotation and corn-soybean-wheat with cover crops yields different w for four target cropping
systems (Table 1.2). The calculation of Table 1.2 is presented in Appendix 2. Here we do not
consider the chisel plow tillage method.

Table 1.2 Soil erosion reduction rate of different cropping systems
Cropping systems

Soil erosion reduction rate (w)

25

Table 1.2 (cont’d)
Corn-soybean rotation

0

Corn-soybean with cover crops

50.0%

Corn-soybean-wheat with cover crops

70.6%

Corn-soybean-wheat with cover crops 70.6%
and reduced input

Soil erosion volume in cropland in Michigan was 11.9 G kg/year (USDA 2003).
Corn-soybean rotation takes up 22% of total area of cropland in Michigan (Padgitt et al 2000)
and 28% of total soil erosion from cropland (Appendix 1). Therefore, the soil erosion from
conventional corn-soybean rotation qs

0

=11.9 * 28%= 3.3 G kg/year.

Then the estimated Δqs ranges from 0 to about 2.3 G kg/year (Table 1.3).

Table 1.3 Soil erosion reduction rates of different cropping systems
Cropping systems

Soil erosion reduction in G kg/year
(Δqs)

Corn-soybean rotation

0

Corn-soybean with cover crops

1.65

Corn-soybean-wheat with cover crops

2.33

Corn-soybean-wheat with cover crops and
reduced input

2.33

4.1.2 Drain dredging cost
There are no dredging cost functions that are suitable to use in our situation, therefore
assumptions have to be made to impose a certain functional form. We assume that dredging
cost is proportional to soil erosion when other variables are constant. Then the dredging cost
26

function where other variables are constant except the soil erosion variable becomes

esdredge(qs |Z) = l qs
Where l is the coefficient that needs to be estimated
Then equation (19) becomes

Δesdredge= l (qs1- qs0 ) = l Δqs

(34)

The coefficient l can be interpreted as the dredging cost for a unit of soil erosion as well
as the marginal cost for a unit increase in soil erosion. Previous literature has provided
information on the dredging cost per unit of soil erosion. Ribaudo (1986) estimates the
dredging cost of both navigable waterways and drainage ditches as $40,800,000 in Lake
States (Wisconsin, Michigan and Minnesota) in 1983. The source of this dredging cost
estimate comes from average annual dredging costs by the Corps of Engineers for
maintaining navigable waterways. The total annual soil erosion of the three states is
181,000,000 tons. Then the average dredging cost per ton of soil erosion would be
40,800,000/181,000,000= $0.23. This value is equivalent to $0.50 in 2010 as in purchasing
power (US Department of Labor 2010).
With l = 0.50 and

Δqs estimated in Table 1.4, Δesdredge ranges from 0 to 1 million

dollars per year.

Table 1.4 Reduced dredging cost of different cropping systems
Cropping systems

Reduced dredging cost in USD
(Δesdredge)

Corn-soybean rotation

0

Corn-soybean with cover crops

826,000

Corn-soybean-wheat with cover crops

1,170,000

27

Table 1.4 (cont’d)
Corn-soybean-wheat with cover crops 1,170,000
and reduced input

4.1.3 Flood damage cost
Less soil erosion will lower the probability of local flooding. The probability of
flooding will influence the expected cost of flood damage cost each year. Because we did not
find any quantitative study on soil erosion and the probability of flooding, instead we assume
that flood damage cost is proportional to soil erosion when other variables are held constant.
Therefore the flood damage cost change in equation (20) is

Δesflood= m (qs1 - qs0) = m Δqs

(35)

The coefficient m stands for the flood damage cost for a unit of soil erosion as well as the
constant marginal flood damage cost for a unit increase of soil erosion. Ribaudo (1986)
estimates annual flooding damage due to sediment of $59,100,000 (in 1983 dollars) in three
Lake States. The estimate is based on Clark’s estimate on flood damages caused by soil
erosion in increased flood heights due to channel aggradations, increased flood volumes due
to sediment loads, direct sediment damages to urban and nonurban areas and reduced
agricultural productivity (Clark 1985). The total annual soil erosion of the three states is
181,000,000 tons. Then the average flooding damage per ton of soil erosion would be
59,100,000/181,000,000=0.33. It is equivalent to $0.72 in 2010 in purchasing power (US
Department of Labor 2010).
With m=0.72 in equation (35), the flood damage cost reduction ranges from 0 to about
$2 million each year (Table 1.5).

28

Table 1.5 Reduced flooding damage cost of different cropping systems
Cropping systems

Reduced flooding damage cost in USD
(Δesflood)

Corn-soybean rotation

0

Corn-soybean with cover crops

1,190,000

Corn-soybean-wheat with cover crops

1,680,000

Corn-soybean-wheat with cover crops and 1,680,000
reduced input

4.2 Phosphorus and eutrophic lakes
4.2.1 Phosphorus runoff function
Most of the phosphorus runoff from land is in the form of soil solid matters. Therefore,
we assume that phosphorus runoff from cropland is proportional to the soil erosion rate. Then
the percentage change in phosphorus runoff due to different cropping systems becomes

Δqp/ qp0=Δqs/ qs0 =w
where

qp0

(36)

denotes the phosphorus run off in the conventional corn-soybean rotation

cropping system,

Δqp

denotes the phosphorus runoff reduction after adopting low input

cropping systems.
The percentage change in phosphorus runoff is the same as the soil erosion reduction
rate in Table 1.2.

4.2.2 Reduction in eutrophic lakes
Many hydrological models estimate how phosphorus loading to lakes and rivers would
change the trophic status of lakes. However, the parameters of the hydrological models are
usually estimated locally. They are not suitable for the scale up estimates. Then we need to
29

simplify the hydrological process of phosphorus in lakes. First, we assume that total
phosphorus (TP) concentration in eutrophic inland lakes is uniformly distributed between the
range of (α, β) (β>α). If a lake’s TP falls below α, then it becomes mesoeutrophic and
recreational quality in the lake will significantly improve. Lakes with TP above β are
hypereutrophic lakes. TP in those lakes will be reduced by switching to alternative cropping
systems but will never fall below α. They stay eutrophic afterward. Recreational quality will
not significantly improve in those lakes. So we only consider those within (α, β) and estimate
how many of them will fall below α after changing cropping systems.
Further, we assume that reduced phosphorus input will lead to the same percentage
decrease, denoted by r, of TP in all inland lakes.

Figure 1.2 Distribution change of eutrophic lakes after a uniform phosphorus reduction

P concentration falls in
those eutrophic lakes
P concentration range shrink
by r
Eutrophic
Mesoeutrohic
α

Hypereutrophic
β

TP concentration

In Figure 1.2, the bar denotes the distribution of lakes by their TP concentration. TP
concentration increases from the left to the right. TP concentration below α is mesoeutrophic,
which is the healthy state. TP between (α, β) are eutrophic. TP higher than β is
hypereutrophic. Under the uniform decrease assumption, a decrease in TP concentration by r
will switch the TP concentration distribution of all lakes from a range of (α, β) to somewhere

30

below α and β. Then the dotted part of the distribution will fall below α and become
“healthy”. The purple part can be calculated as

Δesrecreation= rα / [(β-α) (1-r)] * Neutro0

(37)

0

Where Neutro is the number of eutrophic lakes in Michigan currently

r can be decomposed into the product of four ratios
r= (Δqp/ qp0 )* (qp0/ΔTPfert0 )* (ΔTP fert0/ Δ TP) *(Δ TP/ TP0 ) (38)
Where

ΔTPfert0
Δ TP
TP0

is the phosphorus input to lakes that comes from fertilizer in cropland

is the phosphorus input to lakes from all sources
is the phosphorus stock in lakes at the status quo

Δqp/ qp0 is the percentage change in phosphorus input from corn-soybean rotation
cropland, which is w (equation 36).

qp0/ΔTPfert0

is the percentage contribution of

corn-soybean cropland phosphorus input to total cropland phosphorus input. This ratio is the
same as the soil erosion contribution of corn-soybean cropland to all cropland in Michigan,
which is 28%.

ΔTP

fert

0

/ Δ TP is the percentage contribution of cropland phosphorus

input to annual phosphorus input from all sources. Cropland fertilizer accounted for 54.8% of
the annual inputs of phosphorus to the Great Lakes (Correll 1998).

Δ TP/ TP0

is the

flow-to-stock ratio in lakes. This ratio will change dynamically and the hydrological dynamic
varies according to lake depth, volume, discharge of the outlet and time (Ahlgren et al 1988,
Chapra 1980). It is very difficult to estimate one overall ratio for all inland lakes in Michigan.
Therefore, we have to assume that this ratio will range from 0 to 1. In the following
calculation, we will estimate the upper bound of r, where this ratio is 1. Then from equation
31

(38) we estimated r in target cropping systems.

Table 1.6 Percentage reductions in total phosphorus concentration of different cropping
systems
Cropping systems

% reduction in TP concentration
(r)

Corn-soybean rotation

0

Corn-soybean with cover crops

7.6

Corn-soybean-wheat with cover crops

10.8

Corn-soybean-wheat with cover crops and 10.8
reduced input

TP criteria for eutrophication in inland lakes in Michigan is between 20-50μg/L, above
which are hypereutrophic and below which are mesotrophic and oligotrophic (Bednarz 2007).
Then β=50, α=20.
Now Michigan has over 11000 inland lakes. A survey by the US Geological Survey and
Michigan Department of Environmental Quality showed that of the 730 public access inland
lakes sampled from 2001 to 2005, 27% were eutrophic and 4% were hypereutrophic (Bednaz
2007). Then

Neutro0 = 11000 * 27% = 3000. Substitute α, β, Neutro0 and r into equation

(37), then Δesrecreation ranges from 0 to 240 (Table 1.6).
Table 1.7 Reduced eutrophic inland lakes of different cropping systems
Cropping systems

Reduced eutrophic inland lakes
(Δesrecreation)

Corn-soybean rotation

0

Corn-soybean with cover crops

164

32

Table 1.7 (cont’d)
Corn-soybean-wheat with cover crops

240

Corn-soybean-wheat with cover crops and 240
reduced input

4.3 Nitrate and drinking water quality
4.3.1 Nitrate leaching reduction
The long term ecological research of Kellogg Biological Station at Michigan State
University measures the

nitrate

leaching rate

by

switching

from conventional

corn-soybean-wheat rotation

(qn0, C0) to low input cropping systems (qn1, C1) (Table

1.8). However, the KBS LTER treatments are not completely consistent with the cropping
systems in Table 1.1. The only treatment consistent with Table 1.1 is corn-soybean-wheat
rotation with cover crops and reduced input, which is System 4 in Table 1.1. Alternatively, we
need to look for appropriate second hand field experiment data on nitrate leaching of other
cropping systems in Table 1.1.The study from Strock et al (2004) compares the nitrate loss in
conventional corn-soybean cropping system and corn-soybean rotation with cover crops in
Minnesota. However, due to soil, climate and other site specific characteristics, it is not
appropriate to compare the nitrate loss rate from Strock et al’s with that from KBS LTER.
Instead, the relative scale, the percentage reduction of nitrate compared with conventional
practices, is an appropriate option because the relative scale eliminates the effect of other site
specific characteristics except treatment variables.

33

Table 1.8 Nitrate leaching reduction of low input cropping systems in the KBS LTER
experiments (Syswerda et al, 2010).
Cropping systems

Nitrate Leaching reduction in kg/ha/yr
1
(qn )

Table 1.8 (cont’d)
Conventional
rotation

corn-soybean-wheat 62

Corn-soybean-wheat with cover crops 24
and reduced input
Corn-soybean-wheat with no tillage

42

Corn-soybean-wheat with organic input

19

Let s denote the percentage reduction of nitrate leaching of different cropping systems
compared with conventional corn-soybean cropping system. Then

s = (qn1 - qn0) /qn0 =Δqn/ qn0
Where

(39)

qn0 denotes the nitrate leaching rate from conventional corn-soybean cropping
1

system and qn denotes the nitrate leaching rate from low input cropping systems.
Estimates of s are presented in Table 1.9. From Stock et al’s study (2004), the
corn-soybean with cover crops can reduce nitrate loss by 13%. Note that the percentage
reductions of corn-soybean-wheat cropping systems in Table 1.9 are computed with
conventional corn-soybean-wheat rotation as the baseline because the baseline cropping
system in KBS LTER experiment is corn-soybean-wheat rotation. Compared with
conventional corn-soybean rotation, the addition of wheat in rotations can reduce nitrate
leaching (McDowell and McGregor 1980). As a result, the percentage reductions of these
three low input cropping systems from KBS LTER should be higher if using conventional
corn-soybean rotation as a baseline. Therefore, the reduction estimates in these three cropping
34

systems in Table 1.9 can serve as conservative estimates. There are currently no appropriate
field experiment data for us to estimate the nitrate leaching reduction rate of
corn-soybean-wheat with cover crops.

Table 1.9 Nitrate leaching reduction rate in different cropping systems in KBS research
Cropping systems

Nitrate leaching reduction rate
(s)

Corn-soybean rotation

0

Corn-soybean with cover crops

13%

Corn-soybean-wheat with cover crops

Unknown

a

Corn-soybean-wheat with cover crops 61% b
and reduced input
Corn-soybean-wheat with no tillage

32%

Corn-soybean-wheat with organic input

69%

b
b

a. Cited from Strock et al 2004
b. Calculated from Syswerda et al 2010

4.3.2 Drinking water quality
Similar to the problem of phosphorus concentration modeling, the hydrological models
on nitrate concentration in ground water were local and did not fit our goal of the state wide
estimates. Then again, we need to make assumptions to simplify the drinking water quality
function. First, we assume that unsafe private drinking water wells have nitrate concentration
within a range (γ, δ) (δ>γ) and they are uniformly distributed within this range. Wells with
nitrate below safe standard γ are safe for drinking. δ is the maximum nitrate concentration in
the region.
35

Further, we assume that reduced nitrate leaching from corn-soybean cropland would
lead to the same percentage decrease, denoted by t, in nitrate concentration in all private
water drinking wells.
Figure 1.3 Distribution changes of nitrate wells after a uniform nitrate reduction
Nitrate concentration
falls in unsafe wells
Nitrate concentration range shrinks by
t
Wells above the MCL
Wells safe for drinking
γ

δ

Nitrate concentration increases

In Figure 1.3, the bar denotes the distribution of drinking water wells ordered by nitrate
concentration. Nitrate concentration increases from the left to the right. Nitrate concentration
below the maximum contaminant level (MCL), denoted by γ, is safe for drinking. Wells with
nitrate higher than γ are considered having potential health risk. Under the uniform decrease
assumption, a decrease in nitrate concentration by t will switch the range of nitrate
concentration in those lakes from a range of (γ, δ) to a lower range at somewhere below γ and
δ. The dotted part is the percentage that unsafe wells would become safe after changing
cropping systems. The purple part can be calculated using equation (40)

Δesdw= Munsafe0 * tγ/ [(δ- γ) (1-t)]

(40)

0

Where Munsafe is the number of unsafe wells currently
Similarly, t can be decomposed into the product of four ratios

t = (Δqn/ qn0 )* (qn0/ΔNTcrop0 )* (ΔNTcrop0/ ΔNT) *(Δ NT/ NT0 )
(41)
36

Where

ΔNTcrop0 is the nitrate input to groundwater that comes from fertilizer in cropland
Δ NT is the nitrate input to groundwater from all sources
NT0 is the nitrate stock in groundwater currently
Δqn/ qn0

is s (equation 39).

qn0/ΔNTcrop0

is the percentage contribution of

corn-soybean cropland nitrate input to total cropland nitrate input, which is assumed to be the
same as the ratio of the corn-soybean rotation area to total cropland area (22% as estimated
0

by Padgitt et al 2000). ΔNTcrop

/ ΔNT is the percentage contribution of cropland nitrate

input to annual nitrate input from all sources. In the Midwest Corn Belt area, nitrogen
fertilizer from cropland contributed to minimum 55-60% of nitrate in lake water in peak
season. For the whole year, the minimum percentage ranged from 0 to 60% (Kohl et al 1981).

Δ NT/ NT0 is

the flow-to-stock ratio in groundwater. This ratio is highly variable in

different area and there is no overall estimate for the Michigan area. Therefore, we have to
assume that this ratio will range from 0 to 1. In the following calculation, we will assume this
ratio is 1 in order to get the maximum estimate of t. Then from equation (40) we can estimate
t (Table 1.10).

Table 1.10 Percentage reduction in nitrate concentration in different cropping systems
Cropping systems

Percentage reduction in nitrate concentration
(t)

Corn-soybean rotation

0.0%

Corn-soybean with cover crops

1.7%

Corn-soybean-wheat with cover crops

Unknown

37

Table 1.10 (cont’d)
Corn-soybean-wheat with cover crops and 8.1%
reduced input
Corn-soybean-wheat with no tillage

4.3%

Corn-soybean-wheat with organic input

9.2%

The maximum contaminant level (MCL) for nitrate/nitrite in drinking is 10 mg/l to
prevent observable health effects. Babies under 6 months old who drink water containing
nitrate/nitrite in excess of the MCL could develop methemoglobinemia, commonly known as
the blue baby syndrome (EPA 1976). Several regional studies show that the highest nitrate
concentrations in wells in Michigan is about 20mg/L and about 1-5% of drinking water wells
contain nitrate above the MCL (Kim et al 2002, Aichele 2000, Thomas 2002). Then γ=10,
δ=20. There are about 1.1 million private drinking water wells in Michigan (MDEQ 2008).
0

We can estimate that Munsafe ranges from 11,000 to 56,000. Then substitute the values into
equation (40), we can estimate Δesdw (Table 1.11)
Table 1.11 Reduction in unsafe wells in different cropping systems
Cropping systems

Reduction in unsafe wells
(Δesdw)

Corn-soybean rotation

0

Corn-soybean with cover crops

196-978

Corn-soybean-wheat with cover crops

Unknown

Corn-soybean-wheat with cover crops and 986- 4929
reduced input
Corn-soybean-wheat with no tillage

498- 2490

Corn-soybean-wheat with organic input

1129-5643

38

4.4 Greenhouse gas emission and global warming
4.4.1 Greenhouse gas emission
Instead of estimating the greenhouse gas emission function, we adopt the field
experiment results of the KBS LTER on the global warming potential of several cropping
systems (Robertson et al 2000). The KBS LTER treatments are not completely consistent
with cropping system profiles in Table 1.1 but unfortunately there are no appropriate second
hand data for our estimates of the corn-soybean with cover crops and corn-soybean-wheat
with cover crops. We have to leave these two cropping systems to future studies. We calculate
the reduction

ΔqGHG

as the global warming potential reduction from conventional

corn-soybean system to low input cropping systems (Table 1.12).

Table 1.12 Global warming potential reduction of different cropping systems compared with
conventional practices
Cropping systems

Global warming potential reduction in kg/ha/year
of CO2 equivalent (ΔqGHG)

Corn-soybean rotation

0

Corn-soybean with cover crops

Unknown

Corn-soybean-wheat with cover crops

Unknown

Corn-soybean-wheat with cover crops 510
and reduced input
Corn-soybean-wheat with no tillage

1000

Corn-soybean-wheat with organic input

730

4.4.2 Scale up global warming effect

39

The scale up effect of greenhouse gas emission reduction is calculated by equation 42.

Δesgw= a (qGHG1- qGHG0 ) = a ΔqGHG
Where a is the area of the corn-soybean rotation cropland;
warming potential from conventional corn-soybean cropping system;

(42)
qGHG0
qGHG1

is the global
is the global

warming potential from low input cropping systems.
There are 2,030,000 acres of corn-soybean rotation cropland in Michigan in 2008
(NASS 2008), which is equivalent to 820,000 ha. Then we can estimate Δesgw (Table 1.13).

