'13." n

l1.
,

__, . .
:. 3.-..J2l.

 

 

11:2?
'3.
.

it‘s-1? “‘
1:239

3!

0 .4... "-14- "I?

X ....
M ”m“.

‘v u-.. w

 

4;

Eff-$23.

.m-

'32:???

,

3“:r~""'" ...
4:2}; ‘
0‘

 

1w" ‘05

llllillliﬂlil'llﬂlllvl‘llilll Willi ;

3 1293 01051 0125

This is to certify that the

dissertation entitled

THE VALUATION OF GROUNDWATER POLLUTION POLICIES:
THE DIFFERENTIAL IMPACTS OF PREVENTION AND REMEDIATION

presented by

James D. Caudill

has been accepted towards fulﬁllment

of the requirements for;
Agrlcultural

Ph D . degree in Economics

yea/74

 

Major professor

Date /0%7/7/ Q?

MS U is an Afﬁrmatiw Action/Equal Opportunity Institution 0-12771

 

 

 

 

 

LIBRARY
Mlchlgan State
Unlverslty

 

 

 

me .4 3mm BOXtomnovothbMMMyumd.
TOAVOID FINES Munonorbdonddodm.

DATE DUE DATE DUE DATE DUE

    

 

 

 

 

 

 

Ila: O f
1 we 4. 5 ‘999

 

 

 

I
I
[LES ..

HE-
---
--
I II | l

MSU loAn Mmdlvo Action/E“ Opponunhy Inﬂation
W

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

uni-9.1

TI-IE VALUATION OF GROUNDWATER POLLUTION POLICIES:
THE DIFFERENTIAL IMPACTS OF PREVENTION AND REMEDIATION

By
James D. Caudill

Dissertation

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Agricultural Economies

1992

ABSTRACT

THE VALUATION OF GROUNDWATER POLLUTION POLICIES:
THE DIFFERENTIAL IMPACTS OF PREVENTION AND REMEDIATION
By

James Caudill

Groundwater pollution is a subject of continuing policy debate at the local, state,
and private levels. Past efforts in the economic evaluation of groundwater pollution
policies have typically focused on the health benefits of pollution control. Emphasis has
been on health costs Of pollution and material costs of remedial action as the basis for

estimating pollution control beneﬁts.

A potential weakness of these studies is their focus on mortality and material
costs. Two potentially important reasons exist as to why the exclusive focus on mortality
may result in incomplete or erroneous benefit estimates. First, values not directly
related to the impact of groundwater pollution on household health may be a signiﬁcant
source of beneﬁts. Altruistic, aesthetic and moral concerns may, in a given context,
represent motivations with signiﬁcant explanatory power regarding variations in
household bids for pollution control policies. Second, a considerable divergence may
exist between values Obtained from the mortality benefits approach and those derived

from the households’ perception of health risk from groundwater pollution.

To address these ideas, a research study was developed comprised of both
theoretical and empirical components with the main objective of examining the
relationship between beneﬁts and their underlying motivations and determinants. A
contingent valuation mail survey of over 2000 randomly selected Michigan households

was completed.

An analysis Of the empirical data shows two important results. First, values not
directly related to the health impact Of groundwater pollution are a signiﬁcant source of
policy beneﬁts. A comparison of subsets characterized by households with a high
concern for the possible non-health related environmental impacts of groundwater
pollution with households having a low level of concern showed that high concern
households typically had policy bids twice as large as low concern households. Second,
household perceptions of groundwater pollution health risks are systematically related to
a number of qualitative risk characteristics. The signiﬁcance of these characteristics
indicates that they are important both in being able to capture household risk
perceptions in a fairly consistent manner and providing a systematic link to the economic

trade-Offs household’s are willing to make in a given context.

ACKNOWLEDGEMENTS

I am most indebted to my dissertation advisor, John Hoehn, for his time, effort,
and support. Al Schmid, Luanne mm and Eileen van Ravenswaay provided timely and
constructive input.

I am grateful to the Department Of Agricultural Economies and the Agricultural
Research Station for ﬁnancial support. The departmental staff, especially Pat Neumann,
Nicole Alderman, Ann Davis, and Sherry Rich have always been extremely helpful and it
has been a pleasure knowing and working with them over the years.

Finally, I would like tO acknowledge the instrumental role several people played
in helping me maintain a prom perspective regarding the graduate school dissertation
process... Marian Popp, Cynthia Burton, Laura Geis and Jeff Wilson all played pivotal
roles in helping me maintain my sanity and self-respect. As a very small token Of my

appreciation, this dissertation is dedicated to them.

TABLE OF CONTENTS

Page
LIST OF TABLES ........... ‘ ....................................... viii
LIST OF FIGURES ................................................... x

CHAPTER 1 THE VALUATION OF GROUNDWATER POLLUTION POLICIES:
THE
DIFFERENTIAL IMPACTS OF PREVENTION AND REMEDIATION . . . . 1

Groundwater Usage .............................................. 2
Groundwmer Pollution ........................................... 3
The Health Beneﬁts of Groundwater Pollution Control ................... 4
Rancher ................................................. 4
Sharefkin et al. and Shechter ................................. 6
Implications for Policy Beneﬁt Estimates ........................ 6
Toward a More Comprehensive Framework ........................... 7
Nonhealth Beneﬁt Estimates ................................. 8
Alternative Motivations Behind Perceptions of
Policy Beneﬁts .......................................... 9,
Research Objectives ............................................. 10
Organization of the Dissertation ................................... 11
CHAPTER 2: AN ECONOMIC VALUATION MODEL OF GROUNDWATER
PROTECTION POLICIES ....................................... 13
An Economic Model Of Groundwater Pollution ........................ 14
Household wtp for Groundwmer Pollution ...................... 16
The Welfare Effects Of an. Alternative Groundwater Policy ......... 17
Implications Of the Theoretical Framework ........................... 20

CHAPTER 3: GROUNDWATER RISK PERCEPTIONS IN ECONOMIC
VALUATION ....................................................... 23

iv

Risk Perception Studies .......................................... 24

F ischhoff et al. ........................................... 24
Gardner et al. ............................................ 26

Kraus and Slovic ......................................... 27
Implications Of the Empirical Work ........................... 28
Descriptors Associated with Perceptions of Choice ..................... 31
Voluntariness ............................................ 32
Control ................................................. 33
Descriptors Representing Knowledge ................................ 33
Knowledge .......................................... '. . . . 34
Newness ................................................ 35
Descriptors Associated with Severity Of Outcome ...................... 35
Seriousness of Outcome .................................... 36
Immediacy of Health Effects ................................ 36
Catastrophic Potential ..................................... ' 37
Fear/Dread ............................................. 37
Summary Of Risk Descriptor and WTP Hypotheses ..................... 38
CHAPTER 4 CONTINGENT VALUATION DESIGN AND PROCEDURES ..... 40
Development Of a Contingent Valuation Questionnaire .................. 40
The Use of Focus Groups in CV Questionnaire Development ............. 41
Focus Group Objectives .................................... 41

Focus Group Development ................................. 41

Focus Group Session Format ................................ 42
General Findings ............................................... 43
Groundwater Knowledge ................................... 44
Sensitivity to Wording ..................................... 47
Variation in Underlying Groundwater Values ................... 48
Strength Of Prior Beliefs .................................... 49
Contingent Valuation Survey Format ................................ 51
Presentation Medium ...................................... 51
Description Of Policy Impacts ................................ 54
TheMethOdofProvision.......................... ......... 56

The Method Of Payment ................................... 57

Value Elicitation ..... ‘ .................................... 58
Independent Variables ........................................... 60
Risk Perceptions ......................................... 60
Q-altruism and Bequest Perceptions 62
Demographic and Water-use Characteristics 62

CHAPTER 5 QUESTIONNAIRE RESPONSES ............................ 63

Survey Response Rates .......................................... 63
Survey Responses .............................................. 64
Socio-economic Characteristics .............................. 64
Water Use .............................................. 66
Water Quality Perception ................................... 67
Risk Perceptions ......................................... 71
Environmental Perceptions ....................................... 75
Policy Bids .............. ' ...................................... 78
CHAPTER 6 DATA ANALYSIS ....................................... 81
Identiﬁcation of Independent Variables .............................. 81
Principal Components Analysis of Independent Variables .......... 83
Independent Variables and Testing of Hypotheses ...................... 88
Risk Variables ........................................... 88
Q-altruism and Bequest Risk .............. ‘ .................. 90
SociO-economic Variables ................................... 91
Regression Analysis ............................................. 93
Results ............... ' ........................................ 94
Underlying Variable Effects Averaged Across Policies ............. 95
Underlying Variables Speciﬁed with Interactive
Policy Variables ....................................... 100
Final Bid Equation Using Interactive Policy Variables ............ 104
CHAPTER 7 IMPLICATIONS OF EMPIRICAL RESULTS AND RESEARCH
NEEDS ........................................................... 107
Signiﬁcance Of Risk Perceptions .................................. 107
Implications for Valuation ................................. 107
Implications for Policy .................................... 109
Signiﬁcance Of Q-altruism and Bequest Risk Perceptions ............... 113
Implications for Valuation ................................. 113
Implications for Policy .. . . . . . .. ........................... 114
Future Research Needs ......................................... 116
Risk Perception Variables ................................. 116
Perceptions in Different Contexts ........................... 117
Role Of Information ...................................... 118

Quantitative and Qualitative Risk Measures ................... 118

APPENDICES

APPENDIX 1 SURVEY DESIGN AND IMPLEMENTATION - ......... 120
APPENDIX 2 QUESTIONNAIRE TEXT ................. l ......... 1 37
APPENDIX 3 SURVEY RESPONSE RATES ....................... 150 '
APPENDIX 4 SURVEYRESPONSES ............................ 156
APPENDIX 5 PHONE SCRIPT ................................. 17o

LIST OF REFERENCES ............................................. 171

Vii

LIST OF TABLES

Table 3.1 Comparison Of Selected Studies Showing Correlation (r)

between Risk Perceptions and Risk Factors ................................. 29
Table 3.2 Comparison of Risk Factors Identiﬁed by Factor Analysis

in F ischhoff et al. (F) and Krous and Slovic (KS) ............................ 30
Table 33 Summary of Factor Hypotheses ....................................... 39
Table 5.1 Cross Tabulation Of Water Source and Water Testing ...................... 67

Table 5.2 Cross Tabulation for Rural Sample Between Water Source
Degree to Which Tap Water is Perceived Safe to Drink ....................... 68

Table 5.3 Cross Tabulation for Urban Sample Between Water Source and
Degree to Which Tap Water is perceived Safe to Drink ....................... 69

Table 5.4 Cross Tabulation for Rural Set Of Tap Water Safety
Perceptions and Degree of Conﬁdence that Tap Water is Safe
from Effects Of Groundwater Pollution .................................... 70

Table 5.5 Cross Tabulation for Urban Set Of Tap Water Safety
Perceptions and Degree Of Conﬁdence that Tap Water is Safe
from Effects Of Groundwater Pollution .................................... 70

Table 5.6 Cross Tabulation Between Well Testing and Degree of
Conﬁdence that Tap Water is Safe from Groundwater Pollution ................. 71

Table 5.7 Cross Tabulation of Well Testing and Perceived Seriousness
Of Health Threat from Current Levels of Groundwater Pollution

in County ........................................................... 73
Table 5.8 Cross Tabulation Of Well Testing and Perceived Fatalness

Of Current Levels of Groundwater Pollution in Michigan ...................... 73
Table 5.9 Cross Tabulation of Q-altruism Perceptions and Education

Level .............................................................. 78
Table 5.10 Mean Policy Bids ................................................. 80
Table 6.1 Principal Components Analysis of Risk Variables .......................... 87

viii

Table 62 Principal Components Analysis Of Bequest and q-Altruism
Variables ........................................................... 89

Table 63 Independent Variables Used in Policy Bid Estimation ...................... 92
Table 6.4 Estimated Coefficients for Bid Equation with Impact

of Underlying Variables Averaged Across Policies ............................ 96
Table 6.5 Estimated Coefﬁcients for Bid Equations with Interactive

Policy Variables ..................................................... 101

Table 6.6 Estimated Coefﬁcients for Final Bid Equations with
Interactive Policy Variables ............................................ 105

Table 7.1 Policy Bids Resulting from Evaluation of Variable at
Mean Positive and Negative Deviation from Mean, Other

Variables Constant .................................................. 110
Table 01 Survey Return Characteristics ....................................... 151
Table C-2 Return Rates for Questionnaire Mail-outs .............................. 153
Table 03 Survey Returns by Week ........................................... 154
Table D-l Sample Demographics (percentage of usable responses

in speciﬁed category) ................................................. 157
Table D2 Water Use Information (percent of usable responses

in each category) .................................................... 160

Table D3 Respondent Water Quality Perceptions (percent Of usable
responses in each category) ............................................ 161

Table D4 Respondent Risk Perceptions: Severity (percentage Of
usable surveys in speciﬁed category) ..................................... 162

Table D6 Respondent Risk Perceptions: Knowledge (percentage of
usable surveys in speciﬁed category) ..................................... 164

Table D-6 Respondents Risk Perceptions: Choice (percentage of
usable surveys in speciﬁed category) ..................................... 166

Table D7 Respondents Environmental Perceptions (percentage Of 7
usable responses in speciﬁed category) ................................... 167

LIST OF FIGURES

: Front Cover Of Questionnaire ...................................... 126

: Back Cover of Questionnaire ...................................... 127

': Cover Letter for Initial Mail-out .................................... 129

: Post Card Reminder ............................................. 133

': Second Follow-up Letter .......................................... 135

: Third-Follow-up Letter ........................................... 136

CHAPTER]

THE VALUATION OF GROUNDWATER POLLUTION POLICIES:
THE DIFFERENTIAL IMPACTS OF PREVENTION AND REMEDIATION

Groundwater pollution is a subject Of continuing policy debate at the local, state and
federal levels. Fueling the concern are the importance Of groundwater as a water supply,
particularly in rural areas, and increasing evidence Of groundwater pollution in regions
across the US. (Pye et al., 1987; US. Environmental Protection Agency, 1984; Nielsen and
Lee, 1988). Past efforts in the economic evaluation Of policies to address groundwater
pollution have typically focused on the health beneﬁts Of pollution control (Rancher, 1983;
Sharefkin et al., 1984; and Schechter, 1985 a, b). The frameworks used in these studies rely
on the health costs of pollution and material costs of remedial action as a basis for
estimating groundwater pollution control beneﬁts.

Two potential problems are evident. First, these studies derived health costs using
mortality probability and an estimate Of the value of a statistical life. However, there may
be a considerable divergence between an estimate of beneﬁts Obtained from a mortality
based probabilistic approach, and beneﬁts estimated from an explicit consideration of
people’s perceptions Of the mortality and morbidity risk from groundwater pollution. A
substantial body of research, both in psychology and economics, suggests that people
frequently perceive health risks in a much diﬁ'erent manner than might be expected from
a technical risk assessment based on mortality probabilities and the economic value Of a

statistical life.

2
Second. people may value groundwater services which are unrelated to any direct

health eﬁect of groundwater pollution. Existence and bequest value may be two values
which people Obtain from groundwater. In certain situations, these values may be a .
signiﬁcant source Of policy beneﬁts, and excluding them from the evaluation process may
result in inefficiencies in both the level and mix Of groundwater pollution policies.

This Chapter develops a rationale for the development and empirical testing of a
more comprehensive economic valuation framework for groundwater pollution policies

which explicitly takes into account household perceptions Of groundwater pollution risk

Groundwater Usage

Groundwater is an important source Of freshwater for households in the U.S. and
Michigan. Groundwater comprises about one-fourth Of all freshwater used annually in the
U.S. (U.S. Geological Survey, 1985). Most groundwater is used for agricultural purposes.
However, a substantial portion of household freshwater needs for both the U.S. and
Michigan is supplied by groundwater. Thirty percent of the U.S. population relies on
publicly supplied groundwater, while 21 percent depends on private household wells (Solley
et al., 1983). Groundwater withdrawals have been steadily on the rise. In the U.S.,
withdrawals from public systems increased 164 percent from 1950 to 1980, and withdrawals
from private wells increased 60 percent (Solley et aL, 1983).

Groundwater provides a substantial portion Of Michigan’s freshwater needs. Fifteen
percent Of all households obtain groundwater from public systems, while 28 percent depend
on private wells (Solley et al., 1983). Consequently, nearly 43 percent of Michigan’s

households rely on groundwater as a water source.

3
Groundwater Pollution

Groundwater pollution is caused by both naturally occurring and anthropogenic
substances that are potentially harmful to humans, plants and animals. Naturally occurring
pollutants include dissolved minerals, salts and radionuclides such as radon (Pye et al. 1987).
Anthropogenic sources include: (a) industrial surface impoundments and injection wells,
(b) landﬁlls, (c) septic tanks, (d) land spreading Of sewage, (e) accidental chemical spills and
leaks, (f) farm production, (g) lawn and garden chemicals, (h) sub-surface oil and gasoline
storage tanks, (i) storage areas for road de-icing chemicals, (j) mining activities and (k)
intensive groundwater pumping which in certain areas may lead to saltwater encroachment
(Ward et al., 1985).

The impact of groundwater pollution depends on a variety Of factors. These include:
(a) the type and extent Of groundwater use, (b) the source and chemical composition of
contaminants, and (c) the contaminant concentration level in the water (U.S. Geological
Survey, 1984). The magnitude Of resulting health impacts will depend on (a) the duration
Of expowre, (b) route Of exposure, and (c) the age and physical condition Of exposed
individuals (Environ Corporation, 1983).

Pollution poses a greater hazard in areas where water use patterns and hydro-
geological conditions make the area more vulnerable. For example, groundwater pollution
from agricultural chemical use is a concern in rural areas where farm chemicals are applied
extensively. Approximately 22 percent Of wells sampled in the southern portion of
Michigan’s Van Buren County had nitrate concentrations equal to or exceeding the federal
water quality standard of 10 mg/l (U.S. Geological Survey, 1985). A non-random sample
Of 1476 private rural wells throughout Michigan showed that 8 percent exceeded nitrate
drinking water standards (Michigan Farm Bureau, 1989). A sampling of 50 wells in
Michigan, which were identiﬁed ex ante to be at high risk from agricultural chemical

4

contamination, showed 24 percent as exceeding drinking water standards for nitrates. Four
percent of the wells had pesticide concentrations exceeding EPA Health Advisory levels for

Atrazine (Olson, 1989).

The Health Beneﬁts of Groundwater Pollution Control

Depending on the type of contaminant and length and magnitude Of exposure,
groundwater pollution can adversely affect health in several ways. For example, ingestion
Of nitrates by infants may result in methemoglobinemia, a condition characterized by a
reduction in the oxygen carrying capability Of the blood (National Academy Of Sciences,
1979). High nitrate ingestion has also been associated with stomach cancer (National
Academy Of Sciences, 1979). Heavy metals such as lead, mercury, iron, cadmium and nickel
have been linked to a variety of adverse health ailments such as birth defects, increased
frequency Of miscarriages, renal dysfunction and developmental disabilities (Craun, 1985).

Several economic studies have developed frameworks to estimate the health beneﬁts
of groundwater pollution control (Rancher, 1983, 1986; Shareﬂtin et al., 1984; and Shechter,
1985 a,b). These frameworks use the health costs of pollution and material costs of

remedial action as the basis for estimating the beneﬁts of groundwater pollution control. '

Rancher

Raucher’s two studies (1983, 1986) address two different approaches to the
estimation of health beneﬁts: the materials cost approach and the expected beneﬁts
approach. The materials cost approach (1986) assesses the policy costs necessary to reduce
the health impacts of a groundwater pollution episode. An estimate is made both Of the
number Of excess cancer deaths which would occur in the absence of the policy and policy

costs per cancer avoided.

5
The expected beneﬁts approach (1983) describes the value of reducing mortality risks

from groundwater pollution. This approach to the estimation Of groundwater policy beneﬁts
and its implications for a more comprehenSive estimate Of policy beneﬁts forms the basis
for the dissertation research. Because the expected beneﬁts approach explicitly incorporates
beneﬁts into the framework, the subsequent discussion will focus on this approach.

Rancher (1983) uses the expected beneﬁts approach to estimate the. beneﬁts
associated with several hypothetical situations. These situations diﬁer in terms Ofa variety
Of physical and economic parameters, including contaminant plume size and growth, time
horizon, discount rate and the type of groundwater use. Health beneﬁts are based on the
hypothetical assumption that 1.43 excess deaths per year would occur among the exposed
population in the absence Of any pollution control policies. From previous reSearch on the
value Of a statistical life, a ﬁgure of $1,000,000 is used as the value Of avoiding one excess
mortality. As a result, annual health beneﬁts of a policy which was certain to eliminate the
mortality threat from groundwater pollution equals $1.43 million. Mean annual per capita
health beneﬁts are $14.30 based on an exposed population Of 100,000. To these health
beneﬁts are added the avoided material costs Of remedial action to Obtain a measure of the
expected beneﬁts Of a groundwater policy.

The main ﬁndings of Rancher’s analysis indicate that beneﬁt estimates are sensitive
to a number Of physical and economic parameters. Physical parameters include local
hydrological conditions, water consumption and land-use patterns and waste disposal
practices. Economic factors include the time horizon and the discount rate. The analysis
shows that the expected beneﬁts of groundwater pollution control are greatly dependent on
plume size, rate of growth and whether the water is used as drinking water or for
agricultural irrigation. For example, expected beneﬁts in a situation characterized by a
small, slow growing plume in groundwater used for agricultural purposes, is $146,600.

6
Expected beneﬁts in a situation characterized by a large, fast growing plume in groundwater

used for drinking water is, other things equal, $887,200. The speciﬁc conditions which
characterize the situation under study will have a considerable impact on estimates of

groundwater policy beneﬁts.

Sharendn et al. and Shechter

Shareﬂtin et al. (1984) and Shechter (1985 a,b) extend Rancher’s framework by
incorporating physical groundwater models and dose-response models to analyze speciﬁc
groundwater pollution episodes. An estimate or range Of estimates is Obtained of the value
Of a statistical life. This number is multiplied by the incremental reduction in mortality risk
due to the implementation of a particular policy to Obtain an estimate of health beneﬁts.

Sharefkin et al. estimate an aggregate mortality risk by summing incremental
mortality risks across wells and across chemical contaminants for the area under study.
Using a range Of $100,000 to $1,000,000 for the value Of a statistical life and multiplying
these values by the high and low aggregate mortality estimates, health beneﬁts were
estimated to be in the range Of $176 million to $2.02 billion for an exposed population of
one million. A similar approach is used by Shechter.

Implications for Policy Beneﬁt Estimates

Three potential problems exist with regard to the past focus on mortality and
material policy costs. First, there may be considerable divergence between values Obtained
from the expected beneﬁts approach and those derived from people’s perception of the
mortality risk from groundwater pollution. A substantial body of research, both in
psychology and economics. suggests that people frequently perceive mortality risk in a much
different manner than would be expected from a technical risk assessment. If such is the

7

case, a question arises as to the appropriate source Of information to be used to estimate

policy beneﬁts: technical risk assessments or consumer perceptions Of the situation or some

combination of the two. Integration of information Obtained both from technical risk .

assessments and the study Of consumer risk preferences would seem to Oﬁ'er the best chance
that the design and implementation Of environmental risk policies would result in a pareto
better situation. However, this integrated approach is predicated on a systematic
relationship beaveen perceptions and behavior. More speciﬁcally, are risk perceptions,
including both probabilistic and non-probabilistic components, systematically related to
economic behavior? If not, then any divergence between technical risk assessments and
household perceptions is rendered mute. If there is such a relationship, then household
preferences may be an important consideration in estimating the economic beneﬁts of risk
reducing policies.

Second, aside from mortality risk, morbidity risk may be an important component
of health effects but is typically not considered as a possible source of policy beneﬁts.

Third, values not directly related tO the impact of groundwater pollution on a
consumer’s or households health may still be a signiﬁcant source Of policy beneﬁts. People
may value the knowledge that future generations have access to unpolluted groundwater.
Consumers may value the knowledge that the groundwater resource itself is free of pollution
or that animals and plants which depend on it are not aﬁected by pollution, regardless of
any current or future direct use. These values may be signiﬁcant and excluding them from

valuation process may result in the misallocation Of resources.

Toward a More Comprehensive Framework
Rancher (1983, 1986), Sharetkin et al. (1984) and Shechter (1985 a,b) acknowledge

possible shortcoming Of relying solely on health values as a measure Of policy beneﬁts.

8
Rancher notes that his conceptual framework omits intrinsic beneﬁts Of groundwater

protection. He deﬁnes intrinsic beneﬁts as "values that society may place on groundwater
protection independent of an aquifer’s current use value or near-term costs of
contamination“ (p. 32.3, 1983). The main component of these beneﬁts is existence value,
which he deﬁnes as 'the willingness to pay for the knowledge that a particular level of
environmental quality exists, regardless Of any present or anticipated use by the individual“
(p. 323, 1983). Rancher notes that these intrinsic values associated with groundwater may
be signiﬁcant.

Shareﬁtin et al. (1984) and Shechter (1985) both note the potential signiﬁcance of
household perceptions Of risk associated with groundwater pollution. Sharetkin et al. note
that 'it is risk perceptions that matter to individuals and that determine individual valuations
Of risk‘ (p. 1782).

Nonhealth Beneﬁt Estimates

Edwards (1988) is unique in attempting to estimate non-health beneﬁts of
groundwater pollution control. The Objective Of the study was to determine the economic
beneﬁts Of a policy which eliminated or reduced the probability of nitrate groundwater
pollution. Health risks Of nitrate pollution were not considered as a component Of policy
beneﬁts. 'The list Of [policy] beneﬁts included wanting a cost-effective supply Of water for
personal use and protecting groundwater for use by future generations, but excluded direct
health risks“ (p. 477). Direct health effects were not considered because the state and
county “systematically monitor nitrate levels in each of the public wells in order tO prevent
dangerous exposures to nitrate” (p. 477).

A major ﬁnding is that values motivated by altruistic concerns such as bequest value
are a signiﬁcant source Of beneﬁts from groundwater pollution control. Annual per

9
household willingness to pay (wtp) for groundwater protection ranged from $364 to $1437

depending on the probability of a successful policy. Total wtp was comprised of three
components: (a) personal use value, (b) bequest value, and (c) Option value.

Personal use value is the value people place on the availability of a low cost potable
water supply. Personal use value comprised between 10 and 30 percent of total wtp.
Bequest value is the value to individuals Of knowing that future generations have access to
clean groundwater. Bequest value contributed between 70 and 90 percent of total wtp.
Option value is a risk premium people would be willing to make to ensure future supply of
clean groundwater. Option value was estimated to be 1 percent or less of total wtp.

The comparatively large bequest values estimated by Edwards bring up the question
of their source. High values could be an artifact of methodoloy. For example, ambiguities
in Edwards’ questionnaire may have resulted in direct health values, and possibly other non-
health values besides bequest value, to be included in estimated beneﬁts. In addition, the
range of probabilities of nitrate contamination presented to respondents were comparatively
high, ranging from 25 percent to 100 percent. Nevertheless, it is clear that policy impacts
on non-direct health values may be a signiﬁcant source of beneﬁts. Understanding the
possible motivations behind values associated with non-direct health effects of groundwater

pollution will result in more accurate and comprehensive beneﬁt estimates.

Alternative Motivations Behind Perceptions of Policy Benefits

Perceptions ’of policy beneﬁts may arise from several motivations, including: (a)
household concern over the potential health eﬁect of exposure to groundwater pollution; (b)
concern over the health eﬁ'ects of groundwater pollution on other households and future
generations; (c) the effect of groundwater pollution on a household’s use of resources which

depend on groundwater as a water source, completely aside from any potential health

10
effects; and (d) the eﬁ'ect of groundwater pollution on a household’s non-use values

associated with groundwater.

The health risk from groundwater pollution presents a particular type of health risk.
Household health risk perceptions may be the result of the subjective assessment of the risk
in terms of a number of economic factors and risk characteristics. Household perceptions
of these characteristics may affect their perception of the seriousness of the risk, and in
turn, their valuation of groundwater pollution policies.

Aside from the direct health threat of groundwater pollution, households may be
affected by groundwater pollution in three ways: ' (a) by using resources dependent on
groundwater, for example the recreational use of a lake or river that is recharged by
groundwater; (b) altruistic concern over the potential health effects of groundwater pollution
on others; and (c) altruistic concern with respect to the existence value Of groundwater.
This may include perceptions of groundwater as an integral part of a pollution-free
ecosystem, the intrinsic value in groundwater remaining unpolluted, and that pollution (or
the act of polluting), completely aside from any possible effects on humans, plants or
wildlife, violates a moral principle or social norm. These non-direct health related effects

may be a sigliﬁcant source of beneﬁts from pollution control policies.

Research Objectives
The dissertation research has three Objectives. The ﬁrst is to develop a valuation
framework that considers alternative sources of value in addition to direct health effects.
As existing research indicates, the relatiOnship between health and non-health values is not
well understood. Previous research has focused primarily on the direct health impacts of
groundwater pollution. Yet, as Edwards’ study shows, indirect non-health policy impacts

may comprise a signiﬁcant part of policy beneﬁts.

11

The second Objective is to examine the relationship between beneﬁt values and their
underlying motivations or determinants. These motivations include: (a) such conventional
economic concerns as income, education, and other demographic characteristics; (b) values
associated with household perceptions of groundwater pollution health risks; and (c)
household attitudes toward values derived from altruistic concerns. These motivations will
be examined and linked to a utility theoretic model. Hypotheses regarding the signiﬁcance
and Sign of these relationships will be derived.

The third objective is to develop and implement an empirical study to test the
derived hypotheses. A contingent valuation mail survey is used to obtain the necessary data.
The data include measures of: 1) perceived health and environmental risk, 2) willingness
to pay for policies which. reduce groundwater pollution, and 3) sociO-demographic
characteristics of respondents in the sample. The population sampled are Michigan
households statewide, categorized as urban or rural. This is discussed in greater detail in

Chapter 5.

Organization of the Dissertation
Chapter 2 develops an analytical framework, and examines the implications of
economic and environmental factors on the valuation of alternative groundwater pollution
policies. Chapter 3 speciﬁes how subjective assessments of the health risk from groundwater
pollution may aﬁ'ect policy bids. Chapter 4 discusses the development of a contingent
valuation survey to empirically test the relationship between household perceptions and wtp

for groundwater pollution. policies. Chapter 5 discusses the questionnaire responses.

12
Chapter 6 includes an econometric analysis of the questionnaire responses. Finally, Chapter

7 discusses the implications of the research and identiﬁes future research needs.

CHAPTERZ

AN ECONOMIC VALUATION MODEL OF
GROUNDWATER PROTECTION POLICIES

Groundwater pollution may be controlled using very different policy instruments.
Some policies seek to prevent pollution at the source while others intervene with
remedial action at points of potential human exposure. These diﬁerent instruments
provide diﬁerent services to those affected by groundwater pollution. Pollution
prevention protects natural resources, aesthetic services, and human health. Remedial
action is typically aimed only at protecting human health (Rancher 1983). Since they
provide different services, the economic values of prevention and remedial policies are
likely to diverge.

This chapter develops an economic model to link consumer preferences to the
economic valuation of groundwater policies; More speciﬁcally, economic value is
associated with the policy services provided. Consumer preferences regarding both
health related services and non-health related environmental services are shown to be
key determinants of economic value. In situations where households perceive signiﬁcant
environmental services from groundwater in addition to health considerations, prevention
policies may be the most economically efficient response to groundwater pollution. In
cases where health effects of groundwater pollution are the primary concern, remedial
policies may be the most economically eﬂicient response. Finally, expressions for

l4

marginal policy beneﬁts are derived and show the potential importance Of household

perceptions in estimating household wtp for groundwater pollution policies.

