AN
EXPERIMENT
IN
DETERMINING
WILLINGNESS
TO
PAY
FOR
NATIONAL
WATER
QUALITY
IMPROVEMENTS
Robert
Cameron
Mitchell
&
Richard
T.
Carson
EPA
Report
#
June
1981
DRAFT
AN
EXPERIMENT
IN
DETERMINING
WILLINGNESS
TO
PAY
FOR
NATIONAL
WATER
QUALITY
IMPROVEMENTS
by
Robert
Cameron
Mitchell
Richard
T.
Carson
Resources
for
the
Future
1755
Massachusetts
Ave.,
N.
W.
Washington,
D.
C.
20036
(
Grant
#
R
806906010)

Alan
P.
Carlin
Project
Officer
Office
of
Strategic
Assessment
and
Special
Studies
RD676,
West
Tower
919
U.
S.
E.
P.
A.
401
M
St.,
N.
W.
Washington,
D.
C.
20460
DISCLAIMER
Although
prepared
with
EPA
funding,
this
report
has
neither
been
reviewed
nor
approved
by
the
U.
S.
Environmental
Protection
Agency
for
publication
as
an
EPA
report.
The
contents
do
not
necessarily
reflect
the
views
or
policies
of
the
U.
S.
Environmental
Protection
Agency,
nor
does
mention
of
trade
names
or
commercial
products
constitute
endorsement
or
recommendation
for
use.
TABLE
OF
CONTENTS
Page
Executive
Summary
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
,
.
,
.
.
.
.
.
.
.
.
.
i
Preface,...............................
vi
Chapter
1.
The
Willingness
to
Pay
Method,
Consumer
Surplus
and
Water
Quality
Benefits
.............
1­
1
This
Study
.........................
1­
4
Benefit
Cost
Analysis
...................
1­
6
Public
Goods
and
Public
Policy
...............
1­
8
Consumer
Surplus
......................
1­
13
Surplus
vs.
Variation
...................
1­
16
The
Surplus
Measures:
Definitions
.............
1­
16
Criteria
for
Choosing
Between
the
Hicksian
Surplus
Measures
....................
1­
21
2.
The
Macro
Approach
to
Willingness
to
Pay
Studies
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
2­
1
The
Macro
Willingness
to
Pay
Approach
...........
2­
1
The
Micro
Willingness
to
Pay
Approach
...........
2­
4
Comparison
.........................
2­
10
The
RFF
Macro
Approach
...................
2­
15
3.
Research
Procedures
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
,
.
3­
1
Context
...
...
..
..
Water
Pollution
Ladder
and
Value
Levels
3­
2
3­
5
Payment
Vehicle
......................
3­
10
4.
Control
for
Biases
.
.
.
.
.
.
,
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
4­
1
General
Biases
.......................

Strategic
Bias
.......................

Hypothetical
Bias
.....................

The
Relationship
Between
Strategy
and
Hypothetic
Bias
....................

Survey
Research
and
Market
Simulation
...........

Context
Correspondence
...................

Instrument
Biases
.....................

Starting
Point
Bias
....................

Payment
Vehicle
Bias
....................

information
Bias
.....................

Order
Bias
.........................
Sampling
Biases
......................

Sample
Bias
........................
4­
2
4­
2
4­
16
4­
21
4­
24
4­
30
4­
33
4­
33
4­
40
4­
40
4­
41
4­
44
4­
44
Response
rate
.......................
4­
46
Interview
Biases
......................
4­
49
Item
Nonresponse
Bias
...................
4­
49
Interview
Procedure
and
Interviewer
Biases
.........
4­
58
5.
Experimental
Benefits
Estimates:
Overall,
Use
and
Intrinsic
.................
5­
1
Illustrative
Estimates
..................
Amounts
by
Version
....................
Combined
Amounts.
.....................
Effect
of
Knowing
Amount
Being
Paid
............
Explanation
of
Willingness
to
Pay
for
Water
Quality
.....................
Model
Specification
....................
Estimation
.........................
Intrinsic
and
Recreation
Benefits
.............
5­
2
5­
2
5­
4
5­
6
6.
Conclusion
and
Recommendations
................

Overcoming
Item
Nonresponse
Bias
..............
Intrinsic
Benefit
Estimate
.................
Other
Refinements
and
Techniques
..............

References
.............................
5­
7
5­
7
5­
15
5­
23
6­
1
6­
4
6­
7
6­
8
R­
1
Appendices
A­
I
RFF
Research
Instrument
.
.
.
.
.
.
.
.
.
.
.
.

A­
II
The
Water
Quality
Ladder
by
William
J.
Vaughan
.

A­
III
Derivation
of
Public
Goods
Expenditures
.
.
.
.

A­
IV
Preceeding
Questionnaire
on
Environmental
and
Energy
Issues
.
.
.
.
.
.
.
.
.
.
.
.
.

A­
V
Sampling
Plan
for
the
RFF
Survey
,
.
.
.
.
.
.
.

A­
VI
A­
VII
Cases
Eliminated
from
Analysis
.........

Construction
of
ENVINDEX
............

A­
VIII
A
New
Constructive
Test
for
Heteroskedasticity
by
Richard
T.
Carson
and
William
J.
Vaughan
,
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.

.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
A­
I­
1
A­
II­
1
A­
III­
1
A­
IV­
1
A­
V­
1
A­
VI­
1
.

A­
VII­
1
A­
VIII­
1
1.1
1.2
1.3
1.4
1.5
1.6
4.1
4.2
4.3
4.4
4.5
4.6
Test
for
Starting
Point
Bias
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
4­
39
4.7
Distribution
of
Resources
for
the
Future
Survey
on
Key
Demographic
Variables
.
.
.
,
.
.
.
.
.
.
.
.
.
,
.
.
.
.
.
.
4­
48
4.8
LIST
OF
TABLES
Page
Types
of
Consumer
Surplus
Measures
for
Contingent
Valuation
Studies
.
.
.
.
.
,
.
.
,
.
.
.
.
.
.
,
.
.
.
.
.
.
.
1­
14
Types
of
Property
Rights
From
the
Consumer's
Perspective
.
.
.
.
1­
20
Role
of
Property
Rights
in
Determining
the
Relationship
of
the
WTP,
WTA,
Compensating
and
Equivalence
Dimensions
of
Consumer
Surplus
Measures
for
WTP/
A
Surveys
Valuing
Environmental
Public
Goods
,
,
,
.
.
.
.
,
.
,
.
.
.
,
.
,
,
.
1­
23
National
Benefits
of
Clean
Water
.
.
.
.
.
.
.
.
.
.
.
.
.
,
.
,
1­
29
Conditions
Under
Which
People
May
Hold
Long
and
Short
Personal
Option
Values
for
Environmental
Goods
.
.
.
.
.
.
.
.
.
1­
35
National
Benefits
of
Clean
Water
Measured
by
the
RFF
Survey
.
.
.
1­
46
Potential
Biases
in
Willingness
to
Pay
Studies
.
.
.
.
.
.
.
.
.
4­
1
Percent
of
Those
Giving
Various
Levels
of
Payment
Who
Belong
to
Certain
Demographic
and
Attitudinal
Categories.
,
.
.
.
4­
14
Types
of
Realism
and
Strategic
and
Hypothetic
Bias
.
.
.
.
.
.
4­
22
Study
Design
for
EPA
Water
Pollution
Benefits
Study
and
Number
of
Cases
,
,
.
.
,
.
.
.
.
.
.
.
.
.
.
.
.
,
.
.
.
.
.
4­
36
t
Tests
of
Means
for
Paired
Comparisons
Between
Versions
A,
B,
C,
by
Income
and
Level
of
Water
Quality
.
.
.
.
,
.
.
.
,
.
4­
38
Logit
Regressions
Relating
Background
and
Attitudinal
Variables
to
Certain
Types
of
Willingness
to
Pay
Responses
for
Fishable
Water
.
.
.
.
.
.
.
.
.
.
.
.
,
.
.
.
.
,
4­
55
5.1
5.2
5.3
5.4
5.5
5.6
5.7
5.8
5.9
Page
Mean
Amounts
Willing
to
Pay
Annually
Per
Household
for
Boatable,
Fishable
and
Swimmable
Water
Quality
in
the
United
States
by
Version
and
Income
Levels
.
,
.
.
.
.
.
.
5­
3
Income
Ranges
for
the
RFF
Survey
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
5­
9
Correlation
Matrix
for
Variables
Used
in
the
Regression
Equations
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
5­
16
OLS
Regression
of
Demographic
and
Attitudinal
Variables
on
Willingness
to
Pay
Amounts
for
Three
Levels
of
National
Water
Quality
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
,
.
.
.
.
.
5­
18
Adjusted
Regression
of
Demographic
and
Attitudinal
Variables
on
Willingness
to
Pay
Amounts
for
Three
Levels
of
National
Water
Quality
,
,
.
.
.
.
.
.
.
.
.
.
.
.
.
.
5­
21
Price
Flexibility
of
Income
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
,
.
5­
22
User
OLS
Intrinsic
Benefit
Estimate
for
Fishable
Water
Quality
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
,
,
.
.
.
.
5­
25
Boat,
Swim,
Fish
OLS
Intrinsic
Benefit
Estimate
for
Fishable
Water
Quality
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
5­
26
Percentage
of
Fishable
Water
Quality
WTP
Benefits
Estimated
as
Intrinsic
by
Income
Level
.
.
.
.
.
.
.
.
,
.
.
.
.
5­
28
5.10
t
Ratios
for
Regression
of
Use
on
Willingness
to
Pay
for
Fishable
Water
Holding
Income
Constant
.
.
.
.
.
.
.
.
.
5­
30
6.1
Direction
of
Biases
in
the
RFF
Survey
.
.
.
.
.
.
.
.
.
.
.
.
.
.
6­
2
LIST
OF
FIGURES
1.1
1.2
2.1
2.2
3.1
3.2
3.3
4.1
4.2
5.1
5.2
Page
Collective
Optimization
of
the
Quantity
of
Public
Good
Provided
.
.
.
.
.
.
.
.
.
.
.
.
,
.
.
.
.
.
.
.
1­
12
Hicksian
Surplus
Measures
.
.
.
.
.
.
.
.
.
.
,
,
.
.
.
.
.
.
.
1­
18
Questionnaire
­
Bidding
Game
for
Estimation
of
Recreationists'
Demand
for
Abatement
of
Aesthetic
Environmental
Damage
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
2­
7
Typology
of
Willingness
to
Pay
Studies
.
.
.
.
.
.
.
.
.
,
.
.
2­
19
Water
Quality
Ladder
Card
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3­
7
Payment
Cards
for
Versions
A,
B,
C,
D
For
People
with
Family
Incomes
of
$
10,000­
14,999
.
.
.
,
.
.
,
3­
17
Payment
Cards
for
Income
Levels
I­
IV
for
Version
A
.
.
.
.
.
.
3­
20
Distribution
of
WTP
Amounts
for
Fishable
Water
For
Versions
A,
B,
C
Combined
Including
Zero
Amounts
.
.
.
.
.
.
.
4­
8
Distribution
of
WTP
Amounts
for
Fishable
Water
For
Versions
A,
B,
C
Combined
Excluding
Zero
Amounts
.
.
.
.
.
.
.
4­
11
What
People
Are
Willing
to
Pay
Each
Year
Per
Household
For
Different
Levels
of
National
Water
Quality
.
.
.
.
.
.
.
.
5­
5
Actual
and
Predicted
WTP
Amounts
for
National
Swimmable
Water
Quality
by
Census
Region
.
.
.
.
.
.
.
.
.
.
.
5­
32
i
Executive
Summary
The
purpose
of
this
research
is
to
develop
a
methodology
for
using
direct
interview
survey
techniques
to
estimate
national
benefits
from
freshwater
water
quality
improvements.
In
particular,
this
study
has
developed
a
method
for
estimating
the
intrinsic
benefits
of
water
quality,
a
class
of
benefits
which
include
option,
existence,
and
bequest
benefits
among
others.
The
method
also
measures
consumer
recreational
benefits,
but
does
not
estimate
industrial,
commercial
or
drinking
water
benefits.

To
accomplish
our
purpose
we
adapted
the
contingent
valuation
or
willingnessto
pay
(
WTP)
survey
method
for
use
in
a
national
survey.
We
first
developed
and
pre­
tested
a
research
instrument
which
measures
how
much
people
are
willing
to
pay
each
year
in
taxes
and
higher
prices
for
national
water
quality
of
three
levels
which
we
defined
as
"
boatable,"
"
fishable,"
and
"
swimmable"
(
Chapter
3).

This
instrument
was
then
further
tested
in
a
full
scale
survey
where
it
was
administered
by
professional
interviewers
to
1576
people
comprising
a
nationwide
probability
sample.
For
experimental
purposes,
four
equivalent
sub­
samples
were
asked
different
versions
of
the
instrument.
We
performed
extensive
analysis
on
the
resulting
data
to
determine
the
extent
to
which
the
biases
associated
with
WTP
surveys
were
present
(
Chapter
4).
With
one
exception,
the
item
nonresponse
rate,
the
results
are
very
favorable.

Because
the
purpose
of
our
empirical
work
is
to
test,
validate
and
further
develop
the
methodology,
we
do
not
attempt
to
derive
national
estimates
from
these
data.
We
do,
however,
develop
illustrative
estimates
for
our
cases
which
suggest
aggregate
benefits
within
the
range
of
current
national
expenditures
on
water
pollution
control
(
Chapter
5).
Our
technique
for
estimating
ii
intrinsic
benefits
involves
identifying
those
respondents
who
do
not
use
water
for
"
in­
stream"
recreation
and
using
their
WTP
amounts
as
surrogate
for
intrinsic
benefits,
Our
calculations,
again
illustrative
rather
than
definitive,
suggest
intrinsic
benefits
comprise
roughly
40­
60
percent
of
the
overall
WTP
benefits
(
Chapter
5).

On
the
basis
of
these
empirical
tests
and
our
concurrent
work
on
several
important
theoretical
and
conceptual
issues
relevant
to
water
benefits
analysis
(
Chapters
1­
2),
we
conclude
that
the
use
of
a
national
survey
to
measure
water
benefits
(
including
intrinsic)
is
a
feasible
undertaking
We
specify
the
changes
in
our
pilot
instrument
and
its
administration
which
will
enable
it
to
perform
this
task
at
acceptable
levels
of
reliability
and
validity
(
Chapter
6).

The
following
are
some
of
the
major
findings
of
this
study
in
more
detail:

In
the
course
of
this
project
a
number
of
theoretical
and
conceptual
problems
inherent
in
the
direct
interview
survey
method
were
clarified
and
further
developed.
In
particular,
work
was
done
on
consumer
surplus
measures
(
p.
1­
13ff),
property
rights
(
1­
21ff)

and
the
classification
of
different
types
of
benefits
resulting
from
water
quality
improvements
(
1­
46ff).
A
number
of
conceptual
problems
arose
which
were
closely
integrated
with
the
theoretical
issues.
These
revolved
around
ill­
defined
property
rights
and
the
unworkability
of
willingness
to
accept
compensation
questions,

WTA.
Our
conclusion
was
that
theoretical
considerations
and
survey
considerations
must
both
be
considered
in
the
design
of
WTP
iii
instruments.
Thanks
to
the
recent
work
of
Randall
and
Stoll
(
1980)

and
Brookshire,
et
al.
(
1980),
however
,
we
show
that
any
theoretical
impurity
resulting
from
the
balancing
of
these
two
considerations
need
not
bias
the
results
as
the
correct
theoretical
measures
are
derivable
from
the
appropriate
survey
measures.
Our
conclusions
on
this
question
are
summarized
in
Table
1.3
on
p.
1­
23.

The
most
innovative
aspect
of
this
study
is
the
development
of
a
"
macro"
WTP
approach
to
benefits
estimation.
Previously,
WTP
surveys
had
been
used
primarily
to
assess
willingness
to
pay
for
locally
defined
goods
("
micro").
For
reasons
specified
in
the
report,

water
quality
benefits
lend
themselves
to
macro
WTP
estimation
at
the
national
level,
however.
Our
macro
approach
represents
the
first
time,
to
our
knowledge,
that
a
national
sample
was
surveyed
for
benefits
estimation
purposes
on
their
willingness
to
pay
for
a
public
good.
The
development
of
this
macro
approach
required
the
design
of
several
specialized
research
instruments
such
as
the
water
quality
ladder
(
A­
II)
and
non­
localized
benefits
questions.