Table 1.13 Global warming effect of different cropping systems
Cropping systems

Reduced GHG in kg of CO2 equivalent
per year
(Δesgw)

Corn-soybean rotation

0

Corn-soybean with cover crops

Unknown

Corn-soybean-wheat with cover crops

Unknown

Corn-soybean-wheat with cover crops 419,150,000
and reduced input
Corn-soybean-wheat with no tillage

821,862,000

Corn-soybean-wheat with organic input

599,960,000

5 Discussion
Five categories of ecosystem service changes were quantified as for the whole state.
They were reduction in annual drain and river dredging cost, flood damage cost, nitrate in
private drinking water wells, eutrophic lakes and global warming potential. Other ecosystem
service changes were not quantified because of insufficient information in existing literature.
40

Further scientific studies are necessary to estimate the scale up effect of perceivable air
quality, health risks and population of animals and insects from low input cropping systems.
The maximum dredging cost saving and flood damage cost saving by switching to low
input cropping systems were estimated to be $964,000 and $1,380,000 for the whole state
respectively. The benefit from reducing soil erosion related damage is significant. However,
compared with the 10 million population in Michigan, the reduced annual dredging cost and
flood damage cost per capita were only $0.11 and $ 0.13 respectively, which were too small
to significantly affect people’s household budget.
The reduction of high nitrate wells above the MCL needs further discussion. High
nitrate in drinking water might cause adverse health effects such as higher probability of
cancer but those associations were not deterministic (Weyer et al 2001, Fan and Steinberg
1996, Townsend et al 2003). The only health risk that has been widely recognized was the
blue baby syndrome which generally occurred among babies under six months old (MDEQ
2003). However, not only nitrate in drinking water but infection and inflammation could
cause the disease (Avery 1999). Moreover, the blue baby syndrome is rare nowadays.
Historical data before 1951 showed that there were only seven reported blue baby syndrome
cases in wells with nitrate above MCL in Michigan during 1945-1950 (Walton 1951). The
1

probability for a case occurring at a well above MCL is only at the 1 in 100,000 scale. Now
there are about 33,000 wells above the MCL in Michigan and the nitrate concentration in
drinking water is lower than it was at the 1950s, so it can be roughly estimated that there
would be one case occurring every three years. In addition, the fatality rate of the blue baby
syndrome was 7-8% (American Public Health Association 2000). If it were to be measured, it
1

There were seven reported cases of blue baby syndrome from 1945-1950. At that time,
water sample drawn from Michigan showed that 7% of sampled wells were above the MCL
(Walton 1951). At average, there were 1.4 cases every year and there were roughly
1,100,000*7%=77,000 wells above the MCL in Michigan. So the probability that a case
occurred at a well above the MCL was 1.4/77,000=1.8/100,000.
41

can be estimated with other methods such as benefit transfer.
Another potential environmental problem caused by excess nitrogen is hypoxia in
coastal areas, which is known as dead zones. The outlet of the Great Lakes to the ocean is
from the Gulf of St Lawrence. The bottom water zone of Laurentian Channel in the Gulf of St
Lawrence is suffering from hypoxia for decades but the major cause of hypoxia is the ocean
circulation patterns (Gilbert et al 2005). Moreover, the riverine nitrogen export per square
kilometer from the Great Lakes- St. Lawrence region to the North Atlantic are the second
lowest in North America and the third lowest among 13 regions of the North Atlantic Basin.
Nitrogen fertilizer takes up over 50% of total anthropogenic nitrogen inputs (Howarth et al
1996). The majority of the nitrogen fluxes to the Laurentian Channel is deposited as organic
sediments rather than in the water column and the bottom water remains nitrogen deficient
(Thibodeau et al 2010). It implies that the nitrogen input in the bottom water does not
contribute much to the aquatic productivity nor the hypoxia problem in the bottom water of
the Gulf of St Lawrence. Therefore, we did not quantify the impact of low input cropping
systems on coastal hypoxia problem.
The estimated reduction in eutrophic lakes was significant. The estimates were based on
strong assumptions. First, we assumed that there would be a uniform decrease in phosphorus
concentration in all lakes. As a matter of fact, there should be a dramatic decrease in lakes
near agricultural area. There should be an unbalanced reduction in phosphorus concentration
among lakes in agricultural area and those in nonagricultural area. On the other hand, the
eutrophic lake distribution in Michigan was concentrated in agricultural areas (Bednarz 2007).
It means that eutrophic lakes in agricultural areas were possibly benefiting more in
phosphorus reduction from low input cropping systems. So there should be more eutrophic
lake reduction if the spatial impact was taken into consideration. The estimates we got from
the uniform decrease assumption were conservative. Due to limited time and resource, we

42

could only start with the simplest model in this paper. However, it would be worthwhile to
continue with this framework and go further to consider the spatial aspect of this issue.
The estimates of global warming potential were relatively straight forward. The 2000
emission level in Michigan was 180 billion kg of CO2 equivalent (U.S. Public Interest
Research Group Education Fund, 2006). No tillage cropping system achieved the highest
reduction in global warming potential of 822 million kg annually. It was about 0.5% of the
2000 emission level. The reduction was significant because the corn-soybean rotation only
took up only 22% of total cropland in Michigan and the majority of the greenhouse gas
emission came from fossil fuel.
In conclusion, of the five ecosystem service changes we estimate, four of them are
significant except the health benefit of reduction of high nitrate wells. The estimates are
summarized in Table 1.14. Reduced dredging cost and reduced flood damage cost were
measured in dollar terms so the economic benefit of these two ecosystem service changes is
explicit. On the other hand, the reduction in eutrophic lakes and greenhouse gas are expressed
in physical terms. To estimate the economic benefits of these two ecosystem service changes,
further research is required and it will be elaborated in Chapter Two.

43

Table 1.14 Summary of estimates of five ecosystem services changes of low input cropping systems

Cropping systems

Reduced dredging
cost in USD

Reduced flooding
damage cost in USD

Reduced eutrophic
inland lakes

Reduction in unsafe Reduced GHG in kg
wells
of CO2 equivalent
per year

(Δesdredge)

(Δesflood)

(Δesrecreation)

(Δesdw)

(Δesgw)

Corn-soybean
rotation

0

0

0

0

0

Corn-soybean with
cover crops

826,000

1,190,000

164

196-978

Unknown

Corn-soybean-wheat
with cover crops

1,170,000

1,680,000

Unknown

Unknown

Corn-soybean-wheat
with cover crops and
reduced input

1170000

1,680,000

986- 4929

419,150,000

Corn-soybean-wheat
with no tillage

Not calculated

Not calculated

240

498- 2490

821,862,000

Corn-soybean-wheat
with organic input

Not calculated

Not calculated

240

1129-5643

599,960,000

44

APPENDICES

45

Appendix 1.1 Calculating the percentage of soil erosion from corn-soybean rotation land to
soil erosion from total cropland in Michigan

Corn-soybean rotation takes up 22% of total area of cropland in Michigan and below is
2

a portfolio of major crops in Michigan . Using the RUSLE model, their relative soil erosion
rates to the corn-soybean rotation (by changing only the C factor) are calculated in Column
(2).
Table 1.15 Calculating weighted relative soil erosion rate of different crops
Crop Name

Area
(acres)

Soil erosion relative Weights by area
to CS land

Weighted relative
soil erosion

(1)

(2)

(3)=(1)/subtotal
of (1)

(4)=(2)*(3)

Wheat

433783

0.15

0.161

0.0237

Oats

63485

0.12

0.024

0.0028

Dry bean

259026

0.85

0.096

0.0819

Sugar beets

180054

0.41

0.067

0.0275

CS rotation

1760000 1

0.653

0.6527

Subtotal

2991772

1.000

0.789

The weighted average soil erosion rate of this portfolio is 0.789 relative to that of
corn-soybean rotation cropland. There are 8 million acres of cropland in Michigan. The above
portfolio takes up 2.7 million of them. We assume that the rest 5.3 million has the same soil
erosion rate of the portfolio, then the average soil erosion rate of total cropland in Michigan is
0.789 relative to corn-soybean rotation cropland. Then the percentage of soil erosion from CS

2

NASS. (2008). "Michigan County Estimates." From
http://www.nass.usda.gov/Statistics_by_State/Michigan/Publications/County_Estimates/index
.asp
46

rotation out of total cropland is
CS area * CS soil erosion rate/ (Total cropland * cropland erosion rate)
=(176000 * 1)/ (8,000,000*0.789) = 28%

47

Appendix 1.2 Calculation of soil erosion reduction rates of different cropping systems

The calculation is based on the RUSLE online soil erosion assessment tool, Institute of Water
Research, Michigan State University, http://www.iwr.msu.edu/rusle/
On the RUSLE website, click on “Calculate soil erosion”, then “Agricultural Land” then
select “Kalamazoo” county where KBS is located. For slope, slope length and soil, just select
the first choice applicable. As long as these factors remain the same in all cropping systems,
the soil erosion reduction rate compared with the conventional cropping system does not
change.
1. Calculating soil erosion rate of the corn-soybean conventional cropping system.
Select “Corn for grain” at year 0, “Soybean” at year 1 and so forth until year 5. For tillage
practices select “spring plow” each year. Then the average soil erosion rate is 0.34
ton/acre/year.
2. Calculating soil erosion rate of the corn-soybean rotation with cover crops
Click on “Click here to enter C factor manually”. At that page, enter “Corn for grain” at year
0, “Soybean” at year 1 and so forth until year 5. For tillage practices select “Chisel plow”
each year. To find an appropriate C factor, click the link at the bottom. To be consistent with
the KBS treatment, alfalfa is chosen as the cover crop. The C factor for corn, grain after
alfalfa at year 1 with spring plow is 0.1 and for soybean after alfalfa with spring plow is 0.11.
Enter 0.11 at year 1, 0.1 at year 2 and so forth. Then the average soil erosion rate is 0.17
ton/acre/year.
3. Calculating soil erosion rate of the corn-soybean-wheat rotation with cover crops
Similar to the way we calculate the soil erosion rate of the corn-soybean rotation with cover
crops. We plug in the rotation each year and enter chisel plow as tillage practices. The C

48

factor for wheat after alfalfa is not available. So we have to leave it blank. Then the average
soil erosion rate is 0.1 ton/acre/year.
Reduced input of pesticide and fertilizer does not have significant impact on the soil erosion
rate. Neither the PSNT has significant impact on soil erosion rate. Therefore we assume that
System 1 (corn-soybean rotation with PSNT) has the same erosion rate of the conventional
(corn-soybean rotation without PSNT) and that System 4 (corn-soybean-wheat rotation with
cover crops and reduced input) has the same erosion rate of System 3 (corn-soybean-wheat
rotation with cover crops). Then we can calculate the soil erosion reduction rates in Table 1.2

49

REFERENCES

50

REFERENCES

Ahlgren, I., Frisk, T., & Kamp-Nielsen, L. (1988). Empirical and theoretical models of
phosphorus loading, retention and concentration vs. lake trophic state. Hydrobiologia,
170(1), 285-303.
Aichele, S. S. (2000). Ground-water quality atlas of Oakland County, Michigan: United
States Geological Survey.
American Public Health Association. (2000). Drinking Water Quality and Public Health.
Angers, D. A., Bolinder, M. A., Carter, M. R., Gregorich, E. G., Drury, C. F., Liang, B. C.,
Voroney, R. P., Simard, R. R., Donald, R. G., Beyaert, R. P., & Martel, J. (1997).
Impact of tillage practices on organic carbon and nitrogen storage in cool, humid soils
of eastern Canada. Soil and Tillage Research, 41(3-4), 191-201.
Avery, A. A. (1999). Infantile methemoglobinemia: reexamining the role of drinking water
nitrates. Environmental Health Perspectives, 107(7), 583–586.
Baker, L. W., Fitzell, D. L., Seiber, J. N., Parker, T. R., Shibamoto, T., Poore, M. W.,
Longley, K. E., Tomlin, R. P., Propper, R., & Duncan, D. W. (1996). Ambient Air
Concentrations of Pesticides in California. Environmental Science & Technology,
30(4), 1365-1368.
Bayer, C., Mielniczuk, J., Amado, T. J. C., Martin-Neto, L., & Fernandes, S. V. (2000).
Organic matter storage in a sandy clay loam Acrisol affected by tillage and cropping
systems in southern Brazil. Soil and Tillage Research, 54(1-2), 101-109.
Bednarz, R. (2007, Oct 2007). Overview of Michigan’s Surface Water Monitoring Program –
Inland Lakes. Paper presented at the MiCorps Third Annual Conference.
Borin, M., Giupponi, C., & Morari, F. (1997). Effects of four cultivation systems for maize
on nitrogen leaching 1. Field experiment. European Journal of Agronomy, 6(1-2),
101-112.
Bouwman, A. F. (1996). Direct emission of nitrous oxide from agricultural soils. Nutrient
Cycling in Agroecosystems, 46(1), 53-70.
Brye, K. R., Andraski, T. W., Jarrell, W. M., Bundy, L. G., & Norman, J. M. (2002).
Phosphorus Leaching under a Restored Tallgrass Prairie and Corn Agroecosystems.
Journal of Environmental Quality, 31, 769-781.
Carpenter, S. R., Caraco, N. F., Correll, D. L., Howarth, R. W., Sharpley, A. N., & Smith, V.
H. (1998). Nonpoint pollution of surface waters with phosphorus and nitrogen.
Ecological Applications, 8(3), 559-568.
Chapra, S. (1980). Application of the Phosphorus Loading Concept to the Great Lakes. In C.
Loehr, Martin, C.S. and Rast, W. (Ed.), Phosphorus management strategies for lakes
(Vol. 135-152). Ann Arbor: Ann Arbor Science Publishers.
51

Clark, E. H. (1985). The off-site costs of soil erosion. Journal of Soil and Water
Conservation, 40(11), 19-22.
Correll, D. L. (1998). Role of phosphorus in the eutrophication of receiving waters: A review.
Journal of Environmental Quality, 27(2), 261-266.
Costanza, R., Ralph, A., De Groot, R., Farber, S., Grasso, M., Hannon, B., Limburg, K., &
Naeem, S. (1998). The value of ecosystem services: putting the issues in perspective.
Ecological Economics, 25(1), 67-72.
Coye, M. J. (1986). The health effects of agricultural production: II. The health of the
community. Journal of Public Health Policy, 7(3), 340-354.
Daniels, J. L., Olshan, A. F., & Savitz, D. A. (1997). Pesticides and childhood cancers.
Environmental Health Perspectives, 105(10), 1068-1077.
Daroub, S. H., Ellis, B. G., & Robertson, G. P. (2001). Effect of cropping and low-chemical
input systems on soil phosphorus fractions. Soil Science, 166(4), 281-291.
Dick, W. A., Blevins, R. L., Frye, W. W., Peters, S. E., Christenson, D. R., Pierce, F. J., &
Vitosh, M. L. (1998). Impacts of agricultural management practices on C
sequestration in forest-derived soils of the eastern Corn Belt. Soil and Tillage
Research, 47(3-4), 235-244.
Drinkwater, L. E., Wagoner, P., & Sarrantonio, M. (1998). Legume-based cropping systems
have reduced carbon and nitrogen losses. Nature, 396(6708), 262-264.
Ecobichon, D. J. (1996). Toxic effects of pesticides. In C. D. Klaassen & J. Doull (Eds.),
toxicology: the basic science of poisons (5 ed., pp. 643–689). New York.
Safe Water Drinking Act,

(1976).

Eskenazi, B., Bradman, A., & Castorina, R. (1999). Exposures of Children to
Organophosphate Pesticides and Their Potential Adverse Health Effects. Environment
Health Perspect, 107(suppl 3), 409-419.
Fan, A. M., & Steinberg, V. E. (1996). Health Implications of Nitrate and Nitrite in Drinking
Water: An Update on Methemoglobinemia Occurrence and Reproductive and
Developmental Toxicity. Regulatory Toxicology and Pharmacology, 23(1), 35-43.
Fan, A. M., Willhite, C. C., & Book, S. A. (1987). Evaluation of the nitrate drinking water
standard with reference to infant methemoglobinemia and potential reproductive
toxicity. Regulatory Toxicology and Pharmacology, 7(2), 135-148.
Farber, S., Costanza, R., Childers, D. L., Erickson, J. O. N., Gross, K., Grove, M., Hopkinson,
C. S., Kahn, J., Pincetl, S., Troy, A., Warren, P., & Wilson, M. (2009). Linking
Ecology and Economics for Ecosystem Management. BioScience, 56(2), 121-133.
Fenske, R. A., Kedan, G., Chensheng, L., Fisker-Andersen, J. A., & Curl, C. L. (2002).
Assessment of organophosphorous pesticide exposures in the diets of preschool
children in Washington State. Journal of Exposure Analysis and Environmental
Epidemiology, 22, 21-28.
52

Fenske, R. A., Kissel, J. C., Lu, C., Kalman, D. A., Simcox, N. J., Allen, E. H., & Keifer, M.
C. (2000). Biologically Based Pesticide Dose Estimates for Children in an
Agricultural Community. Environmental Health Perspectives, 108(6), 515-520.
Fisher, B., Turner, R. K., & Morling, P. (2009). Defining and classifying ecosystem services
for decision making. Ecological Economics, 68(3), 643-653.
Food and Agriculture Organization. (2003). Explore on-farm trials for adapting and adopting
good agricultural practices. Rome: FAO.
Gilbert, D., Sundby, B., Gobeil, C., Mucci, A., & Tremblay, G.-H. (2005). A
seventy-two-year record of diminishing deep-water oxygen in the St. Lawrence
estuary: The northwest Atlantic connection. American Society of Limnology and
Oceanography, 50(5), 1654-1666.
Glotfelty, D. E., Seiber, J. N., & Liljedahl, A. (1987). Pesticides in fog. Nature, 325(12),
602-605.
Gray, D. H., & Leiser, A. T. (1989). Biotechnical slope protection and erosion control.
Malabar: Kreiger.
Hall, S. J., Matson, P. A., & Roth, P. M. (1996). NOx emissions from soil: implications for
air quality modeling in agricultural regions. Annual Review of Energy and the
Environment, 21, 311-346.
Halvorson, A. D., Wienhold, B. J., & Black, A. L. (2002). Tillage, nitrogen, and cropping
system effects on soil carbon sequestration. Soil Science Society of America Journal,
66(3), 906-912.
Hecnar, S. J. (1995). Acute and chronic toxicity of ammonium nitrate fertilizer to amphibians
from southern Ontario. Environmental Toxicology and Chemistry, 14(12), 2131-2137.
Hoehn, J., & Randall, A. (1989). Too many proposals pass the benefit–cost test. American
Economic Review, 79, 544-551.
Howarth, R., Billen, G., Swaney, D., Townsend, A., Jaworski, N., Lajtha, K., Downing, J.,
Elmgren, R., Caraco, N., Jordan, T., Berendse, F., Freney, J., Kudeyarov, V.,
Murdoch, P., & Zhao-Liang, Z. (1996). Regional nitrogen budgets and riverine N
& P fluxes for the drainages to the North Atlantic Ocean: Natural and human
influences. Biogeochemistry, 35(1), 75-139.
Howarth, R. B., & Farber, S. (2002). Accounting for the value of ecosystem services.
Ecological Economics, 41(3), 421-429.
Hutciiinson, G. E. (1957). A treatise on limnology, v. 1 (Vol. 29, pp. 1451-1462).
Institute of water research Michigan State University. (2008). RUSLE online soil erosion
assessment tool. from http://www.iwr.msu.edu/rusle/
Jolejole, M. C. B. (2009). Trade-offs, incentives and the supply of ecosystem services from
cropland. Unpublished Master thesis, Michigan State University, East Lansing.