An Economic Model of Groundwater Pollution
Rancher’s analysis (1983) suggests that groundwater pollution is characterized by
a sequence of uncertain events. Pollutants originate at some source and pass through an
environmental medium (in this case the groundwater system) resulting in some level of

actualorpotentialaposrlreasshowninthediagrambelow:

 

 

Contaminant ’ Environmental Human
Source 1 Medium 1 4» Exposure
Preventive Remedial
Policies Policies

The pollution sequence of release leading to environmental effects and exposure
leading to health effects suggests a two-stage model of groundwater policy. The ﬁrst
stage involves prevention of groundwater pollution. Preventive policies control pollution
at the source. Preventive policies use source reduction and various types of barriers to'
intervene between the source and the environmental medium, as shown in the diagram
above. Consequently, prevention protects both the environmental medium and human
health. For simplicity, pollution is viewed as a threshold event: groundwater is polluted
if the concentration of some contaminant exceeds a certain level; it is not polluted if the
concentration falls below a given level. This View of pollution as a threshold event is
embodied in the maximum contaminant levels implemented under Federal drinking
water regulations (HOCFR Part 141).

15

A common feature of pollution strategies is that sucwss is not certain. Policy
success, i.e. preventing or reducing pollution below a certain threshold level, is uncertain.
Typically, from a regulatory perspective, the probability of policy success depends, mm
aha, on the amount of public and private investment in a particular strategy or
combination of strategies. The probability of a successful prevention policy is therefore
noted denoted a', - I" (c,), Bap/ac, > 0, where c, is the level of public and private
expenditure on prevention.

Remedial regulatory action decisions are made at the second stage of the two-
stage pollution sequence. As the diagram indicates, remedial action policies intervene
only after environmental pollution has occurred. Remedial action would include ﬁltering
or treating water used for human consumption, providing alternative drinking water
sources, and other means of avoidance. Remedial action tends to focus on reducing
human exposure in order to protect human health. Even if human exposure occurs,
health effects remain uncertain and are potentially remediated through medical
treatment. As with preventive strategies, the likely success of remedial action depends
upon the level of both public and private investment. The probability of successful
remedial action is denoted an - an (on) where on denotes the level of expenditures on
remediation and Ban/61g, > 0.

The discussion above indicates that two types of services are provided to
households depending on the policy: (1) health services provided by both preventive and
remedial action policies, and (2) environmental services provided by the prevention
policy only. The following section develops a simple model which focuses on those
features with potentially important implications for policy valuation.

16
Household Willingness to Pay for Groundwater Pollution Policies

Household willingness to pay (wtp) represents the households’ valuation of policy
services. A household’s willingness to pay for prevention and remedial action depends
on the intended policy outcomes and a policy’s effectiveness in attaining these outcomes.
The model of willingness to pay, therefore, begins by describing a household’s well-being
as a utility index v(m,s,q,e) where m is household income, 3 is a vector of socioeconomic
characteristics describing the household, q is an index of non-health environmental
services, and e is an index of exposure to groundwater contaminants. For simplicity, it is
assumed that groundwater contamination has no effect on m or s.

The suwess or failure of each stage of groundwater policy has a different
outcome and a diﬁerent welfare effect. Let q° be the initial level of non-health
environmental services and e° be the initial level of almosure. If prevention is successful,
environmental services are q0 and exposure to contaminants is e°. Household well-being
is therefore v,(m) - v(m,s,q°,e°). If prevention fails but remedial action prevents
exposure, the outcome is a lower level of environmental services q’, but no additional
exposure to contaminants, e°. Household well-being is vc . v(m,s,q‘,e°) with failed
prevention and successful remedial action. Finally, if both prevention and remedial
action fail, policy outcomes are q1 and e1 and household utility is vl3 - v(m,s,q‘,e‘). The
utith index VB is implicitly an expected utility index conditioned on the health effects
due to exposure as perceived by the household. Hence, exposure and perceived health
effects are discussed below as qualitatively equivalent issues.

The three utility indexes conditioned on policy outcomes are state dependent
indexes where utility depends upon the policy state. Following Graham (1981), the
expected utility Of a given expenditure on prevention and remedial action is formulated

by combining the state dependent utility indexes with the probability of a given policy

l7
outcome. This yields, ex ante, expected household utility from an initial prevention and

remedial action policy as

(1) 0° = In Vp(m) + (1-fm)[fao Vc(m) + (l-fm)V2(m)]

where an, is the household’s subjective probability of the effectiveness of the initial or
existing prevention policy and a” is the household’s subjective probability that the initial
or existing remedial action policy will be successful.

Equation (1) describes the household’s baseline level of well-being. It depends

upon both the effectiveness of prevention and remedial action.

The Welfare Effects of an Alternative Groundwater Policy

The welfare effects of an alternative groundwater policy are derived by
comparing the baseline utility with the expected utility obtained from an alternative
groundwater policy. This alternative level of utility is

(2) “i ' 7?: WU“) + (I‘Tm)['mivc(m) + (l'tmydmn

where an is the household’s subjective probability that the new policy is successful in
preventing aquifer pollution, and an is the household’s subjective probability that the
remedial action component of the new policy is successful.

The welfare effect of a policy change may be measured by the Hicksian
compensating welfare measure. The Hicksian compensating measure is the maximum
amount of income that an individual is willing to give up, or pay, in order to get the

policy change. This maximum amount would leave an individual indiﬂ'erent between the

18
baseline level of utility and that obtained after paying for the alternative. Algebraically,

the Hicksian compensating measure, t', for a policy that oﬁ'ers both a change in

protection and remedial action is

(3) “o ‘ 7?: VKm‘t.) 4' (I'THH'M Vc(m't.) + (l'fkl)v5(m't.)]

A positive t' represents a household’s maximum willingness to pay for a groundwater
policy. that improves household well-being. If an alternative policy worsens household
well-being t' is negative and the absolute value of t' is the household’s minimum
acceptable compensation for voluntary acceptance of the policy change.

At t', the marginal beneﬁt of additional prevention and remedial action can be
derived by taking the total diﬂ'erential of equation (3) with respect to an and t', and rm
and t'. Taking this approach, the marginal benefit of additional prevention is

(4) mb, " dt°/da',, ' Vr/E“- ' [7m vc + (l'ﬂu) val/Eu,

where v, - v,(m-t'), vC s vc(m-t°), v3 - vB(m-t'), and Eu. is expected marginal utility of
income derived from equation (3). Adding and subtracting vc to equation (4) and

rearranging yields the marginal benefit of additional prevention as

(5) mb, ' (VP ‘ Vc)/E“- + (1‘7 luxvc ”IQ/E“:-

The right-hand side of equation (5) separates the marginal beneﬁt of prevention
into two terms. The ﬁrst term on the right-hand side is the marginal benefit of

protecting environmental services. It arises due to the ecological, intrinsic, or bequest

19

value of protected groundwater services. The second term on the right-hand side is the
health benefit of protection. This health beneﬁt depends not only upon the welfare
effect of exposure, vC - v8, but also upon the probability the remedial action fails to
prevent exposure, (14”). Hence, the health beneﬁt attributable to prevention increases
with the probability of failed remedial action. The health beneﬁt of prevention declines
with the probability of successful remedial action.

The marginal beneﬁt of additional remedial action is

(6) mba ' dtvd‘tm ' (1 ' THXVC ‘ “EVE“:

Equation (6) shows that the marginal beneﬁt of remedial action is the product of (a) the
probability that prevention fails and (b) the welfare effect due to exposure. The beneﬁt
of remedial action increases as the probability of prevention failure increases and
decreases as the probability of successful prevention increases.

It is useful to examine the marginal beneﬁts of prevention in a case where a
household perceives no ecological, intrinsic, or bequest services from groundwater. In
this case, v, is equal to vC and the ﬁrst term in equation (5) equals zero. Without the
environmental effect, the marginal beneﬁt of prevention is similar, but not identical, in
content to the marginal beneﬁt of remediation action. This can be shown by equating
the two marginal beneﬁts expressions-an eﬂ'icient policy mix will equate marginal
beneﬁts across policies. This is shown in equation (7).

(7) mb, - mbn '(1'791XVC'VQE‘1- =(1-wmxvc-9e/Eu.

20

Both marginal beneﬁt expressions are based on the utility gain from avoiding exposure,
(vc - vE)/Eu.. The crucial difference between the two expressions in equation (7) is,
respectively, that the utility gain from' avoiding exposure through prevention is
discounted by the probability that remedial action fails, while the utility gain from
avoiding exposure through remedial action is discounted by the probability that

prevention fails.

Implications of the Theoretical W

The framework developed in this chapter has several implications for
groundwater policy valuation. First, the framework presented here shows the effect of
sequential policy implementation on the measurement of policy beneﬁts. The
discounting of the marginal beneﬁts of prevention and remedial policies is a reﬂection
that the two policies are substitutes for one another, insofar as health beneﬁts are
concerned. This is consistent with the ﬁnding by Hoehn (1989) that because of
substitution effects, the sum of policy impacts valued separately will be yeater than the
same impacts valued together as one policy.

Second, with respect to the estimation of health beneﬁts, Raucher’s mortality-
based approach‘ to estimating policy beneﬁts is essentially a special case of the more
general framework described in this chapter. Raucher’s approach is appropriate when
mortality is the only outcome important to people and there is agreement on the
”correct“ mortality probability. In the absence of these conditions, health beneﬁt
estimates based on the general framework are more appropriate.

Third, the conceptual framework shows the effect on marginal policy beneﬁts of

an explicit consideration of the environmental services provided by a prevention policy.

21

In cases where environmental services are an important source of beneﬁts to people, the
approach used by Raucher will underestimate policy beneﬁts.

Fourth, the derived expressions for the marginal beneﬁts of prevention and
remediation show the potential importance of household’s subjective perceptions in
aﬁecting their wtp for groundwater pollution policies. Equations (4) and (6) specify the
marginal beneﬁts from prevention and remediation respectively. Holding Eu. constant,
three terms will affect the magnitude of policy beneﬁts: (a) the perceived probabilities
of a successful policy; (b) the health beneﬁts component, (vcva); and (c) the
environmental beneﬁts component, (vp-vc). These terms imply that as perceptions
change (or differ across households), wtp will change accordingly. To take a simple
example, suppose that household A believes that exposure to groundwater pollution will
result in death while B believes that exposure will result in a mild stomach upset. Other
things equal, vEB > v“. As a result, the term (vC-vE) and consequently wtp, will be
greater for household A, holding the other arguments of the expression constant.

Perceptions, especially baseline perceptions, may differ across households for a
variety of reasons, including, for example, differences in: (a) past experiences with and
knowledge of groundwater services and pollution, (b) education, (c) income, (d)
environmental concern, and (e) attitudes towards regulatory activity.

An important question for policy valuation is the extent to which household
perceptions of policy services are related to perceived policy beneﬁts as measured by
wtp. The empirical testing of the relationship between health and environmental risk
perceptions and wtp for groundwater policies necessitates the development of variables
which are an accurate reﬂection of household risk perceptions. Chapter 3 addresses this

question in greater depth.

NOTES

Raucher’s approach to estimating the health beneﬁts from reductions in the .
probability of groundwater contamination is based on reductions in mortality
from groundwater contamination exposure. Raucher (1983, 1986) deﬁnes the
expected benefits of a policy as the change in expected damages from
groundwater contamination due to a change in the probability that contamination
will occur. Expected damages is a weighted function of (a) the cost of the most
economically efficient response to the contamination incident and (b) the cost
incurred if contaminated water were used in the same manner as before the
contamination incident (p. 320-1, 1983). Excess annual mortality due to
groundwater contamination among the exposed population is used as an estimate
of the value of the health effects of groundwater contamination. The value
attached to each excess death is derived from a range of values speciﬁed in the

value of a statistical life literature (p. 324, 1983).

CHAPTER3

GROUNDWATER RISK PERCEPTIONS
IN ECONOMIC VALUATION

The theoretical framework developed in the previous chapter identiﬁed three
elements of a choice situation under risk: (a) the outcomes that might occur, v,, v0 and
v,; (b) the probability that these outcomes will occur; and (c) the ability on the part of
the household to aﬁect the outcome and probability (in this case the payment of t').
These three elements are similar to what Rescher identiﬁes as the elements of a risky
situation (1984).

An important implication is that household perceptions of the probabilities and
outcomes in a groundwater risk situation may be signiﬁcant determinants of the beneﬁts
associated with groundwater pollution policies. The importance of perceptions in
aﬂecting economic behavior has been documented in psychological research, which has
shown that typical citizens are unlikely to perceive risk in the abstract terms of an
economic model (e.g.: TVersky and Kahneman, 1981; Slovic et al., 1984; Slovic, 1987).
Households appear to view both probabilities and utility outcomes in terms of a multi-
dimensional set of descriptors Households are also likely to vary in their risk
perceptions. Consequently, to elicit policy values that are accurate reﬂections of
consumer preferences, it is necessary to ﬁnd ways to specify probabilities and outcomes

in terms that are meaningful to ordinary citizens .

24

The objective of this chapter is to examine the likely relationships between risk
perceptions and willingness to pay for groundwater pollution policies. To develop the
rationale for an explicit consideration of risk perceptions in policy valuation, the chapter
is organized as follows: (a) three empirical studies dealing with risk perceptions
(Fischhoﬁ et al., 1978, Gardner et al., 1982, Kraus and Slovic, 1988) are examined; (b)
the implications of the empirical work for policy valuation are addressed; and (c) risk
descriptors and their hypothesized relationship to individuals’ willingness to. pay for

groundwater protection policies are speciﬁed.

Risk Perception Studies
Three risk perception studies are discussed as to their general methodology and
ﬁndings regarding the empirical signiﬁcance of qualitative risk descriptors.

Fischhoﬂ et al.

F ischhoff et al. (1978) developed a psychometric study of attitudes towards
technological risk. Seventy-six members of the League of Women Voters were asked to
make quantitative judgements concerning perceived risk, perceived beneﬁt ‘ and
acceptable risk for 30 diﬂ'erent activities and technologies. The responses were used to
see if any systematic relationship existed between perceived risk, perceived beneﬁt and
acceptable risk.

Nine risk descriptors were studied. According to Fischhoﬁ et al., each of these
risk descriptors had been hypothesized in previous studies (e.g., Lowrance, 1976) as
inﬂuencing risk perceptions. The risk descriptors included (1) voluntariness, (2)
immediacy of effect, (3) scientiﬁc knowledge about risk, (4) exposed knowledge about

risk, (5) control over risk, (6) newness, (7) chronic-catastrophic, (8) common-dread, and

25

(9) severity of consequences. Participants rated the 30 activities and technologies on
each of these nine descriptors. The risk descriptors were rated on a scale of l to 7, with
the end points representing extremes on a continuum. For example, 1 on the scale for
voluntariness was labelled I'voluntary" and 7 was labelled 'involuntary' (Fischhoﬂ' et al.,
1978, p. 133).

Means were calculated for each risk descriptor for each technology. In, a sense,
there were 30 observations (each technology being an 'observation') for each descriptor
with each observation being a mean derived from 76 individual responses. The means
were used in three ways: (a) to ascertain the correlation (r) between risk descriptors
and perceived 158“; (b) to ﬁnd possible underlying dimensions of risk through factor
analysis; and (c) specifying perceived risk as a function of perceived beneﬁts and risk
descriptors, and analyzing this relationship using regression analysis.

WW Only “”0 descriptors were sieﬁﬁmntly
correlated (r) with perceived risk, common / dread and severity of consequences, with
correlations of 0.64 and 0.67 respectively. A number of risk descriptors were
intercorrelated. For example, voluntariness was signiﬁcantly correlated with seven other
risk dacriptors (r coefﬁcients ranging from .54 to .76), and newness was signiﬁcantly
correlated with ﬁve risk descriptors (r coefficients from .53 to .83). The degree of
intercorrelation indicated that they could potentially be represented by a few underlying
dimensions.

W A principal components factor analysis was completed for the
nine risk descriptors. M underlying variables were identiﬁed (loadings in parentheses).
Factor 1 is identiﬁed as representing technological risk. It included voluntariness (.89);
immediacy (.70); knowledge of exposed (.88); knowledge of science (.88); control (-.83);

new (-.87); chronic (.62); and common (.67). This factor is ”associated with new,

26

involuntary, highly, technological [hazards], which have delayed consequences for masses
of people” (p. 147). An «ample would be nuclear power. The second factor represents
severity, and includes chronic (.55); common (.60); and severity (.91). This factor is
correlated with familiar hazards whose consequences are likely to be fatal, such as
general aviation and hand guns. These two factor variables accounted for 80 percent of
the total variability.

mm The next step analyzed the relationship between perceived
risk and possible «planatory variables including perceived beneﬁts and factors 1 and 2.
Estimated coefficients for the independent variables were not given, but the variables as
a group were signiﬁcant (multiple R score of .67 and F -6.96 signiﬁcant at the .005 level.

Gardner et al.

While the Fischhoff et al. approach elicits mean risk desuiptor judgements across
30 different technologies, two other studies look at a single technological risk and elicit
individual risk descriptor responses. Gardner et al. (1982) elicit people’s risk perceptions
towards nuclear power. A component of the study, in a manner similar to Fischhoff et
al., addresses individual risk perceptions of nuclear power by eliciting responses to
questions on perceived risk and several risk descriptors. Zero order correlations for
perceived risk and the risk descriptors dread, catastrophic potential and scientiﬁc
knowledge were obtained. Perceived risk had a correlation coefficient of .63 with
catastrophic potential, .44 with dread, and .33 with scientiﬁc knowledge. Gardner et al.
conclude that these ﬁnding show that “perceived risk rating consisted with [the risk]
characteristics...“ (p. 189). Gardner et al. give two reasons to explain the comparatively
low coefﬁcients in the study: (a) the use of individual responses to risk questions as
opposed to respondent means used in Fischhoﬁ et al.; and (b) the lack of independence

27
between the Fischhoff et al. observations in that risk data for each of the 30 technologies

came from the same respondents (p. 191).

Kraus and Slavic

The research by Kraus and Slovic (1988) is comprised of two studies. The ﬁrst
study develops risk perception regression models for six technologies which are
compared to previous models using group means and diverse technologies. The second
study takes one of these technologies, railroads, and investigates the relationship between
perceived risk and risk descriptors using individual responses. Since the second study is
comparable to Fischhoﬁ et al., the focus will be on the results of the second study.

For the second study, respondents were asked to rate 49 railroad hazards on
seven 11-point scales. One of the scales was overall risk, with l :- not risky and 11 =-
extremely risky. Six risk scales similar to those used in F ischhoff et al. were included;
voluntariness, societies’ knowledge about the risk, control, dreadedness, catastrophic
potential, and newness. A principal components analysis of the risk characteristics
identiﬁed two descriptors representing basic risk dimensions. Factor 1 is identiﬁed
strongly with voluntariness (.93), control (.90) knowledge (.84) and dread (.76). The
second factor is identiﬁed strongly with catastrophic potential (.90) and newness (.85).

A correlation matrix showed that overall risk was signiﬁcantly correlated (r) with
the following risk descriptors: Catastrophic potential (.85), newness (.72), and dread
(67). The other risk descriptors, voluntariness, knowledge and control were not
Significantly correlated.

28
Implications of the Empirical Work

The studies by Fischhoﬁ et al., Gardner et al., and Kraus and Slovic show that
risk descriptors are systematically related to consumer risk perceptions. Although these
studies are not directly comparable, since they «amine diﬁerent technologies using
diﬂerent methodology, the fact that certain risk descriptors appear to be signiﬁcant in
each of the studies suggests that risk descriptors may have a signiﬁcant degree of
«planatory power concerning consumer risk perceptions both within and across
technologies.

Table 3.1 shows the correlation between risk perceptions and risk descriptors for
each of the three studies. The data shows that only one to three descriptors in any one
study have correlation coefficients above 0.6 with respect to consumer risk perceptions.
These descriptors include severity, dread, catastrophic potential and newness, which are
signiﬁcant components of the two risk dimensions identiﬁed by principal components
analysis in both Fischhoﬂ' et al. and Kraus and Slovic.

Table 3.2 shows the components of each of the two measured risk factors for
each study. Severity is a major component of the second factor in Fischhoﬁ et al., dread
is an important component of factor 1 for both studies, catastrophic potential is an
important component of factor 1 in Fischhoff et al. and factor 2 in Kraus and Slavic, and
newness is major component of factor 1 in Fischhoff et al. and in factor 2 in Kraus and
Slovic.

Other descriptors may be important components of underlying risk dimensions
yet not highly correlated with perceived risk individually. For «ample, Table 3.2 shows
that voluntariness and control are major components of factor 1 for both studies but in
neither study are they correlated with perceived risk. This may be «plained by the fact
that Fischhoﬂ' et al. found both voluntariness and control to be highly intercorrelated

29

Table 3.1 Comparison of Selected Studies Showing Correlation (r) between
Risk Perceptions and Risk Factors1

 

 

Risk Factor Fischhoff et al. Gardner et aL Kraus and Slavic
Seriousness .67 - -
Dread/Fear .64 .44 .67
Catastrophic Potential .30 .63 .85
Newness .05 - .72
Control -.04 . - -.07
Known to science -.17 .33 -.03
Known to «posed -.20 - -
Immediacy -.07 - -
Voluntariness .08 - .25

 

1. Fischoﬁ et al. use perceived risk, deﬁned as 'the risk of dying as a
consequence of this activity or technology." (p. 131, 1978). Gardner et al.
include the risk of injury and illness along with the risk of death (p. 184).
Kraus and Slovic use "overall level of risk' which refers to the overall
riskiness of a hazard' (p. 447).

30

Table 3.2 Comparison of Risk Factors Identiﬁed by Factor Analysis in Fischhoﬁ
et al. (F) and Kraus and Slovic (KS).

 

 

 

Factor 1 Factor 2
F KS F is
Voluntariness .89 Voluntariness .93 Seriousness .91 Catastrophic so
Knowledge of Dread 60 Newness .85
exposed .88 Control .90
Knowledg of Knowledge Catastrophic 55 Dread .50
scientists .88 (Society) .84
Newness -.87 Dread .76 Immediacy -.45 Knowledge -.34
Control -.83 Catastrophic -.22 Knowledge of .
exposed -.39 Control .20
Immediacy .70 Newness do Scientiﬁc
knowledge -.28 Voluntariness .07
Dread .67 Control -.24
Catastrophic .62 Newness .14
Seriousness .11 Voluntariness .(B

 

 

31
with other risk descriptors. Voluntariness was found to be signiﬁcantly correlated with

seven other risk descriptors while control was signiﬁcantly correlated with six.
Consequently, these risk descriptors, and perhaps others, may not be individually
correlated with risk perceptions but may be so as part of a group of risk descriptors.

The studies suggest that an explicit consideration of risk descriptors may be
beneﬁcial in the valuation of groundwater protection policies. As the above studies
suggest, risk descriptors, either individually or in a group, potentially aﬁect risk
perceptions. To the «tent that willingness to pay is aﬁected by risk perceptions, risk
descriptors may help «plain the variation in individual wtp for groundwater protection
policies.

The idea that risk descriptors are systematically representative of risk perceptions
and are economically meaningful will be developed by specifying that risk perceptions
can be represented by speciﬁc risk descriptors, which are conceptually similar to concepts
that are characteristic of an economic choice situation, speciﬁcally, the outcome and the
likelihood of the outcome occurring. The oﬁective is to show that the subjective
perception of these two elements within the context of speciﬁc risk ducriptors will be

aﬁect wtp in a predictable or testable manner.

Descriptors Associated with Perceptions of (holes
Consumer perceptions of the degree of choice available when confronted with
potential or actual exposure to groundwater pollution may aﬁect their overall risk
perceptions of groundwater pollution. Two characteristics of choice are: (a)
voluntariness, the «tent to which people voluntarily, «pose themselves to risk; and (b)
control, the ability to reduce or eliminate damages once pollution has occurred.

32
This section «amines choice descriptors identiﬁed by Fischhoﬁ et al. (1978).
Hypotheses will be derived concerning the relationship between these descriptors and

wtp.

Voluntariness

‘Fischhoﬂ' et al. (1978) asked respondents to assess an activity or technology in
terms of its voluntariness. Respondent’s were asked the question (p. 131), 'Do people
get into these situations voluntarily?‘ A dictionary deﬁnition of voluntary is free to
choose (Webster’s New World Dictionary, 3rd College Edition, 1988). For the present
discussion, this implies that an individual only submits to risk if, on the whole, the
beneﬁts outweigh the costs.

In one sense, voluntariness may be negatively correlated with wtp since there is
no need to «pose oneself to voluntary risk unless one is made better off by it. On the
other hand, if one is «posed to involuntary risk, by deﬁnition this implies that based on
their own preferences, individuals would have chosen not to be «posed to the risk.
However, whether a consumer is better or worse oﬁ' depends on how they perceive the
resultant level of risk. Even if exposure to the risk is involuntary, if the marginal
beneﬁts of the risk level are perceived to be greater than the marginal costs, consumers
may not be willing to pay for a risk reducing policy. In this case, increased perceptions
of involuntariness are negatively correlated with wtp for risk reducing policies.
Alternatively, if the marginal costs of the involuntary «posure to risk are perceived to be
greater than the beneﬁts, perceptions of involuntariness will be positively correlated with
wtp for risk reducing policies. In general, the relationship between perceptions of

involuntariness and wtp will be ambiguous.

33

Control

Control refers to the ability to affect the probability of a detrimental event
occurring given continued «posure. Fischhoff et al. (1978) ask respondents 'if you are
«posed to the risk of each activity or technology, to what «tent can you, by personal
skill or diligence, avoid death while engaging in the activity?” (p. 133). Alternatively,
control is the technical feasibility of personally controlling the level of health risk one is
«posed to. In one sense, controllability from the perspective of the household is likely
to be associated with cost. High likelihood of controllability can be associated with
perceptions that given reductions in risk can be achieved at relatively low cost.
Perceptions that groundwater pollution is relatively uncontrollable at the household level
are likely to be associated with perceptions of the high cost of household action to
reduce risk. Consequently, perceptions of high costs associated with individual
household action motivate perceptions of uncontrollability. As perceptions of
uncontrollability increase, wtp for groundwater pollution policies (collective action)
increase. Alternatively, as perceptions of controllability (low perceived cost of household

action) increase, policy wtp decreases.

Descriptors Representing Knowledge
Knowledge descriptors can be categorized as two types: (a) descriptors which are
representative of the degree to which scientiﬁc «perts and «posed individuals are
knowledgeable about the risks from a speciﬁc technolog, and (b) the degree to which

society is «perienwd with a risk.

34

Knowledge
Knowledge refers to both the knowledge exposed individuals have about the risks

they face and scientiﬁc knowledge about the risk. Fischhoﬂ‘ et al. (1978) ask
respondents, 'To what «tent are risks known precisely by the persons who are «posed
to the risks?“ and 'To what «tent are the risks known to science?“ (p. 133).

One deﬁnition of knowledge is the fact or condition of knowing something with
familiarity gained through experience or association (Webster’s New World Dictionary,
3rd College Edition, 1988). Knowledge in the psychological literature appears to mean
how well the individual knows the full dimensions 'of risky situation. If they know it well,
it is understood and the distribution of outcomes and likelihoods is known with some
certainty. A risk is not well known if the distribution of outcomes and likelihoods is very
uncertain.

An individual’s degree of knowledge about the risks from a speciﬁc activity will
inﬂuence optimal risk levels. The more knowledge possessed by «posed individuals the
greater the probability they will make choices consistent with their preferences. For
«ample, suppose a consumer can inﬂuence mortality probability by taking some level of
action a. To the «tent possible, individuals choose a level of at which maximizes their
welfare. The level of a chosen directly corresponds with a given level of optimal risk.
The greater the knowledge, the more likely the chosen level of at, say e°. will approach
the utility maximizing level of a with full information, a‘. With full information, a° - a‘.
With less or more uncertain knowledge, the probability that at is less than or greater than
a‘increases. Inthiscasethemaximized utilitylevelwhen a°doesnotequala‘willbe
less then when the equality holds.

Perceptions of scientiﬁc knowledge about health risks from groundwater pollution

may also affect risk perceptions. However, it is unclear whether the relationship

35
between scientiﬁc knowledge and wtp is positive or negative. The relationship will be
negative if people perceive that the scientiﬁc community has little knowledge concerning
the «tent, magnitude or cause of groundwater pollution, they may have a signiﬁcant wtp
for groundwater protection policies. Alternatively, the relationship will be positive if
people believe that little knowledge on the part of the scientiﬁc community will translate

into ineffective policies. If such is the case, then wtp would decrease.

Newness

Newness refers to how long society has known about a particular risk and how
much experience society has with it (Kraus and Slovic, 1988). Fischhoff et al. (1978) ask
respondent’s “Are these risks, new novel ones or old familiar ones?“ (p. 133). Newness is
conceptually similar to knowledge in that it represents perceptions of the degree or
depth of «perience with a given risk. The greater the perceived «perience, the greater
the likelihood that society has learned to “live with“ a particular risk in the sense that (a)
its causes are known, (b) people know when they are at risk, and (c) people can take
action to prevent or eliminate potential damage from «posure. Consequently, the
greater groundwater pollution risk is perceived to be a new risk, the higer the wtp for

groundwater protection policies.

Descriptors Associated with Severity of Outcome
If groundwater pollution occurs, there is some positive probability that exposure
will occur. If «posure occurs and damage results, peoples’ perception of the severity of
the damages will inﬂuence their overall level of perceived risk. Descriptors included in
this category include: (a) seriousness of outcome, (b) immediacy of outcome, (c)
catastrophic potential, and (d) dread/fear.

36
Seriousness ofOutconre

Seriousness of outcome refers to the individuals perception of the health outcome
of «posure to groundwater pollution. Fischhoff et al. ask respondents “When the risk .
from the activity is realized in the form of a mishap or illness, how likely is it that the
consequences will be fatal?“ (p. 133).

Conceptually, seriousness of outcome can be represented by a continuous scale
with “no health effects“ on one end and “death“ on the other. The utility associated with
the «posure state, v“, will be a function of the perceived seriousness of the «posure
outcome. The more serious the perceived outcome, the lower the utility associated with

the state of «posure.

Immediacy of Health Effects

Immediacy of effect refers to whether the effects of «posure are immediate or
delayed. F ischhoff et al. ask respondents 'To what «tent is the risk of death immediate,
or is death likely to occur at some later time?“ (p. 133).

Perceptions of immediacy may affect perceived marginal beneﬁt derived from a
groundwater policy. For «ample, the consumer maximizes «pected utility by choosing a
level of payment that equates the MB of a reduction in r with the marginal costs of
obtaining that reduction. In comparing two risk situations, suppose that two risks are
identical in terms of probabilities and outcomes «cept that, with «posure, death occurs
in time period t for one risk and time period t + 1 in the other. The present value of
MB for a given discount rate r for the two risks are MB/(1+r)‘ and MB/(l-t-r)‘“,
respectively. The present value of the MB of a reduction in risk is higher the sooner
death occurs. In policy terms, this implies that the beneﬁts of avoiding a death is greater

intimetthanint+ 1. Thepreferredlevelofriskassociatedwithtwillbelowerin

37
comparison with the level of preferred risk associated with death occurring further in the

future at time t + 1. The more immediate the health eﬁ'ects, the lower the utility

associated with the «posed state.