One
clearly
advantageous
aspect
of
the
macro
approach
is
that,
if
correct
sampling
procedures
are
used,
individual
willingness
to
pay
for
water
quality
can
be
directly
and
reliably
aggregated
to
the
national
level,
The
sampling
techniques
used
to
accomplish
this
aggregation
were
implemented
in
the
survey
used
in
this
project
and
are
described
in
Chapter
4
(
4­
22ff)
and
Appendix
V.
iv
Our
pretest
showed
the
traditional
bidding
game
format
resulted
in
respondent
fatigue
and
a
serious
starting
point
bias
problem.
To
overcome
this
problem
we
developed
the
anchored
payment
card
(
3­
14ff).

To
test
for
bias
induced
by
the
payment
card,
its
format
was
systematically
varied
and
three
versions
of
the
instrument
were
administered
to
separate
sub­
samples.
As
this
experiment
showed
no
evidence
of
bias,
the
payment
card
is
a
promising
technique
for
WTP
studies
which
wish
to
avoid
the
bidding
game
format.

Strategic
and
hypothetic
biases
are
of
concern
to
economists
who
desire
to
use
benefits
derived
from
willingness­
to­
pay
surveys.
Our
major
conclusion
here
is
that
strategic
and
hypothetic
are
not
opposite
sides
of
the
same
bias
as
had
commonly
been
assumed
in
the
WTP
literature,
but
comprise
two
separate
and
distinct
potential
biases.
Table
4.3
on
p.
4­
22
shows
the
relationship
of
the
two
biases
and
which
WTP
question
characteristics
are
necessary
to
minimize
their
effects.
We
further
suggest
and
apply
to
our
data
several
tests
for
the
presence
of
strategic
and
hypothetic
biases.
These
tests
suggest
that
strategic
bias
is
not
a
problem
in
our
study.
Our
findings
with
respect
to
hypothetic
are
mixed
because
of
an
item
nonresponse
problem.
However,
regression
equations
estimated
in
Chapter
5
strongly
suggest
that
those
respondents
who
did
answer
the
WTP
items
did
not
do
so
in
a
random
fashion;
one
of
the
requirements
for
the
absence
of
hypothetic
biases.
v
The
item
nonresponse
problem
consists
of
a
high
level
of
no
response
to
the
WTP
questions
(
38
percent)
and
a
relatively
high
level
of
zero
amounts
(
16
percent).
This
problem
may
be
attributed
to
the
circumstances
of
the
interviews
(
it
was
not
possible
to
provide
the
interviewers
with
special
training
or
instructions
for
this
test
as
would
be
the
case
in
a
full
scale
implementation
of
the
method
and
the
WTP
questions
were
asked
after
respondents
had
answered
a
half
hour's
worth
of
questions
for
another
study)
and
the
question
wording
(
a
too
strong
incentive
was
offered
to
the
respondents
to
say
water
quality
wasn't
worth
anything
to
them)

(
4­
49ff).
Improvements
in
the
method,
as
suggested
in
Chapter
6,

should
reduce
this
problem
to
manageable
proportions.
Recommendations
are
made
for
weighting
procedures
(
6­
6)
which
can
adjust
for
the
remaining
missing
data.

In
order
for
WTP
benefit
estimates
to
be
credible,
a
theoretically
sound
predictive
model
must
be
constructed
and
tested,
We
have
hypothesized
the
primary
determinants
of
willingness
to
pay
amounts
for
water
quality
to
be:
income,
water
use,
and
environmental
attitudes.
To
measure
these
and
several
secondary
determinants,

we
chose
items
from
the
long
environmental
survey
which
preceded
the
WTP
survey.
Econometric
estimation
of
this
model
(
5­
15ff)
.

strongly
supports
our
theory,
The
estimates
are
robust
and
highly
significant
(
Table
5.5,
p.
5­
21).
A
special
test
for
heteroskedasticity
appropriate
for
equations
with
both
interval
and
dummy
data
was
developed
for
this
estimation
and
successfully
implemented
(
Appendix
VIII).
vi
Preface
This
study
represents
one
product
of
several
which
Resources
for
the
Future
has
prepared
under
a
Cooperative
Agreement
with
the
United
States
Environmental
Protection
Agency
for
"
Methods
Development
for
Assessing
Economic
Benefits
of
Water
Pollution
Water."
The
particular
methodological
approach
which
we
adopt
in
this
report,
a
macro
willingness
to
pay
survey,

emerged
as
we
studied
the
problem.
It
builds
on
a
tradition
of
innovative
research
using
the
willingness
to
pay
methodology
which
extends
back
to
the
1960s
and
which
has
flourished
during
the
1970s
as
economists
have
grappled
with
the
challenging
task
of
measuring
benefits.
Our
use
of
the
method
diverges
from
this
young
tradition
in
several
important
respects,
however,

and
in
this
sense
is
innovative
and
experimental.
In
the
course
of
changing
our
methodology
we
also
have
found
it
necessary
to
address
a
number
of
generic
methodological,
conceptual
and
theoretical
issues
pertaining
to
benefits
estimation.
The
fruits
of
our
thinking
on
these
issues
is
also
contained
in
this
report.
In
this
area,
too,
we
are
building
on
the
work
of
our
predecessors.

To
state
a
truism:
benefits
estimation
is
a
difficult
and
challenging
enterprise.
Several
years
ago,
Robert
Haveman,
commenting
on
a
paper
which
analyzed
60
benefit
studies,
declared:
"
To
me,
the
situation
is
...

extremely
discouraging,
because,
in
my
view,
what
has
passed
for
benefit
estimates
in
these
studies
forms
a
catalog
of
what
not
to
do
in
costbenefit
analysis"
(
Haveman,
1975).
In
our
endeavor
to
avoid
joining
this
infamous
roll
of
abortive
or
misguided
benefit
studies
we
hewed
as
close
vii
as
possible
to
the
six
methodological
criteria
set
forth
by
A.
Myrick
Freeman
III
in
The
Benefits
of
Environmental
Improvement
(
1979a;
10­
12)

and
to
his
dicta:

Part
of
the
art
of
benefit
analysis
involves
sensitivity
to
the
gap
between
the
ideal
and
the
available
and
knowing
how
much
confidence
to
place
in
the
estimates
being
generated.
(
1979a;
13).

To
help
the
reader
to
evaluate
the
extent
to
which
we
have
succeeded
in
this
task
we
provide
as
much
information
as
possible
in
this
report
about
the
possible
biases
in
our
method
and
how
we
have
tried
to
overcome
them.

In
the
case
of
the
major
problem
which
we
encountered,
item
nonresponse,

we
describe
in
detail
the
procedures
which
we
believe
can
resolve
the
problem
in
a
future
application
of
the
method.

The
structure
of
our
report
follows
from
this
approach.
In
the
first
two
chapters
we
discuss
crucial
theoretical
and
conceptual
matters.
Chapter
3
describes
our
research
instrument.
The
next
chapter
describes
the
potential
biases
which
threaten
the
validity
and
reliability
of
our
findings
and
our
success
in
dealing
with
the
problems
they
present.
Finally,
in
Chapter
5
we
present
our
findings.
Ever
mindful
that
benefit
estimates
take
on
a
life
of
their
own,
however
weak
their
methodological
and
conceptual
basis
may
be,
we
offer
our
findings
only
for
what
they
are:
experimental
data
to
test
a
method.
Our
findings
are
suggestive,
but
only
a
full
scale
application
of
a
revised
instrument
can
produce
estimates
of
sufficient
reliability
to
use
for
policy
purposes.
The
final
chapter
discusses
the
nature
of
the
revisions
we
propose.
viii
With
the
necessary
disclaimer
that
we
alone
are
responsible
for
the
work
reported
here,
we
wish
to
gratefully
acknowledge
the
assistance
of
our
colleagues
here
at
RFF
and
elsewhere.
We
have
benefitted
from
discussions
with
Ralph
d'Arge
and
David
Brookshire
of
the
University
of
Wyoming,
George
Tolley
of
the
University
of
Chicago,
Alan
Randall
of
the
University
of
Kentucky
(
while
on
leave
at
Chicago)
and
Alan
Carlin,
our
project
monitor
at
EPA.
At
RFF,
Raymond
Kopp
and
Michael
Hazilla
offered
us
much
useful
counsel
on
statistical
and
econometric
problems
as
has
William
J.
Vaughan.
Vaughan
also
prepared
our
water
quality
ladder,
helped
us
refine
our
theoretical
and
conceptual
ideas
and
collaborated
with
Carson
on
a
much
needed
constructive
test
for
heteroskedasticity.
Clifford
S.

Russell
was
extremely
generous
with
his
time
and
skillful,
as
always
in
raising
awkward
questions
and
wielding
an
editor's
pen.

Robert
Cameron
Mitchell
Richard
T.
Carson
June
15,
1981
Chapter
1
THE
WILLINGNESS
TO
PAY
METHOD,
CONSUMER
SURPLUS
AND
WATER
QUALITY
BENEFITS
In
valuing
environmental
amenities,
benefit­
cost
analysts
try
to
ascertain
what
individuals
would
be
willing
to
pay
and/
or
would
have
to
be
paid
for
a
particular
public
investment
in
a
world
where
markets
were
pervasive.
In
such
a
world
the
prices
for
marketed
goods
would
convey
information
sufficient
to
ascertain
what
"
the
gainers
and
losers
from
some
public
investment
will
consider
equivalent
in
value
to
their
respective
gains
and
losses"
(
Brookshire,
et
al.,
1979:
33).
Since
a
world
like
this
does
not
exist
for
public
goods
such
as
the
quality
of
the
nation's
freshwater
streams,
rivers
and
lakes
­­
the
subject
of
this
report
­­
the
shape
of
the
demand
curves
for
these
goods
cannot
be
determined
directly
and
economists
have
been
forced
to
develop
techniques
to
infer
the
value
of
these
goods.

According
to
Freeman
(
197!&
4)
there
are
three
approaches
to
determining
the
values
individuals
place
on
improvements
in
environmental
quality
when
markets
fail
or
are
nonexistent:
(
1)
holding
a
referendum
on
proposals
for
alternative
provisions
of
environmental
quality,
(
2)
using
market
data
for
substitutes
or
complements
of
the
environmental
quality
being
studied
in
order
to
determine
the
demand
curve
for
the
environmental
quality,
and
(
3)
direct
questioning
of
individuals
about
the
value
of
environmental
quality
to
them
personally.
The
first
method,
referendums,
have
not
been
used
in
determining
national
policy
on
any
environmental
quality
issues
and
few
legislatures
run
on
platforms
of
specific
provision
of
an
environmental
amenity.

The
second
method
is
the
determination
of
the
demand
curve
for
environmental
quality
indirectly
through
its
relationship
with
a
market
good.
This
technique
has
been
used
extensively,
particularly
in
the
area
of
recreation.

Examples
of
the
indirect
estimation
technique
include:
(
1)
the
determination
of
substitutability
in
household
production
functions,
1­
2
(
2)
the
travel
cost
method
which
assumes
that
a
complementary
position
exists
between
travel
cost
and
enjoyment
of
environmental
quality
and
(
3)
hedonic
pricing
which
assumes
that
environmental
quality
is
not
a
pure
public
good
and
that
a
consumer
can
substitute
(
trade)
market
goods
to
obtain
more
or
less
of
the
environmental
amenity.
(
Property
values
are
usually
used).

Each
of
these
three
methods
of
using
market
generated
data
has
limitations
which
are
unique
to
the
method.
In
addition
they
all
suffer
from
the
common
inability
to
estimate
the
demand
for
benefits
which
are
strongly
separable
in
utility
functions,
2
a
characteristic
which
severely
limits
their
utility
for
water
benefit
estimations.
Freeman
(
1979Q),
for
example,

suggests
that
environmental
amenities
which
are
not
directly
associated
with
private
good
consumptions
are
separable
from
a
utility
function
standpoint.

Existence
value
certainly
meets
this
criteria
and
thus
is
probably
a
separable
component
of
a
consumer's
utility
function,
Cicchetti
and
Freeman
(
1971)

argue
that
some
forms
of
option
value
are
probably
strongly
separable.
Hence
most
of
the
water
pollution
control
benefits
we
will
later
define
as
"
intrinsic"

and
which
are
a
primary
subject
of
this
report
are
not
capable
of
being
estimated
by
means
of
these
techniques.

1
See
Brookshire,
et
al.
(
1979),
Freeman
(
1979a)
and
Feenberg
and
Mills
(
1980)
For
critiques.

2Strongly
separable
utility
functions
take
the
form:

where
X
and
Y
are
subsets
of
marketable
goods
and
Q
is
the
public
good.
Changes
in
Q
have
no
effect
on
the
marginal
rates
of
substitution
of
any
of
the
marketable
goods.
For
a
discussion
of
separability
condition
in
utility
functions
with
respect
to
public
goods
see
Freeman
(
1979a:
70­
78)
or
Mahler
(
1974).
1­
3
The
third
approach,
which
is
the
one
employed
in
this
study,
uses
the
direct
technique
of
asking
people
in
surveys
what
they
are
willing
to
pay
or
to
accept
for
specified
levels
of
the
public
good.
The
use
of
surveys,
as
Brookshire,
et
al.
have
argued
at
length,
allows
the
analyst
to
shortcut
the
problems
inherent
in
the
indirect
method
by
"
positing
a
world
of
pervasive
markets
in
a
form
totally
consistent
with
theoretical
models
of
valuation
for
public
goods"
(
1979:
28).
Most
uses
of
the
WTP
method,

including
ours,
limit
themselves
to
hypothetical
markets
where
no
money
or
goods
actually
change
hands.
In
a
couple
of
intriguing
instances,
however,

researchers
have
used
the
method
in
the
context
of
a
simulated
market.
One
case
involved
subjects
paying
the
amount
they
bid
to
see
a
closed
circuit
TV
program
(
Bohm,
1972);
the
other
one
measured
hunters'
willingness
to
accept
money
for
Canadian
geese
hunting
permits
by
paying
them
the
amounts
they
were
willing
to
accept
in
exchange
for
a
surrender
of
the
permit
(
Bishop
and
Heberlein,
1980).
The
simulated
market
technique
has
little
applicability
to
most
environmental
goods
because
it
requires
exclusion
from
the
benefit
(
not
seeing
the
TV
program;
surrendering
the
hunting
license),
a
situation
which
is
inconsistent
with
how
public
goods
such
as
air
and
water
are
actually
provided
or
how
it
is
possible
to
provide
them
in
an
experimental
situation,

3We
use
WTP
for
convenience,
as
the
method
properly
refers
both
to
people's
willingness
to
pay
(
WTP)
for
a
public
good
or
their
willingness
to
accept
(
WTA)
compensation
for
the
imposition
of
a
public
bad.
1­
4
This
Study
The
objectives
of
this
study
are
to
design
and
validate
a
method
which
can:
1)
measure
the
national
benefits
of
freshwater
water
pollution
control
to
consumers
and
2)
determine
what
portion
of
these
benefits
come
from
in­
stream
recreational
values
(
e.
g.
boating,
fishing)
and
what
portion
from
the
intrinsic
or
non
in­
stream
recreational
values
(
existence,

option,
aesthetic,
etc.).
Very
little
empirical
work
has
been
done
on
the
latter
objective
and
no
previous
study
has
measured
the
former
using
the
WTP
method.
Our
method
employs
a
questionnaire
to
ask
a
national
sample
what
they
are
willing
to
pay
for
national
4
water
quality
of
specified
levels:
boatable,
fishable
and
swimmable.

We
adopt
the
willingness
to
pay
method
because
it
is
the
only
one
of
the
three
valuation
methods
which
can
be
used
to
estimate
intrinsic
benefits.
It
has
the
significant
added
advantage
that
willingness
to
pay
results
obtained
for
a
national
probability
sample
of
respondents
may
be
straightforwardly
blown
up
to
give
national
benefit
estimates.
Studies
using
an
indirect
method,
when
based
on
specific
sites,
present
a
problem
in
this
regard,
for
aggregation
from
single,
or
even
a
few,
sites
to
the
nation
as
a
whole
involves
problems
of
definition
and
computation.
(
See,

for
example,
the
companion
report
by
Vaughan
and
Russell
under
this
cooperative
agreement.)
This
method
is
not
without
its
problems
too,

which
we
will
discuss
at
considerable
length
insubsequent
chapters.
For
All
the
previous
uses
of
the
WTP
method
to
estimate
the
benefits
of
environmental
public
goods
were
limited
to
local
or
regional
studies.
For
reasons
we
will
discuss
in
subsequent
chapters,
the
characteristics
of
national
water
quality
and
its
benefits
are
such
as
to
make
a
national
WTP
survey
a
feasible
and
desirable
undertaking.
1­
5
the
moment
we
should
simply
note
that
the
methodological
requirements
for
a
successful
WTP
survey
are
formidable.
Not
only
must
the
instrument
describe
the
hypothetical
market
in
a
manner
which
meets
the
requirements
of
economic
theory,
it
should
also
be
understandable
to
respondents
with
less
than
a
high
school
education.
The
sampling
and
field
work
must
meet
high
standards,
and
the
sample
size
should
be
large
enough
to
permit
reliable
estimates.
The
fit
between
the
respondent's
experiences
and
the
hypothetical
situations
described
in
the
questions
must
be
close
enough
to
render
the
situations
meaningful
to
the
respondents.