53

Joyce, B. A., Wallender, W. W., Mitchell, J. P., Huyck, L. M., Temple, S. R., Brostrom, P. N.,
& Hsiao, T. C. (2002). Infiltration and soil water storage under winter cover cropping
in California's Sacramento Valley. Trasactions- American Society of Agricultural
Engineers, 45, 315-326.
Kim, M.-J., Nriagu, J., & Haack, S. (2002). Arsenic species and chemistry in groundwater of
southeast Michigan. Environmental Pollution, 120(2), 379-390.
Kohl, D. H., Shearer, G. B., & Commoner, B. (1971). Fertilizer Nitrogen: Contribution to
Nitrate in Surface Water in a Corn Belt Watershed. Science, 174(4016), 1331-1334.
Lee, S., McLaughlin, R., Harnly, M., Gunier, R., & Kreutzer, R. (2002). Community
Exposures to Airborne Agricultural Pesticides in California: Ranking of Inhalation
Risks. Environmental Health Perspectives, 110(12), 1175-1184.
Loomis, J., Kent, P., Strange, L., Fausch, K., & Covich, A. (2000). Measuring the total
economic value of restoring ecosystem services in an impaired river basin: results
from a contingent valuation survey. Ecological Economics 33, 103-117.
Majewski, M. S., Foreman, W. T., Goolsby, D. A., & Nakagaki, N. (1998). Airborne
pesticide residues along the Mississippi River. Environmental Science and
Technology, 32(23), 3689-3698.
McCauley, L. A., Anger, W. K., Keifer, M., Langley, R., Robson, M. G., & Rohlman, D.
(2006). Studying health outcomes in farmworker populations exposed to pesticides.
Environment Health Perspectives, 114(6), 953-960.
McDowell, L. L., & McGregor, K. C. (1980). Nitrogen and phosphorus losses in runoff from
no-till soybeans. Transactions of the American Society of Agricultural and Biological
Engineers, 23(3), 0643-0648.
Michigan Department of Environmental Quality Water Division. (2003). Nitrate in drinking
water: information for private well owners. Retrieved. from
http://www.michigan.gov/deq/0,1607,7-135-3313_3675_3690-9161--,00.html.
Michigan Department of Environmental Quality Water Division. (2004). Michigan lakes:
assessment of water quality (No. FS 2004-3048): US Department of the Interior,
United States Geological Survey.
Michigan Department of Environmental Quality Water Division. (2004). State of Michigan
source water assessment program report. Retrieved. from
http://www.michigan.gov/deq/0,1607,7-135-3313_3675_3693-110933--,00.html.
Michigan Department of Environmental Quality Water Division. (2008). Water Quality and
Pollution Control in Michigan. Retrieved. from
http://michigan.gov/deq/0,1607,7-135-3313_3686_3728-12711--,00.html.
Millennium Ecosystem Assessment. (2005). Millennium ecosystem assessment: Integrated
Assessment through the Millennium Ecosystem Assessment.
Mosier, A. R., Halvorson, A. D., Peterson, G. A., Robertson, G. P., & Sherrod, L. (2005).
Measurement of Net Global Warming Potential in Three Agroecosystems. Nutrient
54

Cycling in Agroecosystems, 72(1), 67-76.
Natural Agricultural Statistics Service. (2008). National Agricultural statistics Michigan
County Estimates. from
http://www.nass.usda.gov/Statistics_by_State/Michigan/Publications/County_Estimat
es/index.asp, retrieved on Dec 5 2010
Natural Resources Conservation services. (2009). Organic Matter in Soil. from
http://www.nrcs.usda.gov/feature/backyard/orgmtrsl.html, retrieved on Dec 5 2010
Padgitt, M., Newton, D., Penn, R., & Sandretto, C. (2000). Production Practices for Major
Crops in U.S. Agriculture, 1990-97 (No. SB969): Economic Research Services
USDA.
Palm, C. A., Alegre, J. C., Arevalo, L., Mutuo, P. K., Mosier, A. R., & Coe, R. (2002).
Nitrous Oxide and Methane Fluxes in Six Different Land Use Systems in the Peruvian
Amazon. Global Biogeochemical Cycles, 16(4), 21-21.
Paul, E. A. (1997). Soil organic matter in temperate agroecosystems: Long-term Experiments
in North America: CRC.
Pimentel, D., Allen, J., Beers, A., Guinand, L., Linder, R., McLaughlin, P., Meer, B.,
Musonda, D., Perdue, D., & Poisson, S. (1987). World Agriculture and Soil Erosion.
Bioscience, 37(4), 277-283.
Pimentel, D., Harvey, C., Resosudarmo, P., Sinclair, K., Kurz, D., McNair, M., Crist, S.,
Shpritz, L., Fitton, L., & Saffouri, R. (1995). Environmental and Economic Costs of
Soil Erosion and Conservation Benefits. Science, 267(5201), 1117-1123.
Pimentel, D., & Kounang, N. (1998). Ecology of soil erosion in ecosystems. Ecosystems, 1(5),
416-426.
Poudel, D. D., Horwath, W. R., Mitchell, J. P., & Temple, S. R. (2001). Impacts of cropping
systems on soil nitrogen storage and loss. Agricultural Systems, 68, 253-268.
Power, J. F., & Schepers, J. S. (1989). Nitrate contamination of groundwater in North
America. Agriculture, Ecosystems & Environment, 26(3-4), 165-187.
Randall, A. (2002). Valuing the outputs of multifunctional agriculture. European Review of
Agriculture Economics, 29(3), 289-307.
Reganold, J. P., Elliott, L. F., & Unger, Y. L. (1987). Long-term effects of organic and
conventional farming on soil erosion. Nature, 330, 370-372.
Reicosky, D. C., Kemper, W. D., Langdale, G. W., Douglas, C. L., & Rasmussen, P. E.
(1995). Soil organic matter changes resulting from tillage and biomass production.
Soil Conservation Society of America, 50(3), 253.
Reicosky, D. C., & Lindstrom, M. J. (1993). Fall tillage method: effect on short-term carbon
dioxide flux from soil. Agronomy Journal, 85(6), 1237.
Ribaudo, M. O. (1986). Reducing soil erosion: offsite benefits. Washtington DC: US
55

Department of Agriculture.
Robertson, D. M. (1996). Sources and Transport of Phosphorus in the Western Lake
Michigan Drainages: US Department of the Interior.
Robertson, G. P. (1999). Standard soil methods for long-term ecological research: Oxford
University Press, USA.
Robertson, G. P., Paul, E. A., & Harwood, R. R. (2000). Greenhouse Gases in Intensive
Agriculture: Contributions of Individual Gases to the Radiative Forcing of the
Atmosphere. Science, 289(5486), 1922-1925.
Robertson, G. P., & Vitousek, P. M. (2009). Nitrogen in Agriculture: Balancing the Cost of
an Essential Resource. 34(1), 97-125.
Sainju, U. M., Singh, B. P., & Whitehead, W. F. (2002). Long-term effects of tillage, cover
crops, and nitrogen fertilization on organic carbon and nitrogen concentrations in
sandy loam soils in Georgia, USA. Soil and Tillage Research, 63(3-4), 167-179.
Scholtz, M. T., Voldner, E., McMillan, A. C., & Van Heyst, B. J. (2002). A pesticide
emission model (PEM) Part I: model development. Atmospheric Environment, 36(32),
5005-5013.
Sharpley, A. N., McDowell, R. W., & Kleinman, P. J. A. (2001). Phosphorus loss from land
to water: integrating agricultural and environmental management. Plant and Soil,
237(2), 287-307.
Smith, K. R. (2006). Public Payments For Environmental Services From Agriculture:
Precedents And Possibilities. American Journal of Agricultural Economics, 88 (5),
1167–1173.
Soil Science of America. (2009). Glossary of Soil Science Terms. from
https://www.soils.org/publications/soils-glossary
Strock, J. S., Porter, P. M., & Russelle, M. P. (2004). Cover Cropping to Reduce Nitrate Loss
through Subsurface Drainage in the Northern U.S. Corn Belt (Vol. 33, pp.
1010-1016).
Syswerda, S.P., B. Basso, G.P. Robertson, S. Hamilton. (2010). Agricultural management
intensity affects long-term nitrate loss. (submitted).
Thibodeau, B., Lehmann, M. F., Kowarzyk, J., Mucci, A., Gélinas, Y., Gilbert, D., Maranger,
R., & Alkhatib, M. (2010). Benthic nutrient fluxes along the Laurentian Channel:
Impacts on the N budget of the St. Lawrence marine system. Estuarine, Coastal and
Shelf Science,90(4), 195-205.
Thomas, M. A. (2002). The effect of residential development on ground-water quality near
Detroit, Michigan. Journal of the American Water Resources Association, 36(5),
1023-1038.
Townsend, A. R., Howarth, R. W., Bazzaz, F. A., Booth, M. S., Cleveland, C. C., Collinge, S.
K., Dobson, A. P., Epstein, P. R., Holland, E. A., Keeney, D. R., Mallin, M. A.,
56

Rogers, C. A., Wayne, P., & Wolfe, A. H. (2003). Human health effects of a changing
global nitrogen cycle (Vol. 1, pp. 240-246).
U.S.PIRG. U.S. Public Interest Research Group Education Fund. (2006). The Carbon Boom:
National and State Trends in Carbon Dioxide Emissions since 1960. Washington DC.
UNFCCC. (2010). Glossary of climate change acronyms. from
http://unfccc.int/essential_background/glossary/items/3666.php#C
US Census Bureau. (2004). United States Summary: 2000 Population and Housing Unit
Counts.
US Department of Labor. (2008). US Department of Labor Bureau of Labor Statistics CPI
inflation calculator. from http://www.bls.gov/data/inflation_calculator.htm
USDepartment of Agriculture. (2003). Natural Resource Inventory. Retrieved. from
http://www.nrcs.usda.gov/technical/NRI/2003/statereports/table1.html
http://www.nrcs.usda.gov/technical/NRI/2003/statereports/table4.html.
Van den Berg, F., Kubiak, R., Benjey, W. G., Majewski, M. S., Yates, S. R., Reeves, G. L.,
Smelt, J. H., & Van der Linden, A. M. A. (2004). Emission of Pesticides into the Air.
Water, Air & Soil Pollution, 115(1-4), 195-218.
Venterea, R. T., Burger, M., & Spokas, K. A. (2005). Nitrogen Oxide and Methane
Emissions under Varying Tillage and Fertilizer Management. Journal of
Environmental Quality, 34(5), 1467-1477.
Voldner, E. C., & Schrueder, V. V. H. (1990). Long range atmospheric transport and
deposition of toxaphene (pp. 223): CRC.
Wagner-Riddle, C., & Thurtell, G. W. (1998). Nitrous oxide emissions from agricultural
fields during winter and spring thaw as affected by management practices. Nutrient
Cycling in Agroecosystems, 52(2), 151-163.
Walton, B. (1951). Survey of the literature relating to infant methemoglobinemia due to
nitrate contaminated water. American Journal of Public Health, 41, 986-996.
West, T., O., & Post, W., M. (2002). Soil organic carbon sequestration rates by tillage and
crop rotation: A global data analysis (Vol. 66). Madison, WI: Soil Science Society of
America.
Weyer , P. J., Cerhan, J. R., Kross, B. C., Hallberg, G. R., Kantamneni, J., Breuer, G., Jones,
M. P., Zheng , W., & Lynch, C. (2001). Municipal Drinking Water Nitrate Level and
Cancer Risk in Older Women: The Iowa Women's Health Study. Epidemiology, 12(3),
327-338.
Williams, E. J., Hutchinson, G. L., & Fehsenfeld, F. C. (1992). NOx and N2O emissions
from soil. Global Biogeochemical Cycles, 6(4), 351-388.
Wright, R. O., Lewander, W. J., & Woolf, A. D. (1999). Methemoglobinemia: etiology,
pharmacology, and clinical management. Annals of Emergency Medicine, 34(5),
57

646-656.
Zahm, S. H., & Ward, M. H. (1998). Pesticides and childhood cancer. Environment Health
Perspectives Supplements, 106(Suppl 3), 893.

58

CHAPTER 2 MICHIGAN PUBLIC’S WILLINGNESS TO PAY FOR REDUCTIONS
IN EUTROPHIC LAKES AND GREENHOUSE GAS DUE TO CHANGES IN LAND
USES: A CONTINGENT VALUATION STUDY

Abstract
Chapter 2 focuses on the economic valuation of two ecosystem services from
widespread adoption of the low input cropping systems discussed in Chapter 1. The
contingent valuation survey method is used to estimate the Michigan public’s willingness to
pay for reductions in eutrophic lakes and greenhouse gas emission. In 2009, a contingent
valuation mail survey was sent to 6000 households in Michigan resulting in a 40% response
rate. The contingent valuation question is designed as a dichotomous choice referendum with
income taxes as the payment vehicle. Respondents’ willingness to pay is modeled as a
function of program cost, the scope of the reduction in eutrophic lakes and GHGs, attitude
variables and demographics. We also model the “spike probability,” the probability of zero
willingness to pay, as endogenous. The models are analyzed using random effect probits.
Respondents’ willingness to pay depends on the scope of reductions in eutrophic lakes,
attitude towards global warming, age, income, education and whether they are voters. The
scope effect of greenhouse gas reduction exists only among those who were concerned about
global warming. Those unconcerned about global warming are not willing to pay anything for
a reduction in GHG. The results also show that the spike probability is endogenous; that is,
the probability of zero willingness to pay is determined by the scope of environmental
changes, income and attitude towards global warming. Finally, sample willingness to pay is
$175 per household per year to get either 170 fewer eutrophic lakes or a 0.52% GHG
reduction from the 2000 level.

59

1 Introduction
1.1 Agriculture as a managed ecosystem
Agriculture is one of the important managed ecosystems where interactions of human
activities and natural resources are intensive and ecosystem service outputs are strongly
affected by management decisions (Swinton et al 2007, Zhang et al 2007). As a managed
ecosystem, agriculture provides various ecosystem services and disservices to human society.
It provides food, fiber and fuel. Apart from the provisioning ecosystem services, agriculture
also provide regulating and supporting ecosystem functions such as water retention (Fawcett
et al 1994, Blevins et al 1983), nutrient cycling (Wyland et al 1996), crop pollination
(Kremen et al 2004), soil formation and conservation (Brussaard 1997). It also contributes to
biodiversity (Tscharntke et al 2005), global warming (Mosier et al 2005) and other cultural
and aesthetic purposes. Nevertheless, it outputs disservices if it is not managed properly by
agricultural practices. For example, conventional intensified agriculture would degrade soil
fertility in the long run compared with organic farming (Maeder et al 2002). Frequent tillage
would increase soil erosion on farmland, reduce soil organic matter as well as more
greenhouse gas emissions (Robertson et al 2000). High rate of chemical fertilizer use would
cause nitrate pollution and eutrophication of waters as excess chemicals run into ground
water and surface waters (Spalding and Exner 1993). Pesticide use would harm biodiversity
by killing beneficial insects and animals (Maeder et al 2002). Toxic chemical use in
agriculture would pose health risks to people (Rosenstock et al, 1991).
On the other hand, flows of ecosystem services and disservices from agriculture can be
managed by land management decisions. In contrast with the examples that generate
disservices previously mentioned, organic farming would increase the soil organic matter and
improve soil structure and fertility. No-tillage farming would have better water retention and
soil erosion. In addition, no-tillage cropping system can reduce 85% of the global warming

60

potential than the conventional tillage counterparts (Robertson et al 2000). Helpful farm
management tools can decide the optimum use of chemical fertilizer to reduce nitrogen and
phosphorus runoff from land, which would improve ground water pollution and surface water
eutrophication. Reduced use of pesticide would help preserve biodiversity as well as reduce
health risks.
Scientific literature has shown that some low-input cropping systems are able to supply
several ecosystem services as well as maintaining comparable yields to conventional
cropping systems. The Long-term Ecological Research (LTER) of row crop agriculture by the
Kellogg Biological Station at Michigan State University has been studying the long term
ecosystem services of some of those low-input cropping systems. The low input cropping
systems tested in KBS’ research are combinations of one or more of the following farming
practices:
•

Multiple crops rotations

•

The use of cover crops

•

No tillage

•

Use of organic fertilizer

•

Reduced rate of nitrogen and phosphorus fertilizers

•

Reduced rate of pesticide use
These low input cropping systems are able to generate several ecosystem service flows

such as better carbon sequestration, less nitrate leaching to water, phosphorus retention in soil
and less soil erosion (Robertson et al 2000, Syswerda et al 2010, Daroub et al 2001).

1.2 Rationale for the study
1.2.1 Policy and research gaps and challenges of the topic
The United States has offered public payment programs to encourage environmentally

61

friendly land management practices since the 1930s. Examples were Conservation Reserve
(CRP) and Environmental Quality Incentives (EQIP) programs. The 2002 Farm Bill
strengthened the support for more environmentally friendly farming practices. It authorized
the Conservation Security Program which offers financial and technical assistance to eligible
private landowners who wish to adopt conservation land management practices (USDA 2004).
Those public programs have increased flows of one or more ecosystem services from
agricultural land. However, most of these public payment programs were supply-side based
which aimed to compensate the land owner’s cost of changing land uses and land
management practices. The demand side is largely ignored in these public payment programs.
The public is paying for environmentally friendly land management activities rather than the
actual environmental outcomes (Smith 2006). Similarly, the ongoing multifunctional
agriculture policy reform debates in Europe have little ideas of what the public demand and
the optimal public good supply (Hall et al 2004).

Ignoring the public’s demand for

environmental outcomes reduces the economic efficiency of these public payment programs
for agriculture.
In the demand side story of the public payment programs, valuation of ecosystem
services remains essential. However, valuation of ecosystem services is a very challenging
task although some non market valuation techniques have been well developed. One
challenge is the complexity of the environmental goods to be valued. First, the ecosystem
services flows from agricultural land have multiple dimensions and the dimensions are
expanding as more and more ecosystem services are recognized to be important to human
beings. Besides the water regulation, water retention, soil formation and fertility, soil
retention, more and more ecosystem services such as climate regulation, landscape,
biodiversity, pollination are being identified as important ecosystem outputs from agriculture.
Carbon sequestration function of land is raising more and more attention because of its

62

important role in mitigating global warming. The number of dimensions of the environmental
outputs is expected to increase with the advances of scientific research. Second, it is very
difficult to set up the quantitative relationships between ecosystem service outputs and land
management decisions. Human-nature interactions are extremely intensive in agricultural
ecosystems. Environmental outputs can be considered as functions of natural inputs and
human management factors. However, these ecological functions are extremely difficult to be
identified because of the complexity of an ecosystem. One management factor change would
involve one or more ecological changes and those ecological changes would interact with
each other. The final “outputs” from the system are the sum of all relevant ecological changes
and their interactions. Scientific literature on ecosystem services from agriculture and
environmentally friendly farming practices have provided important experimental results on
the quantities of one or more measurable outputs from changing farming practices such as
carbon sequestration, nitrogen retention, phosphorus runoff, soil fertility. Those studies were
helpful to address the second problem we just mentioned. However, most of them are onsite
outputs which would go offsite into other ecosystems such as lakes, rivers, atmosphere and
then finally to the general public. Therefore, they are “intermediate” environmental outputs
rather than the endpoint outputs received by the public. The translations from the
intermediate outputs to the endpoint outputs require challenging efforts in examining the
interactions in other ecosystems.
Because of the challenges above, the valuation of ecosystem services flows from
agriculture is rudimentary and further research is required. Despite the difficulties, there are
efforts to value ecosystem services from freshwater ecosystems, wetlands and farmland
amenities. Most of them try to value a set of ecosystem service outputs from the ecosystem.
However, due to the complexity of the good being valued, some studies did not have clear
definitions of the good. It was vaguely defined as “environmental improvements” or

63

“environmental amenities” or a specific public program (Gurluk 2006, Dubgaard 1994,
Bergstrom et al 1985, Drake 1992, Beasley et al 1986, Bowker and Didychuk 1994, Halstead
1984, Pruckner 1995). Relevant ecosystem services were not presented to the public. Some
studies did decompose the good into several ecosystem services outputs and presented the
good to the public but the descriptions of ecosystem services were mostly qualitative (Loomis
et al 2000, Holmes et al 2004, Xu et al 2003, Stevens et al 1995). If the public was not given
sufficient information of relevant ecosystem services, there were potential pitfalls for
misevaluation. They might overvalue or undervalue certain ecosystem services.

1.2.2 Objective of the study
This study explores the willingness of Michigan households to pay for environmental
services. Specifically, we apply the contingent valuation to estimate changes in a set of
ecosystem services that come from switching to low-input cropping systems from
conventional corn-soybean cropping systems in Michigan. The scientific basis for the
changes in ecosystem services for these systems comes from research conducted by the
NSF-Funded Long Term Ecological Research site on row crop agriculture. In order to address
challenges in valuing ecosystem services, we tried to decompose the good into multiple
ecosystem service outputs from low input cropping systems and derive the marginal value of
each output separately. Specifically, although low-input cropping systems offer many
environmental benefits, two are particularly noteworthy to members of the general public,
reduced lake eutrophication and mitigation of greenhouse gas emission. The two ecosystem
services were presented quantitatively to the public in the CV questionnaire.

1.3 Layout of the chapter
The paper is organized as follows. In the next section, valuation literature on relevant

64

ecosystem services from agriculture is reviewed and analyzed. Then the third section comes
with elements in contingent valuation study design and how those elements fit in the
ecosystem service valuation background. In the fourth and fifth section, the Probit model of
contingent valuation and the empirical model are presented respectively. The layouts of
survey design and implementation results are given in the sixth section. Model estimation
results and WTP analysis are discussed in the final part.