Catastrophic Potential

Catastrophic is deﬁned as having the character of a momentous tragic event
ranging from «treme misfortune to utter overthrow or ruin (Webster’s New World
Dictionary, 3rd College Edition, 1988). Fischhoﬂ' et al. ask respondent’s “Is this a risk
that kills people one at a time or a risk that kills large numbers of people at once
[catastrophic]?“ (p. 133).

Household perceptions of groundwater pollution «posure having catastrophic
consequences is similar to believing that the occurrence of a pollution episode results in
multiple «posures. Regardless of whether it is believed that the health effects of such
exposures are manifested as illness or death, perceptions of pollution «posure as
catastrophic entail perceptions of a number of households affected, as opposed to a
single affected household. In this case, as catastrophic perceptions increase, the utility

associated with the state of «posure decreases.

Fear/Dread

Dread is deﬁned as intense fear, especially of something that may happen
(Webster’s New World Dictionary, 3rd College edition, 1988). Fischhoff et al. ask
respondents, “Is it a risk that people can think about reasonably calmly, or is it one that
people have a great dread for?“(p. 133).

It has been previously noted that knowledge, newness, control and voluntariness

affect optimal risk. Fear/dread may in part be comprised of these descriptors in the

38
sense that an individual’s perceived degree of dread with respect to a given activity may
be related to their judgement of the risk as: (a) new, (b) a great deal of uncertainty
associated with the risk, (c) little or no control, and (d) being «posed involuntarily. If
the concept of dread is associated with these descriptors, then dread would aﬁect

acceptable risk and wtp.

SunnaryofRiskDescrlptorandWTPHypotheses ,

Table 3.3 summarizes the hypotheses developed in this section which address the
relationship between risk descriptors and consumer wtp for risk reducing policies. The
hypothesis regarding the choice descriptors voluntariness and control is ambiguous since
wtp will depend on the perceived beneﬁts and costs of risk exposure characterized by
degrees of voluntariness and controllability.

The direction of the relationship between knowledge risk descriptors and wtp
depends on what consumers associate with a perceived lack of knowledge. A perceived
lack of knowledge with respect to the health eﬁects of «posure to groundwater pollution
may be associated with the need for a “better safe than sorry“ approach. In essence,
groundwater pollution should be addressed, even though knowledge regarding health
eﬁ'ects is perceived to be low, just in case future information shows health eﬁects to be
signiﬁcant. Alternatively, if the perceived lack of knowledge is associated with the
inability to design and implement a successful policy, then perceptions of knowledge will
be positively correlated with wtp.

The hypothesis regarding severity descriptors and wtp is unambiguous; as
perceived severity increases, wtp for risk reducing policies decrease.

39

Table 3.3. Summary of Factor Hypotheses

 

 

Factor Description Hypothesis
1. Choice Includes: a) voluntariness: a. If perceived mc «ceeds mb
the «tent to which of involuntary «posure, the
individuals freely choose to greater the perceived
be exposed to risk; and b) involuntariness the higher
control: the ability to aﬁect the wtp.
the probability of a
detrimental event occurring b. If perceived mb «ceeds mc,
given continued exposure. ambiguous association.
2. Knowledge Includes: a) scientiﬁc a1. Perceptions of degree of
knowledge of the health risks knowledge are negatively
of «posure; b) knowledge of correlated with wtp: As
«posed individuals of health perceptions of lack of
risks of «posure; c) newness, knowledge increase, wtp
how long society has known increases.
about a risk and how much
«perience it dealing with it. a2. If perceived lack of
knowledge is associated with
the unlikelihood of a
successful policy, then
perceptions of the degree of
knowledge are positively
correlated with wtp: as
perceptions of the degree of
knowledge increase, wtp
increases.
3. Severity Includes: a) seriousness of a. Perceptions of severity are

outcome; b) immediacy of
health effects; c) catastrophic
potential; and d) fear/ dread.

positively correlated with
wtp: As perceived severity
increases, wtp increases.

 

CHAPTER4

CONTINGENT VALUATION DESIGN AND PROCEDURES

This chapter develops a contingent valuation (CV) survey to elicit necessary information
to test the relationship between household perceptions and wtp for groundwater
pollution policies. The chapter will outline and «plain the development of a

questionnaire using information obtained in a series of focus groups and pretests.

Development of a Contingent
Valuation Questionnaire

The goal of the empirical study is to obtain information on the tradeoﬁs
individuals are willing to make, in order to reduce the threat of groundwater pollution.
A contingent valuation mail survey for obtaining economic information on groundwater
pollution policies was used for the dissertation research. A major advantage of using
contingent valuation is the ability to present consumers with a choice situation that
directly focuses on a particular value (or set of values) of interest (Cumming et al.,
1986). This attribute is of particular importance in situations where markets for the
commodity of interest either do not «ist, or are poorly developed (Bishop and
Herberlein, 1979).

The usefulness of a contingent valuation survey in the policy evaluation process
depends on the «tent to which the survey is constructed, to have a reasonable degree of

success in obtaining accurate valuation information. The n«t section will discuss the

40

41

development of a CV questionnaire on groundwater pollution in terms of its component

parts.

TheUseofFocus Groups inCVQuestionnaireDevelopnlent
A series of focus group sessions were organized to assist in the development of a
questionnaire effective in representing the choice cont«t to consumers. An in-person
pretest and mail pretest evaluated the questionnaire for comprehension, clarity, and
relevance. The following sections will address the objectives, development, format and

general ﬁnding of these sessions.

Focus Group Objectives

The focus group sessions had two objectives. The primary objective was generally
to «plore the topic of groundwater pollution with consumers, focusing on (a) how
individuals perceived groundwater pollution and (b) the type of vocabulary consumers
used in thinking about and discussing the issue (Fischhoff and Furby, 1988).

The second objective was to present participants with alternative policy
descriptions and costs to elicit their reaction and criticism. Using the information gained
from the focus groups, a mail contingent valuation survey was developed and. reﬁned.
This approach was undertaken in order to construct a survey which was understandable
and convenient to answer, yet of sufficient detail to elicit required information to test the

hypotheses developed in Chapters 2 and 3.

Focus Group Development

Four sessions were held over a three week period in April, 1990. Two-hour

sessions were held on 11 April and 19 April, 1990. Two two-hour sessions were held

42
simultaneously on 24 April, 1990. Potential participants were contacted by dialing

random computer generated telephone numbers for the Williamston, Michigan area.
This area was chosen because of its dependence on groundwater, its proximity to the
campus of MSU, and its combination of suburban, rural and farming demographics.

The caller was identiﬁed as a researcher at Michigan State University who was
working on a project that dealt with social issues (environmental issues for the 19 April
session) currently faced by Michigan citizens. A prepared script was read to the
individual, who was then invited to attend a meeting at Michigan State University, where
each would have an opportunity to «press his or her opinion about these issues.
Individuals would be paid $50 to attend (See Appendix E for t«t of phone script.).

For the 11 April session, a total of 308 calls were made to obtain 10 participants.
Of the total number of calls, 137 (44.4 percent) were valid numbers, including both
residential and commercial numbers. Of the valid numbers, 61 (44.5 percent) answered.
Of people contacted, 10 (163 percent) agreed to take part in the focus group session.

For the second session on 19 April, a total of 127 calls were made. Of these, 61
(47.7 percent) were valid numbers. Of the valid numbers, 31 (50.8 percent) answered.
Of those contacted, 10 (32.3 percent) accepted.

For the third and fourth sessions on 24 April, a total of 228 calls were made. Of
these calls, 98 (42.9 percent) were valid. Of the valid numbers, 40 (40.8 percent)

answered. Of those that answered, 10 (25 percent) accepted.

Focus Group Session Format
The focus group sessions were held at the Kellogg Center, a conference center on

the Michigan State University campus. Participation of respondents who agreed to come

43
was 100 percent (one person brought his spouse). The sessions typically lasted two hours

with a 10- minute break at the end of the ﬁrst hour.

The sessions were structured so that participants’ perceptions and opinions
regarding groundwater and groundwater pollution were elicited in a systematic and
comprehensive manner. The topics consisted of: a) ranking of social issues (11 April
and 24 April sessions only), b) making of environmental issues, c) general perceptions of
the physical characteristics of groundwater, d) perceptions of groundwater pollution, e)
assessment of groundwater pollution effects,_ f) prevention of pollution effects, g)
perceptions of the mortality risk from groundwater pollution, and h) risk factor
assessment. This last component entailed the participant’s subjective assessment of
groundwater pollution risk in terms of the psychological risk characteristics «amined in
Chapter 3.

The above components were addressed by asking the group a set of speciﬁc
questions dealing with that particular component. Typically, the discussions were wide
ranging and enthusiastic. The purpose of the session was to obtain information on
people’s perceptions and ways of thinking about groundwater and groundwater pollution.
It is likely that for most people, the topic is one which they have not often thought about
in detail, and with which they have little direct «perience. Open and free discussion was

encouraged.

General Finding
A variety of opinions and perceptions were brought forth, which proved to be
very useful in designing an appropriate questionnaire. General ﬁnding from the
sessions can be summarized as follows: a) a wide variation exists in knowledge about

groundwater and groundwater pollution, b) groundwater perceptions are sensitive to the

44
wording used to describe the problem, c) there is a large variation in underlying values

regarding gomWater, and d) the strength of people’s prior beliefs concerning
groundwater pollution and policies has impOrtant implications for policy development.

Groundwater Knowledge

The focus groups show a wide variation in people’s knowledge about groundw&er
and groundwater pollution. The initial portion of each session was devoted to eliciting
their thoughts about groundwater, groundwater pollution, effects of groundwater
pollution and prevention of effects.

Wm. Most people were at least superﬁcially familiar with
the concept of groundwater. The descriptive nature of the term was helpful, for as one
participant put it, “It means the water in the ground. It’s sort of obvious.“ People
generally agreed that groundwater is water in the ground that can usually be pumped to
the surface and used as a source of fresh water. However, when more speciﬁc questions
were asked, the responses varied.

For «ample, some individuals thought that groundwater «isted within 18-20 feet
of the surface. Others said that depth does not really matter; whether it is just a few
feet or hundreds of feet it is still groundwater.

Opinions also varied concerning the rate of ﬂow of groundwater. Some
individuals said that groundwater is essentially like a pond; that it accumulates in a
particular area and moves very slowly or not at all. Most conairred that groundwater
does move, but slowly, such as “I think it’s just the underground water in pools that
moves slowly through the earth.“

Consistently throughout the sessions, groundwater was thought of as water in the

ground available for human use. The following comments were typical: a) 'To me it

45

means any of the water that’s available that you can pull up to use either for irrigation or
for drinking“; b) “Water that you pump; either the city does or you pump personally out
of a well to drink and for the basics of living“; and c) “I think it’s just the underground
water in pools that moves slowly through the earth. That’s my vision of it and with
wells tapping into the undergound reservoirs.“

It is clear that most people have not given much thought to groundwater, nor do
they have precise knowledge about its physical attributes, even those who rely on their
own wells for water. One person «plained, 'Water must run under the ground but I
never thought about it before. I always thought it was a lake down there that my well
went into. I just keep sucking it up and don’t think about it.“

W People were asked about sources of pollution and how
common groundwater pollution is in Michigan. The most frequently mentioned sources
include industrial and manufacturing waste products, household lawn and garden
chemicals, agricultural chemicals, underground water storage tanks for gasoline and oil,
salt storage areas for road de-icing, dumping of used motor oil, and septic tanks. There
was great variation in perceptions of how common groundwater pollution is. Some
individuals thought mmﬂater pollution is common in urban areas due to the degree
of development and manufacturing, but relatively uncommon in rural areas. Others
thought that rural areas are worse because of the «tent of lawn and garden and
agricultural chemical use.

No general consensus was reached on how «tensive the pollution problem is, but
most people feel that the problem is getting worse. A representative sentiment is, “...I’ll
tell you one thing, I don’t know how rare or common it is but I’ll tell you one thing I

know, it’s getting commoner and commoner.“

46

W The most frequently cited adverse health effect
of exposure to chemical groundwater pollution is cancer. Others frequently mentioned

are birth defects, miscarriages, and liver and kidney disease. Several participants stated .
they did not know what the health effects might be.

When asked how common death from groundwater pollution is, the majority of
people said death is an uncommon occurrence. Non-fatal health eﬁects are judged to be
relatively more common. However, several individuals believe that health effects from
groundwater pollution were rare. For «ample, one person stated, “I think it would be
rare [death] because I think the system basically is monitored" enough where if something
is that serious to cause death that it is brought to the public’s attention and something
is done about it.“ A different person had a similar viewpoint, “I don’t know that I really
believe that there is a lot of adverse health effects from groundwater contamination.
Further, I think if there is groundwater contamination it tends to be pretty local and it is
not a widespread thing.“

W The {06118 in “138 portion 0f the
session concerned the type of actions that could be taken to prevent negative effects of
groundwater pollution and what the appropriate role and level of government is. A
majority of the participants mentioned increased monitoring and information programs
about groundwater pollution. A number of people want increased enforcement of
«isting laws, especially with regard to large industrial sources of pollution. However,
quite a few are skeptical of whether the necessary funds would be available.

Most people think the emphasis should be on the prevention of groundwater
pollution. To understand how people think about remedial policies, the topic of point-
of-use ﬁltration systems was brought up by the focus group leader. This idea received

47

little support because few people think ﬁlters would be effective in removing
contaminants from the water.

Most people seem to favor state or county governments for the implementation
and administration of groundwater pollution policies. As one person who favors state
government put it, “Local government doesn’t care and the feds are too far away.“ Most
of the support for state or county involvement is due to the perception that state
agencies: (a) are familiar enough with the speciﬁcs of a particular problem to be
effective; (b) are capable of having a suﬁicient budget to address the problem; and (e)
have the required technical expertise. Local government is perceived to be lacking (b)
and (c) and the federal government is lacking (a).

Sensitivity to Wording

During the course of the focus group sessions people were asked if the terms
“contamination“ and “pollution“ have similar or different meaning. A slight majority felt
that they have different meaning. The rest indicated that the terms meant more or less
the same thing. Overall, the term pollution has a more general, less severe connotation.
Generally, polluted groundwater is thought of as still being usable; that chemicals might
be present in the groundwater but the water is still usable, if not for drinking at least for
other non-consumptive uses. Contamination is perceived to mean that the water is unﬁt
for any use. The following comments are typical: a) “When you think of something
contaminated, you can’t use it“; b) “When it gets to the degree that you can’t use it
anymore, then it’s contaminated“; c) “If something is just polluted I would think how
polluted is it? Is it still safe to drink? When I hear the word contaminated I would say
no, you can’t use it without some way of purifying it or doing something to it“; and d) “I

guess pollution to me is something that’s polluted. It’s being overwhelmed by something

48

that’s not natural to it and it may or may not necessarily result in contamination per se.
Contamination to me would be something that I don’t want to have any contact with

because it will be harmful to me. Pollution doesn’t necessarily do that for me...“

Variation in Underlying Groundwater Values

Chapter 2 describes various uses of groundwater and how people might
conceivably be affected by its pollution. The focus groups showed that everyone is
concerned about the health effects of groundwater pollution on their own household.
Additionally, most seem to be concerned about their community at large and future
generations. The importance of other groundwater services varies greatly across
Participanm

A number of people were concerned about the impact of pollution on animals
and plants. It is apparent that most were not aware of the potential for groundwater
pollution to affect plants and wildlife. However, during ensuing diswssions, participants
talked of their concern over the potential impact of groundwater pollution on plants and
wildlife.

Less prevalent, but still signiﬁcant were people’s concern over what might be
termed “system integrity.“ System integrity has two components. One is the ecosystem
where people value the knowledge that the components of an ecosystem are unpolluted,
completely aside from any current or future use. The other component is that the act of
polluting violates a person’s moral system of values, completely aside from any damage
that actually occurs. For «ample, one person stated, “Making any part [of the
ecosystem] worse, there is really no reason for us to do that. We should try to at least, if
we can’t make it better, to keep it as good as it is now.“ Another person stated, “It really
doesn’t matter if anything is harmed or not, people shouldn’t be allowed to pollute.“ In

49

response to this statement, another person added, “Well, I’m not ready to give away an
entire years’ paycheck just so that some pollution which isn’t harming anything can be
cleaned up. But I think that within some reasonable limit pollution shouldn’t be allowed
even if there is no apparent effect on anything.“ In other words, it is simply wrong to
pollute.

The discussions in the focus groups did not directly address the question of
protecting groundwater for its own sake, to the «tent that this means something
different from pollution’s effects on animals and wildlife and system integrity. This is not
to say that intrinsic groundwater services are not important, rather that it appears
diﬁicult to separate the intrinsic groundwater services from ecological system services
provided from groundwater.

To summarize, the level of concern over different groundwater values, from most
common to least common, is: (a) health of own household, (b) their own community at
large, (c) future generations, (d) animals and wildlife, (e) system integrity, and (f)
groundwater itself.

Strength of Prior Beliefs

People’s prior beliefs may affect how policy services and eﬂ'ectiveness are
perceived, and in turn how the policy is valued. A major hypothesis of the research is
that non-health groundwater values are a signiﬁcant component of the total value people
place on groundwater services.

Three general policies were initially hypothesized to be of relevance to Michigan;
information, monitoring and enforcement. Information refers to the generation of
scientiﬁc information on the «tent, magnitude and characteristics of groundwater

pollution in Michigan. Monitoring refers to the sampling and chemical analysis of

50

groundwater used for consumption (private and public wells) and aquifers adjacent or
close to known or suspected sources of pollution. Enforcement refers, to regulatory
programs to prevent «posure to pollution and to prevent the commence of groundwater
pollution.

The initial speciﬁcation of these policies reﬂected: a) policies which were being
considered both by Michigan and other states (State of Michigan 1987, 1988 and
Henderson et al., 1987) and b) previous research studies on alternative policies for
addressing groundwater pollution (Pye et al., 1987 and Libby and Kovan, 1986).

The initial hypotheses about relevant groundwater policies for Michigan were
strengthened in the focus groups. A signiﬁcant number of people are supportive of
monitoring programs and obtaining more precise information on groundwater problems
in Michigan. Support is less widespread for regulatory programs. A number of people
think regulation should be a major part of any type of groundwmer policy while others
are not convinced that it is necessary until more scientiﬁc information is obtained. Still
others think there is no need for any type of program at the present time. Reasons
given are: (a) not wanting their tax bills to increase for any reason and (b) the existence
of more. urgent problems than groundwater pollution on which to spend tax dollars.

A previous section discussed perceptions of the appropriate role and level of
government to address groundwater pollution problems. It was found in the focus
groups that while there might be a high degree of support for a particular policy,
individual perceptions vary as to how effective federal, state, and local governments
would be in implementing the policy. For «ample, some people felt that laws might be
enacted but not implemented because of budgetary constraints. Others questioned the
expertise or motivation of the level of government responsible for the policy. For some

individuals these perceptions were true for all levels of government, while for others

51

certain levels of government (e.g. local) evoked different perceptions of eﬁectiveness.
Consequently, the questionnaire was «plicitly designed to account for individual’s prior
beliefs about program effectiveness.

Contingent Valuation Survey Format
Hoehn~ (1987) identiﬁes the objective of a contingent valuation format as “... to
set up an «change situation in which an individual may price policy impacts“ (p. 412).
The design of such a format involves consideration of ﬁve components: (1) presentation
medium; (2) description of policy impacts; (3) method of provision; (4) method of

payment; and (5) value elicitation (Hoehn 1987, p. 413).

Presentation Medium

Presentation medium refers to the manner in which responses are obtained from
survey participants. Three general methods are used: personal interviews, telephone
interviews and mail surveys. Mitchell and Carson (1989) identify three characteristics
which should inﬂuence the choice of survey method (p. 109). First, the more oompl«
and unfamiliar the valuation scenario, the more important it is to provide photographs,
charts and other types of visual references. Second, asking respondents to place values
on policies places a high demand on their attentiveness to the survey. It is important to
choose a method which can motivate people to make the necessary eﬁ'ort. Finally, '

“...the need to «trapolate from the sample to make beneﬁt estimates for

populations necessitates the use of survey methods which support

techniques to compensate for missing data.“ (p. 109).

Based on these criteria, the inoperson interview would be the method of choice,

other thing equal. The face-to-face interview allows for the answering of questions and

the use of visual aids to help respondents understand compl« valuation scenarios. It

52

also allows for respondent-interviewer interaction which may help motivate respondents.
Finally, in-person interviews support missing data techniques (p. 110).

A signiﬁcant drawback to in-person interviews is the «pense. For a given cost,
much larger sample sizes can be drawn with mail and telephone surveys. Consequently,
a trade-off «ists between choosing a survey method which has the desirable
characteristics and cost. However, Mitchell and Carson (1989) and Dillman ( 1978) note
that well designed mail surveys perform adequately compared with in-person interviews.
In addition, mail surveys have the added beneﬁt of reduced cost for a given sample size
or an increased sample size for a given cost (and increased degrees of freedom for
statistical analysis). An additional reason to consider mail surveys is that due to the
comparatively low cost, sample sizes can be chosen on the basis of research needs
instead of strictly in terms of cost considerations (Dillman, 1978).

The focus groups were beneﬁcial in identifying ways for the potential weaknesses
of a mail survey to be overcome. First, mail surveys limit the amount of information
that can be conveyed to respondents. However, one of the objectives of the survey was
to evoke responses based on an interrogative examination of respondents’ prior beliefs.
Consequently, a minimum amount of information was needed regarding the method of
provision and method of payment. Policy objectives were described but impact
measurements were based on respondents’ perceptions rather than on a technical
projection or «pert guess.

Second, mail surveys limit the ability to control the sequencing of information
and responses. To address this, the questionnaire was designed to be relatively free of
the need for sequencing. There was no need to reveal certain types of information in a

sequenced manner. The fact that respondents ﬁll in the risk perception questions would

53

be adequate to ensure that this portion of the questionnaire was considered as part of
the overall response.

Third, unfamiliar technical information concerning the valuation scenario may
have to be conveyed to respondents. A lack of understanding on their part may result in
the need for enumerator assistance. One difﬁculty in developing the questionnaire is the
lack of available information on the extent of groundwater pollution in Michigan. This
was dealt with by basing the valuation on people’s prior beliefs rather than including a
lengthy explanation of available information, its strengths and weaknesses. This
approach ﬁts: a) the lack of consensus among scientiﬁc «perts regarding «tent and
effects of groundwater pollution; b) constraints that are typical of mail surveys that limit
the amount of information given to respondents; and c) the likelihood that choice of
policy (e.g., information, monitoring, and enforcement) will, in fact, be based on widely
varying perceptions rather than a consensus.

Finally, since contingent values are to be conditioned primarily on prior
knowledge, a major portion of the questionnaire is designed to conduct people through
their beliefs in a Socratic manner. This interrogative method encourages respondents to
think through various aspects of the groundwater pollution problem based on
perceptions, and also elicits responses to questions that serve as independent variables in
the analysis of CV responses. To address this, the questionnaire was designed so that
the conveyance of compl« information was not required. Questions and wording were
carefully checked in the focus group sessions and in pro-tests. Focus group and pre-test
participants were debriefed for any doubts, questions, and negative responses.

Fourth, in mail surveys there is a lack of control over non-responses. However,
there is a trade-off in forcing a response versus eliciting a response. An in-pcrson survey

may allow the enumerator to obtain a response after several iterations but this may

54
result in an answer being given just for the sake of giving an answer, as opposed to being

truly indicative of a respondent’s preferences. A mail survey may result in a greater
number of non-responses, but those that are answered likely reﬂect the respondents’ true
preferences. The focus groups were used to evaluate survey questions for
comprehension and to identify questions that were perceived to be trivial or diﬂicult.

Fifth, mail surveys suffer from biased sampling and low response rates. In
response, the mail survey sample was obtained from an «tensive statewide database on
Michigan households developed by a marketing research ﬁrm. The ﬁrm did a systematic
sample of the population of Michigan heads of household who were listed in telephone
directories and auto registration records. Since a signiﬁcant number of households are
not listed in telephone directories, the use of auto registration records results in a
signiﬁcant number of households being contacted that would not have been otherwise.

To address low response rates, the total design method of mail surveys was used
(Dillman, 1978). This method essentially entails the use of a variety of techniques in the
development, printing and mailing out of questionnaires to minimize item non-response

and maximize the likelihood that surveys will returned.

Description of Policy Impacts

This component provides the respondent with a description of the potential
impacts of a given policy. It is the respondents’ only direct source of information about
the valuation situation they face. As such, if the description is incomplete, misleading, or
inconsistent with actual alternatives, inaccurate valuations may result (Hoehn, 1987).
Hoehn (1987) states that “The challenge is to develop a policy description that isboth

technically accurate and intelligible in terms of routine experience“ (p.413).

55

A major reason for concern over groundwater pollution is the potential effect on
human health (State of Michigan 1987, 1988, Pye et al., 1987, and Henderson et al.,
1987). Chapter 2 identiﬁed two general types of policies which address groundwater
pollution; preventive and remedial policies. Preventive policies aﬁect both health and
non-health values from groundwater pollution where remedial policies affect only health
values. Initially, both types of policies were included in the valuation scenario presented
to respondents in anticipation that the difference in value (if any) between a preventive
and remedial policy would be attributable to the value which individuals placed on the
non-health aspects of groundwater use. More speciﬁcally, deﬁne wtp’ as willingness to
pay for a prevention policy and wtp“ as willingness to pay for a remedial policy. If wtpP
- f(health services, non-health services) and wtpll - f(health services), then (wtpP-wtp“)
is the value of the non-health services of groundwater.

The accuracy of the term (wtp’ owtp") in representing the value of non-health
groundwater services depends on the ability of the questionnaire to convey suﬁ'icient
information on the alternative impacts of the two policies. However, in the focus group
sessions, it was discovered that individuals have difﬁculty attributing non-health impacts
to prevention policies. People have difficulty making a link between non-health values
and the policy impacts of prevention policies. However, in a different cont«t, non-health
impacts were found to be of signiﬁcant importance to a number of individuals.
Consequently, non-health values could not be reliably estimated by inferring such values
only from the wtp estimates for preventive and remedial policies.

One alternative method for evoking getting at non-health values is to simply ask
individuals about their wtp for a policy which only affects non-health values. When
presented with such a question (which dealt with q-altruism ) in the focus groups, most

people had difficulty in understanding the purpose or rationale of the policy.

56

Consequently, variables which measure held values are used as independent variables to
infer the impact of these held values on wtp in a CV analysis.

A major objective of the survey was to elicit respondents’ perceptions rather than '
convey a detailed (and perhaps controversial) policy impact description. Therefore, the
valuation questions speciﬁed what the policy was supposed to achieve and how it was
going to achieve it, but did so in a very general manner. This forced individuals to rely
on their prior beliefs. In light of this, a relevant set of policies and objectives needed to
be developed. The focus groups were an «cellent source of information and feedback in
developing the set of policies eventually used in the ﬁnal questionnaire.

Additionally, the questionnaire needed to be structured to obtain relevant
perceptions. The set of questions prmding the valuation questions were in part
developed with this in mind. The focus groups and in-person pre-test participants

provided very useful information in developing these questions.

The Method of Provision

The method of provision refers to the manner in which the policy is implemented
(Hoehn, 1987). The particular method of providing the policy may affect the
respondent’s valuation of the policy. This poses no problem if the purpose of the study
is to «amine the effect on wtp of alternative methods of provision. However, if this is
not the purpose of the study, then measurement bias may be introduced (Mitchell and
Carson, 1989). The source of bias is derived from consumer attitudes towards the
agency, organization, or level of government responsible for implementing and
administering the policy. Other thing equal, some organizations may evoke a higher
wtp than others. This might be the case if an agency is perceived to be more competent,

57
less wasteful of tax dollars, and more knowledgeable about the issue at hand (Cumming

et al., 1986).

Participants in the focus groups indicated that their main concern is the
appropriate level of government for policy administration. Concern was expressed over
the ﬁnancial capability and technical «pertise of local government to effectively deal
with groundwater pollution. Federal government is perceived to have the necessary
expertise and ﬁnances to effectively deal with the problem, but concerns were voiced
over its familiarity with local conditions. In further discussions, people’s attitudes about
state and county government in general are either neutral or fairly positive in regard to
their ability to effectively administer groundwater pollution policies. A speciﬁc agency or
organization was not speciﬁed in developing the policy questions. This was done because
the focus groups indicated that the speciﬁc agency within a particular level of
government is peripheral information from a respondent’s point of view, whereas the

level of government itself is always a concern.

The Method of Payment

The method of payment refers to the manner in which policy costs are passed on
to individuals aﬁected by the policy (Hoehn, 1987). A number of studies have found
that the type of payment mechanism employed in a questionnaire can inﬂuence wtp
(summarized in Cumming et al., 1986, and Mitchell and Carson, 1989). This does not
represent a problem if the objective of the study is to «amine the effects of alternative
payment vehicles on wtp, or if it is known in advance that, if implemented, policy costs
will be collected via a speciﬁc mechanism. A problem occurs in generalizing the results
obtained in a given study to other situations where a diﬂ'erent payment vehicle is more

appropriate or is less controversial (Mitchell and Carson, 1989). As a result, the recent

58

trend has been to frame the payment mechanism in general terms, such as a general rise
in taxes or prices, whichever is more appropriate (Hoehn, 1987, Mitchell and Carson,
1989).

For the groundwater questionnaire, the option of higher taxes was chosen as the
method of payment, primarily because it is a way of paying for public policies which is
familiar to people. Discussion in the focus groups indicated that there was not an overly
negative reaction to higher taxes as a means of paying for a policy. However, this is
dependent on perceptions of the need for the policy, whether the policy would actually
achieve its objectives, and whether the tax revenue would be applied directly for its
intended purposes. A typical comment along these lines is, “If I thought there was an
actual need for groundwater protection, I wouldn’t mind paying higher taxes [within
reason] as long the money isn’t wasted on just fattening up the bureaucrats. I want some
guarantee that the money will actually be used for groundwater protection.“

The questionnaire was designed to address this concern by including questions
on: a) the perceived effectiveness of the policies in accomplishing their stated objectives,
and b) the importance to the respondent of taxes in answering the policy valuation

questions.

Value Elicitation

The value elicitation section is the questionnaire component that obtains value
information from respondents (Hoehn, 1987). The appropriate format for the value
elicitation section depends primarily on, two items (Hoehn, 1987). First, the property
rights inherent in the issue under study will aﬁect the value elicitation format. Second,

there are a variety of methods for eliciting individual estimates of policy values.

59

mm The property rights situation will dictate the appropriate
Hicksian measure of value to use in the questionnaire. If consumers have entitlement to
the initial policy situation, then a Hicksian value measure is required. For the
groundw«er survey, it is assumed that the initial situation is without policy action-
whether this policy action be sampling and monitoring, remedial action, or prevention. If
the policy results in improvements from this initial situation, wtp is the proper format.
For decrements from this initial situation, wta is the valid format. If consumers do not
have property rights to the initial situation, then Hicksian equivalent value measures are
appropriate. For an improvement from the initial situation, wta payment to forego the
improvement is appropriate. For decrements to the initial situation, wtp to avoid the
decrement is the appropriate valuation format (Brookshire et al., 1980).

Hoehn and Randall (1987) deﬁne a satisfactory beneﬁt cost indicator as one that
“identiﬁes at least a portion of the true PPI [potential pareto improvement] proposals as
having positive net value and all non-PPI proposals as having negative net value“ (p.
240). The Hicksian compensating value measure is identiﬁed as a satisfactory beneﬁt
cost indicator for a particular set of CV formats. These formats are diﬁerentiated by the
costs conveyed to respondents including per capita costs, payments proportional to an
individual’s own bid and individually parametric costs (pp. 236-237). The Hicksian
equivalent measure is shown to be an unreliable indicator (Hoehn, 1987).