In
this
chapter
we
discuss
briefly
the
willingness
to
pay
method
of
benefits
in
the
context
of
economic
theory
and
of
the
types/
which
accrue
from
water
pollution
control.
Our
purpose
is
to
clarify
the
theoretical
basis
for
our
measurements
and
to
review
the
literature
on
intrinsic
water
benefits.
1­
6
Benefit­
Cost
Analysis
The
purpose
of
this
study
is
to
estimate
certain
benefits
resulting
from
raising
the
ambient
level
of
fresh
water
quality
in
the
United
States.

These
benefits
are
one
side
of
a
benefit­
cost
analysis
and
may
be
defined
in
terms
of
the
(
Hicks­
Kaldor)
Pareto
optimality
conditions
(
Mishan,
1976a)
5
which
allow
for
the
possibility
that
those
who
gain
in
utility
by
a
change
in
state
can
compensate
those
who
lose
utility
as
a
result
of
the
change.

In
our
case,
where
water
quality
is
assumed
to
be
a
normal
good,
benefits
are
the
largest
amount
of
the
numeraire
the
individual
is
willing
to
pay
to
obtain
a
given
higher
level
of
water
quality,
while
costs
are
the
smallest
amount
that
those
producing
the
water
pollution
are
willing
to
accept
for
reducing
their
pollution
enough
to
achieve
the
specified
level
of
water
quality.
This
can
be
expressed
in
terms
of
utility
for
consumers
and
producers.

for
consumers
6
for
producers
where
the
initial
provision
of
good
W
a
higher
level
provision
of
good
W
Y
=
income
or
all
other
goods
(
numeraire)

B
=
the
amount
of
Y
consumers
are
willing
to
pay
to
obtain
W
Benefit­
cost
analysis
has
long
recognized
that
decision
makers
should
consider
criteria
other
than
economic
efficiency
in
implementing
a
policy,
in
particular
distributional
issues.
These
criteria
are
not
considered
in
this
study.
For
a
discussion,
see
Mishan,
1976a.

6
Since
this
is
a
study
of
benefits
rather
than
cost
we
will
not
consider
production
cost
and
producers
surplus
and
their
associated
problems
(
See
Mishan,
1976a).
1­
7
C
=
the
smallest
amount
of
compensation
that
producers
are
willing
to
accept
to
reduce
their
pollution
enough
to
achieve
W+

These
definitions
can
be
seen
to
be
those
of
the
Hicksian
(
1956)
compensating
measure
of
consumer
surplus,
a
topic
which
we
will
shortly
address
at
greater
length.
Following
Mishan
(
1976a,
1976b)
we
assign
a
minus
sign
to
cost
and
a
plus
sign
to
benefits
and
aggregate
over
consumers
and
producers,

The
standard
benefit­
cost
equations
for
a
change
from
one
state
to
another
can
be
expressed
in
terms
of
the
Hicksian
compensation
measure
as
follows:

Where
B
C
=
Total
benefits
of
the
proposed
change
C
C
=
Total
costs
of
the
proposed
change
CM
=
Hicksian
compensation
measure
The
discussion
thus
far
has
been
deceptively
simple.
We
now
need
to
address
the
complications
which
arise
from
the
special
characteristics
of
public
goods,
the
nature
of
public
policy,
and
the
limitations
of
the
survey
WTP
method.
These
matters
are
a
necessary
background
to
the
resolution
of
the
debate
over
exactly
which
consumer
surplus
measure
is
most
appropriate
for
measuring
the
benefits
of
environmental
amenities.
1­
8
Public
Goods
and
Public
Policy
Public
goods
such
as
national
levels
of
water
quality
are
those
which
once
produced
can
be
supplied
to
everyone
at
zero
marginal
cost
and
whose
enjoyment
by
one
person
does
not
interfere
with
the
enjoyment
of
another.
easily
Furthermore,
individuals
cannot/
be
excluded
from
enjoying
the
benefits
of
the
public
good,
once
it
is
produced.
These
goods
are
normally
produced
as
a
result
of
government
action,
either
by
government
requiring
firms
or
individuals
to
produce
the
goods
or
by
government
directly
subsidizing
this
production
from
tax
revenues.
Once
produced,
public
goods
are
usually
provided
free.
In
the
case
of
water
quality
Congress
declared
its
intent
in
the
Federal
Water
Polltuion
Control
Act
(
1972)
that
all
freshwater
bodies
reach
fishable
and
swimmable
quality
by
1983.
Private
firms
now
have
to
clean
their
water
discharges
to
meet
government
regulations,
and
the
federal
government
subsidizes
the
major
portion
of
a
waste
water
treatment
plant
construction
program
for
local
governments.

For
goods
which
are
provided
through
markets,
individuals
are
always
free
to
optimize
by
trading
along
their
budget
lines
in
order
to
reach
the
highest
indifference
curve
possible.
In
this
situation,
measuring
the
consumer
surplus
is
a
straightforward
problem.
This
is
not
the
case
for
national
water
quality,
however.
First,
since
"
clean
water"
is
a
public
good,
it
is
provided
free
to
citizen
consumers
who
wish
to
boat,
fish,

water
ski
or
simply
contemplate
it.
As
such
it
is
available
at
any
given
time
only
at
the
quality
level
provided
by
government
policy
irrespective
of
whether
some
consumers
are
willing
to
pay
more
for
higher
water
quality.

In
the
case
of
national
freshwater
this
quality
level
consists
of
two
1­
9
factors:
a)
the
ambient
quality
level
(
boatable,
fishable,
swimmable,

etc.)
and
b)
the
amount
of
the
overall
stock
of
freshwater
which
is
mandated
to
reach
a
specified
quality
level,
Thus
if
the
government
had
set
a
boatable
water
quality
standard
for
all
freshwater,
those
who
wished
to
have
a
higher
standard
for
the
nation's
water
(
e.
g.
fishable)
would
have
no
way
7
to
obtain
it
short
of
changing
government
policy.
Even
if
this
were
not
the
case,
it
would
still
make
no
sense
to
use
survey
techniques
to
ask
consumers
how
much
they
were
willinging
to
pay
for
the
quantity
and
quality
of
national
freshwater
they
regard
as
personally
optimal.
Let
us
say
person
A
might
be
willing
to
pay
$
339
a
year
for
national
water
of
fishable
quality
and
person
B
$
400
for
boatable
quality
water.
Once
having
obtained
data
in
this
form,
however,
it
is
impossible
to
aggregate
the
WTP
amounts
to
get
a
national
benefit
estimate
for
any
water
quality
but
the
highest/
level
for
which
WTP
amounts
are
available.
That
is,
we
can
reasonably
count
B's
amount
for
boatable
water
as
the
amount
which
he
would
also
be
willing
to
pay
for
the
higher,
fishable,
level
which
A
7
This
is
an
oversimplification,
of
course.
Many
public
goods,
fresh
water
included,
are
also
available
privately
at
a
cost.
Naturally,
national
water
quality
of
a
certain
level
can
only
be
provided
by
the
government.
But
a
consumer
faced
with
the
absence
of
public
lakes
and
streams
of
fishable
water
quality
in
his
or
her
locality
may
be
able
to
obtain
access
to
private
water
of
that
quality
for
a
fee
of
some
kind.
The
existence
of
numerous
private
swimming
pools,
clubs
and
beaches
attests
to
the
widespread
use
of
privately
supplied
water
for
swimming,
The
availability
and
desirability
of
these
optional
sources
of
water
presumably
influence
the
value
people
place
on
the
public
supply
of
freshwater.
1­
10
regards
as
optimal,
giving
a
total
WTP
amount
of
$
739
for
fishable
water.

We
cannot,
however,
reverse
the
process
and
determine
what
A
would
be
willing
to
pay
for
any
level
of
water
quality
below
his
optimum.
He
might
be
willing
to
pay
most
of
his
$
339
for
water
of
boatable
quality
or
he
might
not
be
willing
to
pay
anything
for
water
of
such
inferior
quality,

The
irrevelance
of
the
consumer's
willingness
to
pay
for
his
or
her
optimal
personal
provision
of
the
public
good
greatly
limits
the
range
of
consumer
surplus
measures
which
are
appropriate
for
the
study
of
national
water
benefits.
David
Bradford
(
1970),
in
an
expansion
of
Samuelson's
(
1954)
early
demonstration
that
the
demand
for
a
public
good
is
the
vertical
summation
of
individual
demand
curves,
takes
these
factors
into
account
in
developing
his
theoretical
framework
for
the
valuation
of
public
goods
in
benefit­
cost
analysis.
This
framework
and
its
subsequent
expansion
by
Randall,
Ives
and
Eastman
(
1974)
has
been
the
theoretical
basis
for
most
of
the
WTP
surveys.
Bradford
makes
the
assumption
that
individuals
choose
between
various
bundles
of
goods
which
may
differ
in
quantity
and
quality
and
proposes
the
concept
of
an
aggregate
bid/
benefit
curve
(
more
recently
referred
to
as
the
total
value
curve)
which
he
defines
as
the
vertical
summation
of
the
individual
bid
curves.
Because
of
this
assumption,
Bradford
was
able
to
demonstrate
that
over
any
relevant
range,

the
aggregate
bid
curve
and
its
corresponding
marginal
bid
curve
(
demand
curve)
need
not
be
continuous
or
downward
sloping.
If
the
aggregate
cost
is
known
and
the
marginal
cost
curve
is
derivable,
the
Bradford
framework
1­
11
resembles
the
traditional
profit
maximization
framework
with
the
optimal
production
occurring
where
the
marginal
aggregate
bid
curve
and
the
marginal
aggregate
cost
curve
intersect
(
See
Figure
1.1).
What
is
being
optimized
here
is
total
welfare
or
utility
rather
than
profits.
This
intersection
can
be
shown
to
be
the
point
where
the
rate
of
commodity
substitution
equals
the
rate
of
technical
substitution
which
is
the
traditional
welfare
economics
position
necessary
for
Pareto
optimality
(
Bradford,
1970;
Henderson
&
Quant,
1971).

Consumer
surplus
is
usually
used
as
the
measure
of
the
aggregate
benefit
curve.

This
caveat
should
be
added.
If
a
unidimensional
scale
(
underlying
metric)
is
unknown
or
does
not
exist,
it
will
be
impossible
to
estimate
the
demand
or
supply
curves
for
the
public
good.
This
means
that
only
specific
levels
of
production
can
be
compared
with
the
initial
level
or
with
other
specified
levels.
This
is,
however,
.
not
as
serious
a
problem
as
it
might
appear
since
policymakers
almost
always
choose
between
a
limited
number
of
alternative
policies,
the
benefits
of
which
can
be
measured
in
the
framework
we
present.
1­
12
Figure
1.1
COLLECTIVE
OPTIMIZATION
OF
THE
QUANTITY
OF
PUBLIC
GOOD
PROVIDED*

*
From
Randall,
Ives
and
Eastman
(
1974).
1­
13
Consumer
Surplus
The
concept
of
consumer
surplus
has
been
the
subject
of
considerable
debate
among
economic
theorists
(
Curie,
et
al.,
1971)
and
among
those
who
use
the
concept
i
benefit­
cost
analysis
it
has
been
the
subject
of
some
confusion
until
recently.
Consumer
surplus
was
at
the
center
of
the
welfare
economics
of
Marshall
and
Pigou.
After
a
period
of
neglect,
it
became
a
point
of
contention
between
two
eminent
theoretical
economists;
Paul
Samuelson
and
John
Hicks.

Samuelson
(
1947)
argued
that
consumer
surplus
was
a
subject
of
"
historical
and
doctrinal
interest,
with
a
limited
amount
of
appeal
as
a
mathematical
puzzle,"
a
view
echoed
more
recently
by
Silverberg
(
1978)
who
charged
that
"
attempts
to
use
consumer
surplus
to
measure
welfare
losses
are
largely
the
application
of
the
inappropriate
to
measure
the
undefinable."
Hicks,
on
the
other
hand,
argued
strongly
that
consumer
surplus
is
useful
to
welfare
economics
and
his
view
has
come
to
prevail
amongst
those
who
conduct
benefit­
cost
analysis.
(
1941,
1943,
1956)
Hicks
/
in
a
series
of
works
beginning
with
The
Revision
of
Consumer
Surplus
(
1941)
and
concluding
with
The
General
Theory
of
Demand
(
1956)

redefined
the
concept
in
an
attempt
to
overcome
the
objections
to
the
Marshallian
version.
He
developed
four
definitions
of
consumer
surplus
which
become
eight
when
both
price
increases
and
decreases
are
taken
into
account.
These
measures
are
set
forth
in
Table
1.1.
The
distinction
between
the
surplus
or
variation
measures
depends
on
whether
the
consumer
is
allowed
in
response
to
price
change
to
adjust
his
or
her
purchases
to
optimize
his
or
her
consumption/(
variation)

or
whether
the
consumer
is
simply
offered
fixed
quantities
of
a
particular
good
(
surplus),
The
second
set
of
distinctions
depends
upon
whether
the
Table
1.1
1­
14
TYPES
OF
CONSUMER
SURPLUS
MEASURES
FOR
CONTINGENT
VALUATION
STUDIES
Hicksian
Measures
Is
Choice
over
quantity
and
quality
provided?
Does
consumer's
initial
level
of
utility
change?
Does
consumer
buy
or
sell
the
good?
1­
15
consumer's
reference
point
is
his
or
her
initial
level
of
utility
or
not.

In
the
compensation
type,
the
individual
moves
along
the
indifference
curve
determined
by
his
or
her
present
utility.
In
the
case
of
the
equivalence
type,
the
individual
moves
from
a
point
on
one
indifference
curve
(
his
or
her
initial
utility)
to
a
point
on
another
indifference
curve,
Thus
the
equivalence
measure
always
represents
either
a
gain
or
a
loss
in
utility.

Since
none
of
these
measures
fulfills
the
need
for
a
single
concept
to
measure
welfare
loss
or
gain
from
various
price
or
quantity
changes,

analysts
have
to
choose
which
of
them
meets
the
requirements
for
their
particular
case.
Mishan,
for
one,
in
a
series
of
writings
(
1947,
1960,

1971,
1976a,
1976b)
argued
that
the
Hicksian
compensation
variation
measure
is
the
appropriate
measure
of
welfare
gain
or
loss
if
a
potential
Pareto
improvement
is
being
considered.
He
further
argued
that
the
variation
form
rather
than
the
surplus
form
is
the
correct
measure
of
consumer
surplus.

Mishan
went
so
far
as
to
drop
all
discussion
of
the
compensation
surplus
measure
in
his
later
works
including
his
influential
book,
Cost­
Benefit
Analysis
(
1971,
2d
ed.
1976a).
The
choices
between
surplus
and
variation,

and
compensation
and
equivalence,
were
much
discussed
during
the
1970'
s
as
analysts
conducting
the
WTP
surveys
tried
to
determine
which
consumer
surplus
measure
is
most
appropriate
for
the
case
of
non­
marketed
environmental
goods,
the
property
rights
for
which
are
ill
defined
and
which
are
provided
to
consumers
in
fixed
quantities.
The
appropriateness
of
measures
involving
paying
for
the
good
(
WTP)
versus
accepting
compensation
for
it
(
WTA)

was
also
discussed
and
tested
empirically
during
this
period.
We
conclude
1­
16
from
our
review
of
these
discussions
and
experiments
that
the
most
appropriate
measure
of
consumer
surplus
for
WTP
surveys
is
the
compensation­
surplus
WTP
measure
and
that
when
methodological
considerations
preclude
the
use
of
questions
in
this
form,
the
equivalence
surplus
WTP
measure
should
be
used,

Surplus
vs.
Variation
Let
us
address
the
surplus
vs.
variation
choice
first.
Mishan
relegated
implicitly
the
surplus
form
to
the
dust
heap,
a
position
taken/
by
others,
most
recently
Daniel
Feenberg
and
Edwin
S.
Mills
in
their
book
Measuring
the
Benefits
of
Water
Pollution
Abatement
(
1980).
As
we
have
shown
above,
however,
our
case
of
well
defined
levels
of
water
quality
fits
the
model
of
lumpy
goods
which
Randall
and
Stoll
(
1980)
have
shown
require
the
use
of
Hicksian
surplus
measures.
Since
our
case
is
typical
of
many
environmental
amenities,
the
surplus
measures
are
appropriate
for
most
WTP
surveys
because
only
they
measure
people's
willingness
to
pay
for
fixed
quantity/
quality
bundles
of
public
8
goods.