2 Current research gaps
2.1 Gaps in valuing ecosystem services from agriculture
There are two basic approaches to value a set of ecosystem service outputs from an
ecosystem. One is to value each ecosystem service output and add the values together. One
example is Costanza’s paper in Nature (Costanza et al 1997) to estimate the economic value
of the world’s ecosystem services. One critique to this approach is that it will lead to double
counting of benefits (Serafy 1998). The summation method is problematic in economic sense
that individuals are facing budget constraints and have to make tradeoffs between different
ecosystem services. Simply adding up the values of each ecosystem services will lead to an
overestimate of the set of outputs (Hoehn and Randall 1989). Another approach is the holistic
approach which takes the set of ecosystem services as one good and estimates its economic
value. The method addresses the double counting issues and the overestimate issues of the
former approach (Loomis et al, 2000). The limitation of the holistic approach is that it
requires valuing the set of ecosystem services with one single valuation method at one time.
One appropriate method for valuing one ecosystem services may not be the same case for
another. For example, the contingent valuation method which is appropriate to value the non
use value of some ecosystem services is not appropriate to value of food production increase,
which is essentially a market good. In its application to empirical studies, there are no studies

65

that value a complete set of ecosystem services outputs with one single method. More
commonly, they value more than one ecosystem services outputs at one time. The good being
valued has multiple dimensions.
Attempts have been made in valuing the ecosystem services from other ecosystems
using the holistic approach.

Some studies value a set of ecosystem services of rivers

(Gurluk 2006, Loomis et al 2000, Holmes et al 2004, Xu et al 2003) and considerable efforts
have been put in valuing ecosystem services from wetlands (reviews from Boyer and Polasky
2009, Brander et al 2006, Brouwer et al 1999, Woodward and Wui 2000). The contingent
valuation method is applied in most of those studies. Some studies value ecosystem service
benefits from an actual developmental project (Gurluk 2006). Ecosystem service outputs are
not decomposed into multiple dimensions but were concentrated in one dimension named
“improved environmental quality”. The definition of the good is vague and was not presented
clearly. The underlying assumption is that the survey respondents have sufficient knowledge
of what they will get and how much they will get. This assumption is rarely sufficient in
practice. Others decompose the set of ecosystem services into multiple dimensions and
present all dimensions to the respondents in the survey (Loomis et al 2000, Holmes et al 2004,
Xu et al 2003, Stevens et al 1995). Those studies have clear definitions of the good being
valued but the levels of each dimension are qualitative. Respondents are told about different
ecosystem services and how they would change in the hypothetical voting scenario but the
descriptions of changes are qualitative. Some do give quantitative descriptions of possible
environmental changes but they are not the direct measurements of final ecosystem service
outputs (Loomis et al 2000). It means that respondents have to make assumptions on the
quantities of each ecosystem service outputs in order to get an exact understanding of the
good. This insufficient information problem may lead to inconsistent estimates of their true
WTPs (Boyle 2003).

66

Valuation literature on ecosystem service from agriculture focuses on valuing farmland
amenities. The farmland amenities being valued in most of the studies are essentially a
package of various ecosystem services such as improved environment, landscape (Drake
1992), aesthetic and cultural values (Bergstrom et al 1985). The hedonic pricing model (Irwin
and Bockstael 2001, Irwin 2002) and the contingent valuation survey (Dubgaard 1994,
Bergstrom et al 1985, Drake 1992, Beasley et al 1986, Bowker and Didychuk 1994, Halstead
1984, Pruckner 1995) are two most commonly used methods to derive the economic value of
farmland amenities. The hedonic model derives the marginal value of farmland amenities
from observed real estate prices. The economic values from the method is the total economic
values of the set of ecosystem services from a piece of farmland but no further decomposition
of the this value into single ecosystem service values is possible. The contingent valuation
method valued people’s willingness to pay for farmland amenities by setting up hypothetical
market scenarios. By decomposing the good into multiple dimensions (the term “attributes”
used in the CV literature), the CV method is able to derive the marginal WTP for each
dimension as well as the WTP for the whole good. However, in previous studies, each
dimension of the ecosystem service outputs was not quantified and valued individually
(Bergstrom et al 1985, Beasley et al 1986). Respondents are asked to vote on land use change
programs but relevant ecosystem services outputs from the programs are not given in the
survey. Similarly, those studies assume that the public have sufficient information on the
categories and quantities of ecosystem services they would get from the whole package of
“environmental amenities” when performing the valuation task. However, this assumption is
debatable. As a matter of fact, the actual quantities of outputs from an ecosystem required
long term scientific research (Magnuson 1990). The scale-up impacts of ecosystem service
change on the general public required even more extensive scientific research on related
ecosystem interactions. Moreover, the value derived from the studies is the total value of a

67

piece of farmland rather than that of one single ecosystem service from the land. The
differences in WTP estimates in those models cannot be quantitatively explained by
differences in ecosystem service outputs, which should be a very important factor in
determining the economic values of farmland.

2.2 Gaps in spike models in contingent valuation
The spike model was introduced to deal with the zero willingness to pay problem in
contingent valuation (Kristrom 1997). A spike probability parameter was introduced to model
the probability of an individual of having zero or negative WTP. The spike probability can be
modeled as one single parameter (Werner 1999, Kristrom 1997) or it can be modeled as a
function of covariates (Strazzera et al 2003, Berrens et al 2001, An et al 1996, Hu 2006). The
covariates used in previous studies were demographic variables. None of those studies
considered the scope effect of environmental change on the spike probability, that is, how the
scale of environmental changes would affect the spike probability. However, from the
comments from our survey, some respondents complained about the scale of environmental
changes and chose not to get it even if it was free. So it is a reason to investigate whether the
scope effect also exists in people’s spike probability, apart from their WTP.

3 Contingent valuation methodologies in valuing ecosystem services
The contingent valuation method sets up a hypothetical market for the public to buy an
environmental good which would increase their utility or accept payments for an
environmental good if it decreases their utility. The willingness to pay or willingness to
accept is considered as an appropriate measure of the economic value of the good.
The biggest challenge of the contingent valuation method in valuing ecosystem services
is to define the “environmental good” to be measured. The good in this case is a set of

68

multiple ecosystem service changes by changing farming practices. It usually has more than
one dimension and each dimension of ecosystem service change has a certain range where the
change can be achieved by changing farm management decisions. Another challenge in
defining the good is that the ecosystem service changes that the public experiences are
different from those that farmers can offer through changes in practices. The environmental
changes that the farmers offer are on-site intermediate changes that would go off-site into
other ecosystems such as waters, soil and the atmospheres and are transformed into other
kinds of ecosystem services when the general public receives them at the end. For example,
farmers reduce on farm nitrogen fertilizer use and it has benefits like less nitrate leaching into
ground water and less NOx emissions into the atmosphere. When these benefits come to the
general public, what they experience are safer drinking water, better perceivable air quality
and reduced health risk. The nitrogen fertilizer use, nitrate leaching and NOx emissions are
finally transformed as the final products of drinking water quality, air quality and health risk
reduction. The set of final products are directly experienced and consumed by the public.
Therefore, it is the environmental good that we would like to describe to the public and ask
them to place their value on. We should calibrate the final products with the intermediate
products that farmers offer. Preliminary studies on the science of the ecosystem service
changes are necessary to determine the possible dimensions and ranges of those ecosystem
service changes.
In the description of the environmental good in the survey, detailed and specific
information is necessary, especially when people were not aware how they may benefit from
the change. The survey should provide the range of changes people are asked to value, how
those changes are going to affect their life (Boyle 2003). These rules should apply to the
description of multiple ecosystem service changes as well because the good has multiple
dimensions and is complicated, rather than one single environmental change. Questions to

69

probe respondents’ understanding of the changes are necessary in assessing the validity of the
valuation. (Boyle 2003, Arrow et al 1993) Specified baseline and measure of changes will
help individuals to make tradeoffs between benefits from various ecosystem services (Toman
1998, Loomis et al 2000).
The design of provision mechanism should balance between credibility and protest. It
should be credible to respondents. If it is not real enough, some people may not take the
valuation question seriously. However, if the provision mechanism is realistic, some people
may protest to the whole scenario just because they dislike this real provision mechanism.
(Boyle 2003). When designing the provision mechanism in ecosystem service changes from
agriculture, some politically sensitive issues should be taken into consideration. Agricultural
subsidies for farmers have been a controversial issue especially when the public are worrying
the huge deficit in the government budget. The idea of “paying farmers to reduce pollution”
may generate protests among those who are against agricultural subsidies.
The selection of payment vehicle should also consider the tradeoff between credibility
and payment vehicle rejection (Boyle 2003). Moreover, the selection of payment vehicle
should be incentive compatible. Some of the ecosystem service changes from agriculture are
semi public goods such as better drinking water quality but many are pure public goods such
as global warming mitigation, reduced soil erosion, lake water quality etc. Therefore, the
payment vehicle should avoid the strategic behaviors of free riders. On the other hand, tax as
the payment vehicle may exclude the free rider behaviors in the survey and makes the
scenario realistic enough to vote on. However, during economic recession when our survey
was completed, raised tax is a very sensitive issue and may generate considerable protests.
The time frame of payment should match the benefit period of the respondents so that
they don’t have to discount the cost and benefit themselves (Boyle 2003). To match the two
periods is not straight forward in valuing ecosystem services from agriculture but can be a

70

very complicated issue. Some issues are noteworthy. One is that there is time lag for actual
ecosystem service changes to happen. The lags may vary from one or two years or more than
twenty years to take effect. For example, nitrate concentration in ground water is a long term
evolving process that may take decades. A longer time lag may happen between the reduction
of greenhouse gas reduction and global temperature decline. Another challenge is the
accumulative effect of actual ecosystem service changes. The scale of benefits may change
overtime and the change may not be linear. Due to the complexity and uncertainty with actual
ecosystem service changes, it is very difficult to give a time frame that completely matches
the benefit period in the survey. Instead, it can be largely simplified as an infinite period and
the changes would happen as long as the payment was in place.
Because the majority of the ecosystem service changes from agriculture are public
goods, the dichotomous choice referendum is more incentive compatible and better avoids
free riding or strategic overbidding tendencies than other choices such as open ended
payments and payment card (Bateman and Turner 1993, Boyle 2003). The anchoring effect is
also minimal among all response choice format choices (Boyle 2003). However, the
yea-saying bias may occur in the dichotomous choice response format and results in the fat
tail of the WTP distribution (Desvousges et al 1993). But the yea-saying bias can be reduced
by providing respondents who vote ‘no’ opportunities to show support to the program
(Blamey et al 1999).
Follow-up questions after the WTP question are recommended in order to better
understand people’s responses to the WTP question (Arrow et al 1993). Respondents may
have uncertainties when they first deal with complicated ecosystem service valuation tasks,
so in the mail survey, closed ended follow-up questions on their reasons for the WTP question
may induce the respondents to vote differently from their true intention. Open ended
follow-up questions give respondents opportunities to express their ideas freely without

71

inducing them to some pre-assumed reasons. Lower protest rate may be achieved when
allowing people to express ambivalent and other opinions (Blamey et al 1999, Ready et al
1995). However, it may suffer from low response rates of that question. Certainty follow-up
questions asking how certain the respondents are with their answers are helpful in ecosystem
service valuation because respondents may be uncertain in valuing complex environment
goods. It also has nice properties in data estimation and lower protest rate (Champ and
Bishop 2001).
The identification of protesters is important in the CV survey of ecosystem service
valuation because the goods to be measured contain politically sensitive issues such as global
warming and agricultural subsidies. If a considerable amount of protesters were not excluded
from the sample, the estimated WTP would be underestimated. Screening questions for
protesters and misleading responses includes (Boyle 2003, Arrow et al 1993)
•

Whether they doubt the feasibility of the proposed action

•

Whether they refuse to accept the component of the scenario (provision mechanism,
payment vehicle)

•

Whether they feel that their votes will have no significant effect on the outcome of the
hypothetical referendum

•

Whether they believe in the scenarios

•

Whether they understand the CV question

•

Whether they take the CV question seriously
The above screening questions in the mail survey may undermine the CV scenario.

Therefore, they should be carefully worded and put in the right section of the survey.
The attribute based method (ABM) is used when valuing multiple ecosystem service
changes. The ABM method is to estimate the economic value of an environmental good by
dividing the good into several attributes and varying the levels of the attributes in different

72

survey versions (Holmes and Adamowicz 2003). The ecosystem service changes to be valued
have multiple dimensions in nature, so ABM turns out to be the appropriate method. By
changing the quantities of the good presented to the respondents, the ABM can test the scope
effects in the CV survey which results from inability of the respondents to know the
difference in quantity changes (Carson and Hanemann 2005).
The CV survey to value ecosystem service changes should reduce the hypothetical bias
as much as possible. There are several choices to solve the problem. Cheap talk design which
gives a short speech on the hypothetical bias problem to the respondents to reduce
hypothetical bias in lab experiments CV (Cummings and Taylor 1999) and some mail CV
survey (List et al 2006) but failed in the other (List 2001). It did not reduce WTP for
knowledgeable consumers (Lusk 2003, List 2001). Moreover, it is too long for respondents to
read through in mail survey. Certainty follow-up question is able to identify certain
boundaries of dichotomous choices (Champ and Bishop 2001). Similarly, polychotomous
choice format was able to set up certain boundaries of WTP by allowing respondents to
choose from a range of intensity of their votes (Ready et al 1994).
Finally, in the CV survey, sampling population should be relevant to the benefits and
costs of the environmental good being valued (Carson 2000). The beneficiaries of most
ecosystem service from agriculture are the general public. Therefore, the sample should be
drawn from all residents.

4 Theoretical model
In the neoclassic economic analysis framework I’m using, the consumer behavior is
motivated by utility maximization, which is defined by

73

Max U ( X , q )
X

S .t. I ≥ rX
q≥0

(1)

Where U denotes consumer’s utility, X denotes the quantities of market goods
consumed, q denotes the environmental quality, r denotes the price of market goods and I
denotes the consumer’s income. By solving equations (1), we can derive the indirect utility
function. The duality of the utility maximization problem is to minimize consumer’s
0

expenditure, given a certain level of utility U . We can rewrite this problem as

Min E ( X , q ) = rX
X

S .t. U ( X , q ) ≥ U 0
q≥0
The expenditure function

(2)

(

E U 0 , r, q

)

is derived from equations (2), which is

0

the minimum cost to get to utility level U .
If environmental quality changes from

q0

to

q1,

consumers’ expenditure to get to

0

utility level U will change too, then their WTP to pay for the environmental quality changes
can be defined as

(

) (

WTP = E U 0 , r , q0 − E U 0 , r , q1

)

(3)

It is possible that people will have zero or negative WTP. People with zero WTP for a
particular good have the utility curve that won’t change no matter how many of the good they
consume and they would not pay anything for it. Negative WTP means that the increase in
the good would reduce people’s utility. The zero and negative WTP respondents should be

74

distinguished from those who vote “no” to the program but are willing to pay for it at a lower
price. For the latter, the good contributes to their utility and they have a positive WTP. The
conditional estimates of WTP would be lower if zero and negative WTP respondents were not
excluded in the estimation (Haab and McConnell, 2002).
The paper by Kristrom (1997) introduced the spike model to deal with zero and
negative WTP problems in the contingent valuation surveys where the zero and negative
WTP respondents can be identified in the survey. Here we discuss the zero WTP case only
because eutrophic lakes reduction and greenhouse gas reduction are unlikely to reduce
people’s utility, if not to increase it. In the spike model, another parameter p , defined as the
probability of a zero WTP to occur is introduced into the linear utility model. Therefore, the
cumulative distribution function of WTP is

p
if WTP = 0
⎧
G (WTP ) = ⎨
⎩(1 − p )D (WTP ) if WTP > 0
where

D(WTP )

is the distribution of WTP conditional on

(4)

WTP > 0

Considering the spike probability suggested by Kristrom, the theoretical model in the
paper is set to model two level decisions of the respondents, the spike probability and their
WTP conditional on positive WTP.

4.1 WTP conditional on WTP>0
Suppose that conditional WTP is a random variable whose value depends on the

distribution of the stochastic term

u ij . Specifically, we assume WTP is positive therefore

the exponential form is imposed to make sure the WTP function is positive

75

uij ⎞
⎛δ β
γ
⎟
WTPi | WTPi >0 = exp⎜ + Δq j + Z i +
⎜α α
α
α ⎟ (5
⎝
⎠
)
where WTP is the person’s willingness to pay for environmental improvements, Δq
denotes change in environmental quality, Z includes other explanatory variables and u is
the random term. Subscript i denotes the respondent, subscript j represent status j (after
paying for environmental improvement). δ, α, β and γ are fixed parameters,

To choose between the status quo and status j, denote the choice variable as

cost of the improvement as

t i . Yi

Yi

and

equals 1 if the respondent chooses to pay for

environmental improvements and equals 0 if the respondent declines to pay.

P (Yi = 1 | WTP > 0 ) = P{
WTPi > t i }
⎧ ⎛δ β
⎫
u ij ⎞
γ
⎪
⎪
⎟ > ti ⎬
= P ⎨exp⎜ + Δq j + Z i +
α
α ⎟
⎪ ⎜α α
⎪
⎠
⎩ ⎝
⎭
(6)
= P (δ + β Δq j + γ Z i + u ij > α ln t i )

= P (δ − α ln t i + β Δq j + γ Z i + u ij > 0 )
If we assume that

u ij has a N (0, σ 2 )

distribution,

u ij
⎛ δ − α ln t i + β Δq j + γ Z i
P(Yi = 1 | WTP > 0) = P⎜
>−
⎜
σ
σ
⎝

β
γ ⎞
⎛δ α
= Φ⎜ − ln t i + Δq j + Z i ⎟
σ
σ ⎠
⎝σ σ
76

⎞
⎟
⎟
⎠
(7)

Where

Φ(⋅)

is the cumulative distribution function of the standard normal

distribution. This is the basic Probit model for binary responses. From equation (7) we can
derive the log likelihood function and get the maximum likelihood estimates of parameters

δ /σ , α /σ , β /σ

and

γ /σ

.

4.2 Spike probability
Instead of modeling the spike probability as one parameter p, the paper assumes
respondent’s spike probability is a function of both environmental change variables and other
covariates:

p (Δq j , Z i ) = P (WTP = 0) = Φ (a + b Δq j + c Z i )

(8)

Assuming no covariance between people’s conditional WTP and spike probability,
equations (7) and (8) can be estimated independently.
With the spike probability, the probability that a respondent would vote yes to a
program was a product of the probability of voting yes conditional on positive WTP (equation
7) and one minus the spike probability (equation 8), which can be written as

P(Yi = 1) = P(Yi = 1 | WTP > 0) × P(WTP > 0)
uij ⎞
⎛ δ − α ln ti + β Δq j + γ Z i
= P⎜
> − ⎟ × (1 − p(Δq j , Z i ))
⎜
σ
σ⎟
⎝
⎠
β
γ ⎞
⎛δ α
= Φ⎜ − ln ti + Δq j + Z i ⎟ × (1 − Φ(a + b Δq j + c Z i ))
σ
σ ⎠
⎝σ σ
(9)

4.3 Welfare calculation
To simplify notation, we rewrite the conditional WTP function (equation (5)) as
77

uij ⎞
⎛δ β
γ
⎟
WTPi | WTPi > 0 = exp⎜ + Δq j + Z i +
⎜α α
α
α ⎟
⎝
⎠
= exp δ * + β *Δq j + γ *Z i + α *uij

(

)

Then we can compute the median and expectation of the conditional

(10)

WTPi .

(

Median(WTP | WTPi>0 ) = exp δ * + β *Δq j + γ *Zi
i

)

(11)

⎛ *
(α *σ ) 2 ⎞
*
*
⎟
E(WTP | WTPi>0 ) = exp⎜ δ + β Δq j + γ Zi +
i
⎜
2 ⎟
⎝
⎠
(12)
The unconditional mean WTP can be computed as the following

E(WTP ) = (1 − p )E(WTP | WTP > 0)
i
i
i
⎛ *
(α *σ ) 2 ⎞
*
*
⎟
= 1 − p(Δq j , Z i ) exp⎜δ + β Δq j + γ Z i +
⎜
2 ⎟
⎝
⎠

[

In

]

the

fixed

p (Δq j , Z i ) = p .
function of

spike

probability

case,

p (Δq j , Z i )

In the endogenous spike probability case,

Δq j , Z i .

To compute the unconditional median WTP, remember that

G (WTPmedian ) = 0.5
Then

WTPmedian = 0

if

p ≥ 0.5 .
78

is

a

(13)

constant

p (Δq j , Z i )

is a

p < 0 .5 ,

For

⎛ *
⎛ 0.5 − p(Δq j , Zi ) ⎞ ⎞
⎜δ + β *Δq + γ *Z + α *σ Φ−1 ⎜
⎟⎟
WTP
median = exp
j
i
⎜ 1 − p(Δq , Z ) ⎟ ⎟
⎜
j
i
⎝
⎠⎠
⎝
(14)

4.4 Econometric model: random effects probit
The estimation of the model depends on the distribution assumption of random term

uij . Previous discussions are based on the assumption that u ij

is independently and

identically distributed. However, in most CV surveys, more than one program is offered to a
respondent. It means that the actual data from the survey will have more than one observation
for each respondent i. Then the identically and independently distributed (i.i.d.) assumption

of

u ij

may not be satisfied. We need to estimate the model using the panel data

econometric method.
In the panel data case, we assume that the stochastic part has two components.

u ij = μ i + ν ij
where

μi

(15)

is the unobservable individual specific error and

ν ij

denotes the

remainder disturbance. Assume that the latter is i.i.d. with a normal distribution of mean zero.