The approach taken in the groundwater pollution survey is that individuals have
an entitlement to the initial policy situation. Essentially, consumers have a property
right to an initial probability of having access to groundwater of adequate quality. The
policy valuation estimate in this case is deﬁned to be the wtp to obtain increments in the
probability of having future access to unpolluted groundwater. This is a Hicksian

60

compensating value measure and thus is a satisfactory beneﬁt cost indicator as well,
given an appropriate elicitation format as discussed above.

W There are two types of elicitation formats: open ended and
binary. With open-ended questions, respondents are asked to state their actual
maximum wtp or minimum wta for a policy. A binary format asks respondents to accept
or reject a given policy cost. For «ample, policy impacts are described and respondents
are asked to accept or reject the policy based on a speciﬁc cost. This is typical of the
binary choice approach (also referred to as the referendum model [Hoehn and Randall,

1987] and the “take it or leave it“ approach [Bishop and Heberlein, l979,l980]).

Independent Variables
In order to adequately test the hypotheses developed in the previous chapter, a

number of independent variables were measured. These variables address a) risk

perceptions, b) held values, and c) demographic and water-use characteristics.

Risk Perceptions

Thirteen qualitative risk questions were included in the survey. Respondents
were asked ten questions about their subjective assessment of groundwater pollution in
the cont«t of risk factors speciﬁed in Chapter 3. Testing of these questions in focus
group sessions and pretests indicated that the risk factor questions were, in general, both
understandable and interpreted correctly by respondents. Each risk factor question has
four or ﬁve possible responses. Individuals chose the response which most closely
represented their own opinion. Discrete answers were provided in order to make the
choice process easier for the respondent. In the Fischhoﬁ' et aL study, respondents

marked their answers on a scale of 1-7, as in the following «ample:

61

If you are «posed to the risk of each activity or technology, to what
«tent can you, by personal skill or diligence, avoid death while engaging

 

in the activity?
1 2 3 4 5 6 7
uncontrollable controllable

This question was modiﬁed for the groundwater pollution survey as follows:

If groundwater pollution occurs, do you believe the damage can be
controlled through technolog'? (Circle one.)

EASY TO CONTROL

POSSIBLE TO CONTROL
SOMEWHAT' POSSIBLE TO CONTROL
UNLIKELY TO BE CONTROLLED
NOT POSSIBLE TO BE CONTROLLED

99999:"

Discussion in the focus groups showed that marking one of several potential
responses led to less confusion and potentially higher response rates than asking
respondents to mark a point on an interval scale. Since the survey was administered by
mail, every effort was made to make the questionnaire straightforward and easy to
respond to.

Three of the questions deal with perceptions of water quality in the home and
the community. These questions address: (a) consumer perceptions of the safety of
tap/faucet water in the home, (b) consumer conﬁdence that tap/faucet water is safe
from groundwater pollution, and (c) the seriousness of the health threat from current
levels of groundwater pollution in their own county.

The ﬁnal risk perception question addresses the likely adverse health effects from

«posure to groundwater pollution.

62

Q-altrulsm and Bequest Perceptions

From the theoretical framework in Chapter 2, it is hypothesized that
environmental concerns as embodied in q-altruism and bequest risk have have a
differential impact on wtp, depending on whether preventive or remedial policies are
being valued. Five questions were used to measure the intensity of concern with respect
to these perceptions: (a) the threat of youndwater pollution to future generations, (b)
the threat of groundwater pollution to wildlife and plants, (c) whether grouther
should be protected for its own sake or because it is part of a natural system, (d)
whether pollution is bad because it is the wrong thing to do or because it has bad effects,
and (e) whether other reasons exist to protect groundwater from pollution besides the

potential effect on human health.

Demographic and Water-use Characteristics

A number of demographic and water-use characteristics are included as possible
explanation for variations in wtp. Demographic variables include: (a) location of home,
(b) farming as a source of income, (c) rent or own home, (d) number of people in
household, (e) children 16 years or younger in household; (1) age: (g) male or female;
(h) last year of school completed, and (i) household income. Water-use characteristics
include: (a) household water source, (b) type of water used for drinking, and (c) testing

of water.

CHAPTERS

QUESTIONNAIRE RESPONSES

This chapter discusses the responses to the mail survey. The focus will be on a
broad overview of the important questionnaire responses. A detailed summary will be found

in various appendices as noted below.

SurveyResponseRates

Sample households were mailed up to three questionnaires in order to increase the
likelihood that questionnaires would be returned (Dillman 1979). A total of 2020 Michigan
households were sent questionnaires. The statewide sample consisted of two subsamples;
the urban subsample consisted of 673 households, the rural subsample 1347 households.
The rural-urban classiﬁcation is based on whether a county is in a metropolitan statistical
area (MSA) or not. MSA’s "are made up of one or more counties around a large
population center together with adjacent communities which are socially and economically
integrated with the central city” (U.S. Department of Commerce 1988). For the purposes
of this study, counties within an MSA are considered urban and counties outside an MSA
are considered rural. A total of 173 surveys were returned as undeliverable (9 percent of
the total). Consequently, a total of 1847. households received questionnaires.

Returned questionnaires numbered 1229, for a gross response rate of 67 percent.

This ﬁgure includes all returned questionnaires. Of the total returned questionnaires, 16

63

64

were not usable. Nine questionnaires had missing identiﬁcation labels. Without the ID
labels it was not possible to classify the questionnaires as coming from the urban or rural
sample. Four questionnaires were returned unanswered. Three questionnaires had
responses which indicated the questionnaire was not taken seriously by the respondent. This
left a total of 1213 usable questionnaires for a net usable return rate of 66 percent. ”Usable"
means most of the questionnaire was answered. Appendix C has a more detailed discussion

of return rates.

Survey Responses

The questionnaire elicited from respondents four types of information: (a) socio-
economic characteristics, (b) water use information, (c) risk perceptions, (c) environmental

perceptions, and (e) policy bids.

Soda-economic Characteristics

The survey elicited information on expected annual income, highest level of
education completed, age, home location, home ownership, household size, and number of
children under 16 years of age living at home (See Appendix D for tables which display the
full range of responses for these questions). The Mann-Whitney test for testing the null
hypothesis that the distributions of two variables are equal will be used to compare
responses for the two samples. A signiﬁcance level of 95 percent will be used to reject the
null hypothesis that the variable distributions are equal.

W Respondents from both the rural and urban samples have
relatively high incomes. Mean expected annual income for the rural sample is $39,000 and
$47,000 for the urban sample. Approximately 38 percent of the rural sample expect an

annual income of $40,000 or more. For the urban sample, over 54 percent expect an annual

65
income of $40,000 or more. Almost half the rural sample, 45 percent, expect an annual

income of less than $30,000. Less than 30 percent of the urban sample expected similar
incomes. The difference in sample distributions is statistically signiﬁcant.

mum Educational attainment differed signiﬁcantly between the urban
and rural samples. Almost one-third of the rural sample completed high school but did not
continue their formal education. Of the urban sample, only 20 percent of respondents listed
high school as the highest level of formal education completed. About a third of each
sample attended college but did not graduate. Almost one-third (32 percent) of the urban
sample had completed college compared with 22 percent of the rural sample. The
difference in sample distributions is statistically signiﬁcant.

Ag. The two samples differed in the age proﬁle response. The most pronounced
differences were in the 26-35 and 66 and over categories. The rural sample had 16 percent
of respondents in the 26-35 age category. The urban sample had 28 percent in the same
category. In the 66 and over category, 26 percent of the rural sample and 16 percent of the
urban sample fell in this category. Both the rural and urban samples had relatively few
respondents in the 18-25 age category, 4 and 3 percent respectively. The difference in
sample distributions is statistically signiﬁcant.

W The two samples differed signiﬁcantly in selected housing
characteristics. In the urban sample, over 75 percent listed their home location as being in
a city or suburb compared with slightly over 30 percent for the rural sample. Similarly, over
52 percent of the rural sample listed their home location as either a rural area (but not a
farm) or a farm. This compares with slightly over 16 percent for the urban sample.
Additionally, the urban sample had a higher percentage of respondents who rent their
homes, 17 to 11 percent for the rural sample. The differences in the sample distributions

for both home location and ownership are statistically signiﬁcant.

66
W Both samples are similar in household size proﬁles. About half of

the households in each sample (53 and 51 percent for the rural and urban samples,
respectively) consisted of one or two persons. Slightly less than a third of each sample
consisted of households with four or more people. There is no statistically signiﬁcant
difference in the sample distributions.

W A majority of respondents in both samples did not have children
16 or under living at home. Of the rural households, 64 percent did not have children living
at home. The ﬁgure for the urban households is 63 percent. One or two children living at
home characterized 27 and 28 percent of the rural and urban households respectively.
Three or more children characterized 8 and 9 percent of the rural and urban samples,

respectively. There is no statistically signiﬁcant diﬁ‘erence in sample distributions.

Water Use

Three types of water information were collected: (a) the water source, (b) type of
drinking water preferred, and (c) whether the tap or faucet water has been tested. In the
rural sample, 58 percent used a household well and 40 percent relied on publicly supplied
water. In the urban sample, 78 percent relied on publicly supplied water while 22 percent
relied on domestic wells.

In both samples, most respondents rely on water from the faucet. In the rural
sample 93 percent rely on tap or faucet water as do 92 percent of the urban sample. About
6 percent in both samples use bottled water.

The overall percentage of rural respondents who tested their water was about twice
the rate for the urban sample. Of the households in the rural sample, 17 percent had their
tap water tested in the past year. Of the urban sample, 9 percent of the urban households

had their tap water tested. Table 5.1 below shows testing broken down by water source.

67

About 20 percent of the households in each sample which rely on well water had their water
tested. Of households who rely on publicly supplied water, 15 percent of the rural sample

tested their wells compared with 5 percent for the urban sample.

Table 5.1 Cross Tabulation of Water Source and Water Testing

 

 

Rural Urban
All Public Well All Public Well
Tested 17% 15% 20% 9% 5% 20%
Not 83% 85% 80% 91% 95% 80%
tested
N 81 1 326 473 363 279 79

 

The differences in sample distributions for water source and testing were statistically

significant. The difference in sample distributions for source of home drinking water is not

signiﬁcant.

Water Quality Perception

A majority (56 percent) of rural households perceive their tap water to be very safe.
Of the urban households, 42 percent believe that their tap water is very safe. Relatively few
households in either sample believe their tap water relatively unsafe to drink. For both the
rural and urban sample, 3 percent believed their tap water is somewhat unsafe or very
unsafe to drink. Urban households displayed a greater degree of uncertainty than the rural
households over the safety of their water. In the urban sample, 31 percent were not sure
of the safety of their water while 21 percent of the rural sample were not sure.

Rural households are generally more conﬁdent than the urban sample that their tap

water is safe from the effects of groundvmter pollution. Of the rural households, 23 percent

68
were very conﬁdent while only 13 percent of the urban households were similarly confident.

Of the urban households, 40 percent were not conﬁdent or not conﬁdent at all while 32
percent of the rural sample had similar perceptions.

Diﬂerences in the sample distributions for both questions were statistically
signiﬁcant. This implies that urban households in general are less certain than rural
households about the safety of their tap water. A majority of both samples do not believe
that tap water is unsafe to drink. However, a signiﬁcant number of households in each
group (with urban households less conﬁdent then rural households) believe their tap water
is not safe from groundwater pollution. ‘

Table 5.2 shows a cross-tabulation between water source and water safety
perceptions for the rural sample (row sums do not add up to 100% due to rounding). Table
5.3 shows a similar table for the urban sample.

Table 5.2 Cross Tabulation for Rural Sample Between Water Source and Degree
toWhichTapWaterisPerceivedSafetoDrink

 

Somewhat Very

 

All Very safe Somavhat safe Not sure unsafe unsafe
Public Supply 41% 44% 27% 25% 3% 1%
Private well 50% 64% 15% 18% 2% 0%

N 814 459 168 177 19 3

 

69

Table 53 Cross Tabulation for Urban Sample Between Water Source and Degree
toWhichTapWaterisPerceivedSafe toDrink

 

Somewhat Very

 

All Very safe Somewhat safe Not sure unsafe unsafe
Public Supply 77% 42% 24% 31% 2% 1%
Private well 21% 42% 26% 30% 1% 1%
N 359 153 89 111 7 3

 

For the rural sample, 25% of those who rely on a public system are not sure of the
safety of their water compared with 18 percent of those who rely on their own private wells.
For the urban sample, 31 percent of private well users and 30 percent of households who
rely on public supplies were unsure of the safety of their tap water.

How respondents perceive tap water safety is associated with their level of
conﬁdence that their tap water is safe from the effects of groundwater pollution. Tables 5.4
and 5.5 show cross tabulations for the rural and urban sets respectively. The association was
analyzed by testing the independence of the rows and columns (association) of the two
variables with Pearson’s chi-square test (Walpole and Myers 1978). The null hypothesis of
no association is rejected at the 95 percent level. The data indicates that people who rate
the safety of their tap water as high are also conﬁdent that their tap water is safe from the

effects of groundwater pollution.

70

Table 5.4 Cross Tabulation for Rural Set of Tap Water Safety Perceptions and Degree of
Conﬁdence that Tap Water is Safe from Effects of Groundwater Pollution.

 

 

Very Not conﬁdent
All conﬁdent Conﬁdent Not conﬁdent at all
Very safe 56% 40% 53% 7% 0%
Somewhat
safe 21% 4% 55% 34% 7%
Not sure 21% 0% 24% 62% 15%
Somewhat
unsafe 2% 0% 0% 37% 63%
Very unsafe 0% 0% 0% 0% 100%
N 824 189 377 203 55

 

Table 5.5 Cross Tabulation for Urban Set of Tap Water Safety Perceptions and Degree of
Conﬁdence that Tap Water is Safe from Effects of Groundwater Pollution.

 

 

Not conﬁdent
All Very conﬁdent Conﬁdent Not conﬁdent at all
very safe 42% 30% 61% 8% 1%
Somewhat
safe 25% 1% 51% 44% 5%
Not sure 31% 0% 29% 56% 15%
Somewhat
unsafe 2% 0% 0% 43% 57%
Very unsafe 1% 0% 0% 100%

§§

55

71

A chi-square test was also conducted on the association of well testing and the
degree of conﬁdence in the safety of their water from groundwater pollution. There was
no association for the urban sample. For the rural sample the two variables are associated. .
Table 5.6 shows the cross tabulation between the two variables. In general, households who
tested their wells were typically more conﬁdent their drinking water is safe from the effects

of groundwater pollution.

Table 5.6 Cross Tabulation Between Well Testing and Degree of Conﬁdence
that Tap Water is Safe from Groundwater Pollution

 

 

All 00:13am Conﬁdent Not Conﬁdent Not conﬁdent at all
Tested 17% 27% 15% 12% 18%
Not Teaed 83% 72% 85% 88% 82%
N 811 188 366 200 55

 

Risk Perceptions

Respondent risk perceptions corresponding to the severity, knowledge and choice
risk descriptors as speciﬁed in Chapter 3 are discussed below.

5mm. Urban households perceived the seriousness of the health risk from
groundwater pollution to be greater than the rural sample. About 38 percent of the urban
sample believed the health risk from groundwater pollution to be serious or very serious
compared with 23 percent in the rural sample. 0f the rural households, 40 percent believed
the health risk as not serious or not a threat at all. For urban households, 27 percent had
similar beliefs. The difference in sample distributions is statistically signiﬁcant at the 95

percent level.

72
About one-fourth of each sample (24 and 23 percent, respectively, for the rural and

urban samples) greatly feared groundwater pollution. In rural households, 47 percent
somewhat feared groundwater pollution while 52 percent of urban households held similar
views. About one fourth of each group (29 percent for rural households and 25 percent for
urban households) either fear groundwater pollution very little or not at all. There is no
statistically signiﬁcant difference in the sample distributions.

The third element of severity is the perceived likelihood that «posure to
groundwater pollution is fatal. Of the rural households, 15 percent believed that
groundwater pollution exposure would likely or very likely be fatal compared with 19 percent
of the urban households About 40 percent in each sample believed that groundwater
pollution was either unlikely to be fatal or not fatal at all. The difference in sample
distributions is not statistically signiﬁcant.

The consistency of responses to the severity questions was addressed by using
Pearson’s chi-square test. It is used to test the hypothesis that the row and column variables
are independent (Blalock 1979). The 95 percent level of signiﬁcance was used to test the
hypothesis of independence. A cross-tabulation of the three severity variables was
accomplished by considering the variables two at a time. Responses to the questions were
consistent for both samples across all three variables. More speciﬁcally, the more serious
people perceived the health threat from groundwater pollution, the higher the percentage
of respondents who feared or greatly feared the risk of groundwater pollution.

Other associations investigated using Pearson’s chi-square test included the
association of severity variables with water source and testing. There was no association
indicated between the severity variables and water source for either sample. There was no
association between severity variables and water testing for the rural sample. For the urban

sample, there was no association between well testing and perceived fear but an association

73

is indicated between well testing and both perceived fatalness and perceived seriousness of
the threat from groundwater pollution. Table 5.7 shows that, in general, the greater the

perceived seriousness, the higher the percentage of people who have had their wells tested.

Table 5.7 CrossTabulation ofWellTestingandPerceivedSeriousnessofI-IealthThreat
from Current Levels of Groundwater Pollution in County.

 

 

Not a Somewhat Very
All Threat Not Serious Serious Serious Serious
Tested 9% 0% 11% 3% 12% 19;
Not tested 91% 100% 89% ' 96% 88% 81%
N so 13 79 124 84 48

 

Table 5.8 below shows the cross tabulation of well testing and perceived fatalness.
Of those respondents who said current pollution levels would very likely be fatal, 23% had
tested their wells compared with 14 percent of those who believed that current pollution
levels would not be fatal.

Tab1e5.8 CrouTabulationofWellTestingandPerceivedFatalnessofCurremLevels
of Groundwater Pollution in Michigan

 

Unlikely to Sometimes Likely to Veryukelyto

 

All Not Fatal be fatal Fatal be Fatal be Full
Tamed 8% 14% 10% 4% 7% 23%
Not med 92% 86% 90% 96% 93% 77%
N 350 22 122 143 41 ﬂ

 

Km Over one-third of the rural and urban samples (37 and 38 percent
respectively) believed that the health risks of groundwater pollution are known or known
precisely by scientiﬁc experts. For both samples, 46 percent believed health risks to be

74
somewhat known. Less than one-fifth of either sample perceived that health risks were only

suspected or guessed at by scientists or not known at all.

The second component is knowledge of the health risks from groundwater pollution
by «posed individuals. Less than 10 percent of each sample believe that «posed
individuals are knowledgeable about the health risks of groundwater pollution. Conversely,
51 percent of the rural sample and 59 percent of the urban sample perceive that health risks
are only guessed at or are not known at all by individuals «posed to groundwater pollution.
For both of the above knowledge components, the diﬁ'erence in sample distributions are not
statistically signiﬁcant.

The third knowledge component is the concept of newness, or the «tent to which
groundwater pollution is perceived to be a new or old environmental problem. In the rural
sample, about one-third believe groundwater pollution to be a relatively old environmental
problem while 40 percent of urban households hold the same view. In comparison, 26
percent of the rural sample and 23 percent of the urban sample believe that groundwater
pollution is a relatively new environmental problem. The difference in sample distributions
is statistically signiﬁcant at the 95 percent level.

mm. Choice variables include: 1) controllability, the degree to which damage
from groundwater pollution can be controlled by technology, and 2) voluntariness, the
degree to which «posure to groundwater pollution is voluntary. Perceptions of
controllability were fairly evenly divided across possible, somewhat possible and very diﬁ'icult
(36, 30 and 30 percent respectively for the rural sample, 33, 29, and 31 percent for urban
sample). The difference in sample distributions is not signiﬁcant.

A majority of both samples believed that «posure to groundwater pollution risks is
involuntary. For the rural sample, 55 percent believed that «posure is involuntary while

67 percent of the urban sample have the same perception. Only 13 percent of rural

75
households and 9 percent of urban households believe that «posure is somewhat voluntary

or voluntary. The sample distribution difference is statistically signiﬁcant.

Environmental Perceptions

Environmental perceptions refer to perceptions of the eﬁ'ects of groundvmter
pollution aside from direct health effects to the household. The survey questionnaire
elicited responses to ﬁve questions related to environmental perceptions: (1) the «tent to
which households believe that future generations will be harmed by groundwater pollution,
(2) household perceptions of the potential threat of groundwater pollution to wildlife and
plants, (3) belief that groundwater should be protected for its own sake (aside from any
human or non-human impacts), (4) household beliefs as to whether pollution is bad because
it is the wrong thing to do or because it has bad effects, and (5) whether groundwater should
be protected from pollution only when pollution poses a direct health threat to humans.
The ﬁrst two questions relate to a kind of bequest risk motivated by intergenerational and
interpersonal concern (Randall and Stoll 1983). The last three comprise a type of altruism
motivated by both intrinsic and ethical concerns which reﬂect the perception that human
health is not the only reason to protect groundwater from pollution. In general, this is
similar to Randall and Stoll’s q-altruism (1983).

W Both rural and urban households believe future generations would be
harmed by groundwater pollution. Of rural households, 27 percent believe future
generations were likely to be harmed. The urban sample had 34 percent with similar beliefs.
For the rural sample, about 76 percent believe future generations are likely or very likely
to be harmed. For the urban sample about 80 percent hold similar views. The difference
in sample distributions is statistically signiﬁcant.

76
The rural and urban samples have signiﬁcant differences regarding the threat of

groundwater pollution to wildlife and plants. For the urban sample, 21 percent believe the
threat to be very serious compared with 12 percent for rural households. For rural
households, 22 percent believe there is not a serious threat. For urban households it is 17
percent. The difference in sample distributions is statistically signiﬁcant at the 95 percent
level. As with the previous question, urban households perceive the threat to wildlife and
plants as more serious than rural households.

Pearson’s chi-square test indicates an association between perceived seriousness of
the health threat from current levels of groundwater pollution in the respondents’ counties
and perceptions of groundwater pollution’s threat to future generations and wildlife and
plants. Speciﬁcally, the greater the perceived severity, the greater the perceived threat to
wildlife and plants. Both the rural and urban samples show this association.

93m A majority of both rural and urban households believe that other
reasons exist to protect groundwater besides human health. For the rural sample, 69
percent believe that groundwater should be protected because it is part of a natural system
while 68 percent of the urban sample hold a similar belief. Both the rural and urban sample
had 20 percent of respondents who believed that groundwmer should be protected for its
own sake. There was no statistical difference in the sample distributions.

A signiﬁcant majority of both samples believe pollution is bad bemuse it is the wrong
thingtodo, asopposed tobeingbad onlywhen ithasbad effectson peopleandotherliving
or non-living things. For both the rural and urban sample, 75 percent believe that pollution
is the wrong thing to do. Only 14 percent of the rural sample and 17 perwnt of the urban
sample believe pollution is had only when it has bad effects.

As to whether groundwater should be protected from pollution only when pollution
poses a direct health threat, 60 and 62 percent of the rural and urban households

77
respectively strongly disagree. For the rural sample, 24 percent either strongly agree or

agree somewhat with the statement. For the urban sample, 22 percent hold similar
perceptions.

A number of cross-tabulations between the q-altruism variables and water source,
education and income were conducted. Using the chi-square test, no association was found
between water source and q-altruism variables. Education had a varied association with the
q-altruism variables. Neither sample showed an association between wanting to protect
groundwater for its own sake and education. Both samples indicate an association between
perceptions that pollution is the wrong thing to do and education. Table 5.9 shows a cross-
tabulation between q-altruism perceptions and education. Up to the education level of
having attended but not completed college, as the level of education increases, a greater
percentage say that pollution is the wrong thing to do. However, for those who have
completed college, there is a marked increase (about 15 percent for both samples) in those
who respond that pollution is had only if it has bad effects. Both rural and urban samples
show an association between education and perceptions that other reasons «ist for
protecting groundwater from pollution besides human health and education. Generally, as
education increases, a greater percentage of respondents are more likely to respond that
other reasons exist besides human health for protecting groundwater from pollution.

Similar to education, the association between income and the q-altruism variables
is varied. The rural sample shows an association between income and perceptions that
groundwater should be protected for its own sake while the urban sample does not. For the
rural sample, a higher percentage of people in the middle income brackets, as opposed to
those in the lower and higher brackets, perceive that groundwater should be protected for

its own sake. The urban sample shows an association between income and perceptions that

78

Table 5.9 Cross Tabulation of Q-altruisrn Perceptions and Education Level.

 

Education level

 

All Elementary Highschool Somecollege more
rural urban rural urban rural urban rural urban rural urban

 

Wrongthingtodo 76% 75% 65% 47% 75% 70% 83% 88% 70% 71%

ammma

hubadeffccts 14% 16% 10% 20% 12% 13% 11% 11% 23% 25%
Disagreewithboth 5% 2% 6% 0% 6% 6% 4% 1% 5% 1%
Don'tknow 5% 6% 19% 33% - 7% 11% 3% 1% 2% 3%
N 786 353 31 15 332 104 224 120 179 114

 

pollution is the wrong thing to do. As income increases, a greater percentage of people
respond that pollution is bad only when it has bad effects. Finally, both samples show an
association between income and other reasons to protect groundwater from pollution
besides human health. As income increases, a greater percentage of people perceive other
reasons to protect groundwater besides human health.

Policy Bids

Table 5.10 summarizes policy bid means for both samples. The nonresponse rate
for the policy bid questions ranged from 20 to 25 percent. All bids are statistically different
from zero. Policy bids from the rural sample are fairly constant for the scientiﬁc
information policy, the prevention policy and the well protection policy, ranging from $43.53
to $45.07. The rural policy bid for the combined program is substantially higher than the
other three policies and is statistically signiﬁcant at the 95 percent level. The urban policy
bids are also fairly constant across three of the policies (scientiﬁc information policy, the
prevention policy and the combined policy). The bids for the three policies ranged from

79
561.53 to 568.66. The mean urban bid on the well protection program is signiﬁcantly

diﬁerent from the other policies.

In comparing the bids across the samples, two of the policy bids are similar and two
are different. Speciﬁcally, the mean bids for the rural and urban samples are statistically
diﬁerent at the 95 percent level for the scientiﬁc information and prevention policy. The
urban bids are considerably higher for these two policies than the rural mean bids. Unlike
the rural sample, the mean bids for the urban sample for the prevention and well protection
policies are statistically different at the 95 percent level. There is no statistically signiﬁcant
difference in the mean bids across the two samples for the well protection and combined
policies.

The combined policy is a multidimensional policy in that it provides multiple
services. Hoehn shows that the presence of substitution eﬁ'ects across policy impacts will
aﬁect the value of the incremental policy impact (1991). For «ample, with the presence
of substitution eﬁ'ects, the valuation of a single policy which has two impacts will be less than
the sum of the valuation of the two impacts separately. This is seen in the policy bids for
both the rural and urban samples. For households in the rural sample who valued the
groundwater protection policy ﬁrst, the mean bid for the protection policy was $45.07 and
the incremental value of the well protection policy was 53.45. For households who valued
the well protection policy ﬁrst, the mean bid for the well protection policy was $44.24 and
the incremental value of the prevention policy was $20.52.

For households in the urban sample who valued the groundwater protection policy
ﬁrst, the mean bid for the protection policy was $64.52 and the incremental value of the well
protection policy was 5 15.63. For households who valued the well protection policy ﬁrst, the
mean bid for the well protection policy was 534.11 and the incremental value of the

prevention policy was $24.66.

80
Two points are evident. First, a comparison of the combined policy bids and the
bids for the prevention and well protection policy bids when valued separately show the
empirical results presented here are consistent with Hoehn’s ﬁndings. Second, the
incremental values are consistent with the change in services offered. Since both prevention
and well protection provide health services and only prevention provides environmental
services, it would be «pected that the incremental value of well protection would be low

while the incremental value of prevention would be greater.

Table 5.10 Mean Policy Bids

 

 

 

Policy
Mean
(standard Scientific , Wei!
error) Information N Prevention N protection N Combined N
Rural $4353 659 545.07 337 54424 324 $56.47 625

(2.68) (3.64) (4.74) (4.03)

Urban 56153 301 56452 138 534.11 152 $68.66 281
(5.82) (8.19) (4.19) (6.79)

 

CHAPTER6

DATA ANALYSIS

The groundwater questionnaire identiﬁes a number of potential variables which may
useful in analyzing the relationship between risk perwptions and household wtp for
groundwater pollution policies. This chapter provides a rationale and methodolog for
identifying variables to include in the analysis. Testable hypotheses regarding the
relationship between these variables and policy bids are then developed. An econometric
equation is speciﬁed and the estimated coefficients and test results are diswssed.

Identification of Independent Variables

The groundwater questionnaire obtained three general types of information: (a)
household risk perceptions in the cont«t of qualitative risk descriptors, which describe
different dimensions of risk (adapted from Fischhoff et al. 1978), (b) environmental
perceptions which measure bequest risk and q-altruism, and (c) demog'aphic information.
A major consideration is how the information obtained in the survey can be used to test
hypotheses regarding the relationship between perceptions, and socioeconomic characteris-
tics with wtp for groundwater pollution policies.

One way to approach this problem is to look at the objective of this analytical section
in a broad, general sense; to ﬁnd out whether a change in risk or altruistic perceptions

results in a change in policy bids. It is important to note that the surveyvariables in all

81

82

likelihood do not represent the exact motivations responsible for behavior in a given cont«t.
This implies the «istence of certain underlying variables which determine people’s behavior.
The difﬁculty lies in obtaining direct measures of these underlying variables given the
compl«ity of the human cognitive process. By careful attention to wording and through
repeated testing in focus groups, the questionnaire attempts to elicit information about risk
and altruistic perceptions which is indicative of their actual, unobserved perceptions.

One approach to obtaining measures of the unobserved motivations through
measures of observed variables is the use of latent variable models (Loehlin 1987). These
models were developed to address situations where multiple variables, including unobserved
ones, are involved (Loehlin 1987). Latent variable models can be used to develop measures
of the underlying motivations which generate observable phenomenon. In the case of
groundwater, a latent variable model can be used to develop variables for both risk and
altruistic perceptions, which are a measure of the underlying motivations which determine
the behavior of the individual in a given cont«t; e.g., in the formulation of a policy bid.
The underlying motivations also generate responses to such questions as “How greatly do
you fear groundwater pollution?."

One type of latent variable model is principal components analysis. Principal
components analysis is a multivariate statistical method, which is used in the analysis of
correlation coefﬁcients among variables. In essence, principal components analysis is used
to identify measures of underlying variables by analyzing the variation and cross correlation
within an observed variable set (Babbie 1973). This is accomplished through the generation
of variables (factors) that are highly correlated with some subset of the variables of interest
and are independent of one another. Consequently, principal components analysis reduces

an original set of proxy variables to a smaller set of underlying variables.

83

The discussion in Chapter 3 regarding psychological research into risk perceptions
noted the importance of risk as a multidimensional concept. The use of principal
components analysis has the potential to identify from a comparatively large set of answers
to risk related questions a few variables which represent the true dimensions of risk as
people perceive them.

Altruistic concerns as reflected in observed behavior are motivated by a variety of
considerations, including the nature .of the individual’s relationship with their immediate
community and society in general. As with risk, a number of survey questions related to
altruistic concern can be analyzed using principal components analysis to identify a few

variables which represent underlying motivations associated with altruistic concern.