The
Surplus
Measures:
Definitions
This
leaves
four
measures
of
consumer
surplus
as
the
object
of
our
concern.
Before
proceeding
further
let
us
define
these
in
words
and
identify
them
graphically
as
follows:

8
Freeman
(
1979b),
after
correctly
distinguishing
the
variation
and
surplus
measures
according
to
Hicks'
definitions,
inexplicably
ignores
this
distinction
when
he
argues
that
if
people
are
only
offered
fixed
quantities
of
goods
the
compensating
variation
measure
is
equivalent
to
the
compensating
surplus
measure
and
hence
one
only
needs
concern
himself
with
the
variation
measures.
1­
17
WTP
(
CS)
­­
The
maximum
amount
a
consumer
is
willing
to
pay
to
obtain
a
prespecified
level
of
W
(
e.
g.
water
quality)
and
have
his
or
her
utility
remain
the
same
as
it
was
initially,

WTA
(
CS)
­­
The
minimum
amount
a
consumer
is
willing
to
accept
for
having
W
decline
to
a
prespecified
level
without
changing
his
or
her
utility.

WTP
(
ES)
­­
The
maximum
amount
a
consumer
is
willing
to
apy
to
avoid
having
W
lowered
to
a
prespecified
level;
either
the
change
in
water
quality
or
the
payment
will
lower
the
consumer's
utility.

WTA
(
ES)
­­
The
minimum
amount
a
consumer
is
willing
to
accept
to
forego
a
promised
increase
to
a
prespecified
higher
level
of
w.
Either
the
payment
or
a
higher
level
of
W
will
increase
the
consumer's
utility
level.

On
Figure
1.2,
if
the
initial
position
is
A,
and
the
prespecified
improvement
is
Q'
to
Q",
then
WTPc"
is
the
amount
of
Y
represented
by
the
line
segment
AE
VTPe'
is
the
amount
Y
represents
by
the
line
segment
xi?.
The
reduction
in
utility
is
accomplished
by
moving
the
consumer
from
D(
Q"
on
II)
to
A(
Q'
on
I).
The
consumer
is
then
indifferent
between
trading
z
amount
of
Y
to
get
back
to
Q",
the
original
endowment
of
W.

wThCS
is
the
amount
of
Y
represented
by
the
line
segment
CA.
The
consumer
in
this
case
is
moving
along
indifference
curve
II
going
from
Q"
to
Q'
in
exchange
for
AC
of
Y.
WTAeS
represents
an
increase
in
utility,
To
make
this
example
parallel
with
the
the
WTP
es
measure,
the
consumer
will
move
from
A(
Q'
on
I)
to
D(
Q"
on
II)
and
is
asked
how
much
Y
would
it
take
to
move
back
to
his
or
her
original
endowment
of
Q
but
remain
on
indifference
curve
II.

That
quantity
shown
on
the
graph
is
CB.
1­
18
Figure
1.2
HICKSIAN
SURPLUS
MEASURES
1­
19
pay
measures
are
equal
and
that
the
two
willingness
to
accept
measures
are
equal.
Further,
it
is
apparent
that
the
WTA
measures
are
not
income
bounded.

Without
proof
(
which
may
be
found
in
Willig,
1976;
Randall
and
Stoll,
1980;

and
Brookshire,
Randall,
and
Stoll,
1980)
we
cite
the
following
useful
From
Figure
1.2
it
is
readily
apparent
that
the
two
willingness
to
generalizations
about
the
relationship
between
the
measures:
(
1)
WTP
<

Marshallian
Consumer
Surplus
(
M)
<
WTA,
(
2)
for
the
case
of
zero
income
elasticity
of
income
for
the
public
good,
all
of
the
Hicksian
measures
are
equal
and
are
also
equal
to
the
Marshallian
(
M)
consumer
surplus,
and
(
3)
when
income
elasticity
(
price
flexibility
of
income
for
the
good)
is
small
(
generally
less
than
1)
and/
or
WTP
(
WTA)
is
small
relative
to
income
(
generally
5%
or
less)
the
bounds
between
WTP
and
WTA
have
been
rigorously
defined
and
are
usually
less
than
estimation
error.
From
these
findings
we
may
conclude
that
the
two
WTP
or
the
two
WTA
measures
may
be
freely
substituted
for
each
other
and
that
these
measures
will
be
close
to
the
Marshallian
consumer
surplus
observed
from
market
data
and
that
the
WTA
measures
could
be
derived
from
the
WTP
measure
or
vice
versa.
Empirically
the
bounds
between
the
WTA
and
WTP
testable
measures
would
be
/
if
it
were
not
for
respondents'
aversion
to
the
WTA
measures
which
we
discuss
shortly.

9
Price
flexibility
of
income
for
a
good
is
analogous
for
the
income
elasticity
for
a
good
except
that
only
specified
quantities
of
the
good
are
supplied
(
Randall
and
Stoll,
1980).
1­
20
Table
1.2
Implied
Property
Rights
Yes
(
consumer
holds)

NO
(
consumer
does
not
hold)
TYPES
OF
PROPERTY
RIGHTS
FROM
THE
CONSUMER'S
PERSPECTIVE
Legal
Property
Rights
Yes
(
Vested
by
Law)
No
(
Not
Vested
by
Law)

Legal
property
Hypothetical
"
Squatters
Rights"

Non
property
1­
21
Criteria
for
Choosing
Between
the
Hicksian
Surplus
Measures
Now
that
we
have
defined
the
four
types
of
Hicksian
surplus
measures
let
us
consider
them
from
the
standpoint
of
measuring
consumer
surplus
in
WTP/
A
surveys.
10
They
are
formed
by
combinations
of
two
set
of
distinctions:

equivalence
vs.
compensation
and
willingness
to
pay
vs.
willingness
to
accept.
To
determine
which
combination
is
the
correct
measure
for
an
environmental
good
being
valued
in
a
WTP/
A
survey
we
need
to
compare
the
property
rights
posited
in
the
questionnaire
with
the
actual
distribution
of
property
rights
for
that
good.
Before
making
our
argument
we
need
to
distinguish
two
types
of
property
rights.
The
usual
sense
of
property
right
is
a
right
vested
by
law.
In
much
of
what
follows
we
speak
of
property
rights
in
different
sense,
as
the
actual
endowment
of
goods
held
by
a
person,
to
which
he
or
she
can
add
or
subtract
(
Silverberg,
1978).
Freeman
calls
this
"
implied
property
rights"
(
1979b).
Table
1.2
shows
the
relationship
between
these
two
types
of
property
rights,
names
the
categories,
and
locates
the
boatable,
fishable,
swimmable
levels
of
water
quality.

Speaking
now
of
property
rights
(
implied),
the
initial
endowment
or
implied
property
right
defines
the
initial
indifference
curve
that
the
consumer
is
on.
Additions
or
subtractions
of
goods
to
the
consumer's
initial
bundle
of
good
which
are
counterbalanced
(
thereby
preserving
the
same
utility
level)
are
Hicksian
compensation
measures.
Changes
in
the
initial
endowment
or
implied
property
right
which
are
not
exactly
counter
balanced
(
thereby
shifting
the
consumer
to
another
indifference
curve)

are
equivalence
measures.
From
the
standpoint
of
the
individual
10
At
this
point
we
will
temporarily
refer
to
these
surveys
as
WTP/
A
in
order
to
avoid
terminological
confusion.
1­
22
consumer,
if
producers
have
the
right
to
pollute
waterways
then
consumers
must
bribe
them
into
not
polluting
if
the
consumers
desire
better
water
quality.
This
calls
for
a
WTP
measure.
In
the
opposite
case,
where
consumers
own
the
right,
producers
must
bribe
the
consumers
if
they
wish
to
pollute
and
a
WTA
measure
should
be
used.
Compensating
surplus
measures
are
appropriate
when
the
contingent
situation
described
to
respondents
in
a
WTP/
A
study
uses
the
same
distribution
of
property
rights
as
actually
exists
at
the
time
of
the
study.
In
this
case
there
is
no
redistribution
implied
in
the
instrument
and
the
potential
Pareto­
improvement
becomes
the
proper
criterion.
Where
the
instrument
posits
a
property
right
which
differs
from
the
existing
situation,
redistribution
is
implied
and
the
equivalence
surplus
measure
is
called
for
(
Mishan,
Table
1.3
cross­
tabulates
the
existing
and
the
contingent
property
rights
to
show
which
measures
of
consumer
surplus
are
theoretically
correct
for
the
four
combinations.
While
these
distinctions
are
clear
theoretically,
in
practice
they
are
difficult
to
apply
to
WTP/
A
instruments.
We
will
illustrate
this
difficulty
by
discussing
our
choice
of
consumer
surplus
measures
and
why
we
believe
WTP/
A
surveys
are
restricted
to
the
equivalence
and
compensating
WTP
measures.

We
sought
to
measure
the
respondent's
consumer
surplus
for
three
levels
of
national
water
quality:
boatable,
fishable
and
swimmable.
To
identify
the
theoretically
appropriate
consumer
surplus
measure
we
had
to
decide
what
property
right
(
implied)
consumers
presently
have
for
these
environmental
amenities.
The
Federal
Water
Pollution
Control
Act
(
as
amended)
1­
23
Table
1.3
ROLE
OF
PROPERTY
RIGHTS
IN
DETERMINING
THE
RELATIONSHIP
OF
THE
WTP,
WTA,
COMPENSATING
AND
EQUIVALENCE
DIMENSIONS
OF
CONSUMER
SURPLUS
MEASURES
FOR
WTP/
A
SURVEYS
VALUING
ENVIRONMENTAL
PUBLIC
GOODS
Contingent
Property
Right
Specified
in
the
Questionnaire
(
Implied)

Consumers
own
Consumers
do
not
own
Consumers
Own
Consumers
Do
not
own
Compensating
Equivalence
WTA
WTA
Equivalence
WTP
Compensating
WTP
1­
24
endows
the
public
(
individual
consumers)
with
a
legal
entitlement
to
fishable/
swimmable
water
nationwide,
the
goal
specified
in
the
Act
to
be
achieved
by
1983.
Its
Congressional
architects
declared:
"
This
legislation
would
clearly
establish
that
no
one
has
the
right
to
pollute
­­
that
pollution
continues
because
of
technological
limits,
not
because
of
any
inherent
right
to
use
the
nation's
waterways
for
the
purposes
of
disposing
of
wastes"
(
Rosenbaum,
1977:
159).
Feenberg
and
Hills
(
1980),
however,

contend
that
in
practice
property
rights
to
water
quality
are
ill
defined
and
in
a
state
of
flux.
We
agree
and
think
this
is
particularly
the
case
from
the
consumer's
point
of
view.
Many
consumers
are
personally
unaware
of
the
national
goal.
What
they
hear
about
is
national
freshwater
lakes
and
stream
virtually
all
of
which
are
at
the
boatable
level
at
the
present
time,
although
what
they
experience
locally
may
be
of
higher
quality.
In
this
context
and
with
regard
to
the
overall
national
level
of
water
quality
which
is
the
public
good
we
are
valuing,
we
believe
the
implied
property
right
is
such
that
it
is
appropriate
to
treat
freshwater
of
boatable
quality
as
if
the
rights
to
it
are
actually
owned
by
consumers
and
to
regard
rights
to
water
of
higher
quality
as
not
(
yet)
owned
by
them.

When
it
comes
to
deciding
how
to
specify
the
property
right
(
implied)

in
our
questionnaire
theoretical
purity
gave
way,
as
we
believe
it
must,

to
methodological
realism.
In
theory
the
distribution
of
property
rights
(
implied)
for
water
quality,
as
specified
above,
should
be
replicated
in
the
questionnaire.
If
we
did
this
the
consumer
surplus
associated
with
1­
25
boatable
water
over
some
base
(
very
low
quality)
level
would
have
to
be
measured
by
a
compensating
WTA
question
and
swimmable
and
fishable
water
by
a
compensating
WTP
question.
While
we
followed
this
theoretically
desirable
practice
for
the
swimmable
and
fishable
levels
(
Qs.
83,
84),

for
methodological
reasons
we
measured
the
boatable
level
(
Q.
82)
with
an
equivalence
WTP
measure
instead
of
a
compensating
WTA
item.

We
made
this
substitution
because
the
hypothetical
market
presented
in
WTP/
A
instruments
must
accord
sufficiently
with
the
respondents
frame
of
reference,
otherwise
respondents
will
give
meaningless
answers.
Clearly,

asking
our
respondents
how
much
they
are
willing
to
pay
for
higher
(
fishable
and
swimmable)
levels
of
water
quality
than
they
presently
enjoy
meets
the
frame
of
reference
test
especially
as
compared
with
the
alternative
of
asking
them
to
accept
compensation
for
reductions
in
levels
which
they
have
not
yet
received
(
VTA").
The
use
of
XTPcs
is
not
appropriate
for
boatable
water,
however,
since
the
respondents
already
enjoy
national
water
of
that
quality.
It
would
be
inconsistent
to
have
them
pretend
that
national
water
quality
is
non­
boatable
and
to
ask
them
how
much
they
are
willing
to
pay
to
raise
it
to
the
boatable
level.
The
theoretically
appropriate
measure,
tiXcsi
also
fails
the
frame
of
reference
test.
Analysts
who
have
attempted
to
ask
WTA
questions
report
that
an
unacceptably
large
number
of
respondents
respond
to
WTA
questions
by
either
refusing
to
answer
the
questions
or
by
saying
there
is
no
price
they
would
accept
for
the
loss
of
environmental
quality
being
valued.
In
one
study
of
the
value
people
place
on
visibility
in
the
Pour
Corners
region
52
and
51
percent
of
1­
26
two
samples
recorded
infinity
bids
for
the
WTA
questions
(
Eastman,
et
al.,
1974:
581)

In
another
study
of
the
value
of
hunting
to
hunters,
54
percent
refused
to
accept
any
finite
amount
of
compensation
(
Brookshire,
et
al.,
1980:
487).
The
WTA
format
places
respondents
in
a
situation
which
is
both
unfamiliar
and
which
is
perceived
by
many
as
unfair.
People
are
not
accustomed
to
being
offered
compensation
for
environmental
goods
and
apparently
some
feel
offended
by
the
notion.
These
considerations
lead
researchers
who
have
experimented
with
the
WTA
format
to
conclude:
"
We
cannot
recommend
compensation
(
WTA)
games"
(
Eastman,
et
al.,
1974:
583)
and
"
iterative
bidding
formats
for
the
direct
observation
of
WTACSdo
not
appear
to
collect
reliable
value
data"
(
Brookshire,
et
al.,
1980:
488).

Fortunately
the
empirical
consequences
of
yielding
to
methodological
considerations
in
the
choice
of
the
consumer
surplus
measure
are
minor.
Randall
et
al.
(
1980)

and
Stoll
(
1980)
and
Brookshire
have
calculated
rigorous
bounds
for
the
difference
between
WTP
and
WTA
measures.
Using
their
equations
the
WTA
measures
can
be
derived
from
the
WTP
measure
and
the
differences
between
the
two
are
small.
For
example,
using
equations
(
11)
and
(
13)
WTA
­
WTP
of
Randall
and
Stoll
(
1980),
and
assuming
for
illustrative
purposes
the
price
flexibility
of
income
=
.7,
household
income
(
Y)
=
$
18,000
and
WTP
=
$
250,
/
WTA
can
be
derived
from
WTP,
Equation
(
11)
is
solvent
for
M
using
a
quadratic
and
then
substituted
into
equation
13.
The
difference
between
the
WTP
and
the
WTA
measures
is
approximately
$
2.50
or
1
percent
of
WTP.
1­
27
The
Nature
of
Water
Benefits
Water
pollution
has
a
wide
range
of
effects
on
various
types
of
consumers
and
potential
consumers.
Insofar
as
these
effects
are
harmful,

they
impose
"
costs."
Since
the
expense
of
reducing
pollution
involves
another
type
of
"
cost"
we
can
avoid
unnecessary
semantic
confusion
by
calling
the
losses
imposed
by
a
reduction
of
environmental
quality
"
damages,"

and
the
gains
associated
with
reduced
pollution
"
benefits"
(
Freeman,
1979b).