Further assumptions should be made on whether

variable in order to estimate the model. If

μi

79

μi

is a fixed parameter or a random

is treated as a fixed parameter, it is called the

fixed probit analysis. However, estimating

μi

with other parameters introduces the

incidental problem which leads to inconsistent estimates of other parameters (Wooldridge

2002). Therefore, instead of assuming

μi

is a fixed parameter; the random effect probit

analysis treats it as a random variable. Further, it assumes that

other explanatory variables and has a distribution of

distribution of

f

(Y

i1

, Yi 2 , K Yij ,K , YiT )

(

μi

is independent from

N 0, σ u2

)

, Then the joint

conditional on t, q and Z, denoted by

can be written as

f (Yi1 , Yi 2 ,..., YiT | t , q, Z ;θ )
∞

(

)

⎡T
*
*
*
* ⎤
= ∫ ⎢ ∏ f Yij | tij , qij , Z i ; δ ,α , β , γ ⎥
j =1
⎦
∞⎣
× (1 / σ μ )φ (μ / σ μ )dμi

80

(16)

(

f Yij | tij , qij , Z i ; δ * , α * , β * , γ *

)

β
γ
⎞
⎛δ α
= Φ⎜ − ln tij + qij + Z i + μi ⎟
σ
σ
⎠
⎝σ σ

Yij

⎡
β
γ
⎞⎤
⎛δ α
× ⎢1 − Φ⎜ − ln tij + qij + Z i + μi ⎟⎥
σ
σ
⎠⎦
⎝σ σ
⎣

(

= Φ δ − α ln tij + β qij + γ Z i + μi

[

**

(

**

**

**

)

Yij

× 1 − Φ δ − α ln tij + β qij + γ Z i + μi
**

**

**

**

(1−Yij )

)]

(1−Yij )

(17)
where

θ

is the vector of all parameters to be estimated which contains

δ ** ,α ** , β ** , γ **

and

σμ .

Then if we plug in N observation values of

(Y

i1

, Yi 2 ,KYij ,K , YiT ), then we

can derive the log likelihood function for equation (7) and (8) and obtain the maximum
likelihood estimators for the parameters.

5 Model specification: Description of explanatory variables and hypotheses
In the theoretical model described in section 4, there are three large categories of
explanatory variables that would have an impact on respondent’s probability of voting “yes”
and their WTP to pay for the ecosystem service changes. They are ecosystem service changes
variable Δq , cost variable t and other explanatory variables Z .
81

Ecosystem service change variables were determined by the preliminary study which
scaled up the outputs of several ecosystem service changes in Michigan if selected low input
cropping systems were to be adopted. The details of this study were elaborated in Chapter 1.
Chapter 1 concluded that two categories of ES changes merited further data collection
and analysis: eutrophic lakes reduction and reduction in global warming potential. These two
variables constructed the ES change variable vector q. In the empirical model, we divided Z
into four subcategories of variables: version variable, demographic variables, related
experience variables and attitude variables. Each subcategory contains one or more than one
variables. The descriptions of each variable under each category and subcategory are
presented in Table 2.1.

6 Survey design and implementation
The questionnaire had five parts. The first part was the information section about lake
eutrophication and global warming. Respondents were given information about the current
status, how their life would be affected and how land management practices would affect
those environmental qualities. The second part gave respondents the maximum range of those
ES changes that a land management program could bring. The third part was the referendum
CV questions which asked the respondents to vote on three proposed land stewardship
programs. The fourth part was the attitude questions that probed respondents’ attitudes on
various environmental issues. The last part was the demographic questions.

6.1 Survey design
6.1.1 Survey attributes and values
The survey design used the attribute based method (ABM). ABM separates an
environmental good into multiple dimensions and varies in each dimension to form different

82

states of the good as a choice set for the respondents (Holmes and Adamowicz 2003). The
dimensions of the good were called “attributes”, the states of each dimension were “values”
of an attribute.
The first two survey attributes are eutrophic lakes and greenhouse gas reduction. In
Chapter 1 we estimate the quantities of five ecosystem service changes from low input
cropping system, which are reduction in drain dredging cost and flood damage cost, eutrophic
lakes, high nitrate wells and greenhouse gas reduction. The benefit for reducing health risks
of high nitrate wells can be estimated with other methods such as benefit transfer. So
reduction in high nitrate wells is not included in the contingent valuation survey either. The
cost savings in drain dredging and flood damage are already estimated in dollar term, it is
problematic to ask the respondents to place a monetary value on a dollar cost saving in the
contingent valuation survey. As a result, reductions in drain dredging cost and flood damage
cost are not included in the contingent valuation survey either. Finally, the set of ecosystem
service changes to be measured in the contingent valuation survey has two components:
reduction in eutrophic lakes and global warming potential. These two components construct
the ES change variable vector q in the empirical model. The maximum reduction of
eutrophic lakes that low input cropping systems could bring is about 200 and the maximum
reduction of greenhouse gas was 0.6% of the 2000 emission level in Michigan (see results in
Chapter 1). The basic values of the two attributes had five levels: zero change, low change,
median change, high change and double change. The “double change” level was the double
of the maximum reduction.

3

The values are presented in Table 2.1.

Another attribute in the survey design was respondent’s share of cost for the proposed

3

In pretest interviews in the contingent valuation survey, some respondents reported that the
ecosystem service changes were too small although they still passed the scope test. To reduce
the probability of scope insensitivity problem, we doubled the original maximum change as
the new range of the two attributes.
83

land stewardship program. This attribute was the cost variable in the empirical model. The
cost was expressed as the respondent’s own share of increased annual federal income tax. The
levels of the cost were $10, $30, $50, $100, $200 per year. The cost levels were set according
to the data from the questionnaire pretest interviews. From the pretest data, we made the
hypothesis that the WTP of the respondents had a positively skewed distribution with median
WTP at around $50.
The last attribute in the CV survey was the provision mechanism to provide such
ecosystem service changes. To test the hypothesis that the provision mechanisms in the WTP
question has significant effect on respondents’ WTP and our interests in agricultural
ecosystem services, two alternative provision mechanisms were devised. One version of the
questionnaire told the respondents that the land stewardship program was to pay farmers to
adopt environmentally friendly farming practices to achieve the stated ecosystem service
changes (“agricultural-farmer”). In the information section of that version, respondents were
told about the agricultural contribution to lake water quality and global warming. The other
version of the questionnaire told them that the program was to pay general land owners to
change their current land management practices (“land management”). The matching
information section told respondents about the land management contribution to the two
environmental qualities.

84

Table 2.1 Explanatory variables, definitions, unit of measure and levels of survey attributes , survey of 6000 Michigan households, autumn 2009

Variable
Definition
name
Environmental change variables
lake

Eutrophic lakes that would be reduced if the program
were to be implemented

GHG

Unit

Greenhouse gas reduction compared with the 2000
emission level that would be achieved if the program
were to be implemented

Survey
Attribute levels
attributes

Yes

0, 70, 140, 200, 400

%

Yes

0, 0.2, 0.4, 0.6, 1.2

$/year

Yes

10, 30, 50, 100, 200

Yes

0-land management, 1-farmer

Cost variable
cost

The amount of annual tax increase that would be used to
fund the program

Other explanatory variables
Version variable
farm

Whether the questionnaire version is the
agricultural-farmer version or the general land
management version

Demographic variables

85

Table 2.1 (cont’d)
income

The household pretax income of the respondent

$/year

No

age

Age of the respondent

Year

No

gender

Gender of the respondent

No

1-male, 2-female

education

Education level of the respondent

No

1-Some high school or less, 2-high school
diploma, 3-technical training beyond high
school, 4-some college, 5-college degree,
6-some graduate work, 7-graduate degree

LenResi

Expected length of future residency in Michigan

No

1-less than 1 year, 2-1-5 years, 3-5-10 years,
4- more than 10 years

FamSize

Size of the respondent's household

No

farmer

Whether the respondent is a farmer or works on a farm

No

1-yes, 0-no

EnvOrg

Whether the respondent belongs to an environmental
organization

No

1-yes, 0-no

forest

Whether the respondent is a forester or works in forests

No

1-yes, 0-no

voter

Whether the respondent is a registered voter

No

1-yes, 0-no

resi

Whether the respondent considers himself/herself as
Michigan resident

No

1-yes, 0-no

86

Table 2.1 (cont’d)

Related experience variables
fish

Frequencies of the respondent goes fishing in inland
lakes in Michigan

No

1-never, 2-in some years, 3-in most years,
4-every year

swim

Frequencies of the respondent goes swimming in inland
lakes in Michigan

No

1-never, 2-in some years, 3-in most years,
4-every year

boat

Frequencies of the respondent goes boating in inland
lakes in Michigan

No

1-never, 2-in some years, 3-in most years,
4-every year

hike

Frequencies of the respondent hike near inland lakes in
Michigan

No

1-never, 2-in some years, 3-in most years,
4-every year

No

1-yes, 0-no

Attitude variables
gw

Whether the respondent is concerned about global
warming

87

6.1.2 Hypothetical WTP questions design
The contingent valuation (CV) questions used as the policy context a land stewardship
program that would cover land owners’ cost of adopting environmentally friendly land
management practices. The program would achieve different levels of reduction in high
nutrient lakes and greenhouse gas emission, depending on its cost.
The CV question used the dichotomous choice referendum format. Respondents were
asked to vote on three separate land stewardship programs that reduced high nutrient lakes
and greenhouse gas emission at different levels and all three programs cost the same
(however, the cost varied among different versions of the questionnaire). If more than 50%
percent of the voters voted on the program, it would be implemented and they would have to
pay the cost.
The payment vehicle in the CV questions was expressed as a fixed amount of federal
income tax increase. To exclude the externality of the federal income tax, respondents were
told that money raised in Michigan would only be used in Michigan. The payment was
annual and the environmental improvements would be achieved as long as the income tax
was in effect.
If the respondent voted “no” for the WTP question, a follow-up question asked them
whether they would vote for the program if it did not cost them anything. This question was
used to identify the respondents with zero WTP. It is necessary to set up zero or negative
WTP questions because large proportion of zeroes or a negative WTP will have a significant
effect on the estimation of the model.
Another certainty follow-up question asked how certain they were of their answers to
the WTP question. Certainty follow-up questions had lower protest/non response rates
because they allowed respondents to express ambivalence. Certainty follow-up questions was

88

also able to identify certain boundaries of dichotomous choices (Champ and Bishop 2001)
There was an open ended question probing on their reasons for their vote on the WTP
question. It was used to probe on protesters against the payment vehicle, the provision
mechanism and any other elements in the hypothetical scenario. It can be used to identify
respondents with positive, zero or negative WTP.

6.1.3 Experimental design of attributes
The experimental design procedure was used to construct an attribute combination plan
to be presented to the respondents (Holmes and Adamowicz 2003). The full factorial design
allows all the main effect and interaction between attributes to be estimated in the model
(Holmes and Adamowicz 2003). However, the full factorial plan in this case has 54=625
combination. Due to the budget constraint of the survey, a fractional design plan was adopted,
dropping some of the interaction effects between attributes. The main effect experimental
design plan of the four attributes should satisfy several conditions and constraints.
First, for each version of the questionnaire, each respondent was given three referendum
4

CV questions to vote on . So in each version of the questionnaire, the farm variable should
be the same for all three CV questions.
Second, in each version of the questionnaire, the cost variable was the same for all three
CV questions. Having the same cost for the three different programs made respondent’s
valuation task easier because they had to compare the eutrophic lakes and greenhouse gas
changes to only one cost.
Third, in the proposed program to vote on, at least one level of the attributes lake and
GHG is larger than zero. The reason is straightforward: it would make no improvement in the
environment if both eutrophic lakes and greenhouse gas emission did not change; it would
4

Three CV questions are sufficient for panel data analysis on individual’s WTP.
89

just be the status quo.
Fourth, in the same version of the questionnaire, one program does not dominate
another, which means that not both the lake and GHG attribute values in one program are
higher or lower than those in another program.
Fifth, if possible, the numbers should be evenly distributed over the ranges of lake and
GHG reduction.
In addition, there is no substantial correlation between the question sequence within a
questionnaire (i.e., first, second or third) and the design attributes.
Finally, the total correlations between all four attributes are as small as possible to make
sure all combinations are orthogonal.
The initial main effect orthogonal plan of 24 attribute combinations was generated by
SPSS 17.0. Then it was adjusted to satisfy the above seven constraints. To lower the
correlations between attributes, the lake and GHG attribute values were allowed to adjust to
represent a continuous change from zero to double change and not confined within the
original five levels. The continuously changing lake and GHG variables more closely
resemble the actual situation where environmental qualities were changing gradually and
continuously. The final plan had 42 attribute combinations in 14 versions of the questionnaire
(Appendix 15). It had a correlation matrix whose off diagonal correlations were all below
0.05.

6.1.4 Validation questions design
A set of Questions screen for protest and types of misleading responses (Boyle 2003,
Arrow et al 1993) were included in the questionnaire. Those questions aimed to test:
whether respondents refuse to accept the component of the scenario (provision
mechanism, payment vehicle)

90

whether they feel that their votes will have no significant effect on the outcome
of the hypothetical referendum
whether they believe in the scenarios
Those questions were placed in the attitude questions in part four so that they were least
likely to undermine the hypothetical CV questions. The screening questions, together with
people’s comments on the open ended questions, would effectively screen out the above three
types of misleading responses.

6.2 Survey pretest
Five rounds of questionnaire pretest were conducted during May and June 2009. People
were randomly intercepted at campus food court and two shopping malls in the Greater
Lansing area in Michigan. Fifty-one randomly intercepted respondents and four graduate
students at Michigan State University participated in the pretest. The pretest method was the
questionnaire booklet and then the one-to-one semi-structured in-person interview. Both the
questionnaire booklet and the interview script were given to interviewers. Each randomly
intercepted respondent was paid for their time and effort in the pretest.
The primary goal of the survey pretest was to test the validity of the CV questionnaire.
The key validation criteria were (Arrow et al 1993, Boyle 2003, Carson 2000)
respondents understand the CV question
they take the CV question seriously
they understand the environmental improvements described
they understand the valuation tasks
they are sensitive to the scope of environmental changes
After five rounds of pretest and further changes to the questionnaire, the above five
criteria were satisfied.

91

The pretest data on the WTP questions was also used to make hypothesis on ranges of
people’s WTP. The data was used to set up levels of the cost attribute in the questionnaire
design.

6.3 Survey implementation
The survey population is Michigan residents because they are beneficiaries of the
ecosystem service changes to be measured. The sample was purchased from Survey
Sampling International Inc. The sample was randomly selected from Michigan landline
telephone directory by Survey Sampling International Inc. It was proportional to the number
of households in each county. The sample size was 6000.
The survey mailing followed a modified version of Dillman’s tailored design method
(Dillman 2000). We had five contacts with the respondents. Personalized prenotice letters
with a printed, blue ink signature were sent to respondents a week before the questionnaire
packages were mailed out. Then the first round of the questionnaire came two weeks after the
prenotice letter. The questionnaire package contained a personalized cover letter with blue
printed signature, a questionnaire booklet and a business reply envelope. Then personalized
reminder postcards with blue printed signature were sent one week after the questionnaire
package. Four weeks later, the second round of the questionnaire package with a revised
cover letter was sent as the fourth contact. Another reminder postcard was sent four weeks
after the second round questionnaire as the final contact.
There were 480 bad addresses in the sample address list and the bad address rate was
8.0%. Forty respondents from the sample were reported dead and forty respondents had
moved out of Michigan. Forty-nine respondents refused to answer the survey. Final responses
were 2215 for a final response rate of 41% (Appendix 7).

92

7 Data analysis
7.1 Protest votes, misunderstanding and other misleading votes
To probe on people’s reasons for their votes, we put an open ended question “what were
your top two reasons for your vote in Question X” after each voting question. Eighty-five
percent of the respondents who answered the questionnaire wrote some comments in at least
one of the three follow-up questions. There were rich information to analyze their reasons as
well as to identify protests, misunderstanding and other misleading votes. We used the
content analysis method to analyze this qualitative data. From the data we extracted 32
themes that were related to their reasons for the votes (Appendix 5). From these 32 themes,
there were 6 themes that indicated protests, misunderstanding and other misleading votes.
The six themes were “Disagree with government regulation / waste”, “Reject provision
mechanism”, “rejection of tax”, “Reject the regime of program or payment”, “Don't believe
the scenario” and “clear misunderstanding”. Responses that fell within the first four types
(“Disagree with government regulation / waste”, “Reject provision mechanism”, “rejection of
tax” and “Reject the regime of program or payment”,) were considered as protests. Those that
belonged to the last type “clear misunderstanding” were misunderstanding votes. Those who
mentioned the “Don’t believe the scenario” theme did not think that the scenario was true and
it was another type of misleading votes that needed to be screened out.
Taxes and government intervention took up the largest proportion of the protests. About
10.1% of the votes rejected taxes and 6.7% of the votes rejected government intervention.
Other protests against provision mechanism and regime of the tax were low. Protest votes
took up 18.6% of the total. There were only 2.2% disbelief votes and less than 1%
misunderstanding votes. The total of the six themes took up 22% of the total. The protest rate
was moderate, considering the economic recession at the survey time.

93

7.2 Descriptive statistics of explanatory variables
To identify respondents with zero WTPs, another follow-up question “would you vote
for this program if it did not cost you anything?” was asked after a respondent voted “no” on
the voting question. Excluding the protest, disbelief and misunderstanding votes, 16.9 % of
the rest of the votes were “no” answers. These answers implied that about 17% of the
respondents felt no increase in their utility when eutrophic lakes and greenhouse gas emission
were reduced. The proportion of zero WTP is significant; therefore we would expect that zero
WTPs would have a significant impact on model estimation and WTP estimates.
After screening out the protest, disbelief, misunderstanding and zero WTP votes, the
percentage of “yes” votes declined as the cost of the program increased and correspondingly
the percentage of “no” votes increased (Figure 2.1).

The result confirmed our expectations

that the respondents were sensitive to the price or cost of the program.
Other descriptive statistics were presented in Table 2.2. The total number of votes was
4128. Among the 4128 votes in the sample, over half of the votes were “yes”. Further, 17% of
the total respondents had zero WTP. If the zero WTPs were subtracted from the 47% of the
“no” votes, we can see that only 30% voted “no” conditional on being a respondent with
positive WTP.
The “LenResi” variable stands for expected future residency in Michigan (Table 2.1).
The mean of 3.7 (Table 2.2) showed that the majority of respondents in the sample would
continue to live in Michigan for 5-10 years or more. Variable “resi” indicates whether the
respondent considered himself Michigan resident (Table 2.1). The mean of 0.99 (Table 2.2)
showed that 99% of the respondents considered themselves Michigan residents. The statistics
of “LenResi” and “resi” indicated that the majority of the sample were residents and would
continue to live long in Michigan. The results of our survey reflected local residents’
long-term interests and concerns about Michigan.

94

From Table 2.2, we found that the “gender” variable was slightly skewed. More men
than women were in the sample. Similarly, the “age” variable was skewed towards the
middle-to-old aged respondents with a mean age of 55. Family sizes (variable “FamSize”)
were small with an average of 2-3 people. Although not presented in Table 2.2, another
frequency analysis of the sample showed that the household pretax income distribution
(“income” variable) had a right skewed distribution with median income of $42500. It was
consistent with the $48,606 household median income statistics from the Michigan census
data (USDA, 2008).
About 95% of the sample was voters (“voter” variable in Table 2.2). This statistics gave
credit to the validity of the CV scenario. Because most of the sample was voters, the
referendum format in the CV scenario was credible to the respondents and they tended to
treat the hypothetical CV questions more seriously.
The sample had very low proportion of farmers (variable “farmer”), foresters (variable
“forest”) and members of environmental organizations (variable “EnvOrg”). Neither of them
made up more than 10% of the sample.
The respondents showed moderate interests in lake related activities in the sample. The
mean values of variables “fish”, “swim”, “boat” and “hike” were all around 2.3, which means
that average respondents in the sample engaged in recreational activities in inland lakes in
some years or most of the years. On the other hand, the respondents had a neutral to negative
attitude towards global warming (variable “gw”). Only 40% of the sample said they were
concerned about global warming and 60% were not. If people are skeptical towards global
warming, they are less likely to be willing to pay for the greenhouse gas reduction and be
sensitive to the changes of it in the voting questions.