Principal Components Analysis of independent Variables
For both the rural and urban samples, separate analyses were mrried out for the risk

perception variables and the altruistic variables. The set of risk descriptor variables
included: (a) scientiﬁc knowledge of groundwater pollution health risks, (b) «posed
individuals knowledge of groundwater pollution health risks, (c) newness, (d) catastrophic
potential, (e) control, (f) voluntariness, (g) fear, (h) fatalness, and (i) perceptions of the
seriousness of the health threat from current levels of groundwater pollution in the
respondent’s county. The set of altruistic variables included: (a) harm to future generations,
(b) health threat to wildlife and plants, (c) whether groundwater should be protected for its
own sake, (d) whether pollution is bad because it is the wrong thing to do or is bad only
when it has bad effects on other living and non-living entities, and (e) if there are other
reasons besides health effects to protect groundwater from pollution. The analysis proceeds
by computing a correlation matrix of the independent variables and then identifying linear

combinations of the observed variables. These linear combinations are «tracted factors.

84
Extracted factors are identiﬁed in descending order beginning with the factor that accounts

for the largest proportion of total variance in the sample. Two criteria were used to
constrain the number of factors derived from the sample data: (a) total variance «plained 4
by the «tracted factor, and (b) scree plots. The observed variables and derived factors are
standardized to haveamean ofzero andavarianceofone. Ifaparticularfactorhasa
variance of less than one, this implies that it is no better than a single variable in «plaining
total variance. Consequently, the criterion that «tracted factors have a variance greater
than one ensures that an «tracted factor is better than a single variable in «plaining total
variation. Scree plots graphically show the total standardized variance for each factor. This
allows a graphical comparison of the factors. There is usually a sharp break between large
and small factors in terms of the size of their loadings. This break does not necessarily
occur at the point where variance equals one (Cattell 1966). Consequently, both of these
techniques were used to identify factors suitable for «traction.

Scree plots for the risk descriptor analysis for both the urban and rural samples
identiﬁed two factors. Using the total variance criteria, three factors were identiﬁed. The
thirdfactor foreachsample appeared tobeofmarginalsigniﬁcance. For«ample, thethird
factor for the rural sample had a total variance measure (eigenvalue) of 1.08,. compared with
the ﬁrst two factors with measures of 2.35 and 1.51 respectively. For the urban sample, the
third factor had a total variance measure of 1.15 compared with the ﬁrst factors with
measures of 1.92 and 1.38 respectively. Given the marginaiity of the third factor, this initial
«traction phase was repeated with the «tracted factors constrained to two.

Using the total variance criteria, the initial «traction phase identiﬁed two factors
for the altruistic variable set for both the rural and urban sets. For the rural set, the ﬁrst
«tracted factor had a total variance measure of 1.86 with the second factor having a

measure of 1.07. For the urban sample, the ﬁrst factor had a variance measure of 1.66 with

85

the second factor having a measure of 1.14. The scree plots for both samples were
ambiguous. Since the «tracted factors were comprised of the expected variables, it was
judged that constraining the initial «traction phase to one factor would not be necessary.

The n«t step rotates the extracted factors, that is, the factors are transformed to
make them more interpretable (Ferguson 1971). The equamax method of factor rotation
was employed for both the rural and urban samples. This method is a combination of two
other methods; the varimax method, which attempts to minimize the number of variables
which have high loading on a particular factor, and the quartimax method which attempts
to minimize the number of factors needed to «plain a variable (Ferguson 1971).

A pn'ori, a minimum coefﬁcient standard of .50 was set for each factor. Any
variables which comprise a factor and have a coefficient less than .50 were to be deleted and
the initial «traction phase rerun. Inspection of the constrained initial extraction phase for
the risk descriptors analysis for the rural set showed three variables in the second factor to
have comparatively low coefﬁcients. Newness had a coeﬁicient of -.24, control .47 and
voluntariness 55. Even though voluntariness had a coefficient greater than .50, all three
variables were dropped and the factor «traction repeated. A judgement was made that
keeping the control and voluntariness variables together would result in a more consistent
analysis than splitting them apart by including one in an «tracted factor and the other as
a separate independent variable. For the urban set, three variables had coefﬁcients less
than .50. Newness had a coefﬁcient of -.25, voluntariness .48 and control .49. Consequently
these variables were deleted and the initial «traction phase repeated. As with the rural
sample, these variables will be included as separate independent variables in the subsequent
regression analysis.

For the altruistic variables, all coefficients for the rural set were above .50. For the

urban set, the variable associated with perceptions that other reasons «ist besides human

86
health to protect groundwater had a coefficient of .42. Since this variable is conceptually

a component of q-altruism which includes the other two variables in the factor, a decision
was made to keep these variables together as a set which comprises the second «tracted
factor.

Lastly, factor values were calculated for each observation in both samples. In this
manner, the «tracted, rotated factors can be used to represent underlying variables in a
regression equation to estimate a policy bid function.

W Principal components analysis of the risk descriptors for the rural
and urban samples is shown in Table 6.1. The values shown are the rotated factor loading.
TWO underlying variables are identiﬁed for both samples. For the rural sample, the ﬁrst
underlying variable (highlighted in bold) has relatively high coefﬁcients of over .60 for the
ﬁrst four risk variables. The urban sample obtained similar results «cept for the lower
loading on perceived fatalness (LC) and the catastrophic potential of groundwater pollution
(1.d). The underlying variable severity is associated with perceptions of fear, the potential
threat of pollution, the likelihood of exposure being fatal and the catastrophic potential of
groundwater pollution.

The second underlying variable has high loading for two of the three risk descriptors
associated with knowledge and discussed in Chapter 3. This variable for knowledge has
large coeﬁicients for the risk variables representing scientiﬁc knowledge of the health risks
from groundwater pollution and knowledge of the health risks of «posed individuals. The
knowledge variable represents household perceptions of the level of knowledge regarding
the health eﬁects of groundwater pollution possessed both by the scientiﬁc community and

«posed households.

87

 

was.

«he.

03..

owe.

m8:

3%

a

~36

N8.

96.

awe.

:9.

mum.

can.

amp.

85.

N5.

08.

96.

has.

and.

 

 

Se: .83.

:35 .23.

 

omBEos.

€28

Agni... $.nuwaéﬁaﬂoﬁaaum
aggsuio—givéaggibgeb
2883§3338aB§§§§ﬁ§28§ ..

57.35 55.155
giggingggsugo—gigagg
Efgjoﬂgibgehaoaouoajgsﬁoﬂm .o

Agata: 63.156883382 omits-uni:
oauaiaosugaigcaggﬁsguegﬁgg
.soaegaosaoaesiﬁéaaasuaaasaéeazsé s

.Aﬂauae: 3.7.2
3:532:5esun315 .2: 22.3313.8
.2... B 2 :25 2. 53.32 a 83:8 .255...» .o :26. 32.3 3533 a

@3235 ”Ba-15E
gabgunveaaﬁﬁoéaglgogﬁaéaggeoﬁ
nous—.2— ..ogveaeh .3 .93 .356 89.. 325 532— 3 «88:88 .8383.— .9

.62-53 $3....sz €85 .33. a:
ﬁa§§.._e$o_§§&ao..8.§_§za
eon—5.x. 3.23585 .33 3.33:9. 523 2 3.500 .a

33§> “7.2 .—

 

83.; 32.02:

€85 232$

 

83%; a .o as; 9:22.28 wees; 8 23.

88

W Table 6.2 shows the results of principal components analysis
applied to ﬁve variables representing different types of altruism. For both the rural and

urban sample, the ﬁrst factor has high loading on variables representing household
perceptions of the health likelihood of groundwater pollution harming future generations
and the threat to wildlife and plants.

The second factor is associated with altruistic motivations associated with both
intrinsic and ethical concerns regarding groundwater pollution. Intrinsic motivations are
captured in variables LC and 1.d in Table 6.2. Question 1.e in particular and 1.d in addition
are associated with an ethical concern regarding pollution which is over and above concern
regarding health impacts to both human and other living thing. Question l.e in a
senserepresents the idea that, aside from any harmful effects, the act of polluting violates
a held ethical norm.

Independent Variables and Testing of Hypotheses
Three sets of variables were used to analyze the relationship between policy bids and

motivating concerns as represented by independent variables. The three sets of variables

include risk factors, altruism, and socio-economic variables.

Risk Variables

TWO underlying variables and three proxy variables will be used for hypothesis testing.
The underlying variable severity measures the households perception of the severity of
groundwater pollution as a health threat. As perceptions of severity increase, policy bids
will increase. Bids for all four policies are expected to be aﬁected by severity perceptions.

 

a
i

«NV. as.

a

n3.

 

:35 .23.

8.. EN.

N9. g.

"nu. v3.

 

:25 .23.

 

8.9.3290

.3. .23....

5... 1.35 «27.23.. 8 2 3... m8... 2.. a s. 3.33..
in... 2.. a a... 93.2. 2.2.3.6.. a 8.8%... . 22.3 3.3.2 .55... 2. < .o

.gmuaﬁa gauze...
3&3... icﬁuv can 85.. 23.55... 5.3 0.3.. 2...... :5. a .5 “.58....
e. 2.9.... 89.... a .08.. 5.5.5.. 05 no.5 :5 5.3.5.. Son. .5805...

2. 2.5... 3.3.58 a... .8922. o... a... 3038.... .o .29. 33.9.3: ...

.§....§.5 gcuaau. 2.2 as a. a. 3.88... a.
2.5... 3.8.6.595 26...... 3.2.3.5.. .. 3.3%... . 22.3 033.2. .88.... Tc < .o
52.535 £3.15... 32.8 team .2. .1 s
28.... a 5.7. A...» 01.8.59. 95... .5 .35... v.3 05.3.... o. .32.. 5.3..
a A.9... 535.3. a. 5.5.5.. 333.555 .0 .95. 52.3 .5 $335.28 .5333.— A.

accuse... .8322. 3.2.... a. 2 as... as... .5 ... s 3.2....
5.7. 5.3 0.8.. 55.. 9a a .5 3.3:: e. m an... o... .95 5.95.5.5.»
.3 v2.5... 2.. 25.3.03» 9.3... 49...? o. 2.25 o... .0 «5.53.2. 29.3.5: 6

8.2.55 98.3.2 .—

 

8......;> 2.2.8....

€85 03.29

 

2......3> 5.2.20 .2... .83.... .o ”.22.... 2:29.58 3.8.... 2. 2......

90
Consequently, severity, severity l, serverity 2, severity 3, severity 23 will have positive and
signiﬁcant coefficients.

The underlying variable knowledge measures household perceptions of the degree
of knowledge possessed by scientists and exposed individuals. The hypothesis regarding the
relationship between knowledge and wtp will be ambiguous, as discussed in chapter 3.
Similarly, the choice variables, voluntariness and control, are also expected to have an
ambiguous correlation with wtp.

Newness was speciﬁed as a knowledge variable in Chapter 3. The factor analysis did
not identify newness as a component of the knowledge dimension. Consequently, newness
will be tested separately. The newer groundwater pollution is perceived to be an
environmental problem, the less knowledge and experience people have in dealing with the
problem. As with the other knowledge variables, the relationship between newness and wtp
will also be ambiguous.

Q-sltrulsn snd Bequest Risk
Q-altruism is associated with existence value of groundwater. Existence value is
comprised, at least in part, of intrinsic value and ethical concerns. Households who obtain
utility from intrinsic value and from knowing that held ethical norms are not being violated
will have higher bids for policies which prevent aquifer pollution than households who do
not hold those views. The presence of q-altruism will not affect bids for remedial policies,
since these policies do not affect aquifer pollution and therefore do not aﬁect existence
value.
The estimated coefficients on the q-altruism variable will be positive for those
policies which prevent aquifer pollution. The effect of existence value perceptions on bids

for the scientiﬁc policy is less clear. The main objective of the policy is to obtain more

91

precise information about the extent of groundwater pollution and the population at risk
from exposure. While the policy does not directly prevent aquifer pollution or exposure, it
does not seem unreasonable to assume that people would think that the information is being
collected for some constructive purpose; i.e., to formulate and implement progams that
reduce groundwater pollution and potential exposure. Consequently, motivations for paying
some positive amount for either prevention and remedial policies would seem to apply to
the scientific information policy also. The following variables will have positive and
signiﬁcant estimated coefficients: (a) q-altruism. (b) q-altruism 1, (c) q-altruism 2, and (d)
q-altruism 23.

Bequest risk measures the perceived threat of groundwater pollution to future
generations and wildlife and plants. The greater the perceived threat, the greater the policy
bid. Perceptions of both aquifer pollution and exposure risk are measured by this variable.
Consequently, both prevention and remedial policies will be aﬂ'ected by bequest risk. The
following variables will have positive and signiﬁcant estimated coefficients: (a) bequest risk,
(b) bequest risk 1, (c) bequest risk 3, (d) bequest risk 4, and (e) bequest risk 23.

Socioeconomic Variables

No variables will be included in this category, eaqaected annual income and education
level of the household. ‘No previous contingent valuation studies addressing groundwater
pollution problems found income to be a signiﬁcant variable in explaining the variation in
wtp for groundwater protection policies (Edwards 1988, Shultz and Lindsay 1990). Studies
cited by Shultz and Lindsay (1990) showed that households with a high degree of education
had high levels of concern over groundwater pollution. Consequently, the estimated
coefficients on education and income will be positive and signiﬁcant. Independent variable

definitions are summarized in Table 6.3.

 

92

Table 6.3 Independent Variables Used in Policy Bid Estimation

 

 

 

Variable Definition

51 A 0.1 dummy variable where 1 indicates the scientiﬁc information policy.

DZ A 0-1 dummy variable where 1 indiates the prevention poliq.

D3 A 0-1 dummy variable where 1 indicates the remedial poliq.

DB AO-ldummyvariablewherelindicatesthecombinedpoliq.

Severity Avariableidentiﬁedbyprincipslcomponemsanalysisusociatedwiththe
perceivedseverityoftheheshhrlskh'omexpomretogroundwaterpoﬂudon.

Severityi AninteractivevariabledefinedasD,’Sevesitywherei-l,2.3,23
corresponding to dummy policy variables.

Knowlecbe A dimension of risk identified byprincipal components analysissssociatedwlth
the perceived level of existing scientific and personal knowlerhe of
groundwater pollution.

Knowledge , An interactive variable defined as DﬂCnowleQe where l-l, 2, 3, ZS.

Bequest risk Avariableidentiﬁedbyprincipal componentsanalysisusoclstedwiththele/el
of concern for others (both human and non-human) affected by groundwater
pollution.

Bequeariski AnimemctivevariabledefinedssD,‘IIner-beingahmismwhesei-l,2,3,
23.

q—altruisin Avariableidentiﬁedbyprincipslcomponerusanalysisassociatcdwlth
perceived importance of the uistence value of groundwater.

q-altruism, AnirueractvariabledeﬂnedasD,‘Q-sltruism.wherei-1,2.3,3.

Newness AO-l dummyvariablewherelindicstesthathcuehokbpercelvegroundwster
pollution risk as s new or somewhat new environmental problem.

Control AO-lchimmyvariablewherelindicsteshouseholdperceptionsthatthe
damage from groundwater pollution is very difﬁcult or impossible to control
through technology.

Voluntariness AO-ldumrnyvariablewherelindicsteshouseholdperceptionsthstpeople
are involuntarily «posed to groundwater pollution.

Education Iastyearofschoolcompletedonasixpointscalewlthl-nosdlooland
6-last year of college or more.

Income Expected annual income in one year, thousands of dollars.

 

93
Regression Analysis
The objective of the regression analysis is to empirically test the speciﬁed hypotheses
regarding risk perceptions, altruism, and socioeconomic variables with wtp for alternative
groundwater policies. To test the hypotheses, two alternative speciﬁcations of an
econometric model are used. The ﬁrst speciﬁcation is

(6.1) BID-B,D1+B,DZ+ 53D3+64D23+B,Severity
+ [3‘ Knowledge + B, Newness + 3. Control
+ B, Voluntariness + B", q-altruism + 5,, Bequest risk

+ ﬂu Education + 3,, Income 4» e.

where D1, D2, D3, and D23 are policy dummy variables denoting the scientiﬁc, preventive,
remedial and combined policy respectively and e is an error term assumed to be normally
distributed with a mean of zero.

This equation deﬁnes a linear relationship between the dependent and independent
variables. Consequently, a linear regression technique such as 01.5 can be used to estimate
the coefﬁcients. The use of the policy dummy variables allows coeﬂicient estimation to be
derived from one equation instead of using one equation for each policy. By specifying the
independent variables in mean deviations form and suppressing the constant term, the
coeﬂ'icients for the policy dummy variables will equal the sample mean for that partimlar
policy-

The second model is the same as equation (6.1) except that all variables derived
from the principal components analysis are used to create interactive variables which

speciﬁcally link the underlying variables q-altruism, inter-being altruism, severity, and

know

of hi

SCI 0
been
be re

the s

numt
to wk
gene:
some
house

must:

and U1
whim
comp;
signjﬁ,
531 ha:

Comp],

94
knowledge with each of the four different policies. This model shows the speciﬁc impact

of household perceptions as embodied in these variables on each of the policy bids.

The regression analysis for both the rural and urban sets is based on a “restricted“ .
set of usable surveys. This subset consists of questionnaires for which all nonresponses have
been eliminated. In the regression analysis, alternative sets of independent variables will
be tested as to their signiﬁcance in explaining bid variation for groundwater policies. Since
the‘ set of questions not answered in a given survey differed across respondents, analysis
based on the complete sample will result in different sets of variables having different
number of observations. In effect, households will drop in and out of the analysis according
to which questions were answered. Tests of statistical signiﬁcance will be aﬂected and the
generality of results may be limited if respondents to diﬁerent questions differ markedly in
some important characteristic. Consequently, by restricting the urban and rural samples to
households that have responded to a given set of variables, the set of respondents will be
constant across alterative speciﬁcations of explanatory variables.

A comparison was made of the complete and restricted samples for both the rural
and urban sets, with respect to the independent variables used in the regression. The Mann-
Whitney test was employed to test the null hypothesis that the two variables (from the
complete set and from the restricted set) have the same distribution. The only statistically
signiﬁcant diﬂ'erence noted was for the income variable for the rural sample. The restricted
set has a greater percentage of people in the higher income brackets compared with the

complete set.

Results
Tables 6.4, 6.5 and 6.6 show results of the OLS estimation of the bid equations. The

independent variables are identiﬁed in the left-hand column. Table 6.4 shows the estimated

95

Table 6.4 Estimated Coeﬂ'icients for Bid Equation with
Impact of Underlying Variables Averaged Across Policies.l

 

 

Variables Rural2 Urban2
Severity 8.48’ 17.21‘
(264) (5.96)
Knowledge 4.70‘ -9.64‘
‘ (1.94) (4.49)
Bequest risk 11.12‘ 0.98
(2.58) (5.79))
q-altruism 5.06‘ 15.61‘
(1.96) (4.39)
Newness 6.62 -9. 14
(433) (10.06)
Control -2.06 036
(4.04) (9.52)
Voluntariness -2.79 7.12
(3.97) (9.36)
Education 1036‘ 3.25
(1.90) (5.04)
Income 0.24‘ 0.20
(0.07) (0.18)
Adjusted R2 .39 .32

 

‘ Estimated coefﬁcients for policy dummy variables are equal to the mean bids for
that sample when independent variables are in mean deviations form:
Rural: Dis-$45.35; D2235154; D3-S40.l6; D23-457.89.
Urban: D1-S71.22; D2-S73.19; D3-S38.54; D23-S75. 15.

3 The "" signiﬁes the estimated coeﬁicient is statistically different from zero at the 95
per cent conﬁdence level.

I.»

Table 65. Estimated Coemcients for Bid Equations with Interactive Policy Variables‘

 

 

Variables Rural’ Urban’
q-altruism Coeﬁcient (SE) Coeﬁcient (SE)
Science Policy 6.12‘ (3.66) 14.25‘ ( 7.59)
Prevention 81)‘ (4.74) 23.97‘ (11.84)
Well Preotection -2.51 (4.74) 3.32 (10.12)
Combined Policy 6.06‘ (3.36) X10? ( 7.59)
Bequest risk
Science Policy 8.52‘ (4.46) 1.74 (. 9.98)
Prevention 1381‘ (5.88) 3.01 (14.82)
Well Protection 7.74 (ms) 10.73 (14.13)
Combined Policy 13.40‘ (4.47) -0.93 ( 9.98)
Severity
Science Policy 6.43‘ (4.54) 20.19‘ (10.10)
Prevention 5.34 (6.09) 29.16‘ (14.60)
Well Protection 9.39 (6.92) -2.43 (15.11)
Combined Policy 1226‘ (4.54) 17.56‘ (10.12)
Knowledge
Sdence Policy 4.07 (3.29) -4.27 (7.59)
Prevention 6.35 (4.70) -17.81 (11.79)
Well Protection 020 (4.61) -2.80 (10.67)
Combined Policy 6.72‘ (3.29) -15.62‘ ( 7.60)
Newness 6.42 (4.33) -11.98 (10.17)
Control -1.68 (415) -0.2.5 (9.57)
Voluntariness -2.72 (3.98) 7.70 (9.41)
Education 1058‘ (1.90) 3.36 (5.04)
Income 0.23‘ (0.07) 0.19 (0.18)
Adjusted R’ 39 .31

 

1. Esﬁmatedcoeﬁidemsfapolkydummyvariablesareapproximatelyequaltothe
mean bids for that sample:
Rural: D1-S45; D2-352; 03-540; D23-$58.
Urban: D1-S71; 02.573; 03-339; D23-$75.

2. Themsigniﬁestheestimatedcoeﬂidemisstaﬁsﬁcaﬂydiﬁeremﬁomzeroatthe
95percentconﬁdencelevel.

Va

q-a

Net

Cor

V0]:

Edu

Into

Flaw.

7k

97

Table 6.6. Estimated Coeﬁcients for Final Bid Equations with

 

 

Interactive Policy Variables‘

Variables Rural2 Urban’
Severity Coemcient (SE) Coeﬂicient (SE)

Science Policy 6.43 (4.52) 19.70‘ (9.98)

Prevention 5.33 (6.09) 22.75‘ (10.31)

Well Protection 15.13‘ (4.57) -

Combined Policy 1226‘ (454) 16.36‘ (7.48)
Knowledge

Science Policy 4.06 (3.29) -

Prevention 6.35 (4.70) -

Well Protection 0.42 (4.60) -

Combined Policy 6.71‘ (3.29) -14.09‘ (7.46)
q-altruism

Science Policy 6.10‘ (3.36) 14.28' (7.51)

Prevention 8.19‘ (4.74) 24.83‘ (11.62)

Well Protection - -

Combined Policy 6.05‘ (3.36) 20.33‘ (7.51)
Bequest risk

Science Policy 8.52" (4.47) -

Prevention 1458’ (5.80) -

Well Protection - -

Combined Policy 1340‘ (4.47) .
Newness 6.24 (3.29) -
Control -1.69 (4.06) -
Voluntariness -2.74 (3.98) -
Education 1069‘ (1.90) -
Income 0.23‘ (0.07) -
Adjusted R’ 39 32

 

1. Estimated coemcients for policy dummy variables are approximately equal to the mean bids for

that sample:

Rural: D1-S45; D28352; D23-$40; Dzs-ssa
Urban: D1-S71; D2-373; D3-S39; D23-$75.

2. 1‘he'”signiﬁestheestimatedcoefﬁcientisstatisticaﬂydiﬁerentfromzeroatthe95percent

conﬁdence level.

98
coefﬁcients for the speciﬁcation in equation 6.1. In this speciﬁcation the impact of the

independent variables on policy bids are averaged acrom the different policies. In Table 6.5,
interactive variables are used to partition out the effects of selected variables on each type
of policy. Table 6.6 shows the ﬁnal bid equations which are the result of reestimating the

equations in Table 6.5 with insigniﬁcant variables dropped from the equation.

Underlying Variable Effects Averaged Across Policies

Table 6.4 shows the estimated coeﬁ'icients of the speciﬁcation which tests the
averaged effect of the independent variables on policy bids. The estimated coefﬁcients
represent the dollar change in the mean policy bid due to a unit change from the mean
value of the speciﬁed independent variable. For example, the estimated coefﬁcient for the
variable Severity for the rural sample is $8.48. For those households with a one-unit
difference from the mean sample value for this variable, the policy bid would increase or
decrease by $8.48 with respect to the mean policy bid for the sample as a whole, given that
the remaining independent variables are evaluated at their respective means.

mm. For the rural set, severity, knowledge, q-altruism, and bequest risk,
education and income had positive and statistically signiﬁcant coeﬂicients. This finding
conﬁrms (reject the null hypothesis that the estimated coefficient is equal to zero) the
hypotheses regarding these variables.

For the variable severity, the statistical results show that as household perceptions
of the severity of groundwater pollution increase, household policy bids increase. This
follows in straightforward manner from the perceptions that comprise the variable severity:
fear, seriousness of the health threat (including mortality risk) from current levels of

groundwater pollution and catastrophic potential.

99

For the knowledge variable, the greater the perception that scientiﬁc experts and
exposed households know little about the health risks from groundwater pollution, the
greater the household policy bid. One interpretation is that if households feel that little is '
known about either the health effects of exposure to perceived levels of groundwater
pollution or the likelihood of exposure, then it is difﬁcult for households to know when or
if they are at. risk, and what the health consequences will be if exposed. If signiﬁcant
uncertainty is perceived to exist, then households are supportive of policies which will reduce
or eliminate exposure to groundwater pollution. In one sense, this is a kind of 'better safe
than sorry' perspective in response to perceived uncertainty. '

For the bequest variable, the greater the perceived threat to both future generations
from future pollution and to wildlife and plants from current levels of pollution, the greater
the household policy bid. With respect to future generations, the bequest variable explicitly
addresses the households’ perceived level of threat to future generations and also implicitly
addresses the degree of concern for future generations. Since the statistical results show
that perceived threat and wtp are positively correlated, this would seem to imply that in
addition to the perceived threat there is also concern for future generations. A similar
argument holds for the second component of the bequest variable, the perceived threat to
wildlife and plants. The important point here is that the bequest variable incorporates
primarily values derived from the use of the groundwater source by entities other than the
household. Included are both future generations and entities which are aﬁ'ected by
groundwater pollution and which in turn are used by others.

For the q-altruism variable, as the level of importance of q-altruism to the household
increases, household policy bids increase. Alternatively, the greater the importance of
nonuse values to the household, the greater the policy bid. It should be noted that q-

altruism perceptions were hypothesized to be correlated with those policies which provide

100

nonuse environmental services, i.e., policies which prevent the groundwater aquifer from
being polluted. In the speciﬁcation shown in Table 6.4, the estimated coefﬁcients are
derived by averaging the eﬁ'ect of the independent variable across all four policies, one of
which does not provide nonuse environmental services. Consequently, the estimated
coefficient for this variable should be considered as only a broad indication of the
relationship between wtp and q-altruism.

For three variables-newness, control, and voluntariness-the hypotheses cannot be
conﬁrmed (fail to reject the null hypothesis). A linear restrictions test (Kmenta 1988) tested
the hypothesis that the estimated coefficients associated with the three variables as a group
are equal to zero. The test showed estimated coefficients of the group to be signiﬁcant at
the 95 percent level.

Both demographic variables are statistically signiﬁcant and of the correct
hypothesized sign. Both level of education and expected annual income are positively
correlated with policy bids.

W. As with the rural sample, severity perceptions are positively
correlated with wtp. Even though most urban households rely on centrally supplied water
as opposed to domestic wells for rural residents, urban households can still be affected by
groundwater pollution. First, there may be the perception that publicly supplied water may
be contaminated by groundwater pollution and not be detected by routine water monitoring.
Second, households which rely on surface water may still be aﬁected by groundwater
pollution to the extent that wells are used to extract the water as opposed to direct diversion
from the surface water body. Third, almost 20 percent of the urban sample rely on domestic
wells for their freshwater source. Consequently, these households will probably have the

same motivation behind wtp as shown for rural households.

101

The estimated coefficient on knowledge for the urban sample is of opposite sign
from the hypothesized sign. The negative knowledge coefficient means the greater the level
of scientiﬁc knowledge and awareness of exposed individuals with respect to the health
effects of groundwater pollution, the higher the policy bid. One possible explanation is that
respondents in the urban sample associate higher levels of knowledge and awareness with
greater likelihood of an implemented policy successfully achieving its objectives. In other
words, urban respondents may believe that as more is known about a particular risk, the
greater the likelihood that a successful policy can be designed, which in turn results in
correspondingly higher policy bids.

Bequest risk and severity show a relatively high simple correlation of .67.
Consequently, one possible reason for the statistical insigniﬁcance of bequest risk is
multicollinearity. Intuitively, this seems plausible since a signiﬁcant component of bequest
risk as derived from principal components analysis is concern over future generations. In
many households the concern over the threat of groundwater pollution probably extends to
other households as well. The equation was estimated again with severity deleted. In this
case, the estimated coefﬁcient on bequest risk was positive and highly signiﬁcant (62 14.38
with a t-score of 3.26).

The estimated coefficient for q-altruism is statistically signiﬁcant and of the correct
sign. The coefﬁcient for the urban sample is over three times the size of the coefﬁcient for
the rural sample.

Three of the risk factor variables, newness, control and voluntariness are not
statistically signiﬁcant. A linear restrictions test found these variables to be insigniﬁcant as

aset.

102
Neither of the demographic variables are statistically signiﬁcant. However, a linear

restrictions test on education and income as a set showed these variables to be statistically

signiﬁcant.

Underlying Variables Speciﬁed with Interactive Policy Variables

The purpose of the interactive variables speciﬁed in Table 6.5 is to test the effect of
speciﬁc independent variables on speciﬁc policies. Table 6.5 lists four policies for each of
the four interactive policy variables as deﬁned in Table 6.3. Three of the risk factor
variables, newness, control and voluntariness are considered individually, as are the .two
demographic variables education and income. The estimated coeﬁicients for the four
interactive independent variables show the change in the mean policy bid for the speciﬁed
policy due to a unit deviation from the mean value for the speciﬁed independent variable.
For example, under the rural sample heading for the severity variable, the estimated
coefﬁcient for the science policy is 6.43. This means that a unit change in the mean value
for the variable severity results in a change in the mean value of the science policy of $6.43,
evaluating the other variables at their mean.

m. The variable representing severity is not signiﬁcant for three of the four
policies for the rural sample. The combined policy has a positive and statistically signiﬁcant
coefficient. A linear restrictions test on severity 1, severity 2 and severity 3 ﬁnds these
three variables signiﬁcant as a set. Additionally, a correlation matrix of the independent
variables shows that correlation coefﬁcients between bequest risk and severity variables
range from .63 to .68. Removing bequest risk from the analysis results in making all four
of the severity interactive variables positive and statistically signiﬁcant. Consequently, all
the hypotheses regarding severity and policy bids are conﬁrmed for the rural sample.

103
The urban sample shows that three of the four interactive variables are positive and

signiﬁcant. Only the interactive variable associated with the well protection program is not
signiﬁcant. Since most respondents in the urban sample do not rely on domestic wells as
a water source. the insigniﬁcance of the severity variable associated with the well protection
policy seems appropriate.

mm For both the rural and urban sets, only the combined policy coefficient
is signiﬁcant. For the rural sample, the estimated coeﬁicient can be interpreted as meaning
that as household perceptions of the level of both scientiﬁc knowledge and exposed peoples’
awareness of the health effects of groundwater contamination decrease, policy bids for the
combined policy increase. A linear restrictions test shows that for the rural sample, the
estimated coefficients for the variables knowledge 1, 2, and 3 are statistically signiﬁcant as
a set.