The
basis
for
determining
what
is
to
be
regarded
as
a
damage
or
a
benefit
is
individuals'
preferences
about
the
ideal
state
of
the
world.
We
tend
to
assume
a
societal
consensus
about
which
effects
of
a
given
change
in
pollution
should
be
defined
as
benefits
and
which
as
damages,
but
such
a
consensus
is
not
inevitable.
If,
for
example,
a
significant
segment
of
the
population
harbored
an
aesthetic
preference
for
misty
landscapes
they
might
regard
a
reduction
in
air
visibility
from
100
miles
to
40
miles
caused
by
the
operation
of
large
scale
coal­
fired
power
plants
in
the
Southwest
as
a
benefit
rather
than
as
a
damage.
Fortunately,
a
strong
consensus
does
seem
to
exist
as
to
which
environmental
changes
should
be
considered
benefits
and
which
as
damages;
otherwise
benefit
estimation
would
be
even
more
complex
than
is
currently
the
case.
The
consensus
does
not
extend
to
the
amount
of
the
benefit
created
by
a
change
in
an
environmental
good.
Since
this
varies
across
individuals,
"
We
define
the
benefit
of
an
environmental
improvement
as
the
sum
of
the
monetary
values
assigned
to
these
effects
by
all
the
individuals
directly
or
indirectly
affected
by
that
action"
(
Freeman,
1979b:
3).
1­
28
As
the
benefits
associated
with
changes
in
an
environmental
media
such
as
water
are
diverse,
any
attempt
to
estimate
benefits
must
specify
which
benefits
are
to
be
measured
and
which
are
not.
Otherwise
certain
benefits
may
be
inadvertently
left
out
or
others
may
be
overestimated
due
to
double
counting.
There
are
several
lists
of
the
benefits
of
improved
water
quality
in
the
literature
(
Feenberg
and
Mills,
1980;
Freeman,
1979a),
none
of
which
is
full
satisfactory.
Table
1.4
offers
our
categorization
of
water
benefits.
It
builds
on
previous
distinctions
for
the
most
part,
but
adds
a
category
of
non­
direct
use
benefits
which
we
call
"
indirect"
benefits
and
assembles
all
the
nondirect
use
benefits
in
a
single
"
intrinsic"
category.

Direct
use
refers
to
activities
which
currently
use
water
either
by
withdrawal
or
by
instream
use
(
Feenberg
and
Mills,
1980:
8).
Improved
water
quality
in
freshwater
rivers,
streams
and
lakes
can
result
in
a
variety
of
withdrawal
11
benefits.
Industries
which
require
water
of
a
certain
quality
for
their
processes
might
have
lower
water
purifying
costs
and
less
damage
to
equipment
which
uses
water.
Likewise
the
costs
of
purifying
water
for
use
in
washing
agricultural
produce
might
be
lowered.
Drinking
water
benefits
would
occur
if
the
improved
quality
of
raw
water
supply
sources
lowers
the
costs
of
purification
and/
or
reduces
the
health
damage
by
previously
unremoved
pollutants.
(
A
companion
report
to
EPA
under
our
cooperative
agreement
by
Mark
Sharefkin
addresses
the
question
of
drinking
water
benefits.)

Instream
use
benefits
occur
in
two
ways:
via
increased
output
or
lower
costs
in
commercial
fisheries
and
via
the
array
of
activities
­­

11
These
are
comparable
to
what
Freeman
(
1979a)
calls
"
diversion
uses."
1­
29
NATIONAL
BENEFITS
OF
CLEAN
WATER
Table
1.4
Withdrawal
­­­
Industry
Agriculture
Drinking
water
Direct
Use
­­­­
Commercial
Fisheries
Instream
­­­­
Boating
Recreation
Fishing
Swimming
Habitat
based
­­­

Indirect
­­­­

Aesthetic
­­­­­­­

Intrinsic
­­­­­

Personal
­­­­­­

Option
­­­­­­

Intergenerational
Consumptive
recreation
(
i.
e.
duck
hunting)

Nonconsumptive
recreation
(
wildlife
watching
and
photography)

Water
enhanced
recreation
(
i.
e.
picnicing,
camping,
sightseeing,
other)

Other
activitiy
(
i.
e.
commuting
to
work)

Short
term
Long
term
(
bequest)

Existence
1­
30
fishing,
swimming,
boating
and
the
like
­­
by
which
people
use
water
to
recreate.
These
activities
are
very
popular
­­
two­
thirds
of
our
respondents
said
they
participated
in
at
least
one
of
these
activities
during
the
past
two
years
­­
and
attempts
to
estimate
national
direct
use
recreational
benefits
assign
them
a
considerable
portion
of
the
total
benefits
of
water
pollution
control.
Freeman
(
1979a),
gives
a
set
of
best
point
estimates
where
recreation
accounts
for
55%
of
the
total.
Most
studies
of
water
recreation
benefits
either
use
the
travel
cost
or
the
"
participation
model"

approach,
although
Gramlich
(
1977)
used
the
WTP
method
to
estimate
the
benefits
of
achieving
swimmable
quality
water
in
the
Charles
River
Basin
of
Massachusetts.

Water
quality
benefits
extend
beyond
its
direct
use
to
include
various
indirect
and
potential
uses.
We
call
this
category
of
values
intrinsic
since
they
stem
from
the
inherent
characteristics
of
freshwater
bodies.

Although
important,
these
benefits
are
less
tangible
than
the
direct
use
benefits
and
are
consequentially
less
studied.
Freeman's
forty
page
review
of
recent
water
benefit
estimates
devotes
a
mere
two
pages
to
the
handful
of
studies
on
this
topic
and
concludes:
"
This
is
a
very
tenuous
empirical
basis
from
which
to
estimate
national
non­
user
benefits"

(
1979a:
162).
Prior
to
the
present
research
no
study
attempted
to
measure
1­
31
the
overall
intrinsic
benefit
for
water
quality
and
the
existence,

intergenerational
and
option
values
of
water
quality
were
measured
by
only
a
single
regional
study
using
the
WTP
method.
(
This
study
is
described
in
Walsh,
et
al.,
1978
and
Greenley,
et
al.,
1980).

We
divide
intrinsic
benefits
into
three
major
sub­
categories:
indirect,

option
and
existence.
Indirect
benefits
occur
when
water
supports
or
enhances
out­
of­
stream
(
non
water
contact)
activities.
Duck
hunters
and
bird
watchers
who
enjoy
observing
acquatic
species
benefit
from
the
availability
of
marshes
and
lakes
which
provide
the
necessary
habitat
for
these
12
birds.
Fresh
water
is
an
aesthetically
pleasing
setting
for
such
diverse
recreational
experiences
as
picnicking
by
a
stream,
hiking
in
wilderness
areas,
strolling
through
a
New
England
village
located
on
a
river,
or
visiting
the
gambling
casinos
at
Lake
Tahoe.
Aesthetic
benefits
also
accrue
to
people
for
whom
rivers,
lakes
or
streams
serve
as
a
backdrop
to
their
normal
activities.
Although
some
would
list
property
values
here
as
a
distinct
indirect
benefit
category,
we
believe
property
values
should
be
regarded
as
a
surrogate
measure
of
aesthetic
and
recreational
benefits.

Adding
them
to
the
list
would
result
in
double
counting
(
Freeman,
1979b).

12
In
a
recent
paper,
Hay
and
McConnell
(
1979)
review
the
sparse
literature
on
the
value
of
non­
consumptive
wildlife
recreation
and
attempt
to
estimate
the
reduced
form
participation
model
demand
for
such
activities.
For
comments
on
statistical
procedures
see
Vaughan
and
Russell
(
1981)
and
Hayward
and
McConnell
(
1981).
1­
32
Weisbrod
(
1964)
first
identified
option
values
as
an
additional
form
of
benefit
that
must
be
added
to
the
consumer
surplus
measure.
The
essential
nature
of
option
value
is
contained
in
Greenley,
et
al.'
s
definition
(
1980)
of
option
value
as
a
willingness
to
pay
for
the
"
opportunity
to
choose
from
among
competing
alternative
uses
of
a
natural
environment
in
the
future."
We
distinguish
between
option
value
based
on
whether
the
individual
values
the
future
opportunity
to
choose
for
his
or
her
personal
use
(
personal
option
value)
or
the
use
of
future
generations
(
intergenerational
option
value).

Let
us
consider
personal
option
benefits
first.
These
benefits
refer
to
the
value
people
place
on
a
particular
environmental
amenity
on
the
chance
that
they
personally
may
wish
to
use
that
amenity
at
some
time
in
the
future.
Among
the
three
conditions
which
Weisbrod
asserted
must
be
met
for
determining
the
presence
of
option
value
is
that
a
decision
about
supplying
the
amenity
in
the
future
is
about
to
be
made
and
should
that
decision
be
negative
it
would
be
very
difficult
or
impossible
to
reestablish
it
(
Cicchetti
and
Freeman,
1971:
528).
There
are
two
situations
where
this
condition
holds
and
we
distinguish
between
what
we
call
short
term
and
long
term
individual
option
value
on
the
basis
of
these
conditions.
The
first
is
where
present
use
or
failure
to
protect
12a
an
amenity
will
damage
it
irreversibly.
If
the
damage
can
be
reversed
in
the
future
(
at
some
expense
of
course)
and
the
individual
does
not
expect
to
exercise
the
option
in
the
"
near"
future,
the
individual
need
not
make
a
present
choice
between
the
damaging
use
and
non­
use
to
preserve
12a
Our
use
of
irreversibility
extends
to
situations
where
the
damage
could
be
undone
at
a
future
date
but
at
a
much
greater
expense.
1­
33
his
or
her
use
option.
We
define
long
term
option
value,
therefore,
as
the
value
people
place
on
a
good
which
is
regarded
as
facing
possible
irreversible
damage.

What
about
the
situation
where
an
individual
is
uncertain
about
whether
or
not
he
or
she
may
wish
to
use
an
amenity
in
the
relatively
near
future?

Under
certain
conditions
such
an
amenity
will
have
option
value
for
a
person
even
when
it
is
not
threatened
with
irreversible
damage.
We
call
this
short
term
option
value
which
we
define
as
the
price
people
will
pay
to
have
the
option
to
use
a
good
immediately
or,
in
the
case
where
a
period
of
repair
(
e.
g.
pollution
control)
is
required
to
make
the
good
usable,
to
use
the
good
as
soon
as
possible.
Unless
the
person
wishes
to
use
this
(
non­
irreversibly
threatened)
good
as
soon
as
possible,

however,
it
should
have
no
option
for
him
or
her.
For
example,
Lake
W.

is
not
now
swimmable
because
of
seepage
from
septic
tanks
but
if
a
sewage
treatment
plant
were
constructed
it
could
be
made
swimmable
in
five
years,

It
is
not
threatened
with
irreversible
pollution.
If
person
X
wishes
to
have
the
option
to
swim
in
the
lake
as
soon
as
possible
(
e.
g.
five
years
from
now),
he
or
she
has
a
short
term
option
value
for
that
amenity.

If
the
person
has
a
longer
option
time
frame,
however,
it
would
make
no
sense
for
the
person
to
express
a
WTP
option
value
today
since
the
potential
to
clean
the
lake
up
after
a
five
year
effort
will
continue
to
exist.
Put
another
way,
since
the
damage
can
be
reversed
in
the
future
the
individual
need
not
make
a
present
choice
between
the
damaging
use
(
continued
use
of
septic
tanks)
and
a
cleanup
program
to
preserve
his
or
1­
34
her
option
to
swim
in
the
lake.
Instead,
he
or
she
should
use
his
or
her
money
for
other
purposes.
Table
1.5
summarizes
the
conditions
under
which
people
will
hold
long
and/
or
short
term
option
values
for
environmental
goods.

Intergenerational
or
bequest
option
benefits
comprise
the
willingness
of
members
of
the
present
generation
to
pay
to
endow
succeeding
generations
with
some
natural
environment.
Some
individuals
may
place
a
value
on
preserving
such
amenities
as
streams
from
being
essentially
destroyed
by
strip
mining
operations
simply
because
they
would
feel
better
knowing
that
these
streams
would
still
be
available
for
their
children
or
future
generations
to
use
if
they
want
to.
A
parallel
argument
is
made
by
some
that
ecosystems
and
species
should
be
preserved
even
when
they
have
no
present
"
use"
because
the
reduction
of
genetic
diversity
in
this
manner
reduces
the
possibilities
available
to
future
generations
to
use
such
species
in
the
ways
we
are
presently
unable
to
imagine.
This
perspective
has
become
law
in
the
Endangered
Species
Act
and
was
instrumental
in
delaying
the
construction
of
the
Tellico
Dam
in
Tennessee
when
it
was
found
that
the
dam
threatened
an
endangered
species
of
minnow,
the
snail
darter.
1­
35
Table
1.5
CONDITIONS
UNDER
WHICH
PEOPLE
MAY
HOLD
LONG
AND
SHORT
PERSONAL
OPTION
VALUES
FOR
ENVIRONMENTAL
GOODS
No
Irreversible
Threat?

Desire
to
have
option
to
use
Yes
No
good
as
soon
as
possible?
Long
and
Yes
short
term
Short
term
Long
term
No
option
value
1­
36
As
with
the
long
term
personal
option
value,
these
benefits
rest
on
the
assumption
that
the
action
taken
by
the
present
generation
poses
an
irreversible
threat
to
the
environmental
good
in
question.
It
is
worth
noting
here
that
the
benefit­
cost
analysis
procedures
in
current
use
effectively
value
benefits
or
costs
a
generation
or
more
in
the
future
at
zero
by
imposing
real
discount
rates
of
about
5
percent
(
Ben­
David,
et
al.,
1979:
33).

The
only
empirical
study
of
the
option
values
of
water
quality
is
by
a
team
of
economists
from
Colorado
State
University
who
designed
a
WTP
instrument
on
the
basis
of
Henry's
(
1974)
option
value
mode.
Henry's
model
posits
the
"
preservation
of
an
irreplaceable
environmental
asset
facing
an
imminent
irreversible
commitment,
until
such
time
that
sufficient
information
becomes
available
affecting
the
future
option
decision
of
selecting
from
among
alternative
uses"
(
Greenley,
et
al.,
1980:
3).
The
researchers
interviewed
a
sample
of
two
hundred
and
two
residents
of
Denver
and
Port
Collins.
In
order
to
measure
the
recreation,
option,
existence
and
preservation
benefits
of
different
levels
of
water
quality
in
the
South
Platte
River
Basin
(
Northeast
Colorado)
the
respondents
were
asked
a
formidable
array
of
willingness
to
pay
questions
(
twelve
in
all)
using
the
bidding
game
format.
The
personal
(
short
term)
option
value
question
posed
two
alternatives
for
the
Basin.
Alternative
I
featured
a
large
expansion
in
mining
development
which
would
severely
pollute,
in
an
irreversible
fashion,
"
many"
lakes
and
streams.
Under
Alternative
II,

any
decision
to
expand
mining
would
be
postponed
1­
37
until
information
became
available,
sufficient
for
the
respondent
to
make
a
decision
"
with
near
certainty
as
to
whether
it
is
more
beneficial
to
you
to
preserve
the
waterways
at
level
A
(
the
highest
level)
for
your
recreational
use
or
to
permit
mining
development"
(
Greenley,
et
al.,
1980:
13).
Using
an
additional
fraction
of
a
percent
to
the
region's
sales
tax
as
the
payment
13
vehicle,
an
annual
mean
bid
per
household
of
$
22.60
to
postpone
development
was
reported
for
the
177
respondents
who
answered
the
question.
The
study
also
measured
intergenerational
option
benefits
by
asking
the
following
question:

Q.
28
If
it
were
certain
you
would
not
use
the
South
Platte
River
Basin
for
water­
based
recreation
[
which
they
defined
as
including
both
direct
and
indirect
recreational
use],
would
you
be
willing
to
add
cents
on
the
dollar
to
present
sales
taxes
every
year
to
ensure
that
future
generations
will
be
able
to
enjoy
clean
water
at
level
A?
(
Walsh,
et
al.,
1978:
82).

A
bequest
value
of
$
16.97
a
year
per
household
is
reported
for
a
subsample
14
of
24
non­
recreationists.