95

Figure 2.1 Frequencies of votes by program costs

Frequencies of Votes by Program Costs
900
800

Frequencies

700
600
500

No

400

Yes

300
200
100
0
10

30

50

100

Cost

96

200

Table 2.2 Definition, observation, mean, standard deviation, minimum and maximum values
of variables
Variable
ref

Definition
Respondent's vote on the proposed
program. Ref=1 if yes

Obs Mean
4128 0.53

SD
0.50

spike1

Whether a respondent has zero WTP
spike1=1 if has zero WTP

1430

0.17

0.37

0

1

farm

Questionnaire version (farmer/land
management)

1430

0.48

0.50

0

1

lncost

Logarithm of the annual tax increase to
fund the program

1430

3.70

0.99

2.30

5.30

lake

Eutrophic lakes that would be reduced if the 1430
program were to be implemented

169

111

0

400

GHG

Greenhouse gas reduction compared with
the 2000 emission level that would be
achieved if the program were to be
implemented

1430

0.53

0.32

0

1.2

LenResi

Expected length of future residency in
Michigan

1430

3.7

0.69

1

4

resi

Whether the respondent considers
himself/herself as Michigan resident

1430

0.99

0.10

0

1

gender

Gender of the respondent

1430

1.34

0.47

1

2

FamSize

Size of the respondent's household

1430

2.55

1.37

0

9

age

Age of the respondent

1430

54.8

15.3

13

96.5

farmer

Whether the respondent is a farmer or
works on a farm

1430

0.04

0.19

0

1

forest

Whether the respondent belongs to an
environmental organization

1430

0.02

0.13

0

1

EnvOrg

Whether the respondent is a forester or
works in forests

1430

0.08

0.27

0

1

97

Min Max
0
1

Table 2.2 (cont’d)
income

The household pretax income (in $1,000)

1430 68.63

50.24

5

250

education Education level of the respondent

1430

4.27

1.73

1

7

fish

Frequencies of the respondent goes fishing
in inland lakes in Michigan

1430

2.18

1.17

1

5

swim

Frequencies of the respondent goes
swimming in inland lakes in Michigan

1430

2.38

1.14

1

5

boat

Frequencies of the respondent goes boating 1430
in inland lakes in Michigan

2.43

1.11

1

5

hike

Frequencies of the respondent hike near
inland lakes in Michigan

1430

2.25

1.15

1

5

voter

Whether the respondent is a registered voter 1430

0.95

0.23

0

1

gw

Whether the respondent is concerned about
global warming. Gw=1 if concerned

1430

0.40

0.49

0

1

1430

0.21

0.33

0

1.2

gw_GHG Product of gw and GHG

98

7.3 Conditional WTP model
The “lncost” variable was also significant in the model. The higher the program cost,
the less likely people would vote yes. The result was consistent with the frequency analysis of
Figure 2.1. When the cost was getting higher, people with lower WTP than the given cost
would vote “no”.
The coefficient of variable “lake” was significant in the conditional WTP equation. It
was significant because people were concerned about lakes. Descriptive statistics in Table 2.2
showed that the majority had favorable view on lake related activities. When given the
statement “Whether or not I visit lakes, just knowing that they are clean is important to me,”
90% of the respondents chose “strongly agree” or “agree”. Therefore the lakes in Michigan
were likely of both use and non use value to the respondents. The more reduction in eutrophic
lakes, the more likely they would vote yes on the program.
On the other hand, variable “GHG” failed the scope test, and respondents’ votes were
not sensitive to differences of greenhouse gas reduction in the scenarios. From the descriptive
statistics of people’s concerns about global warming we knew that 60% were unconcerned
about global warming. So the scope test probably failed among those respondents. If a
considerable proportion of the respondents were not concerned about global warming, the
“GHG” variable would become insignificant. As a result, an interaction term was added to
probe the relationship between people’s attitudes towards global warming and their WTP. The
“gw” variable was derived from the question “how concerned are you about global warming”.
There were four types of responses “very concerned”, “somewhat concerned”, “somewhat
unconcerned”, “not at all concerned”. A new variable was created, “gw”, which equaled to 1
if the respondents were very concerned or somewhat concerned and equaled to 0 if otherwise.
An interaction term “gw_GHG” was added to the model, which was the product of the “gw”

99

and “GHG”. Both of the “gw” main effect term and the interaction term were significant. The
main effect term “gw” showed that if people were concerned about global warming, they
were likely to vote yes. The interaction term showed that GHG reduction did matter for those
who cared about global warming. The scope test passed in those respondents that were
concerned about global warming.

No interaction term was required for eutrophic lakes

because almost all respondents cared about lakes.
In the demographic variable group, variables “age”, “income”, “education” and “voters”
were significant. The positive sign of “age” means that older people are more likely to vote
yes than younger people. The same positive signs of “income” and “education” variables
mean that people with higher income and higher education tend to vote yes. Voters are more
likely to vote yes than non-voters. Other demographic variables were not significant, which
means that those variables would not have an effect on people’s probability of voting yes or
their WTP. It did not matter to their votes or WTP no matter how long they expected to live in
Michigan, how large their family size was, whether they considered themselves a Michigan
resident, whether they were male or female, whether they were farmers, foresters or
environmental organization members.
We originally made the hypothesis that the provision mechanism in the questionnaire
would affect people’s votes and WTP. To test the hypothesis, we created two versions: the
farm version, where farmers were going to provide the eutrophic lake and GHG reduction,
and the general land version where land owners in general were to provide the changes
(“farm” variable, see descriptions in Table 2.1). However, the “farm” variable turned out
insignificant in the model. The respondents were not sensitive to whether the reduction was
provided by farmers or not. Agricultural-farmer or land management versions of the
questionnaire would not affect their probability of voting yes nor WTP.
To capture potential correlations due to the repeated questions per person (panel data),

100

the parameter ρ in Table 2.3 was the ratio of the variance of individual error to the sum
variance of individual error and remainder disturbance. The larger it is, the more significant
the individual effect across a person’s answers within the data. The ρ in the conditional
WTP model was 0.93 and it was significant in likelihood ratio test. The null hypothesis of

ρ = 0 was rejected. Therefore, there was an individual effect in the data and it cannot be
omitted in model estimation. Compared with the cross sectional probit model, the random
effect probit for panel data was more appropriate to analyze our dataset.
The parameter σu was the standard error of individual error term. It was 3.61. The Wald
chi-square test showed that the model was significant and it had explanatory power on the
dependent variable ref.
The prediction power of the model was desirable. 82.3% of the “yes” votes were
predicted correctly using the model; 69.4% of the “no” votes were predicted correctly. In total,
77.6% of the predictions were correct.

7.4 Spike probability model
Another finding is that the probability that a respondent had a zero WTP (the spike
probability) was endogenous. Variables “lake”, “GHG”, “income”, “gw” were significant in
the random effects probit model for the spike (Table 2.3). The significant coefficients of
“lake” and “GHG” showed that there was scope effect of eutrophic “lake” and “GHG”
reduction on people’s spike probability. Further, the negative sign of both the coefficients
implied that the higher the reduction in eutrophic lakes and GHG, the lower probability the
respondent would have zero WTP. The result was consistent with our intuition. Respondents
were more likely to place no value on the reductions (i.e., not be willing to pay anything for it)
when the reductions were small. The higher the reductions, the more likely people would
place a positive value on it (i.e., the more likely they were willing to pay for it).
101

The spike probability was also significantly affected by people’s concern about global
warming. The negative sign of “gw” showed that people who were concerned about global
warming were less likely to vote against the program. Because they were concerned about
global warming, they were more willing to pay for GHG reduction, which means they had a
positive WTP.
Income was another factor that affected the spike probability. Higher income people
were less likely to have zero WTP. Perhaps because high income people have more
discretionary income, they were more likely to be willing to pay for environmental
improvements.
The “farm” variable was not significant in the spike model either. It means that whether
the reduction was provided by farmers or not did not affect the spike probability. The fact that
the farm variable was not significant in the conditional WTP model or the spike probability
model showed that the provision mechanism difference in this case did not affect people’s
WTP at all. No matter who provided the eutrophic lakes and GHG reduction, neither the
conditional nor the unconditional WTP would change, nor do people’s votes on the
hypothetical programs.
Other demographic variables were not significant except “income”. It means that people
did not vary in their chances of having zero WTP by their expected length of living in
Michigan, their residency, gender, family size, age, education or whether they were farmers,
foresters or environmental organization members. Whether they were voters would not affect
the spike probability either.
The related experience variables “fish”, “swim”, “boat”, “hike” did not have an impact
on the spike probability. They did not affect the conditional WTP equation either. The two
findings implied that the frequencies they used the lakes did not affect their WTP for
eutrophic lake and GHG reduction. They did care about lake quality but how often they went

102

on such recreational activities in lakes did not affect their WTP.
The Wald chi square test was significant, so the spike probability model had
explanatory power on the data. The standard error of individual error term was 2.43. ρ value
was 0.855, which means the variance of the individual error took up 85.5% of the sum of
variance. The likelihood test of ρ=0 was rejected so the individual error was significant and
the random effects probit was the appropriate econometric model.
Although the model turned out to be a good fit to the dataset, it did not do well in
predicting the binary outcome. Using a naïve prediction threshold of 0.5, the model predicted
that none of the sample had zero WTP because the predicted spike probability for each
respondent was lower than 0.5. When the mean share of “ones” in the data is far from 0.5, it
is not unusual for the probit to have a significant power to discriminate yes votes from no
votes yet not yield any predictions that cross the 0.5 threshold.

103

Table 2.3 Estimation results of proposed models

Conditional WTP

Spike probability

N

3433

4128

Groups

1296

1430

Dependent variables (Y)

ref

spike1

Explanatory variables

coefficient

P>|z|

coefficient

P>|z|

farm

-0.0929

0.697

0.0475

0.774

lncost

-1.42

lake

0.00379

<.001

-0.004

GHG

-0.156

0.549

LenResi

-0.0788

resi

***
***

<.001
***

<0.001

-0.383

**

0.032

0.66

0.205

0.107

1.64

0.196

-0.908

0.267

gender

0.176

0.518

-0.0839

0.655

FamSize

-0.0243

0.812

0.0154

0.822

age

0.0297

0.001

-0.00269

0.674

farmer

-0.642

0.324

0.566

0.193

forest

0.775

0.408

0.0453

0.945

EnvOrg

0.575

0.196

0.129

0.686

income

0.0148

<.001

-0.00424

***

***

104

**

0.031

Table 2.3 (cont’d)
**

education

0.181

0.023

-0.00278

0.96

fish

0.214

0.126

0.0989

0.311

swim

0.0858

0.589

0.00115

0.991

boat

0.144

0.386

0.0218

0.853

hike

0.135

0.316

0.0991

0.282

voter

1.71

0.002

-0.340

0.35

_cons

-2.44

0.162

-0.858

0.45

gw

1.51

***

<.001

-1.28

gw_GHG

0.617

*

0.079

***

2

***

***

<0.001

2

**

Wald χ (21)=234.3

Wald χ (19)= 104.6

lnσu

2.57

1.77

σu

3.61

2.43

ρ

0.929

0.855

log likelihood

-1449.1

-1350.7

Likelihood- ratio test of
ρ=0

χ 2 (01) =

2

***

-1382.33

χ 2 (01) =

% of correct prediction 82.3
of Y=1

0

% of correct prediction 69.4
of Y=0

100

% of correct prediction

83.2

77.6

Note: * for P<0.1, ** for P<0.05, *** for P<0.01

105

827.36

***

7.5 Comparison of the fixed spike model and the endogenous spike model in the probability
of voting yes and WTP estimates
In the fixed spike model, we did not use the spike probability model to predict the spike
probability. Instead, we used the observed spike probability, which was the proportion of zero
WTPs in the sample. The fixed spike probability p was 0.169. Then we computed the mean
and median annual WTP under the fixed spike probability according to equation (10) and (11).
The sample mean WTP is $175 per household annually for reductions of 170 eutrophic lakes
and 0.52% GHG emissions from the 2000 level. The number of household in Michigan was
3.8 million in 2000 (US Census Bureau, 2010). If the sample values are representative of all
households in Michigan, then the aggregate annual WTP for the whole state would be $665
million.
The sample mean marginal WTP for one eutrophic lake reduction was $0.45 per
household per year. Multiplied by the household number of 3.8 million, the aggregate
marginal WTP for one eutrophic lake reduction was $1.7 million per year for the whole state.
The sample mean marginal WTP for 1% reduction of GHG from the 2000 level was
$141 per household per year for those who were concerned about global warming. For those
who were not concerned about global warming, their marginal WTP for 1% reduction of
GHG was zero. In the sample, 40% of the respondents were concerned about global warming.
If we assumed that there were also 40% of the households concerned about global warming
in Michigan, there were 1.52 million households were concerned about global warming. Then
we computed the aggregate marginal WTP for 1% GHG reduction by multiplying the
marginal WTP per household by 1.52 million households. It was $214 million per year for the
whole state.
Instead of assuming the spike probability was exogenous and constant, the endogenous

106

spike model examined the scope effect of eutrophic lakes and GHG reduction on the
conditional WTP as well as on the spike probability. The endogenous spike model suggested
that the quantity of eutrophic lakes and GHG reduction would affect not only the conditional
WTP but also their chances of having zero WTPs. From equation (9) and (10), we knew that
the conditional WTP and the spike probability would affect respondent’s unconditional
probability of voting yes and their unconditional WTP. Moreover, model estimation results in
Table 2.3 supported the endogenous spike model over the fixed spike one. It showed that the
spike probability was indeed endogenous. Therefore, it was worthwhile to compare the two
models with their estimates in the unconditional probability of voting yes and unconditional
WTP.
The difference in the unconditional probability of voting yes was illustrated in Figure
2.2. The probabilities were plotted as functions of eutrophic lakes reduction “lake” (2.2a) and
functions of GHG reduction “GHG” (2.2b) respectively. The probability functions displayed
similar patterns in the lake case and the GHG case. Compared with the fixed spike model, the
endogenous spike model had a steeper curve. As the reduction in eutrophic lakes and
greenhouse gas increased, the spike probability decreased in the endogenous spike model
while the fixed spike probability remained constant. The conditional probability of voting yes
also increases faster in the endogenous spike model according to equation (6). The function
of two models intersected each other in the lake case (2.2a). Because the fixed spike
probability was higher than the endogenous one when eutrophic lake reduction was zero and
had a lower slope, the two functions intersected when eutrophic lake reduction was about 40.
The two functions did not intersect in the GHG case (2.2b) because the fixed one started
lower and had a lower slope. It should be noted that in both Graph 2.2a and 2.2b, the
unconditional probability of voting yes was not zero even when the reduction in lake and
GHG was zero. It was because the functions were the cross sections of a multidimensional

107

probability function of all categories of variables including environmental change variables,
demographic variables, cost variables, attitude variables and the interaction term. When other
variables took non zero values, the probability function was not necessarily zero even if one
of the environmental change variables was zero. In other words, people would still vote yes
when eutrophic lake reduction was zero because the GHG reduction was non zero.
The unconditional mean WTPs were plotted as functions with respect to eutrophic lake
reduction (Figure 2.3a) and functions with respect to GHG reduction (Figure 2.3b). The
unconditional mean WTP functions in Figure 2.3 had similar patterns with the unconditional
probability functions in Figure 2.2. When reduction in eutrophic lakes and greenhouse gas
increased, the endogenous spike probability decreased while the fixed spike probability
remained constant, then the unconditional mean WTP increased faster in the endogenous
spike case. Therefore, the unconditional mean WTP functions were steeper in the endogenous
spike model. Similar to Figure 2.2, the endogenous spike function and the fixed spike one
intersected when eutrophic lake reduction was 40 (2.3a) but did not intersect in the GHG case.
It can be inferred that everything else remaining constant, when the eutrophic lake reduction
was 40, the predicted endogenous spike probability was equal to the observed fixed spike
probability. On the other hand, the endogenous spike probability would never be equal to the
fixed spike probability at a positive value of GHG reduction, holding others constant.

108

Figure 2.2 Probability of voting yes as a function of environmental change variables
2.2a. Probability of voting yes with respect to reduction in eutrophic lakes

Probability of voting yes with respect to reduction in eutrophic
lakes

0.9
0.8

Probability

0.7
0.6
0.5

Fixed spike
probability
Endogenous spike
probability

0.4
0.3
0.2
0.1
0
0

40

80

120 160 200 240 280 320 360 400 440 480
Reduction in eutrophic lakes
109

Figure 2.2 (cont’d)
2.2b. Probability of voting yes with respect to reduction in Greenhouse Gas

Probability of voting yes with respect to reduction in
Greenhouse Gas

0.8
0.7

Probability

0.6
0.5
0.4
0.3
0.2

Fixed spike
probability
Endogenous spike
probability

0.1
0
0
0.2 0.4 0.6 0.8
1
1.2 1.4 1.6 1.8
GHG Reduction (% reduction compared to 2000 emission level)
110

Figure 2.3 Unconditional mean WTP estimates as a function of environmental change variables
2.3a. Mean WTP with respect to reduction in eutrophic lakes

Mean WTP with respect to reduction in eutrophic lakes

500
WTP per household (in USD)

450
400
350
300
250
200

Fixed spike probability

150
Endogenous spike
probability

100
50
0
0

50

100

150

200

250

300

350

Reduction in eutrophic lakes
111

400

450

500

Figure 2.3 (cont’d)
2.2b. Mean WTP with respect to reduction in Greenhouse Gas

Mean WTP with respect to reduction in Greenhouse Gas

WTP per household (in USD)

400
350
300
250
200

Fixed spike probability

150

Endogenous spike
probability

100
50
0
0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

Reduction in GHG (in % reduction compared to 2000 emission level)
112

2

8 Conclusions
Ecosystem services from changing agricultural practices were valued by residents.
Specifically, our research found that residents were willing to pay for reductions in eutrophic
lakes and greenhouse gases. Their willingness to pay depended on the scope of lake and
greenhouse gas reduction, attitude towards global warming, age, income, education and
whether they are voters. The higher the reduction, the more they were willing to pay for it.
Older, higher income, high education people had higher WTP. Voters had higher WTP than
non voters. On the other hand, their WTP was not affected by whether the environmental
improvements were provided by farmers or not, nor was it affected by the frequencies with
which they went on recreational activities to lakes.
Although most respondents cared about lake water quality and were willing to pay for
the eutrophic lake reduction, only 40% of the sample said they were concerned about global
warming. Their WTP for greenhouse gas reduction differed by their attitude towards global
warming. Those concerned about global warming were willing to pay a substantial amount
for a unit reduction in GHG while those unconcerned would not pay anything for the
reduction.
The probability of having zero WTP, the spike probability, was endogenous. It was
determined by the scope of environmental changes, income and attitude towards global
warming. Higher income people and those concerned about global warming had lower spike
probability. The scope of environmental changes affected both the spike probability and the
WTP conditional on WTP being positive. When modeling explicitly the endogenous spike,
two scope effects added up. The marginal WTP and the marginal probability of voting yes for
a unit increase in eutrophic lake and greenhouse gas reduction were higher than those in the
conventional fixed spike case. No other study in the literature has modeled spike probabilities

113

as endogenous functions of the environmental changes provided by a program. Our findings
suggest that ignoring the endogenous spike will lead to an underestimation of the benefits of
large improvements.
Michigan residents were willing to pay $175 per household every year to get 170
eutrophic lakes reduction and 0.52% GHG reduction from the 2000 level. They were willing
to pay $0.45 per household every year to get rid of one eutrophic lake and the 40% that were
concerned about global warming were willing to pay $141 per household per year every year
to reduce 1% greenhouse gas. The Michigan public had a large WTP for eutrophic lake
reduction and greenhouse gas reduction. Even though a substantial portion of them were not
concerned about global warming, those who cared about global warming had moderately
high WTP for the greenhouse gas reduction.
Another finding was that despite the poor economy, an income tax payment vehicle did
not generate an excessively high protest rate. Only 10% of the sample was protesting against
increased income tax lending support to income tax increase as an appropriate choice of
payment vehicle in the contingent valuation survey.
The results of the study estimated the public’s demand for two important ecosystem
services from agriculture. These can be coupled with estimates of farmers’ willingness to
supply these services by adopting low-input cropping practices to see if a “market” for these
services could exist.