The estimated coefficient on the combined policy for the urban sample is negative.
This is interpreted as meaning the lower the perceived level of scientiﬁc knowledge and
awareness of exposed individuals with respect to the health effects of groundwater pollution,
the lower the policy bid. Alternatively, the higher the perceived level of knowledge and
awareness, the greater the bid. Similar to the discussion concerning averaged independent
variables, the negative coeﬂicients on knowledge for the urban group might be associated
with perceptions of the likelihood of the policy actually achieving its stated objectives. The
reasoning may be that the greater the level of scientiﬁc knowledge and personal awareness,
the greater the possibility that policies can be designed and implemented with a reasonable
chance of success.

For the urban sample, a linear restrictions test shows that the estimated coemcients

on knowledge 1, 2, and 3 are not statistically signiﬁcant as a set.

104
m For the rural sample, the estimated coefficients on q-altruism are

positive and statistically signiﬁcant with the exception of the well protection program. This

conﬁrms the hypotheses developed for the impact of measures of existence value on policy ,

bids. Since the well protection policy does not protect the groundwater aquifer, policy bids
are not affected by the importance of existence value to the household. Similarly, the
coefficients for the urban sample, again with the exception of the well protection program,
are positive and statistically signiﬁcant. The size of the coefficients for the urban sample
are two to three times larger than the coefficients for the rural sample.

W Three of four bequest risk variables for the rural sample are positive
and statistically signiﬁcant. As with q-altruism, the coefficient on the well protection policy
is not signiﬁcant. Unlike existence value, it was hypothesized that the coeﬂ'icient on the well
protection program would be signiﬁcant. However, this hypothesis is not conﬁrmed. None
of the estimated coeﬂicients for bequest risk for the urban sample are signiﬁcant. A linear
restrictions test on the set of bequest risk variables fails to reject the hypothesis that the
estimated coefficients as a set are zero.

WW Individually, none of these variables are statistically
signiﬁcant for either sample. However, for the rural sample, these variables are statistically
signiﬁcant as a set. The calculated F statistic was 26.20, an indication of the strength of the
signiﬁcance. However, the hypotheses associated with these variables cannot be conﬁrmed
since the sign on the coefficients is unknown. The estimated coefﬁcients for the urban set
are not signiﬁcant as a set. Consequently, the hypotheses associated with these variables
are not conﬁrmed.

mm. For the rural sample, both education and income variables
are positive and statistically signiﬁcant. For the urban sample, neither variable is signiﬁcant

independently. A linear restrictions test shows that the variables as a set are not signiﬁcant

105

as a set. Consequently, the hypotheses associated with education and income for the rural

sample only, are confirmed.

Final Bid Equation Using Interactive Policy Variables

Table 6.6 shows the estimated coefﬁcients for the ﬁnal bid equations with variables
having insigniﬁcant estimated coefficients dropped from the equations.

m. For the rural sample, the estimated coefﬁcient for the well protection and
combined policies are statistically signiﬁcant and of the correct hypothesized sign. For the
urban sample, the estimated coefﬁcients for all three of the included interactive variables-
science information, prevention and the combined policies-are statistically signiﬁcant and
of the correct sign. As hypothesized, severity perceptions are a signiﬁcant motivation in
rural household willingness to pay for well protection policies. However, the estimated
coefficients for the scientiﬁc information policy and the prevention policy are not statistically
signiﬁcant. This contrasts with urban households, who are not typically dependent on
private domestic wells as a water source. For urban households, both the scientiﬁc
information policy and the prevention policy are statistically signiﬁcant and of the correct
hypothesized sign. As previously established, severity is not correlated with well protection
bids for the urban sample.

W. For both rural and urban households, the estimated coefﬁcients for the
knowledge variable are statistically signiﬁcant for the combined policy. As was previously
the case, the sign on the coefficient for the rural sample was as hypothesized, while the sign
for the urban sample was opposite from that hypothesized.

m The estimated coeﬂicients for both samples for all three interactive
variables are statistically signiﬁcant and of the correct hypothesized sign.

 

106
M For the rural sample, all of the estimated coefﬁcients for bequest risk

are statistically signiﬁcant and of the correct hypothesized sign.

W The results for the rural sample are similar to the results previously
discussed. Newness, control and voluntariness are not signiﬁcant individually but are
signiﬁcant as a set for the rural sample. Education and income are statistically signiﬁcant

and of the correct hypothesized sign.

CHAPTER 7

IMPLICATIONS OF EMPIRICAL RESULTS AND RESEARCH NEEDS

The empirical analysis shows a number of perception variables as being both
economically and statistically signiﬁcant in explaining the variation in household bids for
goundwater policies. The significance of health and environmental risk perceptions. for
valuation and policy formulation will be addressed in this chapter. Finally, future research
needs identiﬁed by the theoretical framework and the empirical analysis will be discussed.

Significance of Risk Perceptions
The data analysis showed the empirical signiﬁcance of risk perceptions associated
with severity and knowledge in explaining bid variations. As perceptions of severity and
knowledge change, there is a corresponding effect on policy bids. The implications for
valuation and policy formulation are discussed below.

Implications for Valuation

An important implication of the empirical results is that the use of probability and
a single outcome descriptor, such as death. results in an incomplete and inadequate
description of the risk/ income trade-off situation faced by households. The empirical

signiﬁcance of the association between risk perceptions and wtp for goundwater pollution

107

 

 

108

policies shows the insight gained from an explicit consideration of the underlying motivations
affecting risk perceptions.

In essence, their subjective perception of goundwater pollution risk is expressed
tlnrougln the consideration of the risk in terms of the qualitative risk descriptors. The
empirical results showed that policy bids and risk perceptions are systematically related.
Consequently, household perception of the risk as embodied in their policy bid may vary
dramatically from that predicted by.a beneﬁt estimate typically based on a quantitative
speciﬁcation of mortality probability and a dollar ﬁgure which represents an estimate of the
value of a statistical life.

Given the empirical signiﬁcance of risk perceptions in explaining bid variations, it
would appear that use of the more general valuation framework in the estimation of policy
beneﬁts would be the most appropriate approach for policy evaluation. Mortality based
beneﬁt estimates can be used as a first cut, or if the opportunity for obtaining survey-based
estimates does not exist. However, the research results show the empirical importance of
direct estimation of beneﬁts This would argue for a contingent valuation study as the most
appropriate method (cost and time considerations aside) to estimate policy beneﬁts. In fact,
it would seem that the mortality-based approach would be a third-best approach. A second-
best approach would be to use existing valuation studies, which explicitly account for
household risk perceptions. The ﬁrst-best approach would be to explicitly elicit bids for the
speciﬁc situation under study.

It can be argued that reliance on people’s responses to technically complex risk
issues which are typically based on limited experience is hardly an improvement on mortality
based beneﬁt estimates. However, the issue seems not to be so much the speciﬁcation of
“the“ correct beneﬁts estimate, but more to what extent is society willing to reallocate scarce

resources from one purpose to another? This reallocation is based, at least in part, on

109

consumer preferences which in turn are based on how the situation under consideration is
perceived. Beneﬁt estimates based on consumer preferences may be considerably larger or
smaller than might be expected from a technical risk assessment. It should be noted, .
however, that consumer preferences are not formed in a vacuum. The empirical importance
of preferences points to the need for information, discussion and public debate to increase
the likelihood that informed choices are made by consumers.

The empirical importance of risk perceptions implies that risk descriptors associated
with these perceptions can aid in setting an appropriate context for policy valuation. For
example, policy impacts can be speciﬁed in terms of risk descriptors that have been found
to be of empirical sigliﬁcance to households. Risk perceptions found to be of economic
signiﬁcance, such as severity and knowledge in the goundwater survey, would be used as
a context for describing policy impacts. Explaining the current situation in the context of
these variables, or explaining policy impacts in terms of these variables may provide
information to respondents for policy valuation purposes, which accurately reﬂects
dimensions of risk which are meaningful to people in formulating their perceptions of

goundw&er pollution risks.

Implications for Policy

The econometric analysis showed certain risk perceptions to be systematically related
to bid variations. Table 7.1 shows how policy bids diﬁ’er with respect to whether
independent variable responses are above or below the sample mean for that variable.
Responses to each of the four interactive variables are categorized as either above or below
the mean value for that variable. The households with responses above the mean can be
characterized as perceiving the health risks from goundwater contamination to be more

severe in comparison with households with responses at or below the mean. Households

110

.a as: an ace 6:8 8:358 an a3... .8» ace canoe... =2. "a as;
“B ”—8.: £3.36... ..uE—ﬁeacb :5 ”:33 .m3 ”—2.: deuce—no.3 853 A822. :3... 8 33286 8338.» genes—.5 53v 388 home.—

38553.. 5883..
a seas» 2.. 5s. Bases. 8388 83:8 o... 8.8.85 .3 use“. 52.. 3:358 2859. £83.... 98.. 8.28.8 8. a... gem

 

age:
3 Ba . - 3 3» ea 5 "as: in 898m
3 8a . - men «a on «a use:
a men 3 ea 9.» an “as: 53.3...
a a - - .. - - - 2399
a sea ”.85: amusing
3» «a . - 5 ma ea an 235
an 5 a «a . - - - use: broom

 

58:. 320m :3:— osee< 53:. 328— :32 o>ee< e86 320m 53:. 269< SUE >83— 58:. esoe<

 

mom—oa— 352500 5:88.:— =03 .5202an— 8330580 cote—585 05.5QO 035...;

 

_ 233.50 8333> .850 .502 So: 2:330 ozﬁmoz
. een season ..82 a used; no 8:325 62. mes—=8”. 8:. gen 2 use.

 

1 1 1
with responses below the mean perceive goundwater pollution health risks to be less severe

compared to households with responses at or above the mean. The rest of the interactive
variables can be interpreted in a similar manner. The mean value for all positive (above the
mean) responses and all negative (below the mean) responses for each interactive variable
for each sample set is calculated. The positive and negative mean for each variable is
multiplied by the estimated coefﬁcient for that variable, and added or subtracted as
necessary to the mean policy bid when all variables are evaluated at their respective mean.
The indicated dollar value represents the policy bid of an household with above or below
average perceptions in terms of a particular variable, with all other variables evaluated at
their respective means.

For the rural sample, policy bids for respondents with above average severity
perceptions range from $52 to $67. For below average severity perceptions, policy bids
range from $28 to $48. For the urban sample, the range for respondents with above average
severity perceptions range from $89 to $93. For below average severity perceptions, bids
range from $56 to $61. In considering the scientiﬁc information, goundwater protection
and well protection policies, both the rural and urban samples value the groundwater
protection policy more than the other two policies. Above average severity perceptions for.
the urban sample result in signiﬁcantly geater policy bids compared with the rural sample
for all policies «cept well protection.

Knowledge perceptions, which were identiﬁed as an important underlying dimension
of risk through principal components analysis, are not generally of great economic
importance. The exception is for the combined policy for both the rural and urban samples.
Above average perceptions of the lack of knowledge regarding health risks results in an bid
increase for the rural sample and a bid decrease for the urban sample. In other words, the

greater rural households perceive that both the scientiﬁc community and «posed

 

 

 

112
households know little about the health risks from groundwater pollution, the higher the

policy value. For urban households, the geater the perception that little knowledge of
health risks «ist, the less value placed on policies. One interpretation of this relationship
is that urban households perceive that if the scientiﬁc community hasn’t established the
relationship between goundwater pollution and health eﬁects, then groundwater policies
are less valued. Conversely, as the people perceive the relationship to be known,
goundwater policies are valued more. I

The percentage increase in bids as a result of going from below to above average
perceptions can be quite signiﬁcant. For «ample, the mean bid in the rural sample for the
well protection policy increases 86 percent. For the rural sample, averaged across policies,
going from below to above average severity perceptions increased policy bids by 63 percent;
21 percent for the knowledge variable. For the urban sample, the increases are 58 percent
for the severity variable; 27 percent for the knowledge variable; and 57 percent for the q-
altruism variable.

Related to the use of risk descriptors ill providing a cont«t for valuation is the
design of risk communication progams with the «plicit use of empirically signiﬁcant risk
descriptors as a basis for formulating a communication progam. Aside from policy
evaluation studies, risk communication may be of importance in a variety of situations as
how to avoid or signiﬁcantly reduce certain types of risk; the provision of information to
households for assistance in voting on referendums dealing with policies which affect some
type of technological or environmental risk; and providing information to people for a
particular agencies elicitation of public comments on a proposed project or policy.

The risk descriptor severity is associated with two types of perceptions: (a)
perceptions of the effects of exposure and (b) perceptions of the likelihood of «posure.

The risk descriptor knowledge is associated with both of these perceptions. Consequently,

3C

113

in designing risk communication programs, the empirical results of the survey would seem
to indicate that an emphasis should be placed on communicating the risk within the context
deﬁned by the risk descriptors that have meaning to households. In this way, the multi-
dimensional aspects of risk can better be communicated, as opposed to reliance on some

uni—dimensional risk descriptor, such as probability of mortality from some level of exposure.

Significance of Q-altruism and Bequest Risk Perceptions
The empirical signiﬁcance of household perceptions that other reasons besides direct
health threats «ist for addressing goundwater pollution has a variety of valuation and

policy implications. These implications are discussed below.

Implications for Valuation

The empirical analysis showed that other reasons exist to protect goundwater quality
besides consideration of direct health effects to the household. Four implications are
evident. First, ignoring this aspect of household concern results in an insufficient amount
of pollution prevention. The theoretical framework showed that the marginal beneﬁts of
prevention will be understated if beneﬁts derived from altruistic concern are not included
in the policy evaluation, and households have some degee of altruistic concern regarding
the potential effects of goundwater pollution.

Second, goundwater policies will be biased toward remedial action. The theoretical
framework speciﬁed the necessary conditions for an efficient allocation of expenditures
across preventive and remedial policies. If values derived from altruistic concern are not
considered, an inefﬁcient allocation of expenditures will result.

Third, by ignoring altruistic based values, a given expenditure on protection results

in a smaller gain in well-being than if prevention and remedial action were efﬁciently

114
balanced. A given expenditure can result in a higher level of utility by reallocating

expenditures away from remedial action, toward prevention, until the marginal costs of
prevention are equated with its true marginal beneﬁts.

F ourth, a mortality-based estimate of the value of life may be sufﬁcient for valuation
purposes when altruistic-based values are absent. If people have altruistic concerns, then
sole reliance on mortality-based estimates of health beneﬁts will underestimate the true
marginal beneﬁts.

Impliations for Policy

Both q-altruism and bequest concerns are signiﬁcant «planatory variables for the
rural sample. Table 7.1 shows that for the rural sample, above average perceptions of q-
altruism result in bids ranging from $50 to $60. Below above average q-altruism perceptions
result in policy bids ranging from $40 to $53. For the urban sample, above average q-
altruism perceptions result in policy bids ranging from $82 to $92 with below average
perceptions resulting in bids ranging from $52 to $59. For bequest risk, above average
perceptions for the rural sample result in policies ranging from $67 to $51. Below average
perceptions result in bids ranging from $36 to $45. The average percentage increase in bids
from below to above average perceptions for the rural sample is 24 percent for q-altruism
and 50 percent for bequest risk For the urban sample, the average percentage increase
across policies for the q-altruism variable is 57 percent.

The main implications of the economic signiﬁcance of q—altruism and bequest
perceptions are: 1) that non-users (most of the urban sample) can still have a considerable
economic interest in the resource; and 2) actual users of the resource, (most of the rural
sample), can be motivated to pay for groundwater pollution policies for reasons besides any

actual or potential direct health risks.

 

115

perceptions result in bids ranging from $36 to $45. The average percentage increase in bids
from below to above average perceptions for the rural sample is 24 percent for q-altruism
and 50 percent for bequest risk. For the urban sample, the average percentage increase
across policies for the q-altruism variable is 57 percent.

The main implications of the economic signiﬁcance of q-altruism and bequest
perceptions are: 1) that non-users (most of the urban sample) can still have a considerable
economic interest in the resource; and 2) actual users of the resource, (most of the rural
sample), can be motivated to pay for goundwater pollution policies for reasons besides any
actual or potential direct health risks.

An additional policy implication of the signiﬁcance of existence values is as a
defensible way of giving weight to nonusers where it is deemed that such is necessary or
appropriate. A nongoundwater «ample is the observation that much of the criticism of
public land management agencies, such as the Forest Service and Bureau of Land
Management, for putting too much emphasis on development and not enough focus on
species preservation and the ecological integity of the land, comes from urban households
(Gresham, 1991). At least a part of this emphasis may be «plained by the importance of
q-altruism shown by the urban sample. This is not to suggest that urban households are to
be perceived as more environmentally conscious or operate at a higher ethical plane than
the rural set The important implication is that people who do not use the land directly, as
indicated by the empirical study, have an economically signiﬁcant interest in the land. While
this should not be interpreted as meaning that these concerns are overriding, the empirical
signiﬁcance does indicate that if these values or sentiments are ignored in policy valuation,
then, in an overall sense, too much emphasis will be placed on development and too little

on policies whose objectives are to provide for ecological integity and pollution control.

116

Future Research Needs

A consideration of the theoretical framework and the empirical ﬁndings results in
the identiﬁcation of a number of areas where additional research would aid in
understanding the economic signiﬁcance of risk and altruistic perceptions. Four general
areas for further research include: (a) identiﬁcation of risk variables which get closer to
underlying dimensions of risk and altruism; (b) measurement of risk and altruistic
perceptions and their effect on policy bids in different cont«ts; (c) ascertaining the role of
information in affecting risk and altruistic perceptions, and the resulting eﬁect on wtp; and
(d) a comparison of quantitative and qualitative risk measures, both in «plaining bid
variations and as information to give to consumers as an aid in thinking about the econonnic
trade-offs they would be willing to make in risky situations. The section below discusses

these items in more detail.

Risk Perception Variables

The identiﬁcation of risk perception variables used in the questionnaire were
primarily based on the previous research of Fischhoff et al. The proxy variables for altruism
were developed primarily from information gathered from focus groups. While the
regession analysis found that a signiﬁcant amount of bid variation is «plained by these
variables (and education and income), over 60 percent of the variation is left unexplained.
Further investigation into risk and altruistic perceptions may identify proxy variables which
more accurately reﬂect the underlying dimensions of risk and altruism, which aﬁ'ect people’s

motivations and, subsequently, their observable behavior in a given cont«t.

117
Perceptions in Different Contexts

The research considered in this study found that for a particular resource,
goundwater, risk and altruistic perceptions are important motivations behind policy bids.
An important question for policy valuation is how other types of environmental pollution
are perceived, and how these perceptions are related to policy bids. For «ample, the
economic signiﬁcance of perceptions may vary substantially across different types of
environmental pollution scenarios. Air pollution, surface water pollution, radon, ozone layer
depletion, and the geenhouse effect may each be perceived differently in the cont«t of risk
perception variables. Perhaps newness, control and voluntariness will be economically
signiﬁcant for some subset of pollution scenarios. One question to be answered in this
regard is the robustness of the empirical ﬁndings of the groundwater study. Severity and
knowledge were found to be the most signiﬁcant components of risk perception for the
groundwater study. This may, or may not, be the case in other situations or contexts.
Similarly, does the economic impact of risk and altruistic perceptions vary? For «ample,
the groundwater study found that severity and bequest risk were economically the most
signiﬁcant «planatory factor for the rural sample, while for the urban sample severity and
q-altruism were most signiﬁcant. It is conceivable and even «pected that different valuation

cont«ts will result in different perception components being the most economically
signiﬁcant.

Role of Information
The goundwater research did not provide information as such to survey
respondents. The objective was to obtain base-line measures of risk and altruistic

perceptions and demogaphic characteristics. The basic objective of additional research

118

would be to ascertain the «tent to which perceptions, and ultimately, wtp are affected by
information. The approach to take would depend on the purpose for looking at the irnnpacts
of information. For «ample. is the objective of irnformation provision to measure existing
perceptions, or is it «plicitly to change observable behavior? The latter objective is similar
to Smith’s study (1986) of the effect of command and cajole information types on mitigating
behavior of people potentially exposed to radon pollution. As an «tension of the current
work, the effect of different types of. information (amount, command, cajole, etc.) on risk
and altruistic perceptions, as measured by severity, knowledge, q-altruism and bequest risk
and in turn policy wtp can be considered. A useful result might be to measure the
sensitivity of the economic trade-offs people are willing to make to the type and amount of
information given them. This addresses the question ill the Bayesian context of the strength
of people’s priors. If risk and altruistic perceptions are a weighted function of priors and
new information, a pertinent policy question is how these elements are weighted, and are

they stable over time or are they dynamic?

QuantitativeandQualitativeRiskMeasures

Related to irnformation is the respective roles both quantitative and qualitative risk
measures play in aﬁecting individual’s risk perceptions. These risk measures may be used
t6 assist individuals in the policy evaluation process (including wtp), or as speciﬁc

information whose purpose is to change behavior in some fashion. The economic
Sigraiﬁcance of these risk measures, as measured by their effect on policy wtp, would be an
l'l'l‘l[><>r'tant consideration in policy design and evaluation. More speciﬁcally, two topics seem
especially important: (a) the determination of whether there is a systematic relationship

 

 

119
between quantitative risk measures and wtp for different types of policy, and (b) the

importance or weight people attribute to quantitative and qualitative risk measures.

APPENDICES

APPENDIX A

APPENDIX A

SURVEY DESIGN AND IMPLEMENTATION

The goundwater survey design phase had two objectives: a) to provide
respondents with a straightforward, understandable questionnaire, and b) to increase. the
likelihood that the survey would be completed in a usable manner and returned. The
basic survey design and implementation procedure generally followed the "total design
method [TDM]" (Dillman, 1978). Dillman states

'... to maximize both the quantity and quality of responses, attention must

be given to every detail that might affect response behavior. The TDM

relies on a theoretically based view of why people do and do not respond

to questionnaires and a well conﬁrmed belief that attention to

administrative details is essential to conducting successful surveys“ (p.

viii).

Survey design and implementation can be addressed in three components: a)

question design, b) questionnaire design, and c) implementation of the mail survey.

Question Design
Dillman (1978) identifies two major considerations in designing a question; the
kind of information being sought, and the question structure.
mum Four general types of information are identiﬁed (p.80):
attitudes, beliefs, behavior and attributes. Attitudes reﬂect a person’s normative

assessment of a subject or situation, i.e., what should be or what should not be. Beliefs

120

121
are what people believe to be true or false, or what is and is not (p. 81). In this sense it

is the person’s positive assessment of a subject or situation. Behavioral questions
address what people have done in the past, what they are currently doing or what they
plan to do in the future (p. 83). Finally, attribute questions address personal or
demogaphic characteristies (p. 83).

An important reason for distinguishing between these different types is that the
type of information obtained from a question depends on the wording of the question.
In rewriting questions to increase clarity and comprehension, there is increased
possibility that the type of information obtained from a question may be different then
intended (Dillman, 1978). Consequently, survey information objectives need to be
precisely identified so that changes in question wording do not change the type of
information sought.

Previous sections of this chapter have identiﬁed five types of information
variables included in the goundwater survey: risk perception questions, held value
questions, CV questions and demogaphic questions. These variables an be correlated
to the four question types listed above. Risk perception questions are examples of belief
questions. Held value questions are examples of attitude questions. CV questions are

examples of behavioral questions. Finally, demogaphic questions are examples of
a ttribute questions.

Wm. Dillman identifies four stnnctures based on the response
behavior required to answer the question. These structures are characterized as: a)
Open-ended, b) close-ended with ordered choices, c) close-ended with unordered
1' eernnse choices, and d) partially close-ended (p. 86-87). Open-ended questions require

the respondent to provide an answer in their own words A close-ended questionnaire

With ordered choices provides the set of choices on a continuum which represents a

122
single dimension or characteristic. The respondent then chooses the most appropriate
response.

Close-ended questions with unordered choices refers to choices which are not _
gadations on a scale or continuum. The choices typically represent entirely different
concepts or alternatives (p. 90). A partially close-ended questionnaire combines a set of
ordered or unordered choices with an open-ended response.

For the groundwater survey, close-ended questions were chosen as the primary
question structure. An important rationale in using close-ended questions is that they
are typically less demanding in respondent effort than other types. This was an
important consideration in the goundwater survey, since the focus goup sessions made
apparent the lack of familiarity and experience with groundwater pollution issues.

Close-ended questions with ordered choices were developed to elicit information
on risk perception variables, held values, and demographic values. Close-ended
questions with unordered choices were developed for demographic and held values.
Open ended questions were developed for policy valuation questions.

QuestknnnalreStn-uctnn

Two design considerations for the questionnaire as a whole are length and
physical layout of the questionnaire.

Deciding on questionnaire length involves a trade-off between the amount of
information obtained and the likelihood that the questionnaire will be returned, or that
answers to questions in the later portion of the questionnaire will be answered (Dillman,
1978). Surveys which have used the TDM show a negative correlation between response.
rates and questionnaire length (Dillman, p. 55-56, 1978). However, even for a 14-page

questionnaire, the average response rate was 62 percent (Dillman, p. 56). While these

123
above results are based on a limited number of studies, it appears that lengthy mail

surveys can be developed which result in acceptable return rates.

The groundwater questionnaire ﬁnally developed was 14 pages long with a
minimum of 49 and a maximum of 59 responses required (several questions provided for
multiple responses). While the questionnaire was long, the number of required
responses was relatively small in comparison with the studies cited by Dillman (p. 56). It
was decided that the length of the questionnaire would not be a significant factor in
affecting response rates. The in-person and mail pretests conﬁrmed this initial
assessment. In the debriefing portions of the inn-person pretest, a majority of the
participants stated that the questionnaire length did not seem unduly long. The results
from the mail pretest were consistent with these ﬁndings. The response rate of the mail
pretest was 45 percent after the ﬁrst mailing, and a postcard reminder a week later.
This compares with several surveys of the general public in four states whose response
rates ranged from 34 to 62 percent after a similar number of mailings (Dillman et al.,
1974). It should be stressed that these surveys were not CV surveys. Given the
cognitive demands made on respondents in CV valuation scenarios, plus the unfamiliarity
with goundwater pollution issues, the mail pretest showed that the response rate was,
nevertheless, competitive with other surveys using state-of-the-art mail survey
procedures. The basic objective of the physical layout of the questionnaire is to motivate
the respondent to answer, and return the questionnaire. To accomplish this, the
questionnaire should be constructed so that its various components combine to form an
appealing package with the following characteristics: a) visually appealing, b) conveys to
the respondent the importance of the survey and their response, and c) gives impression
that completing the survey will not take much time (Dillman, 1978). A variety of design

124

elements should be considered for each of the three main components of the survey: a)

the questionnaire itself, b) mailing envelopes, and c) postage.

mm The major design elements to be considered include size, shape,
weight, color, paper quality, cover design and general layout (Dillman, 1978). The size of
the goundwater questionnaire is 8 1/2' by 7', which is legal size paper folded in half.
The pages are formed into an 8-page booklet, printed on both sides, for a total of 16
printed pages.

This reduced size booklet form has two advantages: a) cost saving by reducing
the amount of paper needed, and lighter postage; and b) the reduced size gives the
impression that the questionnaire is shorter than it actually is.

Wlnite, legal size 60-pound bond paper was used in printing the questionnaire.
The lighter 16- and 20-pound bond was insufficiently opaque with two-sided printing.
The increased cost of using 60'pound bond was more than balanced out by the higher
quality look and feel of the questionnaire, using the heavier paper. The questionnaire
was commercially printed, using high quality photo-copying processes.

The physical layout of the questionnaire was designed to provide a visually
appealing format to respondents. Questions were structured so that a large amount of
“white space” (blank space) was used both within an individual question and between
questions (See Appendix C.). The judicious use of white space provides a degee of
visual relief, in comparison with pages of condensed text. Too much white space gives
the appearance of an unprofessional and, incomplete questionnaire, while too little results
in an imposing looking questionnaire which may not get answered (Dillman, 1978).

The objective of the cover page of the questionnaire is to convey the subject

matter of the survey, and to make it sound interesting (Dillman, 1978). The cover page

125

has four components: a) the title, b) a graphic illustration, c) directions, and d) name
and address of the researchers (Dillman, 1978). Figure A-l shows the cover page from
the goundwater survey. The title is shown against a dark backgound, which provides a
visual contrast with the rest of the cover page. The ﬁrst sentence states what the survey
is about. The second tells the respondent that their opinion is important and that their
privacy will be ensured. The third sentence gives a name and address for returning the
survey. The fourth, and last, sentence thanks the respondent for their time and effort.
Constructing the cover page in this manner increases the likelihood the questionnaire is
taken seriously by the respondent and returned.

Figure A-2 shows the last page of the questionnaire which elicits additional
comments from respondents and thanks them for their time and effort.

M. The cover letter is typically the ﬁrst component of the survey
package that is examined by respondents (Dillman 1978). The cover letter has four
components: a) to tell people what the study is about, b) to convince them that the
study is important, c) to convince them that their answers are important to the success
of the study, and d) that their identity and answers will be kept in conﬁdence. Figure A-
3 shows the cover letter used for the ﬁrst mailing of the goundwater survey. The ﬁrst
paragraph speciﬁes what the topic is, and why the research study is being done.

The second paragaph explains that the respondents’ opinions and views are

important to the success of the study.

126

 

 

 

 

 

 

 

Michigan’s Groundwater Choices is a project designed
to ﬁnd out what' Michigan citizens think about
goundwater pollution.

Your opinion is what’s important to us. To ensure
your privacy, the questionnaire is identiﬁed by number
only.

When completed, place the questionnaire in the
business reply envelope and send it to:

James Caudill, Project Director
Department of Agricultural Economics
Michigan State University

East Lansing, MI 48824

We appreciate your time and effort. Thank you!

wnmwmum

 

Figure A-l: Front Cover of Questionnaire

 

 

127

 

 

.C‘IOCOOOO

Thankyoovaymuchforyomhelpandcooperation. Youropinioooo
eachquestionconmtsageatdeal.

Hyouwouldﬁketosharemyadditiondcommentapleasemiteonthis
pasc-

 

Figure A-2: Back Cover of Questionnaire

 

 

128

The third paragraph gives instructions on who is to ﬁll out the questionnaire, and
the length of time it should take to complete the questionnaire.

The fourth paragraph speciﬁes that the survey results will be shared with
governmental agencies and the general public, and how respondents can obtain copies of
the results.

’The ﬁfth paragraph addresses the issue of conﬁdentiality, and that they are
participating in the survey voluntarily (statement required by university regulations).

Finally, the last paragaph provides a phone number for people to call if they
have any questions about the survey. Cover letters were printed on official
departmental stationary with a water-marked university seal.

Cover letters for the second and third questionnaire mailouts will be discussed in
the following section on mail implementation procedures.

When. Two types of questionnaire mailing envelopes are needed.
Mailout envelopes are used to mail the questionnaire to households in the sample.
Return envelopes are used by respondents to mail back the questionnaire upon
completion.

The mailout envelopes are 9 1/2' by 12 1/2' white catalog envelopes with a
gummed flap and no clasp. A departmental return address label was centered on the
front of the envelope. Gummed labels with names and addresses of people in the
sample were then attached to the return address label. The large size of the mailout

envelope, compared with the size of the questionnaire, was necessitated by the

129

August 21, 1990

Name
Address

City, State, Zip
Dear Name:

Midniganfacesmajorgoundwaterdecisions. T'lnesededsionswillslnapeMidn'gsn’sfutureas
splacetoliveandwork. Rigiunow,thereisverylittleinformationsvailableabwtthegoals

Yourhouselnoldisoneofasmallnumberwhiclnisbeingukedtoglveanopinionon
Michigan's groundwater choices. Households were scientiﬁcally selected to give a balanced
andacwratepiaureofstatewidecitizenopinion. Youranswerstotheenclosed
qrestionnairewillensureatruly representative report.