13
They
repeated
each
bidding
game
using
a
second
bidding
vehicle,
an
additional
charge
to
the
respondent's
water
bill.

14
The
researchers
eschew
using
the
intergenerational
option
benefit
amount
for
the
recreators
in
their
sample
because
they
doubt
the
recreator's
ability
to
leave
out
their
personal
recreational
considerations
when
answering
this
question.
Based
on
our
review
of
the
instrument
this
is
the
correct
decision,
but
it
reduces
the
sample
size
so
much
that
the
bequest
estimate
can
only
be
regarded
as
suggestive
(
Greenley,
et
al.,
1980:
15,
33).
1­
38
The
final
type
of
intrinsic
benefit
is
existence
benefits.
In
1967
Krutilla
wryly
commented
regarding
wilderness
that:
"
There
are
many
persons
who
obtain
satisfaction
from
mere
knowledge
that
part
of
wilderness
North
America
remains
even
though
they
would
be
appalled
by
the
prospect
of
being
exposed
to
it"
(
1967;
see
also
Krutilla
and
Fisher,
1979.
Existence
value
is
the
willingness
to
pay
for
the
knowledge
that
a
natural
environment
is
preserved
(
Greenley,
et
al.,
1980:
1)
quite
apart
from
any
use
or
expectation
of
use
by
the
respondent
or
by
future
generations.
The
lone
attempt
to
measure
the
existence
benefits
of
water
quality
is
the
above
mentioned
Colorado
State
study
which
uses
the
following
question:

Q.
27
If
it
were
certain
you
would
not
use
the
South
Platte
River
Basin
for
water­
based
recreation
would
you
be
willing
to
add
cents
on
the
dollar
to
present
sales
taxes
every
year,
just
to
know
clean
water
exists
at
level
A
as
a
natural
habitat
for
plants,
fish,
wildlife,
etc?

They
report
a
mean
figure
of
$
24.98
for
the
24
non­
recreationalists
who
answered
this
question.

Since
the
Colorado
State
study
represents
the
state
of
the
art
in
estimating
option
and
existence
benefits,
a
closer
examination
of
its
methodology
is
relevant
to
our
purposes.
Three
questions
will
be
addressed.
Is
it
methodologically
sound?
How
adequate
are
their
measurements
and
estimation
procedures
for
option
and
existence
values?
How
much
credence
should
be
placed
on
their
annual
benefit
estimate
for
the
South
Platte
River
Basin
of
$
61
million
of
which
$
26.4
million
or
43
percent
is
attributed
to
recreation
benefits
(
both
direct
and,
using
our
terminology,
aesthetic)
leaving
57
percent
attributed
to
option,
existence
and
bequest
benefits?
1­
39
The
study
is
a
useful
methodological
experiment
from
which
we
can
learn
a
great
deal
thanks
to
the
admirably
complete
report
they
wrote
for
their
sponsor
(
Walsh,
et
al.,
1978)
and
which
is
available
through
NTIS.

Unfortunately,
the
study's
flaws
are
such
that
the
researchers'
decision
to
extrapolate
their
findings
without
qualifications
or
reservations
in
the
form
of
aggregate
point
estimates
in
the
report
and
in
a
brief
journal
article
(
Greenley,
et
al.,
1980)
is
unwarranted
and
potentially
misleading.

Since
we
are
primarily
concerned
with
the
study's
approach
to
measuring
intrinsic
benefits,
we
will
only
briefly
mention
the
more
serious
of
its
other
methodological
problems.
These
are:

A
low
response
rate
­­
only
37
percent
of
the
sampled
households
which
received
the
letter
announcing
the
intention
of
the
researchers
to
interview
a
household
member
participated.
According
to
sampling
theory
this
low
a
rate
means
that
the
findings
cannot
be
generalized
to
the
total
population
of
those
areas
which
constitute
the
study's
sampling
frame.

Starting
point
bias.
The
large
difference
in
results
between
their
two
bidding
vehicles
­­
sales
tax
increase
and
increase
in
sewer
1­
40
bills
­­
may
be
attributed
to
the
aggregate
yearly
payment
15
implied
by
the
starting
point
for
each
vehicle.
Furthermore,

the
mean
bids
for
option,
bequest,
and
existence
values
are
16
very
close
to
the
starting
point
for
each
vehicle.
Since
their
questionnaire
involved
so
many
bidding
games,
a
combination
of
respondent
fatigue
and
a
willingness
to
please
the
interviewer
possibly
may
account
for
a
large
portion
of
the
bids.

The
payment
vehicle,
additional
taxes
at
the
regional
level,
is
ambiguous.
Since
water
quality
actually
is
paid
for
in
higher
prices
and
federal
income
taxes
for
the
most
part,
the
respondents
are
already
paying
large
amounts
for
this
purpose.
We
have
no
way
of
knowing
whether
the
respondents
are
15
The
starting
points
for
the
vehicles
were
one
quarter
of
a
cent
increment
in
sales
tax
and
$.
50
a
month
on
the
water
sewer
bill.
Prior
to
bidding
the
respondents
were
informed
how
much
additional
money
they
would
pay
a
year
for
every
one
quarter
cent
increment
in
sales
tax.
(
Walsh,
et
al.,
1979:
29).
The
study
report
does
not
say
whether
an
annual
amount
for
the
water/
sewer
fee
was
calculated
for
the
respondents
but
even
if
it
wasn't
the
respondents
would
be
able
to
calculate
this
easily
themselves
For
the
entire
sample
they
report
an
annual
recreation
value
of
$
18.60
for
the
water
fee
vehicle
and
$
56.68
for
the
sales
tax.
(
Every
respondent
bid
for
recreation
using
each
of
the
vehicles,
total
N
=
174)
The
only
explanation
they
offer
for
respondents'
willingness
to
pay
only
about
one­
fourth
as
much
in
water­
sewer
fees
as
in
sales
tax
was
that
they
"
may
have
perceived
inequities"
in
the
fees
since
everyone,
including
tourists,
would
be
liable
for
sales
taxes
(
Greenley,
et
al.,
1980:
17).
However,
since
the
starting
points
for
the
two
vehicles
"
generated
revenue
of
$
6
per
year
in
water­
sewer
fees
and
$
25
per
year
in
sales
tax
for
a
typical
household
of
four
with
an
average
income
of
$
13,500
per
year"
(
Greenley,
et
al.,
1980:
11),
it
is
more
likely
that
the
difference
results
from
starting
point
bias.

16
In
Table
1
of
Greenley,
et
al.
(
1980)
they
give
the
mean
bid
for
option,
bequest
and
existence
values
for
each
vehicle.
In
every
case,
irrespective
of
vehicle,
the
bids
for
these
values
hover
around
the
starting
point.
The
average
difference
from
the
starting
point
is
17
percent.
(
It
is
true,
however,
that
the
bequest
value
lies
slightly
below
the
starting
point,
while
the
other
two
have
mean
bids
above
the
starting
point,
suggesting
that
people
do
value
bequest
values
less
than
the
other
two.)
1­
41
willing
to
pay
these
amounts
plus
the
additional
amounts
elicited
in
the
bidding
games
or
not.

Of
direct
importance
to
our
present
discussion
is
the
method
used
by
the
Colorado
State
researchers
to
measure
the
option
and
existence
values.

Their
approach
is
additive.
They
ask
separate
questions
for
each
of
the
four
benefit
categories
(
recreation,
option,
bequest
and
existence)
and
add
the
resulting
mean
bids
to
get
a
total
WTP
figure
for
the
Basin's
water
quality.
Since
the
additive
technique
requires
each
benefit
to
be
measured
independently
with
no
overlap,
the
WTP
instrument
must
ensure
that
respondents
bid
on
one
value
at
a
time
and
only
on
that
value.

Otherwise
double
counting
will
occur
biasing
the
total
estimate
upward
and
making
it
impossible
to
derive
reliable
estimates
for
the
component
values.
A
close
scrutiny
of
the
wording
of
the
recreational
and
option
value
questions
in
the
Colorado
State
instrument
raises
serious
doubts
17
about
their
independence.

Here
is
the
wording
of
the
question
they
used
to
measure
recreational
value:

17
In
the
case
of
the
existence
and
bequest
values,
however,
they
recognized
after
the
fact
that
their
survey
"
did
not
ask
users
about
(
these
benefits)
in
such
a
way
as
to
permit
adding
them
to
user's
values"
(
Walsh,
et
al.,
1978:
39).
For
this
reason
they
restricted
their
estimates
of
these
benefits
to
the
very
small
number
of
non­
recreationists.
In
the
discussion
which
follows
we
consider
only
the
recreation
and
option
values,
both
of
which
they
estimated
for
the
full
sample,
although
we
believe
our
criticism
also
holds
for
the
other
two
measures.
1­
42
Suppose
a
sales
tax
was
collected
from
the
citizens
of
the
South
Platte
River
Basin
for
the
purpose
of
financing
water
quality
in
this
Basin.
All
of
the
additional
tax
would
be
used
for
water
quality
improvements
to
enhance
recreational
enjoyment.
Every
Basin
resident
would
pay
the
tax.
All
bodies
of
water
in
the
River
Basin
would
be
cleaned
up
by
1983.
Assume
that
this
is
the
only
way
to
finance
water
quality
improvement.

14.
Would
you
be
willing
to
add
cents
on
the
dollar
to
present
sales
taxes
every
year,
if
that
resulted
in
an
improvement
from
situation
C
to
situation
B?

15.
Would
you
be
willing
to
add
cents
on
the
dollar
to
present
sales
taxes
every
year,
if
that
resulted
in
an
improvement
from
situation
C
to
situation
A?

The
three
water
quality
levels
A
(
best),
B
(
medium)
and
C
(
worst)

were
represented
by
photographs
showing
colored
water
features
associated
with
mine
drainage.
Although
the
wording
says
all
the
additional
tax
would
be
used
"
to
enhance
recreational
enjoyment"
the
question
does
not
explicitly
ask
the
respondent
to
limit
his
or
her
answer
to
recreational
benefits
nor
does
it
inform
the
respondent
that
he
or
she
will
be
presented
with
subsequent
opportunities
to
say
how
much
they
are
willing
to
pay
for
other
(
intrinsic)
values.
Since
the
apportioning
of
water
quality
values
to
precise
categories
is
not
a
familiar
undertaking
for
most
people,
the
form
of
the
question
with
its
emphasis
upon
the
quality
shift
from
C
to
B
and
C
to
A
and
the
use
of
the
pictures
which
depict
aesthetic
degradation
serve
to
create
the
impression
that
the
respondent
is
being
asked
about
water
pollution
in
general.
The
bids
for
the
recreation
question
probably
should
be
regarded
as
the
consumer's
total
willingness
to
pay
for
an
increase
in
water
quality
in
the
area
from
C
to
A.
1­
43
The
option
value
question
has
the
same
weakness.
Although
the
researchers
are
careful
to
specify
the
option
characteristics
in
accord
with
their
theoretical
model,
the
question
is
worded
in
such
a
way
that
the
respondents
could
interpret
it
as
asking
them
to
value
water
quality
of
level
A
while
bearing
in
mind
the
economic
tradeoff
of
foregoing
mining
activity.
(
A
Further
problem
with
the
option
question
is
that
the
respondents
may
not
believe
level
C
to
be
irreversible
since
the
recreational
questions
in
the
interview
told
them
that
level
C
could
be
improved
to
levels
B
or
A.)
The
option
question
is
worded
as
follows:

In
the
near
future,
one
of
two
alternatives
is
likely
to
occur
in
the
South
Platte
River
Basin.
The
first
alternative
is
that
a
large
expansion
in
mining
development
will
soon
take
place,
creating
jobs
and
income
for
the
region.
As
a
consequence,
however,
many
lakes
and
streams
would
become
severely
polluted.
It
is
highly
unlikely,
as
is
shown
in
situation
C,
that
these
waterways
could
ever
be
returned
to
their
natural
condition.
They
could
not
be
used
for
recreation.
Growing
demand
could
cause
all
other
waterways
in
the
area
to
be
crowded
with
other
recreationists.

The
second
possible
alternative
is
to
postpone
any
decision
to
expand
mining
activities
which
would
irreversibly
pollute
these
waterways.
During
this
time,
they
would
be
preserved
at
level
A
for
your
recreational
use.
Furthermore,
information
would
become
available
enabling
you
to
preserve
the
waterways
at
level
A
for
your
recreational
use
or
to
permit
mining
development.
Of
course,
if
the
first
alternative
takes
place,
you
could
not
make
this
future
choice
since
the
waterways
would
be
irreversibly
polluted.

26.
Given
your
chances
of
future
recreational
use,
would
you
be
willing
to
add
cents
on
the
dollar
to
present
sales
taxes
every
year
to
postpone
mining
development?
This
postponement
1­
44
would
permit
information
to
become
available
enabling
you
to
make
a
decision
with
near
certainty
in
the
future
as
to
which
option
(
recreational
use
or
mining
development)
would
be
most
beneficial
to
you?

Whereas
the
"
recreational
value"
questions
(
14
and
15)
ask
the
respondents
to
imagine
that
they
are
at
level
C
or
B
and
to
say
how
much
they
are
willing
to
pay
to
move
to
level
A,
the
"
option
value"
question
asks
them
to
assume
that
they
are
at
level
A
and
asks
how
much
they
are
willing
to
pay
to
remain
at
level
A
instead
of
moving
to
level
C.
In
this
respect,
the
question
is
simply
another
way
of
measuring
the
consumer
surplus
for
level
A
and
we
would
again
anticipate
that
the
respondents'

WTP
amounts
will
reflect
their
total
recreational
and
intrinsic
values
for
water
quality
rather
than
just
the
intended
independent
(
and
additive)

option
value.
Of
course
the
bids
on
this
question
will
be
influenced
by
the
additional
information
conveyed
by
Question
26,
namely:
a)
the
water
quality
change
will
be
irreversible;
b)
keeping
water
quality
at
level
A
involves
economic
tradeoffs
(
jobs
and
income)
and
c)
there
is
pressure
for
such
development.

This
last
point,
which
is
implied
rather
than
stated,
might
lead
the
respondent
to
believe
that
further
mining
activity
is
18
inevitable,
and
therefore
to
give
low
or
zero
bids.

18
These
factors
may
explain
why
the
"
option"
question
received
a
lower
mean
bid
($
23)
than
the
"
recreation"
question
($
57).
1­
45
In
this
study
we
use
an
approach
which
contrasts
with
the
Colorado
State
method
in
several
respects.
First,
we
do
not
attempt
to
measure
the
various
sub­
categories
of
intrinsic
benefits
as
they
did,
although
we
do
obtain
separate
estimates
for
the
overall
intrinsic
benefits
and
for
the
in­
stream
recreational
benefits.
Second,
we
begin
by
ascertaining
the
individuals'
total
consumer
benefits
(
recreational
plus
intrinsic)
through
a
sequence
of
WTP
questions.
Only
then
do
we
apportion
these
total
benefits
to
the
separate
recreational
and
intrinsic
categories
as
the
basis
of
information
which
we
obtained
in
the
interview
about
the
respondents'

recreational
use
or
non­
use
of
freshwater.
Our
process
is
subtractive
rather
than
additive
and
uses
self­
reported
behavior
rather
than
answers
to
specific
WTP
questions
to
distinguish
recreational
from
intrinsic
benefits.
Working
backwards
from
a
total
benefit
figure
has
the
advantage
of
forcing
respondents
to
consider
their
budget
restraints
more
realisitically
than
in
the
case
when
they
are
asked
to
value
a
sequence
of
component
benefits
without
confronting
the
overall
expenditure
involved
in
these
separate
decisions.
Table
1.6
shows
which
of
the
benefit
categories
in
Table
1.4
we
measure
in
this
study.
We
present
our
findings
in
Chapter
5.
1­
46
NATIONAL
BENEFITS
OF
CLEAN
WATER
MEASURED
BY
THE
RFF
SURVEY
Table
1.6
Industry
Withdrawal­­­
Agriculture
Drinking
water
Direct
Use
­­­­
Commercial
Fisheries
Instream
­­­­

Consumptive
recreation
(
i.
e.
duck
hunting)
Habitat
based
­­­
Nonconsumptive
recreation
(
wild­

Indirect
­­­­
life
watching
and
photography)

Water
enhanced
recreation
(
i.
e.
picnicing,
camping,
sightseeing,
other)
Aesthetic
­­­­­­­

Other
activitiy
(
i.
e.
commuting
to
work)
­­­­­

Individual
­­­­­
Short
term
Option
Long
term
­­­­­
Intergenerational
(
bequest)

Existence
Categories
in
italics
are
those
estimated
in
this
report.
The
subcategories
in
the
boxes
are
not
included
in
our
intrinsic
benefits
total
because
the
changes
in
water
quality
which
we
value
are
defined
as
irreversible.
CHAPTER
2
THE
MACRO
APPROACH
TO
WILLINGNESS
TO
PAY
STUDIES
Our
review
of
studies
using
the
willingness
to
pay
method
reveals
two
distinct
research
traditions.
In
one
tradition,
willingness­
to­
pay
questions
are
used
in
national
polls
as
a
measure
of
environmental
concern.