114

APPENDICES

115

Appendix 2.1 Responses in survey period (by week)
Figure 2.4 Responses in survey period (by week)

Responses by week
1000
900
800
700
600
500
400
300
200
100
0
1

2

3

4

5

6

7

8

9

10

11

12

Note: Week 1 started from Aug 3- Aug 9, 2009

116

13

14

15

unknown

Appendix 2.2 Protests, disbelief and misunderstanding votes
Table 2.4 Protests, disbelief and misunderstanding votes
Reasons
rejection

of Government
intervention

% of total votes

Provision
mechanism

Regime of tax

Don't believe the Misunderstood
scenario
the scenario

10.07

6.74

Taxes

1.19

0.56

2.55

Figure 2.5 Protests, disbelief and misunderstanding votes as a percentage of total vote

Protests, Disbelief and Misunderstanding Votes
10.00
8.00
6.00
4.00
2.00

117

Misunderstood
the scenario

Don't believe
the scenario

Reject the
regime of tax

Reject
provsion
mechanism

Reject taxes

0.00

Reject
government
intervention

% of total votes

12.00

0.87

Appendix 2.3 Content analysis process
Preliminary themes to start with
Set preliminary themes to start with. According to the research interests, the preliminary
themes were about protests, disbeliefs and misunderstanding votes.

Training stage
Start the 1st round training of two undergraduate student coders. Each person had the
preliminary themes in hand. Each did the same 100 comments independently (i.e., without
discussing them with one another), to come up with other preliminary themes. The
preliminary themes are pretty narrow and specific.
The supervisor reviewed the themes, re-grouped and reconciled them.

Then the new

theme lists were developed. It was to be based on in later coding. The group had a meeting
and discussed the problems of the first round coding and the theme list.
In the 2nd round training, each coder coded another 300 comments based on the master
list. Each coder was given different sets of comments this time. Every time new themes came
up, new themes were added.
The supervisor re-grouped and reconciled the new themes that came up in the 2nd
round. The group met again to discuss the problems and the updated the theme list.

Self coding stage
1. Coders were assigned same amount of the comments and coded them based on the
updated theme list.
2. Each time new theme came up the theme list was updated and circulated within the
group so that others knew about the change. The supervisor would reconciled and keep track
of the updated theme list. When the whole set of comments was coded once, the theme list

118

was finalized.
3. Coders shifted data and redid the coding a second time based on the final theme list
so that all comments were coded twice.

119

Appendix 2.4 Themes developed from comments
Table 2.5 Themes names, descriptions and example quotes
Theme Name

Description

Example

GHG only

They mention GHG alone

It is good to have land management for both the
economy and GHG

lakes only

They mention lakes or water quality alone

Not enough done on water- it is the first step

Good for the environment

Respondents said 1.the program will do good to the
environment or 2. the environment is important

Land stewardship will work as the problem is
mostly the actions of man - water is more valuable
than gold. For our kids

cost effective

Respondents feel that the cost is reasonable compared to
the improvements that are made.

best plan for both lakes and GHG reductions to get
an increased tax rate

recreational use

Respondents mention the program will improve fishing,
boating, swimming and any other uses or they want to
improve the quality of these activities

I use these lakes and need to help in upkeep.

future generation
consideration/Moral obligation

Respondents want a better environment for their future
generations/or have a moral obligation to improve the
environment

The future of any three sons and their families is
my main concern!!!

120

Table 2.5 (cont’d)
help the economy

Respondents consider the program would help boost the
economy

The potential to create Michigan jobs for
Michiganders

reductions too small

They vote 'no' and mentioned that 1.reductions are too
small, insignificant or 2. they want more reductions

In this program I would like to see a larger
reduction of lakes with excess nutrients.

Good start/better than none

They vote 'yes' but mentioned that 1.changes are small or
2.better than none or 3.good to start to do something

Minimal cost. Would help a little – a starting
point

benefit not worth cost

Respondents think it is not cost effective

too small results assuming everyone in state is
paying 200

can't afford it

Respondents cannot afford the cost although think it is
good

I pay enough income taxes. I have no extra money
now

complain about zero reduction

Respondents don't like zero reduction in either GHG or
lakes

The increase of GHG that nearly fulfills the needed
reduction. I was not happy with the zero
reduction in eutrophic lakes, but would vote yes to
obtain some benefit

121

Table 2.5 (cont’d)
don't believe in GW

respondents don't believe GW or don't think it is a problem Global warming is not that big of an issue, from all
the facts that I’ve found out about of my own
research of the scientists what I

Need more information about
program

Will not vote for program because the respondent wants to
know more about technical details about environmental
program.

you have to explain what type of land management
practices are to be implemented

Rejects premise of environmental Respondents don't think that people should do anything to
improvement
change the environment

Let Nature take its course

Irrelevant

Comments are irrelevant to voting.

stop trash coming from Canada

Other substitutes

Respondents think about other programs that can
substitute the programs

I believe with education we can achieve these goals

122

Table 2.5 (cont’d)
Other issues more important

Believe that other issues, i.e. healthcare or employment,
are more important than GW/lakes

1. Increased taxes. 2. Other issues taking
precedence; i.e economy, having a job, being able
to pay mortgage and monthly bills. 3. The
health care system. 4. Hunger, homeless in our
own state. 5. issues need addressed now: child
abuse, animal welfare

Health concerns

Respondents mention health concerns as a reason for
voting for the program.

Believe in health reason.

environmental improvement not
a concern

Respondents don't care about the environment at all

I don’t want to pay anymore taxes! It just is not
important to me.

don't care GHG

Respondents mentioned they do not care about greenhouse
gas but they might believe global warming is true

GHG taken out of factor

Bad for the economy/Economy is Respondents think the economy is too bad to afford the
bad
program or the cost would harm the economy

The hard economy

Polluters pay

Most farming is Big Business - corporate owned.
I in no way think it is appropriate for government
subsidies @ taxpayer expense.

Polluters (big corporations, farmers, land owners) should
pay, not the public

123

Table 2.5 (cont’d)
Disagree with government
regulation / waste

1. Reject government regulation or 2. mention that tax
money will be wasted by the government

To much money wasted by govt. Money being
spent on something else

Reject provision mechanism

Reject the idea of paying farmers/land owners to reduce
pollution

The government (federal or state) has no right to
spend my tax dollars for farmers or lake property
owners

rejection of tax

opposed to raising taxes to fund the program

I can’t justify being taxed a penny more when
wage increases don’t even batch inflation or are
non existent. 2 we are already overtaxed now for
everything. I won’t vote to pay more.

Reject the regime of program or
payment

Respondents 1. reject it is federal tax or 2. prefer local or
state taxes to fed tax

I believe land management should be done at the
county and state level only. A increased state
income tax only drives more businesses our of
Michigan.
The leviathan (federal government) should not be
fed.

124

Table 2.5 (cont’d)
They don't think 1) the program would reduce the amount
of lakes and GHG given by the tables or 2)the money
would go to the program or 3)the cost remain the same
overtime or 4)the money would be used only in Michigan

Increased taxes – money never seems to end up
where the government says it goes.

clear misunderstanding

Respondents clearly have misunderstanding on the
scenario

Even though we are reducing 6% it may be a
valuable investment in land air, lake quality for
only $30 (the actually we gave in his questionnaire
was $100)

possible misunderstanding

Some of respondent's comments suggest that they might
have misunderstandings but not sure if they did
misunderstand the scenario

The increase of GHG that nearly fulfills the needed
reduction. I was not happy with the zero
reduction in eutrophic lakes, but would vote yes to
obtain some benefit (note: reduction in GHG did
not "nearly fulfills the needed reduction". It
reduces only 1.2% of the needed reduction at the
maximum. However, his comments can be
interpreted as "fulfills nearly all the reduction that
we can achieve from changing land management
practices"

Don't believe the scenario

125

The program would never reduce that many lakes.
The improvement would be a much smaller
fraction even if it works.

Table 2.5 (cont’d)
fund from existing budget

Respondent wants to fund the program by cutting expenses I pay enough in taxes as it is. Attempt to fit this
of other parts in the budget
into the state budged first
I believe if the government can manage / use the
federal income (instead of giving it to banks) or
spend on wars) Tax more appropriately. We should
not need to pay increased amount of federal
income ax to cover this cost.
Again- use of money already in system. No to rise
taxes.

Users pay

Respondent thinks that those who enjoy the benefit of the
improvement should pay, instead of the general population

126

Tax the user of the additional nutrients. 2. does not
make sense for any $ to be spent for user boaters
on lake and homeowners on lakes. Who fertilize
and add nutrients to the lake?

Appendix 2.5 Survey return statistics

Table 2.6 Survey return categories and quantities
Category

Quantity

Returned and answered

2215

Refused

49

Bad address

480

Deceased

40

Moved out of Michigan

40

Total sample size

6000

Response rate= 2215/ (6000-480-40-40) = 40.72%

127

Appendix 2.6 Questionnaire

Environmental Improvements in Michigan
A SURVEY OF YOUR OPINIONS
This survey aims to understand Michigan citizens’ views on various
environmental improvements. There are no right or wrong answers. We need
your views.

Your opinions matter!
By completing this survey, you are helping to inform the design of future
policies that better reflect the views and concerns of Michigan citizens.
128

Introduction to the survey

We would like to know your opinions on several proposed land
stewardship programs. Scientists have concluded that by changing
land management practices, these programs would improve the
following aspects of the environment:
• Lake water quality
• Global warming

ON EACH OF THE FOLLOWING PAGES, PLEASE READ THE
TEXT IN THE SHADED BOX FIRST, AND THEN ANSWER THE
QUESTIONS.

129

Lake water quality
Lakes need nutrients that run off from the land. However, high levels of nutrients in
lake water may cause excess growth of plants living in water. Large quantities of
aquatic weeds and water grass can take over the shoreline and float on the water surface.
High nutrient levels can also cause algal blooms. Algal blooms are population
explosions by microscopic organisms called algae. When algal blooms happen, large
amounts of algae and plants float on the surface of the lakes. Algae use up oxygen in the
water, which can kill fish. The water can smell bad and become cloudy, colored green,
yellow, brown, or red.
Excess plant growth and algal blooms may interfere with recreational activities such as
fishing, boating and swimming. Lake scenery is altered. Some toxic algae pose health
risks to people.
There are about 11,000 inland lakes in Michigan. About 3400 of them (31%) contain
high nutrient levels that promote excess plant growth and are at high risk of algal
blooms.
Land management can significantly affect nutrient input into lakes. Scientists report that
in some areas, about half of the nutrients added to the nearby lakes comes from fertilizer
runoff from land.
1. Have you ever seen an algal bloom in a lake?
No
2. Has the presence of algal blooms affected your
Yes
If yes
recreational activities in or around Michigan
lakes?
No
Yes
3. Have you ever seen excess plant growth in a lake?
No
4. Has the presence of excess plant growth affected your
Yes
If yes
recreational activities in or around Michigan
lakes?
No
Yes
5. Compared to what you previously thought, how serious of a problem are the
consequences of high nutrient levels in Michigan lakes?
More serious than I thought
About the same as I thought
Less serious than I thought
6. Which best describes your level of agreement with the following statement
“nutrients in lakes are likely to affect me”?
Strongly agree
Agree
Neutral
Disagree
Strongly disagree
130

Global warming
The global temperature has increased by more than one degree Fahrenheit during
the past 100 years. Scientists believe that global warming is increasing.

Several gases cause global warming by holding in heat like a greenhouse. These gases are
referred to as “greenhouse gases.”
Global warming has many effects. Ice in the North and South Poles is melting. Sea levels
are expected to rise. Some low areas are expected to become flooded and submerged in
water. Floods, tornadoes, hurricanes and snow storms are expected to happen more
frequently. Tropical diseases are expected to break out in more areas. Some animals and
plants are expected to become extinct.
To keep the long term global temperature increase under 6 degrees, scientists have
concluded that greenhouse gas emissions in 2050 should be reduced by 30% from
the 2000 emission level. If Michigan were to meet this goal, it would have to reduce 56.7
million tons of greenhouse gases (which is equal to the annual greenhouse gases from
driving 10.5 million cars).

Land management can significantly affect greenhouse gas emission. Scientists estimate
that land management contributes to 10% of global greenhouse gas emission.
7. How concerned are you about global warming?
Very concerned
Somewhat concerned
Somewhat unconcerned
Not at all concerned
8.

Compared to what you previously thought, how serious of a problem are the
consequences of global warming?
More serious than I thought
About the same as I thought
Less serious than I thought

Strongly 
Disagree 

Disagree 

Neutral 

 

Agree 

 

Strongly 
Agree 

Please indicate your level of agreement with the following statements.

9.

Global warming is a problem for me now

 

 

 

 

10.

Global warming will be a problem for me in
the future 

 

 

 

 

 

11.

Global warming will be a problem for future
generations

 

 

 

 

 

131

 

How the environment may be improved by changing land management
practices
Land management can significantly affect lake water quality and global warming.
For example, changing land management practices can reduce the amount of nutrients
entering lakes. In addition, changing fertilizer application methods on land can slow
down global warming.
Scientists have evaluated a program to change some land management practices in
Michigan and have determined that the following environmental improvements could be
achieved. The actual amount of the improvements would vary depending on the level
of investment in the program.
Table 1: How much could this type of land management program improve the
environment?

Now
Number of lakes with
excess nutrients
Greenhouse gas (GHG)
reduction needed to
slow global warming

MAXIMUM improvement that
this type of land management
program could bring

After

Reduce 400
(12%)

Need 30%
reduction

3,000

Reduce by 1.2

3,400

Need
28.8%
reduction

12. Prior to reading this, how aware were you that land management practices can
significantly affect both lake water quality and greenhouse gas emissions?
Very aware
Somewhat aware
Not aware
13. Looking at Table 1 above, the largest improvement that can be achieved by
changing land management practices would reduce the number of lakes with
excess nutrients by 400 (down to 3,000 lakes with excess nutrients). How
important would it be to achieve this reduction?
Very important
Somewhat important
Not important
14. Looking at Table 1 above, the largest improvement that can be achieved by
changing land management practices would reduce greenhouse gas emissions by
1.2% in the 30% reduction goal (down to 28.8% reduction needed). How
important would it be to achieve this reduction?
Very important
Somewhat important
Not important
132

Voting on a potential environmental improvement program
Proposed Land Stewardship Programs
The federal government is considering a land stewardship program that would cover
land owners’ cost of adopting environmentally friendly land management practices. The
program would be voluntary. The program would contribute to the environmental
improvements described previously.
Program cost
The program will require funds to operate. The cost of the program is divided into two
major parts.
•
90% goes to cover land owners’ costs of changing land management practices.
•
10% goes to project management and monitoring.
If the program were implemented, you would need to pay an increased amount of
federal income tax annually to fund the program. The money raised would only be
used for this specific program. Money raised in Michigan would be spent in Michigan.

We need your input
In the pages that follow, we will show you three different land stewardship programs
that provide environmental improvements and cost you money. For each of the three
programs, please imagine that you are at the voting booth and have to vote on the
program. Please answer each question as if you were really voting on it.
In a real vote, only one of the programs would be offered. We show you three different
programs because we need your input on all three programs. Although we ask you to
vote on three different programs, please vote on each one as if the other programs were
not available. In other words, when you consider each program, please vote as if it is
the only one being offered.
In a real vote, if at least 50% of the people voted for a program, it would be
implemented. If the program is implemented, you would have to pay the income tax
increase annually. The environmental improvements would be achieved annually for as
long as the tax is in effect.
Important reminders before you vote
The land stewardship program is one of the ways that you could pay to get
environmental improvements. Some people will consider the program valuable and
will vote yes. Others will prefer having the money to save or spend on other things. We
ask for your views. There are no “right” answers other than what you believe is the best
way to vote.
133

Land stewardship program A

Please review program A and then answer questions 15-18. Program A has the following
environmental improvements and cost:

Program A is going
Now

After
to …

3,400

Reduce by 200
(6%)

3,200

Greenhouse gas (GHG)
reduction needed to
slow global warming

Need 30%
reduction

Reduce by 1.2

Need 28.8%
reduction

Your share of the costs
for the program

$ 0 per year in
increased
income tax

Number of lakes with
excess nutrients

$10 per year in
increased
income tax

15. Would you vote for program A if it increased income taxes and your share of the
increased tax was $10 per year?
Yes
16. Would you vote for this program if it did not cost
No
If no
you anything?
Yes
No
17. On a scale of 1 to 10, where 1 means “very uncertain” and 10 means “very
certain”, how certain are you with your answer in Question 15?
1
2 3 4 5 6 7 8 9
10
Very uncertain
Very certain
18. What were your top two reasons for your vote in Question 15?
_________________________________________________________________________

134

Land stewardship program B

Please review program B and then answer questions 19-22. Assume that program A is not
being offered. Program B has the following environmental improvements and cost:

Program B is going
Now

After
to …

3,400

Reduce by 260
(8%)

3,140

Greenhouse gas (GHG)
reduction needed to
slow global warming

Need 30%
reduction

Reduce by 1.0

Need 29.0%
reduction

Your share of the costs
for the program

$ 0 per year in
increased
income tax

Number of lakes with
excess nutrients

$10 per year in
increased
income tax

19. Would you vote for program B if it increased income taxes and your share of the
increased tax was $10 per year?
Yes
20. Would you vote for this program if it did not cost
No
If no
you anything?
Yes
No
21. On a scale of 1 to 10, where 1 means “very uncertain” and 10 means “very
certain”, how certain are you with your answer in Question 19?
1
2 3 4 5 6 7 8 9
10
Very uncertain
Very certain
22. What were your top two reasons for your vote in Question 19?
_________________________________________________________________________

135

Land stewardship program C

Please review program C and then answer questions 23-26. Assume the previous programs
are not being offered. Program C has the following environmental improvements and cost:

Program C is going
Now

After
to …

3,400

Reduce by 340
(10%)

3,060

Greenhouse gas (GHG)
reduction needed to
slow global warming

Need 30%
reduction

Reduce by 0.6

Need 29.4%
reduction

Your share of the costs
for the program

$ 0 per year in
increased
income tax

Number of lakes with
excess nutrients

$10 per year in
increased
income tax

23. Would you vote for program C if it increased income taxes and your share of the
increased tax was $10 per year?
Yes
24. Would you vote for this program if it did not cost
No
If no
you anything?
Yes
No
25. On a scale of 1 to 10, where 1 means “very uncertain” and 10 means “very
certain”, how certain are you with your answer in Question 23?
1
2 3 4 5 6 7 8 9
10
Very uncertain
Very certain
26. What were your top two reasons for your vote in Question 23?
_________________________________________________________________________

136

Your opinion about environmental issues

Strongly 
Disagree 

Disagree 

Neutral 

Agree 

Strongly 
Agree 

Please indicate your level of agreement with the following statements.

 

 

27.

It’s fair that I pay to get better lake
water quality.

 

 

 

 

28.

It’s fair that I pay to get less
greenhouse gas emission. 

 

 

 

 

 

29.

Polluters should pay for their
pollution. 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

30.

31.

The government should finance the
land stewardship programs from
existing funding (leaving less for
other programs).
My votes on the land stewardship
programs would affect whether or
not land management practices
change. 

 

 

32.

Improved lake water quality
benefits me. 

 

 

 

 

 

33.

Improved lake water quality
benefits others.

 

 

 

 

 

34.

Controlled global temperature
benefits me.

 

 

 

 

 

35.

Controlled global temperature
benefits others in Michigan.

 

 

 

 

 

36.

Controlled global temperature
benefits others outside Michigan.

 

 

 

 

 

Whether or not I visit lakes, just
knowing that they are clean is
important to me. 
I would pay more for the programs
if the money went to foresters.

 

 

 

 

 

 

 

 

 

 

I would pay more for the programs
if the money went to farmers.

 

 

 

 

 

37.
38.
39.

137

Questions about you
40. How long do you expect to continue living in Michigan?
Less than 1 year
1-5 years
5-10 years
More than 10 years
41. Do you consider yourself a Michigan resident?
Yes
No
42. What is your gender?
Male

Female

43. Including yourself, how many people are there in your household? _____
(including myself)
44. What is your age? _________________
45. Are you a farmer or do you work on a farm?
Yes
No
46. Are you a forester or do you work in forests?
Yes
No
47. Do you belong to any environmental organizations?
Yes
No
48. How much is the annual pretax income of your household (all sources added
together)?
Less than $10,000
$50,000-74,999
$10,000-14,999
$75,000-99,999
$15,000-24,999
$100,000-149,999
$25,000-34,999
$150,000- 199,999
$35,000-49,999
$200,000 or more
49. What is the highest level of education you have completed?
Some high school or less
High school diploma
Technical training beyond high school
Some college (including AA, AS degrees)
4-year college degree
Some graduate work
Graduate degree
138

50. How often do you go fishing in inland lakes in Michigan?
Never
In some years
In most years
Every year
51. How often do you go swimming in inland lakes in Michigan?
Never
In some years
In most years
Every year
52. How often do you go boating in inland lakes in Michigan?
Never
In some years
In most years
Every year
53. How often do you hike near inland lakes in Michigan?
Never
In some years
In most years
Every year
54. Are you a registered voter?
Yes
No
55.