Thequestionnaireslnouldbefilledoutbyanachnltinyourhouseholdwhosharesinmajor
householddecisions. Byhouuholdwemeanthepeopleinyourhomewhoactasafamily
unitandwhoshareallmajorpurclnuededsions. lfyouonlyslnarerentormortgagecosts
withamommatecompletethequeuionnsireviavingyourselfuthehousehold The
mestionnairewilltakeaboutﬁormminutestoconmlete.

Thereaduofthisruarchwiubednaredwithrepruemaﬁvuinourwdsgwmem
localoffidalaandsllinterestedcitizens Youmsyobtainasummaryofresultsbywriting
'copyofresultsrequested’onthebackoftheretumenvelope,andprintingyournameand
addressbelov'n.

Youmsybeassuredofcompleteconﬁdentiality. Themestionnairelnasanidantiﬁcation

numberformailingpurposesonly. Youindicateyourvoluntaryag‘eementtoparticlpatein
thisreseardnprojectbycompletingandretumingtlnisquestionnaire.

Iwouldbelnappytoanssersnyquestionstlnatyoumighthsve. Pleasefeelfreetowriteor
call My telephone number is 511355-83.

Thankyouforyourlnelp.

Sincerely,

James Caudill.
Projea Director

Figure A-3: Cover Letter for Initial Mail-out

130

desire not to fold the cover letter or questionnaire. To enhance the convenience of
returning the questionnaire, the mailout package was designed so that as it was opened,
the unfolded cover letter would be encountered ﬁrst, then the questionnaire and then the
return envelope. Keeping the cover letter unfolded would, hopefully, decrease the
likelihood that the cover letter would be initially bypassed and the questionnaire perused
before the cover letter is read.

The return envelopes were business reply envelopes printed with the researchers’
name and address. The size of the return envelopes was the same as the mailout
envelopes Again, the rationale is to make it as convenient as possible for the
respondent to return the questionnaire. The use of large envelopes permitted the
respondent simply to place the questionnaire in the return envelope without having to
fold it.

A cost advantage of business reply envelopes is that postage costs are charged
only when the envelope is returned.

W. The questionnaires were mailed using ﬁrst class postage. Using ﬁrst
class postage has two advantages: a) it conveys an image of importance that less
expensive postage does not, and b) ﬁrst class mail receives a high handling priority from
the postal service. Additionally, ﬁrst class mail is automatically forwarded for up to a
year in case of an address change (Dillman, 1978). The goundwater questionnaires
were mailed using ﬁrst class, 25 cent stamps. The mailout envelopes required three
stamps each. First class business reply envelopes cost 53 cents per returned

questionnaire.

Su

1h

13 1
Survey Implementation

The survey implementation has two components: the characteristics of the survey

sample and the logistics of the mailout.

Survey Sample

Since little information exists on how Michigan citizens perceive goundwater
pollution issues, a state-wide survey would have the advantage of gathering opinions from
a cross section of state households. In this way, baseline information could be compiled
on consumer perceptions of gomWater pollution issues.

The population to be sampled included heads of households in Michigan. The
sample size was 2020. This was divided into two sub-samples, urban and rural. Using
the U.S. Bureau of Census deﬁnition for metropolitan statistical area (MSA), Michigan
counties were divided into urban counties (counties which included an MSA), and rural
counties (counties which did not include an MSA). Two-thirds of the total sample came
from rural counties, one-third came from urban counties.

Names and addresses were obtained from a marketing research ﬁrm in Lansing,
Michigan. The names were on a national computerized database which contained names.
from all ﬁfty states. The Michigan database in June 1990 contained 3,151,507
households. This is 98 percent of the total households listed in the most current census
(ref).

There are two sources of names and addresses: a) telephone directories, and b)
auto registration records. A sample of names and addresses are periodically checked for
accuracy, and updated by examining auto registration records and national change of

address records. Awarding to the company, the result is that 92-93 percent of names

132

and addresses contained in the database are current and up to date (Personal
communication Dan Braun, Beurman-Marshall Corporation, East Lansing. MI).

A systematic random sampling of the Michigan database of every kth name, for

both rural and urban counties, was undertaken by the marketing ﬁrm. Six thousand

names were selected. These names were transferred to a 3-1/2-inch microdisk to be

used in the dissertation research. Subsequently, a systematic sampling of both name sets

obtained 1347 names from rural counties and 673 from urban counties, for a total of

2020 names and addresses in the survey sample.

Mallout Logistics

The T'DM requires an initial mailout of the questionnaire with a minimum of
three follow-ups (Dillman, 1978). The follow-ups increase the likelihood that a high
response rate is obtained from the sample.

The initial mailout occurred on Tuesday, 31 July, 1990. Tuesday is preferred as a
mailout day for three reasons: a) it is early enougln in the week for questionnaires to be
forwarded to people who have moved and still receive the questionnaire that same week;
b) the buildup of weekend mail results in Monday being the busiest day of the week,
potentially increasing the likelihood of clerical errors; and c) for the follow-ups, mailing
on Tuesday allows time for possible returns from the weekend mail to be acknowledged,
thus decreasing the number of follow-ups needed to be mailed on Tuesday (Dillman,
1 978),

One week after the mailout, Tuesday, August 7, all households in the survey were

Sent a postcard follow-up. The basic objective of the postcard is to remirnd people about
the questionnaire and the importance of their responses to the study. Dillman states,

I
the post card [sic] follow-up is not written to overcome resistance, but to jog memories

133

and rearrange priorities' (1978, p. 183). The timing of the postcard follow-up attempts
t o convey a sense of importance while not appearing impatient or strident (Dillman,
1 978).

Figure A-4 shows the text of the postcard. The ﬁrst paragaph reminds the
addressee that a questionnaire was sent the previous week, and that they were part of a
scientiﬁcally designed sample of Michigan households.

The second paragaph thanks the person if they have already returned the
questionnaire. If the questionnaire has not been returned, they are reminded about the
importance of their opinions to the success of the research project.

Finally, the third paragaph requests people to call immediately for a replacement

questionnaire, if needed.

 

August 7, 1990

lastweekaquestionnairewasmailedtoyouseekingyourhousehold’sopiniononkey
groundwater issues facing our state. Your household was selected in a scientiﬁcally

designed sample of all Michigan households.

Hyouhaveakadycomplaedandraumedthequecionnaintouaphaseaeceptour
sincerethanks lfnot,pleasedosotoday. Becausetlnestudyinvolvesasmall,
repruentuivemmpkofMidnignreddentankenranelyhnponamthuywrmbe
includediftheresultsareto acalrately representtheopinionsofMidnigancitizens.

lib/somechanceyoudidnotreeeivethequestionnaireoritgotmisplaced,pleasecallme
right now, collect (517.355.2320), and 1 will get another one in the mail to you today.

Sincerely,

James Caudill
Project Director

 

 

 

Figure A-4: Post Card Reminder

134
The second follow-up was sent on Tuesday, 21 August. This was two weeks after

t he postcard follow-up and three weeks after the initial mailout. The purpose of this
f0 llow-up is to convey a more urgent tone than the previous follow-up and to give a .
personal appeal to the reader that their opinions are vital to the success of the survey.
A replacement questionnaire is included.
Figure A-5 shows the text of the second follow-up letter. While the information
in the letter is essentially the same as in the previous cover letter and postcard, a geater

emphasis is placed on personalizing the message.

 

The ﬁrst paragaph addresses the fact that the questionnaire has not been
received. The second paragaph gives a rationale for seeking their opinion in the ﬁrst
place. The third paragaph again emphasizes the importance of their response to the
success of the research. The last three paragraphs respectively: a) give directions on
ﬁlling out the questionnaire, b) address the conﬁdentiality of the questionnaire, and c)
provide an address and telephone number to contact.

The third and ﬁnal follow-up was sent on 18 September. A replacement survey
was included. The tone of this follow-up is similar to the previous follow-ups but is
shorter and more direct irn emphasizing the importance of the readers opinions. Figure
A.6 shows the text of the third follow-up cover letter. The second paragaph, in
particular, shows the emphasis on personalizing the message. For example, the second
and third sentences tell the reader that the accuracy of the survey ﬁnding depends on
their responses, and that their possrhly unique views on goundwater pollution issues are

important to know for the survey.

135

Date (planned mailing date. 3 wks. after lst mailing)

Name
Address

City, State Zip
Dear Name:

About threeweeksagolwrotetoyouseekingyouropiniononmajorgoundwuerisnies
facingourstate. Asol’todsy,wehavenotyetreceivedyourcompletedtpestionnaire.

Ourresearclnunit hasundenakenthisstudybecauseofthebeliefthatdtizenopinionsshould
be taken into account in the formation of public policies about goundwater.

I am writing to you again because of the importance of your response. Your household was

selectedaspartofasrnalLrepresentativesample ofMichigancitizens. Forourstudytotruly
represent citizen opinion, it is essential that you complete and return the questionnaire.

lhaveendosedarephcunemqueaionnaireincasetheoﬂgndhasbeenmisphced The
questionnaire should be ﬁlled out by an adnlt in your household who shares in major
household decisions By household we mean the people in your home who act as a family
unitandwhoshareallmajorpurchasedecisions. lfyouonlysharerentormortgagecosts
with a roommate, complete the questionnaire viewingyourself as the household.

Please remember that your opinion is conﬁdential. Your name and individnal opinion will
not be revealed You indicate your voluntary agreement to participate in this research
project by completing and returning this questionnaire.

If you have any questions about filling out the questionnaire, please feel free to write or call.
My telephone number is 517-355-2320.

Thankyouforyourcooperation.
Cordially,

James Caudill
Project Director

Figure A-5: Second F ollow-up Letter

 

136

Date (planned mailing date, 7 wks after 1st mailing)

Name
Address

City, State Zip
Dear Name:

lamwritingtoyouaboutourstudyofgoundwaterisnies. Wehavenotyetreceivedyour
completedquestionnaire.

The large number of questionnaires returned is very encouraging. However, whether we will
beabletoacarratelydescnbehowMichigancitizens feelontheseissuesdepenthuponyou.
You may have quite different opinions concerning goundwater issues than those who have

alreadyresponded.
Thnnthefmﬁmawwimsdentiﬁallydedglednudyofdtizenopinimhubeen
doneontheseissues. Theresultsareofpartianlarimponancetothemarnycit’nens,private
mdpubﬁcagandeamdhwmakerswhoamconcanedabmumakinggpundwaterdedsiom
thatbestmeettheneedsofMichigan’speople.

Itisforthesereasonstlnatlamsendingyouanothercopyofourquestionnairetoyour
household The questionnaire should be ﬁlled out by an adtlt in your household who shares
in major household decisions.

l'llbelnappytosendyouacopyoftheresultsofourstudy. Simplywriteyourname,
addresaand'copyofrewlurequested‘onthebackoftheretumenvelope.

Your contribution to the succeu of this study will be geatly appreciated.

Sincerely,

James Caudill
Project Director

Figure A-6: Third Follow-up Letter

 

APPENDIX B

 

APPENDIX B

QUESTIONNAIRE TEXT

 

 

 

 

 

 

Mlelnigan'sGromdwa’tarChoices is a project designed to find out what Michigan
cklzons think abort groundwater pollution.

Youopiniornlswtnat‘s important to us. To ensureyourprlvacy. ttnequewonnalre
isldertliedbynumbsronly.

When completed. place the questionnaire in the business reply envelope and sand
I to:

James Caudll. Project Director
Department of Agricdturai Economics
Michigan State University
East Lansing, MI 48824

We appreclue your time and effort Thank you!

mistaking-mum

 

137

 

 

138

Michigan’s Groundwater

Groundwater is water that collects below the surface of the land in porous rocks and sands. Twenty-
cight percent of all Michigan households get their water from goundwater through household wells. 7
Fifty-seven percent of the cities and towns in Michigan use goundwater for part of their water supply.

Where does the tap or faucet water in your home come from? (circle one)

1. A COMMUNITY OR PUBLIC WATER SYSTEM
2. A HOUSEHOLD WELL

3. A SPRING, STREAM, OR RIVER

4. SOME OTHER SOURCE

What do you usually use for drinking water in your home? (circle one)

TAP OR FAUCET WATER

PURCHASED BOTTLED WATER

BOTTLED WATER FROM A FRIEND OR NEIGHBOR
WATER FROM SOME OTHER SOURCE

#ri

In your judgement, how would you rate the tap or faucet water in your home? (circle one)

VERY SAFE TO DRINK
SOMEWHAT SAFE TO DRINK
NOT SURE ABOUT SAFETY
SOMEWHAT UNSAFE

VERY UNSAFE

vswwr

Withinthelastyear,hasthetaporfaucetwaterinyourhomebeentestedforsafetybya
laboratory? (circle one)

1. NO

2. YES
How conﬁdent are you that your tap or faucet water is safe from the eﬁects of goundwater
pollution? (circle one)

VERY CONFIDENT
CONFIDENT

NOT CONFIDENT

NOT CONFIDENT AT ALL

pwwr

 

all that App”)

999Nr

county? (circle one)

vrwwr

experts? (circle one)

9993“!"

pollution? (circle one)

P§FNP

139

During the last year, have you read or seen any news stories about groundwater pollution? (circle

NO

YES, ON TELEVISION

YES, IN A NEWSPAPER

YES, IN A MAGAZINE

YES, SOME OTHER NEWS SOURCE

In your judgement, how serious at health threat is the current level of groundwater pollution in your

VERY SERIOUS
SERIOUS

SOMEWHAT SERIOUS
NOT SERIOUS

NOT A THREAT AT ALI.

‘I‘owhatextentdoyouthinkthehealthrisksofgrounMerpollutionareknownbyscientiﬁc

KNOWN PRECISELY

KNOWN

SOMEWHAT KNOWN

ONLY SUSPECI'ED OR GUESSED AT
NOT KNOWN AT ALL

'I‘owhatenentdoyouthinkthehealthrisksareknownbypeoplewhoareexpoeedtogroundwater

KNOWN PRECISELY

KNOWN

SOMEWHAT KNOWN

ONLY SUSPECTED OR GUESSED AT
NOT KNOWN AT ALL

10. Do you View the risks of groundwater pollution as a new environmental problem or something

rather old? (any; one)

MfPNP

NEW

SOMEWHAT NEW

MIDWAY BETWEEN NEW AND OLD
SOMEWHAT OLD

OLD

._.. A

11.

12.

14.

140

In your judgement, is groundwater quality a good or poor indicator of the general quality of the

environment? (circle one)

ppwwr

A GOOD, GENERAL INDICATOR
SOMEWHAT GOOD

AVERAGE

SOMEWHAT POOR

A POOR INDICATOR

Doyouviewthehealthrisksofgroundwaterpoﬂutionasinjuﬁngpeopleoneatatimeorahrge
numberofpeopleat once,likeacatastrophe? (circle one)

M§PNF

ONE AT A TIME

A FEW PEOPLE AT ONCE

MANY PEOPLE AT ONCE

VERY MANY PEOPLE AT ONCE

A VERY LARGE NUMBER AT ONCE

Where groundwater pollution occurs, do you believe the damage can be controlled through

technology? (circle one)

Mewwr

EASY TO CONTROL

POSSIBLE TO CONTROL

SOMEWHAT POSSIBLE TO CONTROL
VERY DIFFICULT TO CONTROL
NOT POSSIBLE TO CONTROL

Doyouviewtherisksofgroundwaterpoﬂutionassomethingthatpeopleareexposedtovoluntarily

or involuntarily’? (circle one)

PPI”?

VOLUNTARILY
SOMEWHAT VOLUNTARILY
SOMEWHAT INVOLUNTARY
INVOLUNTARY

Is the risk of groundwater pollution in Michigan something that you greatly fear? (circle one)

1.

FEAR GREATLY

2. FEAR SOMIEWHAT

3.
4.

mewmywmm
memnxum.

 

141

16. For a person exposed by drinking water to typical levels of groundwater pollution in Michigan, are
health effects likely to occur soon after exposure or at some later time? (circle one)

M§PNP

NO HEALTH EFFECTS ARE LIKELY
SOON AFTER EXPOSURE
SOMETIME WITHIN A YEAR
SOMETIME WITHIN A DECADE
SOMETIME WITHIN A LIFETME

17. In your best judgement, what health eﬁ'ects are likely to result from typical levels of groundwater
pollution in drinking water in Michigan? (circle all that apply)

pesovewwr

rmnmuﬂmﬁwmumumwv
mummm

nmnummms
numnmmmmmsme

omam
mmmmommammmnmamwum
Pwomummuommmms
srowxcn, INTES'I'INAL DISORDERS
OTHER (511 in)

 

18. Are current levels of groundwater pollution in Michigan likely to be fatal or not fatal to people?

(circle one)

vrwwr

VERY LIKELY TO BE FATAL
LIKELY TO BE FATAL
SOMETIMES FATAL
UNLIKELY TO BE FATAL
NOT FATAL

19. In your judgement, are future generations likely to be harmed by the groundwater pollution that
occursinMichiganoverthenextStolOyearchircleone)

M§PNP

VERY LIKELY TO BE HARMED
LIKELY TO BE HARMED

NOT LIKELY TO BE HARMED
VERY UNLIKELY TO BE HARMED
NOT HARMEDATALL

20- InYmn‘judgement,howseriousisthecurrentlevelofgroundwaterpollutioninMichiganasa
health threat to wildlife and plants? (circle one)

wavwr

VERY SERIOUS
SERIOUS

SOMEWHAT SERIOUS
NOT SERIOUS

NOT A THREAT AT ALL

 

142

21. Some people say that groundwater should be protected from pollution for its own sake. That is,
groundwater should be free from pollution whether or not people, plants, and wildlife are aﬁected.

Other people say that goundwater should be protected from pollution only because it is part of a
natural system-4f pollution affects groundwater it will eventually aﬂect human beings or other living
tma

Which opinion best describes your own? (circle best one)

1. DISAGREE WITH BOTH OPINIONS

2. PROTECT GROUNDWATER FOR 173 OWN 3.410:

3. PROTECT GROUNDWATER BECAUSE IT IS PART OF
A NATURAL SYSTEM

4. DON’T KNOW

22. Some people say that pollution is bad because it is simply the mong thing to do. Other people say
thatpollution isbadonlywhenithasbadcﬁ'ect: onpeople,wildlife,andotherlivingornon-living
things Which opinion best describes your own? (circle one)

1. DISAGREE WITH BOTH OPINIONS

2. POLLUTION IS THE WRONG THING TO DO

3. POLLUTION IS BAD ONLY “’HEN IT HAS BAD
EFFECTS

4. DON’T KNOW

23. Some people say that groundwater should be protected from pollution only when the pollution
pmgsadirect health threat tohuman beings. Doyouagreeordisagreewiththisopinion? (circle
one

STRONGLY AGREE

AGREE SOMEWHAT

DISAGREE SOMEWHAT

STRONGLY DISAGREE

DON'T KNOW

vawwr

143

24. ThetablebdowshoaesonedmemttupeopledthMlcNoanuchyear. It
slmhwmanypeopledleandmedeathratepamﬂbnfulsaaumponedbyme

Michigan Department of Health.

RISK

CATEGORY
Heart Disease
All Cancers
Pneumonia
Motor Vehicle Accidents
Breast Cancer (Female)
5mm
Leukemia

' Fire
Aethma'
Inﬂuenza
Nutritional Deﬁcienciee
Gastric Ulcer
Aircraft Accidents
Tuberculoeie
Viral Hepatitis
Breast Cancer (Male)
Railway Accidents
Handgun Accident!
DOWN
Chicken Pox

NUMBER OF RATE OF
DEATHS DEATH

33.521 3.”
17.69 1,914
2.354 255
1.725 187
1.573 170
748 81

847 70

1Q 18

1Q 18

142 15

123 13

101 11

0.3
0.2
0.1

enumSBESSB
##00000

a. Consideranskcetegorycalled'deathefromo’wldweterpolwon' Thlecategorywodd
WWWMMWBNWWhaM lnyarbeetludgemem-
mbeetguees-wlmewoddthegmldwaterrlskcetegayflhthetahle? (clrcleone)

{

4.

99!“?

I AM VERY SURE WHERE IT WOULD FIT.

I CAN ONLY GUESS WHERE IT WOULD FIT.
GROUNDWATER POLLUTION IS NOT UKELY TO BE A
DECIDING FACTOR IN ANY DEATHS-IT WOULD NOT BE
IN THE TABLE.

DON’T KNOW.

b. lflor2clrcled: Drawallnelromlentondletralghtecroeethetebletoehowyourbest
guessabmnwherethegrmndwaterrlekcetegorywmldm

QWZSJIandalaskyoutovotefororaoalnetthreedﬁererltormmwaterpfoommc
Cons“breachprogramseparately. Itapproved.aproommwoudbecomeapermenentpertot

stateorcountyoavemm

25~Th031medehbanmeedngaspecldproqammg¢mscbmﬂlcﬂmnatbnm
(”WhereorwuwmdpolwmoccunanMmmnypeopleammedwpposem
on'W'IIytol‘lnencethlsprogramwerethrouol'lllllpecldtax.

“hologramcaldbeputoavote. Itthlsplogramcoetyouhouseholdﬂpermondnor
I"0WPayment01812peryear,worldyouvotel‘ororegnlnetthleprogramandtax? (circle

one)

1.

VOTE FOR THE PROGRAM AND TAX

2. VOTE AGAINST THE PROGRAM AND TAX

144
26. Whabﬂnnmetyouwmldpeymwyemhhlolwmxeetowpponﬂwprogmmdescnbed
lnqueetlon25? (fllhanarrmltemaltoorgreeterthanzero).
IWOULD PAY UP TO s PERYEAR

r W v

27. Suppoeeaspechlprooramwempropoeedfaywcmnymatwaldpraeagmuumw
frompolllnlon.

TheprogmmwoudprManyhmthupoltmonhmcmny. It
wodd:
1. Identifysourceeofaotmerpcllwon

2. Take educationd. momendleoﬂactlontopreventhnurepollutlonof
groundwater.

SuppoeethlsMo—panprogramwereputoavote. Worldyouvotefororaoalnstthecounty-
wideprooramandtaxllcosryourhoueeholdﬂ permonthforatotal paymerlofSlZper

yeaf? (circle one)

1. VOTE FORTHE PROGRAM AND TAX
2. VOTE AGAINST THE PROGRAM AND TAX

28. Whubmmoetyouwoudpeymdyemhmmmetpponmeprogramdeecnmd
lnqueetlonzr? (lillnanamountequdtoorweeterthanzero).

I WOULD PAY UP TO S PER YEAR

 

27. mmwbmmmmwmammupwm
”Windearenotmyteeted. Suppoeeaspeclelprogramweresetuphyour
Mtoprotectthehedmapeoplewhomehweeholdwella

mmmummmmhwmmmmwmapwmtopeople
memhmcmm. ltwoud:
1. Teetdhmnellddweﬂshﬂlemlyformmchemlcale. Eechwellwoud
beteetedonceeyear.

2. Immmlammedbyamlddtommanyramﬂu
chemlcelethuareddected.

3. Envoytralnedtechnlclamtomonltorand trialrltalnthellterlngdeviceslntop
workingcondltlon

28.

145

Suppoeethisthree-partprooramwerepmtoavote. Worldyouvotefororaoahstthecounty-
wideprooramandtaxllcostyourhoweholdﬂ pernronﬂwloratotd paymentol$12per
yea/I (circleone)

1. VOTE FOR THE PROGRAM AND TAX
2. VOTE AGAINST THE PROGRAM AND TAX

Whmbmenwetyouwmidpeymayeenhmmeemwpponﬂoproommdescnbed
lnqueetlon27?(ﬁlllnanamourueq1u100rgrenterthanzero).

I WOULD PAY UP TO S PER YEAR

lnmludmtwoddﬂuprmnhqmeﬂonﬂnducethehedﬂlwpoeedby
groundwaterpollutlonhyoucwrm Bthnuchwoudthehedththreetreaﬂybe
reduced?(cirdeone)

1. WTO1mPEFCENT 8. 40TO40PEFCENT
2. ”TOﬂPEmENT 7. aoroaepencem
3. roronpencem 8. 207an
4. ”TOﬂPEmENT 9. 10 TO “RECENT
5. SOTOGBPEmENT 10. oroepencem

lnywludoenmwmiddnprooamhqneﬂmﬂpraeamhmcmmm
hlrtherpolwon? Byhovnuchwoudhmorotmweterpolwonreelyhereducedﬂclrde
one)

1. 90701001590941 8. 40TO40PEFCENT
2. ”TOQPEICENT 7. ”T03PEFCENT
3. TOTOTGPEFCENT 8. aorozepeacem
4. ”TOGPEmENT 0. 10TO19PEFCENT
5. SOTOSOPEIDENT 10. OTOOPEFCENT

 

-Th°prooramducrlbedhmeetlo027protecteqwndweter. Theprooramcoudbe

broadenedtodlrectlyprotectpeoplee’health.
Wmnﬂbmmmmywmmpdmbn
Householdwelsarenotroulnelyteeted. Suppoeeaspeclalproqamweresetuplnyom
countYtoprotectthehealthotpeoplewhoueehouseholdwella.
Unmzxﬂnproqammldprmlanymhcreaeemmupdlwonln
moomty. Invalid:

1. lder'itlymceeotgromdwaterpolutlon.

2. Takeeducatbrld,mgtlatory.andleodacﬂontoprevemmpcuwond
War.

In?

146

mmmmwwemmmmmmmmmpwmm
youcounly. Itwodd:

3. Testallhouseholdwellslnthecomwforharmfumemlcels. Eachwellwodd
beteetedonceayear.

4. lnstaﬂﬂterindevlceelaweterusedbyahwsehddtoremoveanyham
chemicalsthataredetected.

5. Employtralnedteclmlclmeto monitorandmalmalntheflterlngdevlceehtop
working condtlon.

Supposethlsﬂve—panprogramwereputoavote. Worldyouvotelororeoalnstthecounty-
wldeproommandtaxllcostmhandnldﬂpumaﬂHuatotalpaymatdsapa
year? (circleone)

1. VOTEFORTHEPROGRAMANDTAX

2. VOTE AGAINST THE PROGRAM AND TAX

32. Whatbmevaymoetyouwoudpeymeryeenhmmxeetosupponmeproomm
describedthlon31?(Mhanarmlequdtoorgreeterhnzero).

I WOULD PAY UP TO S PER YEAR

 

Wm

31. Theprogramdeecrlbedhqnetlonvprotectepeople'sheelm Theproo'amcoudbe
broadenedtodkeetlyprotectqourmer.

leeqnubn27,UnprooramwoddpraeaﬂopeoplehmmtywllomehaneIldd
wells. Itwoad:
1. Teudhoueeholdwellehthecormiorhenmms. Eachwellwoud
beteetedonceeyeer.

2. Inetalﬂterhodevlceeforwatermedbyehmnehddtoremoveanyhamﬂu
chemicalsthataredetected.

3. Employmmedtedwchmtomorﬂorandmmhmeﬂtmmdevlceehtop
workIrucondltlon

Emprotmmwodddwprmanymkmhammupolmbnhyomcoumy.

4. ldentlyeourceeotgroundwaterpolwon

5. Takeeducatlonnl. rogdatory, arrdleodecdontoprevathuﬂierpdlmlonof
groundwater.

32.

147

Suppoeemisﬂve-partprogramwereputoavote. Worldyouvotefororagalnstthecounty-
wideprogremandtaxlhcostyourhouseholdSZpermonthloretotal peymel'eofsaper

yeaf? (circle one)

1. VOTE FOR THE PROGRAM AND TAX
2. VOTE AGAINST THE PROGRAM AND TAX

WhatletheverymoetyouwotJdpey.peryeer. lnhlghertaxeetoeupporttheprogram
deecnbedlnquealonal? (fillinananmequaltoorgreetertl'lanzero).

I WOULD PAY UP TO S PER YEAR

lnywjudgeMdoeemeprogmmmqueetbanducemehedmthmetpoeedby
groundwaterpollmlonlnyoucoumﬂ Byhorvmuchwoddtheheelththreetreellybe

reduced? (circle one)

1. WTOKDPEMENT 8. 40TO40PEK3ENT
2. wTO&PEFCENT 7. mTO$PEI~K3ENT
3. mronpencsm 8. aorozepencem
4. mTO&PEFCENT 9. 10 TO 10 PEFCENT
5. SOTOSOPEmENT 10. OTOSPEKZENT

34.1nm1mgamdoeeﬂnprogmmhquealm31prdeaqomdwetuhmcoumm

humerpolmon? Byhownulchwoudhxthergromdweterpolwonreellybereducedﬂclrde
000)

1. ”TO1WPEMENT 8. 40TO40PEK3ENT
2. ”TO&PEMENT 7. mTOSPEI-CENT
3. TOTOTDPEFCENT 8. mTOZDPEFCENT
4. NTOQPEFCENT 0. 10TO10W
5. SOTOSOPEFCENT 10. OTOSPEmENT

35. mmmmmnmmmdmmmraumwmmw

taxdeecrlbodlnqueetlonsl?
‘Jtcm— WW

a. Health of your household... VERY mama SOMEWHAT LEAST NOT
mmm murmur mm mm

b. Health ofother houeeholde.. VERY meornm WT LEAST nor
meonrm meonmrr mm mm

c. Health of More generaione. venv «490mm souswmr LEAST NOT
uponrm meonrm uponrm mm

d- Prat-comm

10! he own sake ............... vary mm mm LEAST nor
wonmrr woman wonmrr mm
°- Tax cost of program .......... veer mm mar LEAST nor

WANT WANT WANT WANT

 

148

You and Your Household

Where is your home located? (circle one)

37.

39.

41.

vrwwr

ACITY

ATOWN .

ASUBURB
ARURALAREABUTNOTAFARM
ONAFARM

Is farming an important source of income for your household? (circle one)

1.
2.

YES
NO

Does your household rent or own the home you live in?

1.
2.
3.

RENT
OWN
OTHER

How many peOple, including yourself, are there in your household? (circle one)

1. ONE SK
2. NO SEVW
3. THREE IGEI'
4. POUR NINE
S. FIVE 10.TEN OR MORE

Howmanychildrenléyearsofageoryoungerarethereinyourhousehold?(circleone)

I. HERO F118

2. ONE SIX

3. TWO SW

4. THREE .GEI‘

5. POUR 10.NINE OR MORE
W is your age? (circle one)

1. 18-25 YEARS

2. 26-35 YEARS

3. 36-45 YEARS

4. 46—55 YEARS

5. 56-65 YEARS

6. 66-75 YEARS

7. 76 OR MORE YEARS

 

149

42 Are you female or male? (circle one)

1.FEMALE
2.MALE

43. What was the last year of school that you completed? (circle one)

NO SCHOOL
SOME HIGH SCHOOL

999531”?