In
the
other,
the
questions
are
employed
by
economists
to
develop
benefit
estimates
for
particular
environmental
goods.
We
have
named
these
approaches
the
macro
and
micro,
respectively.
Each
has
advantages
and
disadvantages
for
benefit
estimation.
We
have
experimented
with
a
new
kind
of
macro
approach,
one
which
borrows
heavily
from
methodological
innovations
developed
by
practitioners
of
the
micro
approach.
In
this
chapter
we
describe
these
two
approaches
and
the
rationale
behind
our
synthesis.

The
Macro
Willingness
to
Pay
Approach
Since
1969
at
least
8
different
surveys
have
asked
questions
using
the
"
macro
willingness
to
pay"
(
macro
WTP
type).
The
kind
of
environmental
public
goods
covered
in
these
surveys
range
from
air
pollution
devices
on
new
automobiles
Viladus,
1973)
to
the
more
general
category
of
"
cleaning
up
pollution
now"
(
Gallup,
1971).
They
also
vary
in
how
they
ask
for
the
amount.
Some
questions
are
open
ended,
but
macro
WTP
questions
usually
offer
a
specific
amount
or
a
limited
sequence
of
specific
amounts
for
the
respondent's
judgment.
For
example,
in
1969
a
Harris
poll
for
the
National
Wildlife
Federation
asked
1500
adults
nationwide:

You
are
already
sharing
in
the
costs
brought
to
us
all
by
air
and
water
pollution.
In
order
to
solve
our
national
problems
of
air
and
water
pollution
the
public
may
have
to
pay
higher
taxes
and
higher
prices
for
some
products.
To
get
real
clean­
up
in
your
natural
environment,
would
you
be
willing
to
accept
a
per­
year
increase
in
your
family's
total
expenses
of
$
200?
2­
2
The
question
was
repeated
for
the
amounts
of
$
100,
$
50
and
$
20.
Other
examples
of
macro
questions
include
these
taken
from
national
surveys.

Would
you
be
willing
to
pay
an
additional
$
20
per
year
on
your
electric
bill
in
order
to
cut
down
air
pollution
caused
by
power
plants?
(
Federal
Energy
Administration,
1975
survey).

(
After
asking
people
the
amount
of
their
last
electric
bill
....)
Now
suppose
that
the
only
way
to
stop
the
electric
power
plants
from
polluting
is
to
install
expensive
equipment,
and
this
equipment
made
your
electric
bill
go
up
unless
you
used
less
electricity
than
you
use
now.
How
much
more
would
you
be
willing
to
pay
a
month
to
clean
up
this
form
of
pollution?
$
(
Viladus,
May
1973
survey).

The
past
uses
of
the
macro
WTP
approach
have
the
following
characteristics:

1.
Purpose:
In
these
earlier
uses,
macro
WTP
questions
were
not
intended
to
provide
the
basis
for
benefit
estimates
in
the
strict
sense.
They
were
used
for
the
conventional
poll
takers
purpose
of
measuring
public
concern
about
environmental
goods.
It
is
assumed
that
asking
people
the
amount
of
money
they
are
personally
willing
to
pay
for
pollution
reduction
is
a
more
stringent
test
of
people's
concern
than
questions
which
simply
elicit
concern
without
reference
to
the
cost.
The
relevant
audience
for
these
studies
are
those
who
normally
use
public
polls
on
environmental
issues.

2.
Survey
Method:
The
macro
WTP
questions
were
used
in
social
surveys
conducted
by
professional
polling
organizations.
Because
the
respondents
were
chosen
by
modern
sampling
techniques,
with
sample
sizes
ranging
from
800
to
1500,
the
results
may
be
generalized
to
the
appropriate
sampling
frame
within
a
statistically
determinable
degree
of
accuracy.

The
interviewers
are
trained
adult
workers
under
contract
to
the
polling
2­
3
organization
whose
work
is
subjected
to
independent
checks.
In
each
use
of
the
macro
WTP
questions
they
have
been
just
one
component
of
a
larger
list
of
questions.

3.
Specification
of
the
good
and
procedure
for
ascertaining
WTP:

The
nature
and
geographical
distribution
of
the
environmental
good
is
described
in
general
terms.
People
are
asked
about
"
cutting
down
air
pollution,"
for
example,
with
no
mention
of
where
this
would
happen
or
how
much
"
cutting
down"
is
involved.
No
attempt
is
made
to
vary
the
amount
of
the
good,
to
provide
visual
aids
describing
it,
to
present
the
parameters
of
a
hypothetical
market
in
the
good,
or
to
sepcify
the
geographical
location
which
would
receive
the
environmental
benefit.

4.
Test
for
biases:
The
standard
assumptions
about
the
reliability/

validity
of
survey
research
are
applied
to
the
macro
WTP
questions.
These
assume
that
a
question
is
reliable
if
it
uses
words
which
are
understood
by
all
the
respondents,
is
unambiguous
in
meaning,
is
neutral
in
its
wording,
and
asks
about
a
matter
on
which
respondents
may
be
presumed
to
have
an
opinion.
Validity
is
established
by
judgment
of
whether
or
not
the
description
of
the
environmental
good
in
the
question
appears
to
be
adequate
(
face
validity).
No
attempts
were
made
to
undertake
specific
tests
for
threats
to
reliability
and
validity.
Data
reporting
was
limited
to
presentation
of
the
marginal
results
and
cross
tabulation
by
standard
background
variables.

5.
Sampling
Frame:
The
sampling
frame
for
these
surveys
was
a
large
geographic
area.
Most
were
national
(
the
lower
48
states)
although
macro
WTP
questions
have
occasionally
been
used
in
state
surveys.
2­
4
We
use
"
macro"
as
a
label
for
this
tradition
of
WTP
questioning
because
of
its
focus
on
national
benefits:
No
matter
what
environmental
good
these
questions
solicit
willingness­
to­
pay
amounts
for,
the
money
would
pay
for
supply
of
the
good
across
the
country.
The
micro
approach,
as
we
will
see,
is
interested
in
the
benefits
for
a
specific
geographic
area.
1a
The
Micro
Willingness
to
Pay
Approach
Since
the
Second
World
War,
economists
have
been
increasingly
faced
with
the
need
to
measure
the
use
values
associated
with
natural
resources.

Insofar
as
values
associated
with
goods
are
measured
in
the
market
place
in
terms
of
price,
obtaining
dollar
estimates
for
them
is
relatively
straightforward
But
natural
resources,
including
the
amenities
of
clean
air
and
water,
have
characteristics
which
severely
limit
the
use
of
exchange
to
determine
their
value
for
society.
Because
they
have
the
attributes
of
public
goods
especially
in
that
it
is
difficult
or
impossible
to
exclude
consumers
from
using
them,
they
are
outside
conventional
market
structures.
The
rather
intangible
nature
of
some
of
the
values
these
resources
convey,
such
as
aesthetic
and
existence
values,
means
that
people
are
likely
to
have
difficulty
imagining
the
good
with
precision
and
conceiving
of
a
hypothetical
market
in
those
values.

1
If
the
Grand
Canyon
has
symbolic
national
value
then
the
location
of
the
benefits
is
national
rather
than
local.

1a
Macro
need
not
refer
to
only
national
benefits.
For
instance,
the
benefit
could
be
global
d'Arge
et
al.,
1980)
or
regional
as
in
a
survey
of
WTP
for
air
quality
regulations
in
California
of
a
random
sample
of
all
California
(
if
California
only
generated
and
was
affected
by
the
air
pollution).
At
the
margin
the
distinction
between
macro
and
micro
become
blurred.
2­
5
Economists
have
experimented
with
ways
to
overcome
these
obstacles
in
order
to
simulate
a
market
in
environmental
goods.
Among
the
myriad
of
techniques
developed
over
the
past
three
decades
for
this
purpose
(
see
Wyckoff,
1971;
for
an
overview
as
of
1970)
is
the
use
of
survey
research
instruments
to
ask
people
what
they
are
willing
to
pay
for
such
goods.

Although
Ciriacy­
Wantrup
suggested
such
a
technique
­­
which
he
called
the
"
direct
interview
method"
­­
as
early
as
1947
(
Wyckoff,
1971:
13),
it
apparently
did
not
come
into
actual
use
until
the
1960s
when
Davis
(
1963)

used
questionnaires
to
estimate
recreation
benefits.
Since
that
time
the
technique
has
been
used
repeatedly
by
economists
to
measure
such
things
as
recreational
benefits
(
Binkley
and
Hanemann,
1978;
Darling,
1973,
McKinney
and
MacRae,
1978);
water
quality
benefits
(
Gramlich,
1977;
Walsh,
et
al.,
1978)

(
Davis,
1980);
benefits
of
decreased
risk
from
a
nuclear
power
plant
accident
(
Mulligan,
1978);
aesthetic
benefits
from
foregoing
a
geothermal
power
forthcoming
plant
(
Thayer,
/
);
aesthetic
benefits
of
air
visibility
(
Randall,
et
al.,
1974;
Brookshire,
et
al.,
1976;
Rowe,
et
al.,
1979a
and
b);
and
aesthetic
and
health
benefits
of
air
quality
(
Brookshire,
et
al.,
1979)

In
the
course
of
this
research
the
direct
interview
technique
has
been
refined
and
a
great
deal
of
study
has
been
given
to
its
possible
biases.
Much
of
this
work
has
been
undertaken
by
Randall
and
colleagues
(
Randall,
et
al.,
1974)
and
by
d'Arge,
Brookshire,
Rowe
and
others
from
the
University
of
Wyoming
in
their
series
of
studies
on
the
aesthetic
benefits
of
air
pollution
reduction.
In
1979
the
latter
group
produced
a
major
methodological
study
of
the
technique
for
EPA
(
Brookshire,
et
al.,

1979).
2­
6
Figure
2.1
gives
the
text
of
a
micro
WTP
question.
It
was
used
for
a
1975
study
of
the
aesthetic
damages
of
a
possible
power
plant
near
Lake
Powell
in
Utah
and
illustrates
the
essentials
of
the
micro
approach.
This
approach,
particularly
as
used
in
the
air
pollution
benefits
studies,
differs
from
the
micro
approach
in
a
number
of
important
respects.

1.
Purpose:
The
micro
studies
are
specifically
designed
to
obtain
estimates
of
economic
benefits
by
gathering
data
which
enable
the
fitting
of
a
demand
curve
for
the
value
in
question.
Their
designers
seek
to
gather
data
which
will
be
accepted
as
valid
for
this
purpose
by
their
fellow
economists.

2.
Survey
Method:
The
field
work
for
the
micro
WTP
studies
is
usually
conducted
by
the
researchers
using
student
interviewers
who
are
specially
trained
for
the
study.
The
WTP
questions
are
the
centerpiece
of
the
survey
instrument
which
is
dedicated
solely
to
the
benefits
measure
study.
In
a
number
of
the
past
studies
sample
sizes
have
been
very
small
by
conventional
survey
research
standards;
sub­
groups
which
are
the
focus
of
extensive
analysis
sometimes
consist
of
only
20­
30
cases.
Sometimes
the
descriptions
of
the
sample
frame
and
procedures
are
sketchy
or
lacking
entirely
2
in
the
report
so
it
is
difficult
to
know
whether
the
findings
can
be
generalized
reliably
to
larger
populations
and
what
those
populations
might
be.
In
other
case
(
e.
g.
Rowe,
et
al.,
1979b:
85­
89)
a
representative
rather
than
random
sample
was
used
which
precludes
such
generalization.

2
For
example,
the
interview
dates,
the
response
rate,
and/
or
the
method
of
selecting
the
respondents
may
be
missing.
Figure
1.1
Good
Morning/
Afternoon.
My
name
is
.
I'm
doing
research
for
the
Economics
Department
at
the
University
of
New
Mexico,
as
a
part
of
the
Lake
Powell
Research
Project,
funded
by
the
John
Muir
Institute
for
Environmental
Studies.

This
research
is
designed
to
more
closely
examine
some
of
the
trade­
offs
between
industrial
development
recreation
and
the
environment
in
the
Lake
Powell
area.
In
connection
with
these
objectives.
I
would
like
to
ask
you
a
few
questions
to
see
how
you
feel
about
environmental
quality
and
its
future
In
this
area.

1.
How
many
members
of
your
family
are
here
with
you?
persons.
2­
7
QUESTIONNAIRE
Bidding
Game
for
Estimation
of
Recreationists'
Demand
for
Abatement
of
Aesthetic
Environmental
Damage
2.
What
is
the
expected
length
of
your
stay?
days.

3.
Where
are
you
staying?
(
a)
local
resident.
(
d)
developed
or
semideveloped
campground
(
b)
lodge,
Page
motel
(
e)
remote
(
specify
(
c)
passerby
location)

4.
If
you
don't
mind,
could
you
please
indicate
which
of
the
following
brackets
your
family
income
falls
into:

0
­
4,999
20,000
­
26,999
5,000
­
9,999
25,000
­
39,999
10,000
­
14,999
30,000
­
49,399
13,000
­
19,999
50,000
and
up
There
are
plans
to
construct
a
large
electric
generating
plant
north
of
Lake
Powell.
This
plant
is
expected
to
be
at
least
as
large
as
the
Navajo
Plant
on
the
south
side
of
the
lake.

5.
Have
you
noticed
the
Navajo
Plant
or
its
smokestacks?
yes
no
Depending
on
exactly
where
and
how
a
new
plant
is
constructed,
it
could
have
a
significant
effect
on
the
quality
of
the
environment.
miles
up
and
down
the
lake.
If
the
plane
is
built
near
the
lake,
it
could
be
visible
for
many
If
air
pollution
is
not
strictly
controlled,
visibility
in
the
area
may
be
significantly
affected.

These
photographs
(
show)
are
designed
to
show
how
a
new
powerplant
on
the
north
side
of
the
lake
might
appear.
Situation
A
shows
a
possible
plant
site
but
assumes
that
the
powerplant
would
be
built
at
some
distant
location,
not
visible
from
the
lake
area.
the
lake,
but
emits
very
little
smoke;
In
Situation
B
the
powerplant
is
easily
seen
from
visibility
is
virtually
unaffected.
Situation
C
is
intended
to
show
the
situation
with
the
greatest
impact
on
the
environment
of
recreationists
in
the
area.
It
is
easily
seen
from
the
lake,
and
the
smoke
substantially
reduces
visibility.

Vacationers,
of
course,
spend
considerable
amounts
of
money
and
vehicles,
boats,
camping
and
fishing
gear,
and
time
and
effort
to
equip
themselves
with
for
traveling
to
the
destination
of
their
choice.
It
is
reasonable
to
assume
that
the
amount
of
money
you
are
willing
to
spend
for
a
recreational
experience
depends,
among
other
things,
on
the
quality
of
the
experience
you
expect.
be
expected
to
be
of
greater
value
to
you
than
a
degraded
one.
Since
it
does
cost,
money
to
improve
the
An
improved
experience
would
environment,
we
would
like
to
get
an
estimate
of
how
much
a
better
environment
is
worth
to
you.

First,
let's
assume
that
visitors
to
GCNRA
are
to
finance
environmental
improvements
by
paying
an
entrance
fee
to
be
admitted
into
the
recreation
area.
ments
in
the
area.
This
will
be
the
only
way
to
finance
such
improve­
Let's
also
assume
that
all
visitors
to
the
area
will
pay
the
same
daily
and
all
the
money
collected
will
be
used
to
finance
fee
as
you,
the
environmental
improvements
shown
in
the
photos.

6.
Would
you
be
willing
to
pay
a
$
1.00
per
day
family
fee
to
prevent
Situation
C
from
occurring,
thus
preserving
Situation
A?
$
2.00
per
day?
(
increment
by
$
1.00
per
day
until
a
negative
response
is
obtained,
then
decrease
the
bid
by
25~
per
day
until
a
positive
response
is
obtained,
and
record
the
amount.)

7.
Would
you
be
willing
to
pay
a
$
1.00
per
day
fee
to
prevent
Situation
B
from
occurring,
thus
preserving
Situation
A?
(
Repeat
bidding
procedure).