(Optional) Do you have any comments?
________________________________________________________________________________
________________________________________________________________________________
_________________________________________________________________________________

Thank you! This completes the questionnaire.
We appreciate your help and feedback.
If you have questions, please contact Professor Frank A Lupi at 1- 517-432-3883, by email at
lupi@msu.edu, or by postal mail at 301B, Agriculture Hall, East Lansing, MI 48824.

139

Appendix 2.7 Prenotice letter

(Address)

Dear (First Name, Middle Initial, Last Name, Jr/Sr/II/VI etc),
In a few days you will receive a questionnaire in the mail about Michigan’s environment.
When it arrives, I hope that you will take 15 minutes to fill it out and mail it back. I am
writing to you now since many people like to receive advance notice.
You are one of a small number of people who are being asked to express their opinions on the
issue. It is very important that you reply so that the results best reflect the views and concerns
of Michigan residents.
Thank you in advance for your help.

Sincerely,

Frank Lupi
Professor

140

Appendix 2.8 First round cover letter

July 23, 2009
(Address)
Dear (First name, middle name initial, last name),
I would like to ask you about your views on some possible environmental improvements in
Michigan. By completing the enclosed questionnaire, you will be helping to shape future
policies, programs and research that influence the daily life of people in Michigan.
You are one of a small number of people being consulted for opinions on these matters as part
of a scientific sample of Michigan residents. There are no right or wrong opinions. Because
people are different, I need your opinions for the results to accurately represent the people of
Michigan.
So I ask for 15 minutes of your time to complete the attached questionnaire and return it in
the prepaid envelope.
Your individual views will be completely confidential and your privacy will be protected to
the maximum extent permitted by law. Also, your participation in the survey is voluntary, and
you may refuse to answer certain questions. If you have questions or concerns about your role
and rights as a research participant, would like to obtain information or offer input, or would
like to register a complaint about this study, you may contact, anonymously if you wish, the
Michigan State University's Human Research Protection Program at 517-355-2180, Fax
517-432-4503, or e-mail irb@msu.edu or regular mail at 202 Olds Hall, MSU, East Lansing,
MI 48824.
I would be most happy to answer any questions you might have. You can call me at 1-517432-3883 or email me at lupi@msu.edu.
Thank you for your help.
Sincerely,

Frank Lupi
Professor
Enclosure

141

Appendix 2.9 Second round cover letter

September 4, 2009
(Address)
Dear (First name, middle name initial, last name),
Last month, I wrote to ask your views on environmental improvements in Michigan. To my
knowledge, I have not yet received your completed questionnaire.
We have undertaken this study to ensure that the design of future environmental policies is
informed by the views and concerns of Michigan residents.
I am writing to you again because your input is vital to this study! You are one of the small
number of people drawn from a scientific sampling process. In order for the results of this
study to accurately represent the opinions of all Michigan residents, it is very important that
each person in the sample return their questionnaire.
In the event that your questionnaire has been misplaced, I enclose a replacement.
As I mentioned in the earlier letter, your individual views will be completely confidential and
your privacy will be protected to the maximum extent permitted by law. Also, your
participation in the survey is voluntary, and you may refuse to answer certain questions. If
you have questions or concerns about your role and rights as a research participant, would
like to obtain information or offer input, or would like to register a complaint about this study,
you may contact, anonymously if you wish, the Michigan State University's Human Research
Protection Program at 517-355-2180, Fax 517-432-4503, or e-mail irb@msu.edu or regular
mail at 202 Olds Hall, MSU, East Lansing, MI 48824.
I would be most happy to answer any questions you might have. You can call me at 1-517432-3883 or email me at lupi@msu.edu.
I do hope that you can take a few minutes to complete and return the questionnaire. Thanks
in advance for your cooperation.
Sincerely,

Frank Lupi
Professor
Enclosure
142

Appendix 2.10 First round reminder postcard

143

Appendix 2.11 Second round reminder postcard

144

Appendix 2.12 Steps to generate experimental design

1. Determine the attributes and levels.
2. Start with the orthogonal main effect plan generated by SPSS. There will be one
combination that contains “no change” levels in both lake and GHG.
3. Sort the combinations in ascending order with price first and farmer second. Add the
version numbers to each combination (each version has three combinations and all three
combinations should have same farmer and price levels). Add blank combinations if
necessary (yellow ones). Then there were 10 versions and 30 combinations in the plan.
4. Change the numbers in lake and GHG in the plan to satisfy the following criteria:
In every combination, at least one number in lake and GHG is larger than 0.
Within the same version, one combination does not dominate another.
The total correlations within all variables as well as the sequence number (the order that
the combination is presented in a version, 1st, 2nd or 3rd) are as small as possible.
If possible, the numbers should evenly cover the whole ranges of lake and GHG.

Sometimes, in one version, one combination is far superior to another (for example, a
combination with very high GHG reduction and zero eutrophic lake reduction and the one
with no GHG reduction and low eutrophic lake reduction). In this case, separate the version
into two versions- one with the superior combination and one with the inferior one(s). The
two combinations have the same farmer and price levels. Then add the blank combinations
according to the criteria above.

145

Appendix 2.13 Experimental design plan
Attributes and levels
Table 2.7 Attributes and levels of the contingent valuation design
Attributes

Levels

Agriculture

Farmer

Eutrophic lakes reduction

0

68

136

204

408

GHG reduction

0

0.2%

0.4%

0.6%

1.2%

Price Levels

10

30

50

100

200

General land management

Experimental design plan
Table 2.8 Experiment design plans for each version of the questionnaire
Questionnaire
Version
A

Sequence

farmer

price

lake

GHG

1

0

10

200

1.2

A

2

0

10

260

1

A

3

0

10

340

0.6

B

1

0

10

70

0.6

B

2

0

10

140

0.4

B

3

0

10

200

0.3

C

1

1

10

0

0.4

C

2

1

10

100

0

C

3

1

10

70

0.3

D

1

0

30

200

0.2

D

2

0

30

140

0.4

D

3

0

30

70

1.2

146

Table 2.8 (cont’d)
E

1

1

30

100

0.5

E

2

1

30

70

0.6

E

3

1

30

200

0.4

F

1

1

30

100

0.5

F

2

1

30

0

0.6

F

3

1

30

140

0

G

1

1

30

400

0.4

G

2

1

30

300

0.8

G

3

1

30

200

1

H

1

0

50

140

0.2

H

2

0

50

0

0.5

H

3

0

50

70

0.3

I

1

0

50

140

0.6

I

2

0

50

250

0.2

I

3

0

50

200

0.5

J

1

1

50

400

0.2

J

2

1

50

300

0.4

J

3

1

50

200

1

K

1

0

100

200

1

K

2

0

100

300

0.8

K

3

0

100

400

0

L

1

1

100

100

0.5

147

Table 2.8 (cont’d)
L

2

1

100

200

0.4

L

3

1

100

70

0.6

M

1

0

200

0

0.4

M

2

0

200

200

0

M

3

0

200

70

0.2

N

1

1

200

180

0.8

N

2

1

200

400

0.6

N

3

1

200

120

1.2

148

REFERENCES

149

REFERENCES

An, M. Y., Ayala, R. A., Bank, C., & Ecuador, E. Q. A mixture model of willingness to pay
distributions: Duke University, Department of Economics.
Arrow, K., Solow, R., Portney, P. R., Leamer, E. E., Radner, R., & Schuman, H. (1993).
Report of the NOAA Panel on Contingent Valuation. Federal Register, 58(10),
4601-4614.
Bateman, I., & Turner, R. (1993). Valuation of environment, methods and techniques: the
contingent valuation method. In Kerry. Turner. R (Ed.), Sustainable environmental
economics and management: principles and practice (pp. 120- 191). London:
Belhaven Press.
Beasley, S., William D, Workman, G., & Williams, N. A. (1986). Estimating Amenity Values
of Urban Fringe Farmland: A Contingent Valuation Approach: Note. Growth and
Change, Volume 17 Issue 4 Page 70-78.
Bergstrom, J., Dillman, B., & Stoll, J. (1985). Public environmental amenity benefits of
private land: The case of prime agricultural land. South Journal of Agricultural
Economics 17, 139-149.
Berrens, R. P., Bohara, A. K., Jenkins-Smith, H. C., Silva, C. L., & Weimer, D. L. (2004).
Information and effort in contingent valuation surveys: application to global climate
change using national internet samples. Journal of Environmental Economics and
Management, 47(2), 331-363.
Blamey, R. K., Bennett, J. W., & Morrison, M. D. (1999). Yea-saying in contingent valuation
surveys. Land Economics, 75(1), 126-141.
Blevins, R. L., M. S. Smith, G. W. Thomas, & Frye, W. W. (1983). Influence of conservation
tillage on soil properties. Journal of Soil and Water Conservation, 38(3), 301-305.
Bowker, J. M., & Didychuk, D. D. (1994). Estimation of nonmarket benefits of agricultural
land retention in Eastern Canada. Agricultural and Resource Economics Review,
23(2), 218-225.
Boyer, T., & Polasky, S. (2009). Valuing urban wetlands: a review of non-market valuation
studies. Wetlands, 24(4), 744-755.
Boyle. (2003). Contingent Valuation in Practice. In P. A. Champ, K.J. Boyle, T.C. Brown
(Ed.), A Primer on Nonmarket Valuation (pp. 111-170). London: Kluwer Academic
Publishers.
Brander, L. M., Raymond, J., Florax, G. M., & Vermaat, J. E. (2006). The Empirics of
Wetland Valuation: A Comprehensive Summary and a Meta-Analysis of the
Literature. Environmental and Resource Economics, 33(2), 223-250.
Brouwer, R., Langford, I. H., Bateman, I. J., & Turner, R. K. (1999). A meta-analysis of
150

wetland contingent valuation studies. Regional Environmental Change, 1(1), 47-57.
Brussaard, L. (1997). Biodiversity and Ecosystem Functioning in Soil. Ambio, 26(8), 563-570
Carson, R. T. (2000). Contingent Valuation:A User's Guide. Environmental Science &
Technology, 34(8), 1413-1418.
Carson, R. T., & Hanemann, W. M. (2005). Chapter 17 Contingent Valuation. In J. Vincent
& C. F. Korstian (Eds.), Handbook of Environmental Economics (Vol. 2, pp.
821-936). Amsterdam: Elsevier.
Champ, P. A., & Bishop, R. C. (2001). Donation payment mechanisms and contingent
valuation: an empirical study of hypothetical bias. Environmental and Resource
Economics, 19(383-402).
Costanza, R., d'Arge, R., de Groot, R., Farber, S., Grasso, M., Hannon, B., Limburg, K.,
Naeem, S., O'Neill, R. V., Paruelo, J., Raskin, R. G., Sutton, P., & van den Belt, M.
(1997). The value of the world's ecosystem services and natural capital. Nature,
387(6630), 253-260.
Cummings, R. G., & Taylor, L. O. (1999). Unbiased value estimates for environmental goods:
a cheap talk design for the contingent valuation method. The American Economic
Review, 89, 649-665.
Daroub, S. H., Ellis, B. G., & Robertson, G. P. (2001). Effect of cropping and low-chemical
input systems on soil phosphorus fractions. Soil Science, 166(4), 281-291.
Desvousges, W. H., Johnson, F. R., Dunford, R. W., Hudson, S. P., Wilson, K. N., & Boyle,
K. J. (1993). Measuring natural resource damages with contingent valuation: test of
validity and reliability. In J. A. Hausman (Ed.), Contributions to Economic Analysis.
Amsterdam, The Netherlands: Elsevier Science Publishers B.B.
Dillman, D. A. (2000). Mail and internet surveys: The tailored design method: John Wiley &
Sons Inc.
Drake, L. (1992). The non-market value of the Swedish agricultural landscape. European
Review of Agricultural Econonmics, 19(3), 351-364.
Dubgaard, A. (1994). Valuing recreation benefits from the Mols Bjerge area, Denmark. In A.
Dubgaard, Bateman, I., Merlo, M. (Ed.), Economic Valuation of Benefits from
Countryside Stewardship (pp. 145-166). Vauk, : Kiel, .
Fawcett, S. R., Christensen, B. R., & Tierney, D. P. (1994). The impact of conservation
tillage on pesticide runoff into surface water: A review and analysis. Journal of Soil
and Water Conservation, 49(2), 126-135.
Gurluk, S. (2006). The estimation of ecosystem services' value in the region of Misi Rural
Development Project: Results from a contingent valuation survey. Forest Policy and
Economics, 9, 209-218.
Haab, T. C., & McConnell, K. E. (2002). Valuing Environmental and Natural Resources: The
Econometrics of Non-Market Valuation (2002) Cheltenham. Boston: Edward Elgar
151

Publishing.
Hall, C., McVittie, A., & Moran, D. (2004). What does the public want from agriculture and
the countryside? A review of evidence and methods. Journal of Rural Studies, 20(2),
211-225.
Halstead, J. (1984). Measuring the nonmarket value of Massachusetts agricultural land: A
case study. Journal of Northeast Agricultural Economics Council, 13, 12-19.
Hoehn, J., & Randall, A. (1989). Too many proposals pass the benefit–cost test. American
Economic Review, 79, 544-551.
Holmes, T. P., & Adamowicz, W. L. (2003). Attribute based methods. In P. A. Champ, K.J.
Boyle, T.C. Brown (Ed.), A Primer on Nonmarket Valuation (pp. 171-219). London:
Kluwer Academic Publishers.
Holmes, T. P., Bergstromb, J. C., Huszarc, E., Kaskd, S. B., & Orr, F. I. (2004). Contingent
valuation, net marginal benefits, and the scale of riparian ecosystem restoration.
Ecological Economics, 49, 19-30.
Hu, W. (2006). Use of Spike Models in Measuring Consumers' Willingness to Pay for
Non-GM Oil. Journal of Agricultural and Applied Economics, 38(03), 525-538.
Irwin, E. G. (2002). The Effects of Open Space on Residential Property Values. Land
Economics, 78(4), 465-480.
Irwin, E. G., & Bockstael, N. E. (2001). The Problem of Identifying Land Use Spillovers:
Measuring the Effects of Open Space on Residential Property Values. American
Journal of Agricultural Economics, 83(3), 698-704.
Kremen, C., Williams, N. M., Bugg, R. L., Fay, J. P., & Thorp, R. W. (2004). The area
requirements of an ecosystem service: crop pollination by native bee communities in
California. Ecology Letters, 7(11), 1109-1119.
Kristrom, B. (1997). Spike models in contingent valuation. American Journal of
Agricultural Economics, 79, 1013-1023.
List, J. A. (2001). Do explicit warnings eliminate the hypothetical bias in elicitation
procedures? Evidence from field auctions for sportscards The American Economic
Review, 91(5), 1498-1507.
List, J. A., Sinha, P., & Taylor, M. H. (2006). Using Choice Experiments to Value
Non-Market Goods and Services: Evidence from Field Experiments. Advances in
Economic Analysis and Policy, 6(2), Article 2.
Loomis, J., Kent, P., Strange, L., Fausch, K., & Covich, A. (2000). Measuring the total
economic value of restoring ecosystem services in an impaired river basin: results
from a contingent valuation survey. Ecological Economics 33, 103-117.
Lusk, J. L. (2003). Effects of cheap talk on consumer willingness to pay for golden rice.
American Journal of Agricultural Economics, 85(4), 840-856.

152

Maeder, P., Fliessbach, A., Dubois, D., Gunst, L., Fried, P., & Niggli, U. (2002). Soil
Fertility and Biodiversity in Organic Farming (Vol. 296, pp. 1694-1697).
Magnuson, J. J. (1990). Long-Term Ecological Research and the Invisible Present.
BioScience, 40(7), 495-501.
Mosier, A. R., Halvorson, A. D., Peterson, G. A., Robertson, G. P., & Sherrod, L. (2005).
Measurement of Net Global Warming Potential in Three Agroecosystems. Nutrient
Cycling in Agroecosystems, 72(1), 67-76.
Pruckner, G. J. (1995). Agricultural landscape cultivation in Austria: An application of the
CVM. European Review of Agricultural Econonmics, 22(2), 173-190.
Ready, R. C., Whitehead, J. C., & Blomquist, G. C. (1994). Contingent valuation when
respondents are ambivalent. Journal of Environmental Economics and Management,
29, 181-196.
Robertson, G. P., Paul, E. A., & Harwood, R. R. (2000). Greenhouse Gases in Intensive
Agriculture: Contributions of Individual Gases to the Radiative Forcing of the
Atmosphere. Science, 289(5486), 1922-1925.
Rosenstock, L., Keifer, M., Daniell, W. E., McConnell, R., & Claypoole, K. (1991). Chronic
central nervous system effects of acute organophosphate pesticide intoxication. The
Pesticide Health Effects Study Group. The Lancet(338), 223-226.
Serafy, S. E. (1998). Pricing the invaluable. Ecological Economics, 25, 25-27.
Smith, K. R. (2006). Public Payments For Environmental Services From Agriculture:
Precedents And Possibilities. American Journal of Agricultural Economics, 88 (5),
1167–1173.
Spalding, R. F., & Exner, M. E. (1993). Occurrence of Nitrate in Groundwater. Journal of
Environmental Quality, 22(3), 392.
Stevens, T., Benin, S., & Larson, J. (1995). Public attitudes and economic values for wetland
preservation in New England. Wetlands, 15(3), 226-231.
Strazzera, E., Scarpa, R., Calia, P., Garrod, G. D., & Willis, K. G. (2003). Modelling zero
values and protest responses in contingent valuation surveys. Applied economics,
35(2), 133-138.
Swinton, S. M., Lupi, F., Robertson, G. P., & Hamilton, S. K. (2007). Ecosystem services
and agriculture: Cultivating agricultural ecosystems for diverse benefits. Ecological
Economics, 64(2), 245-252.
Syswerda, S., Basso, B., Robertson, G. P., & Hamilton, S. (2008). Agricultural management
intensity affects long-term nitrate loss. (in prep).
Toman, M. (1998). Special section: forum on valuation of ecosystem services:Why not to
calculate the value of the world's ecosystem services and natural capital. Ecological
Economics, 25(1), 57-60.

153

Tscharntke, T., Klein, A. M., Kruess, A., Steffan-Dewenter, I., & Thies, C. (2005).
Landscape perspectives on agricultural intensification and biodiversity - ecosystem
service management. Ecology Letters, 8(8), 857-874.
United States Department of Agriculture. (2004). Farm Bill 2002. Retrieved. from
www.usda.gov/farmbill/.
United States Department of Agriculture Economic Research Service. (2008). County-Level
Unemployment and Median Household Income for Michigan. from
http://www.ers.usda.gov/data/unemployment/RDList2.asp?ST=MI
US Census Bureau. (2010). State & County QuickFacts- Michigan. from
http://quickfacts.census.gov/qfd/states/26000.html
Werner, M. (1999). Allowing for zeros in dichotomous-choice contingent-valuation models.
Journal of Business & Economic Statistics, 17(4), 479-486.
Woodward, R. T., & Wui, Y. S. (2001). The economic value of wetland services: a
meta-analysis. Ecological Economics, 37(2), 257-270.
Wooldridge, J. M. (2002). Econometric analysis of cross section and panel data: The MIT
press.
Wyland, L. J., L. E. Jackson, W. E. Chaney, K. Klonsky, Koike, S. T., & Kimple, B. (1996).
Winter cover crops in a vegetable cropping system: Impacts on nitrate leaching, soil
water, crop yield, pests and management costs. Agriculture, Ecosystems and
Environment, 59(1-2), 1-17.
Xu, Z., Cheng, G., Zhang, Z., Su, Z., & Loomis, J. (2003). Applying contingent valuation in
China to measure the total economic value of restoring ecosystem services in Ejina
region. Ecological Economics, 44(2-3), 345-358.
Zhang, W., Ricketts, T. H., Kremen, C., Carney, K., & Swinton, S. M. (2007). Ecosystem
services and dis-services to agriculture. Ecological Economics, 64(2), 253-260.

154