44.Whatdoyouexpectyourtotalhouseholdincomqfromallsources,tobeduringthenext12 ,

months? (circle one)

30 to $9.999

310.11!) to 19,999
$20,000 to 29,999
$30,000 to 39,999
$40,000 to 49,999
$50,111) to $9,999
$60,000 to 69,999

SP‘V‘PPN!‘

OOOOOOOOOO

10.
II.
12.
13.
14.

e

SOME OR ALL OF ELEMENTARY SCHOO

LAST YEAR OF HIGH SCHOOL
SOME COLLEGE OR TRADE SCHOOL P
LAST YEAR OF COLLEGE OR MORE ;

 

$70,000 to 79,999
UN) to 89,999
390.11!) to 99,999
8100,“ to 109,999
$11011!) to 149,999
$150,000 to 199,999
900,000 and over

ginkyoumymuch foryourhelpandcooperation. Youropiniononeachquestioncountsagreat

IfYouwouldliketoshareanyadditionalcomments,pleasewriteonthispage.

APPENDIX C

APPENDIX C

SURVEY RESPONSE RATES

Sample households were mailed up to three questionnaires, in order to increase the
likelihood that questionnaires would be returned. Table 01 summarizes the overall survey
response rate. A total of 2,020 Michigan households were sent questionnaires. The
Statewide sample consisted of two subsamples; the urban subsample consisted of 673
households, the rural subsample 1,347 households. A total of 173 surveys were returned as
undeliverable (8.6 percent of the total). Consequently, a total of 1,847 households received
Questionnaires.

Returned questionnaires numbered 1,229, for a gross response rate of 66.5 percent.
This figure includes all returned questionnaires. Of the total returned questionnaires, 16
were not usable. Identification labels were missing on nine questionnaires. As a result, it
Was not possible to classify these questionnaires as urban or rural; they were not used. Four

questionnaires were returned unanswered, and three had responses which indicated they
Were not taken seriously. This left a total of 1,213 net usable questionnaires, for a net
Usable return rate of 65.7 percent. 'Usable' is used here to mean that at least a portion of
the questionnaire was responded to, and the responses were such that the questionnaire

appeared to be taken seriously.

150

151
Table 01 Survey Return Characteristics

 

Michigan Households Surveyed.

Rural: 1,347
Urban: 673
Total: 2,020

Questionnaires undeliverable to addressee: 173 (8.6 percent of total).
Households receiving questionnaire: 1,847

Total returned Questionnaires: 1,229

Gross return rate: 66.5 percent

Unusable Questionnaires: 16

a. ID labels missing: 9
b. Questionnaires returned unanswered: 4

c. Questionable response validity: 3

Net usable returns: 1,213

Net usable return rate: 65.7 percent

 

152

Ten of the returned questionnaires were not from the person to whom the
questionnaire was addressed. Eight of these questionnaires were from spouses or children
of deceased individuals Two were from individuals who lived at addresses where the -
original addressee no longer lived. Since the survey attempts to sample rural or urban
headsof households, and the number is srnalL this substitution of respondents does not
appear to be crucial to the statistical validity of the survey.

The overall response rate is comparable with other surveys, both generally and
speciﬁcally with regard to contingent valuation surveys. Dillman (1978, pp. 21-24)
summarizes the response rates to 38 different mail surveys, which used the total design
method and dealt with a wide variety of topics. The number of responses required for a
given survey ranged from 25 to 238 items. The average response rate was 74 percent, with
no survey obtaining less than 50 percent. However, none the these surveys required
presented respondents with economic valuation questions.

Cameron and Mitchell (1989, p. 281) summarize the return rates from 16 contingent
valuation surveys undertaken between 1974 and 1986. Using a gross response rate (total
number of surveys in the sample divided by the total number of questionnaires completed
and returned), the average for the surveys cited in Mitchell and Carson is 47.6 percent. It
should be noted that not all of these surveys employed the total design method. Two recent
contingent valuation surveys dealing with groundwater pollution are Edwards (1988) and
Shultz and Lindsay (1990). Both used the total design method in implementing the survey.
Edwards obtained a gross response rate of 785 percent, while Shultz and Lindsay 57.6
percent. The Michigan groundwater survey obtained a gross response rate of 60.5 percent.
It should be noted that both the Edwards and Shultz and Lindsay studies focused on a
speciﬁc community, while the Michigan study was statewide. It would seem that higher

response rates would be expected by focusing on areas where currently there is concern over

 

153
groundwater contamination (whether government planning purposes or from public

perception), as opposed to a statewide survey where that may or may not be the case.

Table C-2 summarizes the return rates for each of the three mailouts. The ﬁrst
questionnaire mailout was Tuesday, 31 July 1990. The total number of surveys mailed was
2,020. Of these, 761 were eventually returned. This results in a gross return rate for the
first mailout of 37.7 percent. If all the questionnaires which were post returned are
subtracted from the total number of surveys sent, the return rate increases to 41.2 percent.

The second questionnaire mailout occurred three weeks later on Tuesday, 21 August.
All households which did not return a questionnaire, and for which no postal return had
been received, were sent a second questionnaire. Questionnaires mailed in this sectmd
mailout totaled 1,262. A total of 359 were eventually returned for a return rate of 28.4
percent.

The third and final questionnaire mailout took place four weeks later, on Tuesday,

18 September. A total of 749 questionnaires were mailed, and 109 were returned for a

return rate of 14.6 percent.

TahIeC-z ReturnRatesforQnestionnalreMallonts

 

 

Mailout Date Number Mailed Number Returned Ratel

July 31 2.020 761 37.7
August 21 1,262 359 23.4
September 18 749 109 14.6

 

Rate is calculated by dividing number returned by number mailed out. If all
post returns are subtracted from the number mailed for the July 31 survey, the
return rate is 40.5 percent.

 

154

Table 03 shows the percentage of total questionnaire returns by week for
each of the three surveys. For example, the ﬁrst entry under the July 31 column,
38.6, shows the percentage of total returns for the July 31 mailout that occurred by
the end of the ﬁrst week after the mailout. During the second post mailout week,
41.8 percent of returned surveys were returned. During the third week, 3.7 percent
were returned, with the remaining 15.9 percent being returned after the third week.
The ﬁrst two mailouts are fairly comparable in weekly return rates. The 31 July
mailout had about 80 percent of all returned questionnaires returned by the end of
the second post-mailout week. The 21 August mailout had approximately 78 percent

of its total returns by the end of a comparable time period.

Table 03 Survey Returns by Week

 

 

Week Percentage of Total Return Surveys1
First week 38.6 25.9 33.1
Second week 41.8 52.6 25.4
Third week 3.7 9.2 15.3
After third week _15_,2 .12.} .212

Total 100.0 100.0 1111.0

 

‘ Total Returned surveys for speciﬁed mailout date.

The weekly response rate of the initial mailout is comparable with other
surveys used the total design method. Dillman (1978, p. 185-186) refers to ﬁve
surveys of the general public, conducted in four states, where the response rate up

to the postcard reminder mailout one week after the initial mailout ranged from 19

 

155

to 27 percent. The response rate for the groundwater survey for a comparable
period was 16.0 percent. For the period after the postcard reminder, but before the
second mailout, the incremental response rate ranged from 15 to 25 percent. For the

groundwater survey, the incremental response rate was 23.8 percent.

APPENDIX D

APPENDIX D

SURVEY RESPONSES

This appendix contains a number of tables which detail questionnaire responses.
The following seven tables are included: (1) sample demoographies; (2) water use
information; (3) respondent water quality perceptions; (4) respondent risk perceptions
regarding severity; (5) respondent risk perceptions regarding knowledge; (6) respondent risk
perceptions regarding choice; and (7) respondents environmental perceptions.

Responses are included for both the rural and urban samples. For each sample
responses for both the complete sample set and the restricted sample set are also included.

The presentation order of the tables follows closely the order of the subject matter

presentation in the text.

156

157

Table D-I. Sample Demographics
(percentage elusable responses In speciﬁed category)

 

 

Rural Urban
Complete Restricted Complete Restricted
Expected Annual Income (044)

Under $10.11!) 0.1 02 0.0 0.0
10H!) to 19,999 21.9 17.7 138 12.7
20.000 to 29,999 Bl B5 15.7 14.7
301“) to 39,999 163 19.2 16.3 14.2
4001) to $9,999 27.3 26.9 34.1 34.8
601”) to 99,999 8.0 9.4 143 17.1
over 100,000 3.0 32 5.6 8.4
Non-responses 14.7 - 12.5 -
Mean 39X) 40.48 47.26 48.53
8D. 26.53 26.76 29.09 2751

Education (highest level completed) (043)

No school ( 1) 0.1 0.0 0.0 0.0
Elementary (2) 3.7 2.8 4.1 15
Some high school (3) 9.6 85 8.8 5.4
High school (4) 320 27.7 203 19.6
Some college (5) 30.9 32.8 33.2 35.8
College (6) 22.3 27.1 31.6 36.3
Non-responses 4.4 - 2,7 -
Mean‘ 4.58 4.74 4.81 5.01
SD. 1.06 1.04 1.10 0.96

Responsemeansarebasedonthenumberinparenthesesfollowingtheresponsecategoriesforaparticular
question.

158

Table D-l (continued)

 

 

Rural Urban
Complete Restricted Complete Restricted
AI! (Q41)
1&25 (1) 3.6 4.9 2.7 3.4
26-35 (2) 16.1 18.6 28.2 33.8
36-45 (3) 22.0 23.9 19.7 25.0
46-55 (4) 168 19.4 18.6 16.7
56-65 (5) 15.5 126 14.5 9.8
66 and over (6,7) 26.0 20.7 162 11.3
Non-responses 3.6 - 14 -
Mean 4.12 3.842 3.67 333
SD. 1.68 1.63 158 1.49
Home location (036)

City (1) 25.6 273 48.1 475
Town (2) 16.4 15.8 7.9 83
Suburb (3) 4.9 6.0 27.7 275
Rural area but not

a farm (4) 40.6 38.4 14.1 13.7
Farm (5) 12.5 126 2.2 2.9
Non-responses 3.1 - 1.6 -
Mean 298 2.93 2.14 216

SD. 1.45 1.46 1.23 1.24

159

Table 04 (continued)

 

 

Rural Urban
Complete Restricted Complete Restricted
Ila-e ownership (Q38)
Rent (1) 10.5 10.9 16.9 172
Own (2) 87.7 87.6 81.4 82.4
Other (3) 1.8 15 1.7 0.4
Non-responses 33 - 2.1 -
Mean 1.91 1.91 1.85 1.83
SD. 034 034 0.40 039
Household she (039)
One (1) 16.6 11.9 175 17.6
Two (2) 36.7 382 33.3 275
Three (3) 145 14.9 19.4 20.1
Four or more (4-10) 32.1 34.8 29.8 34.8
Non-responses 3.0 - 21 -
Mean 2.81 2.92 279 2.87
SD. 1.45 1.42 1.40 1.39

Numberofcbllrh'en 16yenrs
oryoungerlnbousebold(040)

None (1) 64.1 612 63.4 55.9
One (2) 11.4 124 12.8 162

Two (3) 163 17.9 15.0 19.1

Three (4) 5.6 . 55 68 7.4

Four or more (5-10) 25 2.9 1.9 1.5
Non-responses 3.7 - 21 -

Mean 1.73 1.80 1.72 1.83

SD. 1.17 1.23 1.10 1.09

160

TebIeD-Z. WatsrUeeInlrr-ation
(pes'ceutolusableresponeesineechcetepry)

 

 

Rural Urban
Complete Restricted Complete Restricted
Water Source (01)
Community or
public water
system (1) 40.9 41.9 77.1 76.0
Household '
Well (2) 57.7 568 215 23.5
Spring, stream,
river (3) 1.0 0.6 0.5 0.0
Other (4) 05 0.6 0.8 0.5
Non-responses
1.0 - 13 -
Mean (SD) 1.61 (0.54) 1.60 (054) 125 (050) 125 (0.47)
Drinking Water
Source (02)
Tap (1) 92.6 923 91.8 93.1
Bottled
Water,
purchased (2)
45 4.7 65 6.4
Bottled water.
other (3) 1.6 15 0.5 0.5
Non-responses
1.1 - 13 -
Mean (SD) 1.12 (0.47) 1.12 (0.48) 1.09 (035) 1.07 (030)
Water Tested (04)
No (1) 827 85.8 912 91.6
Yes (2) 17.3 142 8.8 8.4
Non-responses 2.6 - 22 -
Mean (SD) 1.17 (038) 1.14 (035) 1.09 (028) 1.08 (0.28)

 

 

161

Table D3. Respondent Water Quality Perceptions
(percentolnsableresponseslneechcntepry)

 

Rural Urban
Complete Restricted Complete Restricted

 

 

Tap Water Safety (03)
Very safe (1) 555 53.2 41.9 412
Somewhat safe (2) 20.4 21.6 24.4 26.0
Not sure (3) 21.4 22.6 30.7 31.4
Somewhat unsafe (4) 2.3 2.6 1.9 1.0
Very unsafe (5) 0.4 0.2 0.8 05
Non-responses 0.8 - 1.9 -
Mean (SD) 1.72 (0.90) 1.75 (0.91) 1.96 (0.95) 1.94 (0.90)

Conﬁdeneetbnttnpwnteelssalenomeﬂectaofpoundwaterpollutioumﬂ

Very conﬁdent (1) 22.9 22.0 125 10.3
Conﬁdent (2) 45.8 44.1 46.9 48.8
Not conﬁdent (3) 24.7 25.8 31.9 32.0
Not conﬁdent

at all (4) 6.7 8.1 8.4 8.9
Non-responses 1.3 - 1.3 -

Mean (SD) 2.15 (0.85) 220 (0.87) 2.36 (0.81) 2.39 (0.79)

 

162

TabIeD-4. RespondentRiskPesuptions: Severity

(percentageolnsablesurveyslnspeclﬁedcntegory)

 

 

Rural Urban
Complete Restricted Complete Restricted
Seriousness of health risk an. grounduuser pollution (07)
Very Serious (1)
9.1 9.6 13.6 12.3
Serious (2) 135 11.9 24.4 21.1
Somewhat
Serious (3) 36.7 37.5 35.7 37.7
Not
Serious (4) 33.4 35.0 22.7 27.5
Not a threat
at all (5) 6.4 6.0 3.7 15
Non-responses 24 - 5.1 -
Mean (SD) 3.15 (1.04) 3.16 (1.03) 278 (1.06) 2.85 (1.01)
Deg-es 0! her ofﬁnundwuter pollution (015)
Fear greatly (1) 24.0 235 233 21.6
Fear
Somewhat (2) 472 488 52.0 50.0
Fear very
little (3) 222 no 11.6 245
Fear not at
all (4) 6.7 5.8 4.1 3.9
Non-responses 13 - 08 -
Mean (SD) 211 (0.85) 210 (0.82) 2.05 (0.77) 2.11 (0.78)

 

 

163

Table D-4 (coatlaaed)

 

 

 

Rural Urban
Complete Restricted Complete Restricted
Fatalaeu (018)
Very likely
to be fatal (1) 3.5 3.0 6.7 5.4
Likely to be
fatal (2) 11.7 113 118 10.8
Sometimes
Fatal (3) 443 429 412 40.2
Unlikely to _.
be fatal (4) 33.4 362 342 37.7
Not fatal (5) 72 6.6 62 5.9
Non-responses 33 - 4,0 -
Mean (SD) 3.29 (0.89) 332 (0.87) 321 (0.97) 328 (0.93)
Gmadnterpollatloa laluﬂagpopleoaeatatl-eor-aayatouce

One at a time (1) 14.0 12.8 10.1 93
A few people at
once (2) 40.6 38.2 29.9 333
Mm Mk II
once (3) 325 36.9 342 333
Very mny maple
at once (4) 65 6.8 11.0 113
A very large number
at once (5) 63 53 142 12.7
Non-reeponnee 26 - 1.9 -
Mean (id) 250 (1.02) 254(098) 289 (1.17) 2.85 (1.15)

 

164

‘l‘nbleD-S. wmmw
Mdmﬂemhapadﬂadcamﬂ

 

Rural Urban

Complete Restricted Complete Restricted

 

ScleadﬂchowhﬂedmpolloﬂoahaahrhkﬁQS)

 

Known precisely (1) 4.2 45 8.0 4.9

Known (2) 32.6 348 29.9 30.4

Somewhat known (3) 45.6 45.4 45.6 49.0

Only mspected or

gueued at (4) 15.2 13.0 143 152

Not known at all (5) 23 21 1.6 0.5

Non-responses 1.4 - 2.2 -

Men (5.0.) 2.79 (083) 274 (0.82) 272 (0.87) 2.76 (0.79)
Km .r expend individuals (09)

Known precisely (l) 0.6 0.9 3.0 1.5

Known (2) 92 8.1 52 4.4

Somewhat known (3) 392 38.0 322 30.4

Only wspeaed or

gueued at (4) 427 44.1 46.0 50.5

Not known at all (5) 8.0 8.7 13.4 132

Non-responses 1.6 - 13 -

Mean (SD) 3.49 (030) 352 (0.82) 3.62 (089) 3.70 (0.81)

 

165

 

 

Table 06 (ceatlnuad)
Rural Urban
Complete Restricted Complete Restricted
Wpolletloau aeworolde-vlm-eahl [noble-(010)

Old (1) 103 10.7 14.4 142
Somewhat old (2) 22.3 232 253 21.6
Midway between new
and old (3) 415 40.9 372 39.7
Somewhat new (4) 212 20.9 19.0 19.6
New (5) 4.7 43 4.1 4.9
Non-reepomee 1.3 - 1.1 -
Mean (5.1).) 2.88 (1.01) 2.85 (1.01) 273 (1.06) 279 (1.07)

 

 

166

Teﬂon-6. Mammalian
(pace-”datablawlaapadﬂadcam'y)

 

Rural Urban
Complete Restricted Complete Rained .

 

 

Controllablllty of goundwater pollatloa damp (013)

 

Easy (1) 29 32 4.6 2.9

resume (2) 36.0 33.9 33.1 333 ,
possible (3) 29.1 28.4 29.0 28.9

Very difﬁcult (4) 29.1 318 30.6 32.4 :1
Not possible (5) 2.7 313 30.6 32.4
Non-responses 1.6 - 1.6 -

Mean (51).) 293 (0.93) 2.97 (0.94) 2.94 (0.96) 2.98 (0.94)

Volaatarlaeu 0! exposure to poaadwater pollution rlalta (014)

Voluntary

Somewhat voluntary
Somewhat involuntary

Involuntary

Non-responses
Mean (SD)

2.25
11.0
31.5

55.0

1.3

339 (0.78)

1.5
9.6
34.1

54.8

3.42 (0.73)

1.9
73
24.2

66.6

1.1

355 (0.71)

1.5
6.9
25.5

66.2

356 (0.69)

 

167

Table D-7. Respondents Environmental Perceptlons
(percentage of usable responses in speciﬁed category)

 

 

Rural Urban
Complete Restricted Complete Remixed
Extent towhlchlbtnrepnsrstlonswill bshnr-sdllyuoandwntsrpollutlon (019)
Not harmed
at all (1) 1.4 1.1 1.4 1.5
Very unlikely
to be harmed (2) 65 6.6 6.6 83
Not likely
to be harmed (3) 163 16.0 11.7 12.3
Likely to be
banned (4) 49.4 51.2 462 46.6
Very [Rely
to be harmed (5) 26.5 252 34.2 31.4
Non-responses 3.0 - 1.6 -
Mean (SD) 3.93 (0.90) 3.93 (0.88) 4.05 (0.92) 3%(095)
Threat dp'onndwntsrpollntlon towilrllhnnd plants (020)

Not a threat
at all (1) 2.9 15 2.5 2.9
Not serious (2) 19.1 213 13.1 142
Somewhat serious (3) 402 40.1 36.1 35.8
Serious (4) 25.9 25.4 273 30.4
Very serious (5) 118 11.7 21.0 16.7
Non-responses 20 - 1.6 -
Mean (SD) 325 (0.99) 325 (0.97) 351 (1.04) 3.44(1.02)

 

 

168

Table D-‘I (continued)

 

Rural Urban
Complete Restricted Complete Reariaed

 

mmmummummnuemdemmmmmmm
"vi-ethl-vmn)‘

Protect for own sake 20.0 21.9 19.9 18.6

Protectbeesusepart

 

of natural system 685 69.9 68.0 74.0
Disagree with both

opinions 18 23 2.2 15
Don’t Know 9.7 6.8 98 5.9
Non-responses 24 - 1.6 -

Pollutionlshndllscnnssltlsthswl'oqthlngtodoorllsenuselthasllndeﬁcts(022)l

Pollution is wrong

thing to do 75.4 T7.4 74.9 79.4
Pollution is bad only

when it has bad

effects 142 15.6 16.6 16.7
Disagree with both

opinions 5.0 45 22 25
Don’t know 53 26 63 15
Non-response! 18 - 13 -

 

Means are not relevant for these qrestions.

Table D-7 (anti-and)

169

 

Rural
Complete

Restricted

Urbm
Complete

Rearicted

 

WMhMﬂo-pollutlonoﬂywhsnpoﬂuﬂonposssu

(023)

Strongly agree (1)
Ayee somewhat (2)
Dingo: mm (3)
Strongly disagree (4)

Don't Know (5) '

Non-responses

Mean (SD)

14.6
9.6

11.9
60.1

3.8

1.7

320 (1.12)

115
8.7

12.8
65.0

1.9

337 (1.07)

13.9
7.6

12.0
62.0

4.6

1.1

324 (1.09)

drsethsalththreattohumnns

10.8
6.9

13.2
67.6

15

3380.02)

 

APPENDIX E

APPENDIX 5

PHONE SCRIPT

Hello, my name is (state name). I am a researcher at Michigan State University in the
Department of Agricultural Economies.

We are doing research on some environmental issues that face Michigan citizens. As
part of this» project, we are asking people their opinions on some important
environmental issues in Michigan.

What I would like to do is to invite you to a small group meeting on campus so that we
can get your views on some of these important issues.

The meeting will include about 8-10 people from the Williamston area and will last about
two hours.

If you decide to attend, we will pay you $50 to reimburse you for your time and trouble.
Do you think you might like to attend?

The meeting will be held at the Kellogg Center on the M.S.U. campus. It will begin
(state time and date). We will pay for your parking.

We would like to mail you a formal invitation and directions to get to the Kellogg Center
and which room the meeting will be held. May I have your name and address?

We’ll send your invitation out within the next few days, and thank you very much for
agreeing to attend.

170

' LIST OF REFERENCES

5.... ' w
.5 l

LIST OF REFERENCES

Babbie, Earl. Survey Research Methods. Belmont CA: Wadsworth Publishing Company,
Inc, 1973.

Bishop, Richard C. and Thomas A. Herberlein. Simulated Markets, Hypothetical Markets,
andTravelCostAnalysis: AleruativeMethodsolEstimntingOutdoorRecreation
Demand Staff Paper Series No. 187. Department of Agricultural Economies,
University of Wisconsin, Madison.

Bishop, Richard C. and Thomas A. Herberlein. “Measuring Values of Extra-Market Goods:
Are Indirect Measures Biased?.' American Journal I! Agricultural Economics. 61.
( 1979):926-930.

Brookshire, David; Alan Randall, and John Stoll. "Valuing Increments and Decrements in
Natural Resource Service Flows." AmericanJournnlongricultural Economics.
63. (1980):478-488.

Craun, Gunther F. A Summary ofWaterhorne Illness Transmitted Through Contaminated
Groundwater. Research Triangle Park NC: U. S. Environmental Protection Agency,
1985.

Cummings, Ronald G.; David S. Brookshire, and William D. Schulze, eds. Valuing
Environmental Goods. Totowa NJ: Rowman and Allanheld, 1986.

Dillman, Don A. Mall and Telephone Surveys: The Total Design Method. New York:
John Wiley and Sons, 1978.

Edwards, Steven F. "Option Prices for Groundwater Protection.” Journal 11 Environmental
Economics and Management 15. (1988):475-487.

Environ Corporation. Approaches to the Assessment of the Health Impacts of
GroundwaterContamhtion. Washington DC: Office of Technological Assessment,
1983.

Ferguson, George A. summer We h Psychology and sum. 3rd edition. New
York: McGraw-Hill Book Company, 1971.

Fisher, Ann; Dan Violette, and Lauraine Chestnut. The Value of Reducing the Risks of

Death: A Note on New Evidence.‘ Journal 0! Policy Analyst and Management 8.
(1989):88-100.

171

 

172

Fischhoﬂ, Baruch; Paul Slovic, Sarah Lichtenstein, Stephen Read, and Barbara Combs.
'How Safe is Safe Enough? A Psychometric Study of Attitudes Towards
Technologieal Risks and Benefits." Policy Studies 9. (1978):127-152.

F ischhoﬂ, Baruch and Lita Furby. 'Measuring Values: A Conceptual Framework for
Interpreting Transactions with Special Reference to Contingent Valuation of
Visibility.‘ Journal of Risk and Uncertainty l. (1988):147-184.

Gardner, Gerald T., Adrian R. Tieman, Leroy C. Gould, Donald R. De Luca, Leornard W.
Doob and Jan A. J. Stolwijk. "Risk and Benefit Perceptions, Acceptability
Judgements and Self-Reported Actions Toward Nuclear Power.“ The Journal of
Social Psychology 116. (1982):179-197.

Henderson, Timothy R., J eﬁrey Trauberman and Tara Gallagher. Groundwater
for State Action. Washington, DC: Environmental Law Institute, 1987.

Graham, Daniel A. “Cost-Benefit Analysis under Uncertainty.“ The American Economic
Review 71. (1981): 715.725.

Hoehn, John P. “Contingent Valuation in Fisheries Management: The Design of
Satisfactory Contingent Valuation F ormats.’ Transactions of the American Fisheries
Society 116. (1987):412-419.

Hoehn, John P. and Alan Randall. 'A Satisfactory Beneﬁt Cost Indicator from Contingent
Valuation.“ Journal at Environmental Economics and Management 14. (1987):226-
247.

Hoehn, John P. 'Valuing the Multidimensional Impacts of Environmental Policy: Theory
and Methods.“ American Journal of Agricultural Economics 73. (1991):289-299.

Kmenta, Jan. Elements of Econometrics. 2nd edition. New York: MacMillan Publishing
Company, 1986.

Kraus, Nancy Nighswonger and Paul Slavic. Taxonomic Analysis of Perceived Risk:
Modeling Individual and Group Perceptions within Homogeneous Hazard Domains.“
thAnalysis 8. (1988):435-455,

Libby, Lawrence W. and Jessica T. Kovan. Rural Groundwater Contamination: Impacts
dandPotentialBeneritshomLandUsePlanningandZouing. Agricultural
Economics Staﬁ Paper No. 86-88. East Lansing: Department of Agricultural
Economics, Michigan State University, 1986.

Lichenstein, Sarah, Paul Slavic, Baruch Fischhoﬁ, Mark Layman, and Barbara Combs.
"Judged Frequency of Lethal Events.“ Journal of Experimental Psychology: Human
learning and Memory 4. (1978):551-578.

Loehlin, John C. LatentVarhhle ModekAn Introductlontanctor,Pathand Structural
Analysis. Hillsdale NJ: Lawrence Erlbaum Associates,1987.

 

‘ hr -4 .

173
Lawrence, William W. 0! Acceptable Risk. Los Altos CA: William Kaufman, Inc., 1976.

Michigan, Stateof. PolicyandProgralerectionsforGroundwaterProtectionln
Michigan. Discussion Paper for the Michigan Water Quality Management Plan.
1988.

Michigan Farm Bureau. Nitrate Contamination ofGroundwater hr Michigan: A Survey.
Lansing MI: Michigan Farm Bureau, 1989.

Michigan, Stateof. Repor-toItheHouseAdI-IocCommitteeonGroundwsterProtectlon.
Lansing MI: Michigan State Legislature, 1987.

Mitchell, Robert Cameron and Richard T. Carson. Ushg Surveys to Value Public Goods.
Washington, DC: Resources for the Future, 1989.

Nielsen, Elizabeth G. and Linda K. Lee. The Magnitude and Costs of Groundwater
Contamination from Agricultural Chemicals. Washington DC: U.S. Department
of Agriculture, Economic Research Service, 1987.

Olsen, Larry G. "Agricultural Chemicals Detected in Groundwater in Michigan by MDA
Pilot Survey." Pesticide Notes 2. published by Michigan Cooperative Extension
Service (1989):2-5.

Pye, Veronica 1., Ruth Patrick and John Queries. Groundwater Contamination in the
United States. Philadelphia: University of Pennsylvania, 1987.

Randall, Alan and John R. Stall. "Existence Value in a Total Valuation Framework." in
Robert D. Rowe and Lauraine G. Chestnutt (eds) Managing Air Quality and
ScenlcResourcesatNationalPsrksandWllderuessAreas. Boulder,CO: Westview
Press, 1983.

Raucher, Robert. "A Conceptual Framework for Measuring the Benefits of Groundwater
Protection." Water Resources Research 19. (1983):320-326.

Rancher, Robert. The Benefits and Costs of Policies Related to Groundwater
Contamination." Land Economics 62. (1986):33-45.

Sharefkin, Mark, Mordechai Shechter and Allen Kneese. "Impacts, Costs, and Techniques
for Mitigation of Contaminated Groundwater: A Review." Water Resources
Research 20. (1984):1771-1783.

Shechter, Mordechai. "An Anatomy of a Groundwater Contamination Episode." Journal
of Environmental Economics and Management 12 (l985a):72-88.

Shechter, Mordechai. "Economic Aspects in the Investigation of Ground-water
Contamination Episodes." Groundwater 23. (1985):l90-197.

Shultz, Steven D. and Bruce E. Lindsay. "The Willingness to Pay for Groundwater
Protection." Water Resources Research 26. (1990): 1869-1875.

 

174
Slavic, Paul. "Perceptions of Risk." Science 236. (1987):280-285.

Slavic, Paul, Baruch Fischhaff and Sarah Lichtenstein. "Behavioral Decision Theory
Perspectives on Risk and Safety." Acts Psycholog 56. (1984):18&203.

Solley, Wayne 8., Edith B. Chase and William B. Mann IV. Estimated Use deter in the '
United States in 1980. Geologin Survey Circular 1001. Washington, DC:
Government Printing Office, 1983.

Tversky, Amos and Daniel Kahneman. 'I'he Framing of Decisions and the Psychology of
Choice." Science 211. (1981):453-458.

U.S. Environmental Protection Agency. Natitmnl Stathtieal Assessment 11 Rural Water
Conditions, Executive Summary. Washington DC: U.S. Environmental Protection

Agency, 1984.

U.S. Department of Commerce. County and City Data Book 1988. Washington, DC: U.S.
Government Printing Office, 1988.

us. Geological Survey. Groundwater Hydrolog and Water Quality in Kalamazoo County,
Michigan. Washington, DC: U.S. Geological Survey, 1985.

U.S. Geological Survey. Natiorml Water Summary 1984: Hydrologlc Events, Selected
Water Quality Trends and Groundwater Resources. Washington D.C.: U.S.
Government Printing Ofﬁce, 1985.

U.S. Geological Survey. Basic Groundwater Hydrolog. Water Supply Paper 2200.
(Washington, DC: U.S. Government Printing Ofﬁce, 1984.

Walpole, Ronald E. and Raymond H. Myers. Probability and Statistics for Engineers and
Scientists. New York: MacMillan Publishing Company, Inc., 1978. ‘

Ward, C.H., W.Geiger and PL McCarty. Groundwater Qnaﬂy. New York: John Wiley
and Sons, 1985. '

Weinstein, Milton C. and Robert J. Quinn. "Psychological Considerations in Valuing Health
Risk Reductions." Natural Resources Journal 23. (1983):659-673.

Weinstein, Milton 0, Donald S. Shepard and Joseph S. Pliskin. 'I'he Economic Value of
Changing Mortality Probabilities: A Decision-Theoretic Approach." The Quarterly
Journal 0! Economics 84. (l980):373-396.