8.
(
Answer
only
if
a
zero
bid
was
recorded
for
question
6
or
7
above.)
Did
you
bid
zero
because
you
believe
that:

the
damage
is
not
significant
it
is
unfair
or
immoral
to
expect
the
victim
of
the
damage
to
have
to
pay
the
costs
of
preventing
the
damage
Other
(
specify)

9.
In
your
opinion,
has
visibility,
depth
or
color
perception
in
this
area
been
significantly
reduced
by
air
pollution?
2­
8
3.
Specifications
of
the
good
and
procedures
for
ascertaining
WTP
Because
of
the
importance
of
making
the
situation
as
realistic
and
credible
as
possible,
great
attention
is
given
to
the
description
of
the
environmental
good
in
the
micro
studies.
It
is
typically
described
as
occurring
in
a
specific
locality
(
usually
the
locality
where
the
interviewing
is
taking
place);
a
time
frame
is
specified;
and
an
extensive
verbal
description
of
the
good
is
supplemented
with
pictures
or
other
visual
devices.
A
great
deal
of
care
is
also
given
to
the
procedures
for
eliciting
the
WTP
amount.
The
survey
instrument
describes
a
hypothetical
market
with
a
substantial
degree
of
institutional
detail;
specific,
plausible
means
of
payment
are
specified;
and
contingencies
relevant
to
the
respondent's
valuing
the
good
are
described.
A
common
feature
of
most
of
these
studies
is
the
use
of
a
"
bidding
game"
procedure
to
ascertain
the
dollar
amount
the
respondent
is
willing
to
pay.

The
bidding
game
works
in
the
following
manner:
after
the
hypothetical
market
is
staged
by
means
of
preliminary
questions,
verbal
description,

and
the
use
of
the
visual
aids,
a
particular
good
is
identified
and
the
person
is
asked
whether
he
or
she
is
willing
to
pay
$
x
for
the
good.
If
the
starting
amount
(
e.
g.
one
dollar)
is
agreed
to,
the
interviewer
increases
it
by
a
set
interval
(
e.
g.
50
cents)
until
the
respondent
rejects
an
amount.
The
study
may
then
require
the
interviewer
to
decrease
the
amount
rejected
by
a
smaller
amount
(
e.
g.
$.
25)
until
the
precise
maximum
amount
the
individual
is
willing
to
pay
is
reached.
3
This
procedure
is
usually
repeated
for
several
levels
of
the
good
in
question
so
that
the
demand
curve
can
be
traced
out.

4.
Test
for
biases:
Because
they
are
explicitly
intended
to
provide
benefits
estimates
for
policy
purposes,
micro
studies
attempt
to
obtain
as
close
a
surrogate
as
possible
to
actual
market
behavior.

3
Several
micro
studies
also
used
parallel
procedures
to
ascertain
how
much
respondents
were
willing
to
accept
(
WTA)
in
return
for
the
loss
of
the
environmental
good.
2­
9
The
efficacy
of
bidding
games
used
for
this
purpose
[
to
measure
aesthetic
environmental
improvements]
depends
on
the
reliability
with
which
stated
hypothetical
behavior
is
converted
to
action,
should
the
hypothetical
situation
posited
in
the
game
arise
in
actuality
(
Randall,
et
al.,
1974:
135).

Since
many
economists
are
skeptical
about
the
fit
between
attitudes
and
behavior,
credibility
in
this
regard
is
crucial.
Accordingly,
those
conducting
micro
studies
have
placed
a
great
deal
of
emphasis
upon
testing
for
potential
biases.
In
a
number
of
cases,
most
notably
the
studies
done
by
d'Arge,
Brookshire,
and
their
colleagues,
tests
for
biases
are
built
into
the
study
design
as
when
comparable
samples
are
offered
different
dollar
amounts
as
starting
points
for
the
same
environmental
good
in
order
to
test
for
starting
point
bias.
Strategic
bias
has
been
examined
in
a
similar
manner.

5.
Sampling
frame:
The
environmental
amenities
valued
by
the
micro
WTP
approach
are,
as
we
have
seen,
location
specific.
Those
interviewed
for
these
studies
are
generally
sampled
from
people
who
live
or
recreate
in
the
particular
area.
This
conjunction
of
a
local
good
and
a
local
sample
is
intended
to
reduce
the
artificiality
inherent
in
the
bidding
games
since
people
will
be
bidding
on
a
good
which
they
can
easily
comprehend
and
which
is
of
immediate
concern
to
them.
For
the
South
Platte
River
Basin
(
Colorado)
202
residents
of
Denver
and
Fort
Collins
were
interviewed
(
Walsh,

et
al.,
1978);
for
the
Glen
Canyon
Recreation
Area
the
82
respondents
included
local
residents,
motel
visitors,
developed
campgrounds
visitors
and
remote
campers
(
Brookshire,
et
al.,
1976).
2­
10
Comparison
This
brief
description
of
these
two
ongoing
research
traditions
captures
the
essential
features
of
each
as
they
existed
in
1979
when
planning
for
the
RFF
experiment
began.
Each
has
a
major
strength
and
a
compensating
weakness.

Realism
Of
the
two
approaches,
the
micro
approach
has
been
far
superior
in
its
realism.
People
are
asked
about
a
good
which
they
personally
have
experienced
or
which
they
would
experience
in
that
location
if
pollution
levels
increased.
The
several
values
associated
with
the
good
(
existence,

aesthetic,
health,
etc.)
are
differentiated
and
the
value
chosen
for
measurement
is
described
in
detail
both
verbally
and,
if
possible,
pictorially.
The
payment
vehicle
and
the
hypothetical
market
are
designed
to
match
the
respondent's
experience
as
closely
as
possible.
In
comparison,
the
designers
of
the
macro
questions
have
made
very
little
effort
to
stimulate
a
market
or
to
describe
the
environmental
goods
in
detail.

Generalizability
Realism
is
an
important
factor
in
designing
reliable
and
valid
measures
of
WTP.
But
once
reliable
and
valid
benefit
estimates
have
been
obtained
from
a
set
of
respondents,
for
our
purposes
it
is
necessary
to
aggregate
them
to
obtain
overall
benefits
estimates.
The
great
strength
of
the
macro
approach
with
its
use
of
a
national
sampling
frame
is
the
ease
with
which
the
results
can
be
generalized
to
give
a
national
benefits
estimate.
In
contrast,
it
is
difficult
to
aggregate
micro
study
findings
beyond
the
location
where
the
study
was
conducted
and
it
is
extremely
difficult
to
make
reliable
national
estimates
from
a
series
of
micro
studies.
2­
11
Probability
Sampling
and
Aggregation
Survey
research
has
a
standard
solution
to
the
aggregation
problem
­­

probability
sampling.
If
Gallup
wants
to
predict
the
national
presidential
vote,
he
interviews
1500
people
nationwide
who
are
chosen
by
an
elaborate
sampling
procedure
based
on
statistical
principles.
Providing
his
survey
takes
place
immediately
before
the
vote
and
that
his
interviewers
adhere
to
the
sampling
plan,
he
will
be
able
to
predict
the
vote
with
an
accuracy
of
3
percent.
Good
sampling
requires:
1)
designation
of
the
appropriate
sampling
frame
for
the
population
to
which
one
wishes
to
generalize
(
in
the
Gallup
example
this
is
people
living
in
non­
institutionalized
settings
in
the
lower
48
states),
2)
design
of
a
sampling
plan
which
will
give
every
relevant
person
(
e.!.

(
e.
g.
adult
voters)
a
known
probability
of
inclusion,
and
3)
strict
execution
of
the
sample.,
Once
the
sampling
frame
is
chosen,
the
design
and
execution
of
the
sample
is
straightforward,
although
certain
adaptations
can
be
made
to
a
strict
probability
design
in
the
interests
of
economy
without
undue
bias
resulting
(
see
Sudman,
1976,
for
a
review
of
these
procedures).

The
choice
of
the
sampling
frame
necessarily
depends
upon
the
researcher's
problem
and
purpose.
For
WTP
studies,
it
should
be
the
population
for
4
which
the
researcher
wishes
to
have
an
aggregrate
benefit.
There
are
two
separate
issues
involved
which
complicate
the
choice
of
a
sampling
frame
for
WTP
studies:
a)
which
groups
can
be
presumed
to
"
have"
benefits
that
should
be
included
in
any
comprehensive
measure
and
b)
what
groups
are
relevant
under
different
equity
positions;
i.
e.,
do
only
those
who
pay
get
to
have
their
benefits
counted?
Let
us
suppose
that
he
or
she
wishes
to
4
The
researcher
also
needs
to
define
any
special
sub­
populations
which
are
likely
to
have
an
especially
high
value
for
the
good
in
question,
If
there
are
such
sub­
populations,
he
or
she
may
need
to
oversample
these
people.
Otherwise
they
may
be
too
few
in
number
to
enable
a
reliable
estimate
to
be
made
of
their
benefits.
For
example,
one
in
fifteen
men
in
an
area
may
be
fishermen.
If
300
people
are
sampled
for
a
study
of
water
recreation
benefits
in
an
area
only
20
are
likely
be
to
fishermen
(
0.066
x
150)
When
benefits
are
aggregated
across
the
entire
sample,
the
benefits
for
oversampled
sub­
population(
s)
must
be
weighted
to
reflect
their
proportion
of
2­
12
estimate
noise
pollution
control
benefits.
In
the
case
of
a
village
which
wishes
to
use
WTP
techniques
to
estimate
the
benefits
of
ordering
quieter
garbage
trucks,
which
would
be
paid
for
out
of
village
property
taxes,
the
appropriate
sampling
frame
is
the
residents
of
the
village.
If
noise
regulations
are
a
state
matter
and
their
cost
is
paid
for
by
state
taxes,
then
the
state
population
would
be
the
appropriate
frame.
In
both
these
cases
the
selection
of
the
sampling
frame
is
simplified
because
the
same
population
is
affected
by
and
pays
for
the
public
good
in
question.

Choosing
the
appropriate
sampling
frame
becomes
more
complex
where
the
two
do
not
coincide.
The
table
below
shows
the
four
possible
relationships
between
paying
for
and
using
a
public
good.
Using
our
example
of
the
town
contemplating
the
purchase
of
garbage
trucks,
an
example
of
B
is
visitors
Pay
for
the
Good
yes
no
Use
the
good
yes
A
B
no
D
C
to
the
town
who
would
benefit
from
quiet
garbage
trucks
although
they
wouldn't
pay
for
them
since
they
are
not
subject
to
town
property
taxes.

Position
D
would
include
deaf
residents
and
absentee
property
owners.

Note
that
by
using
the
sampling
frame
of
the
town
residents,
we
include
some
D's
(
town
population
=
A
+
D
minus
absentee
taxpayers).
Sampling
frames
comprised
of
those
who
live
in
political
jurisdictions
responsible
for
public
goods
almost
2­
13
inevitably
include
both
users
and
non­
users.
For
example,
those
who
reside
in
a
city
with
a
public
school
system
include
the
childless,
people
whose
children
are
too
young
or
old
for
public
school,
and
those
who
send
their
children
to
private
schools.
Note
also
that
the
use
of
the
town
population
as
the
sampling
frame
leaves
out
some
D's.
Presuming
that
property
taxes
are
the
source
of
the
town's
revenue,
absentee
landlords
would
not
be
represented
in
a
sample
of
town
residents.
A
different
sampling
frame
consisting
of
property
tax
payers
would,
of
course,
include
them
but
it
would
exclude
renters
5
B
is
an
important
category
for
some
benefits
estimates.
Consider
the
case
of
the
huge
Four
Corners
power
plant
at
Fruitland,
New
Mexico
in
the
Southwest
(
Randall,
et
al.,
1974).
Residents
of
the
area
and
visitors
who
come
to
enjoy
the
scenery
use
the
public
good
of
high
air
visibility
without
paying
the
cost
of
maintaining
it.
This
cost
is
(
would
be)
borne
by
those
in
Los
Angeles
(
and
elsewhere)
who
purchase
their
electricity
from
the
utility
which
owns
the
plant.
Nevertheless,
area
residents
and
visitors
are
a
crucial
sampling
frame
for
a
WTP
study
of
the
aesthetic
benefits
of
local
air
visibility.

A
further
complication
is
introduced
when
we
consider
the
question
of
intrinsic
benefits.
It
may
be
worth
something
to
Los
Angeles
residents
(
D)

who
never
recreate
or
intend
to
recreate
in
the
Four
Corners
area
to
know
that
the
extraordinary
air
visibility
in
that
area
is
untouched
by
the
emissions
of
the
plants
which
provide
their
electricity.
Indeed,
and
here
we
come
to
position
C,
it
may
be
worth
something
to
residents
of
Ohio
as
well.
A
local
or
even
regional
sampling
frame
is
inadequate
if
the
researcher
wishes
to
include
intrinsic
benefits
in
a
national
estimate
of
the
benefits
of
high
visibility
in
the
Southwest.

5
Recognizing,
of
course,
that
renters
eventually
pay
all
or
some
of
the
taxes
imposed
on
landlords.
2­
14
Interrelationship
Between
Generalizability
and
Realism
The
sampling
frame
and
the
realism
of
the
WTP
instrument
are
interrelated
Where
users
and
payers
are
in
the
same
population
(
position
A),

both
the
description
of
the
good
and
the
payment
vehicle
can
be
related
to
their
actual
experience
and
realism
is
enhanced.
People
in
position
B,

may
be
more
unrealistic
in
their
WTP
estimates
than
those
in
A
or
D
because
they
know
they
are
not
paying
for
the
good
and
are
unlikely
to
think
they
will
have
to
pay
for
it
in
the
future.
The
good
may
be
especially
abstract
and
hard
to
imagine
for
those
in
position
D
who
pay
for
the
good
but
who
do
not
use
it.
Thus
the
potential
for
measurement
bias
is
reduced
when
the
sampling
frame
consists
primariy
of
A's.
To
the
extent
that
respondents
anticipate
that
their
answers
will
affect
their
level
of
payment
or
their
level
of
supply
of
the
public
good,
B's
estimates
will
tend
to
overestimate
the
consumer
surplus
and
D's
to
underestimate
it
owing
to
the
effect
of
strategic
bias.

This
description
of
the
strengths
and
weaknesses
of
the
two
research
traditions
as
they
have
been
practiced
to
date
is
summarized
in
the
following
four­
fold
table.

Generalizability
Realism
High
Low
High
Low
micro
macro
The
obvious
goal
for
a
study
of
public
good
benefits
is
to
move
to
the
box
where
the
data
are
both
realistic
and
generalizable.
This
is
a
difficult
task
because
the
two
dimensions
are
somewhat
incompatible,
necessitating
2­
15
tradeoffs
between
degree
of
realism
and
degree
of
generalizability.
Thanks
to
the
experimental
micro
studies
of
the
1970s,
however,
we
have
a
much
greater
knowledge
of
the
properties
of
willingness
to
pay
measures.
For
example,
micro
research
has
shown
us
that
certain
potential
problems
such
as
strategic
bias
are
not
as
much
of
a
problem
as
some
had
thought
(
see
Chapter
4).

Knowledge
such
a
WTP
research
devise
a
macro
as
this
gives
the
researcher
greater
flexibility
in
designing
instrument,
flexibility
which
was
essential
to
our
effort
to
instrument
which
was
workable
yet
sufficiently
realistic
in
its
description
of
water
quality
to
give
US
valid
results.
In
Chapter
4
we
argue
the
need
to
jointly
minimize
the
potential
for
strategic
and
hypothetical
bias.

The
RFF
Macro
Approach
For
public
goods
which
are
mandated
at
the
national
level
and
are
paid
for
by
everyone
in
higher
prices
and
taxes
there
is
a
need
to
obtain
national
benefits
estimates.
The
quality
of
water
in
the
nation's
freshwater
bodies
is
such
a
public
good.
In
1972
Congress
passed
the
Federal
Water
Pollution
Control
Act
Amendments
(
later
amended).
In
this
law:

Congress
has
declared
its
intent
"
that
the
discharge
of
pollutants
into
the
navigable
waters
be
eliminated
by
1985"
and
that
"
wherever
attainable,
an
interim
goal
of
water
quality
which
provides
for
the
protection
and
propagation
of
fish,
shellfish,
and
wildlife
and
provides
for
recreation
in
and
on
the
water
be
achieved
by
1
July
1983."
In
effect,
this
amounts
to
a
commitment
to
make
all
the
nation's
navigable
waters
"
swimmable
and
fishable"
by
1983
and
wholly
free
of
pollutants
in
1985.
(
Rosenbaum,
1977:
158).

The
law
established
a
national
permit
system
for
all
municipal
and
industrial
effluent
discharges
according
to
national
standards
and
mandated
the
use
of
"
best
practicable"
technology
to
control
water
pollution
by
1977
and
the
"
best
available"
technology
by
1983.
Although
it
is
implemented
by
the
states,
the
standards
and
compliance
deadlines
are
set
by
Washington.
2­
16
The
cost
of
this
mammoth
pollution
control
program
is
ultimately
borne
by
all
U.
S.
taxpayers
and
consumers.
The
federal
government
provides
the
construction
monies
for
municipal
waste
treatment
facilities
in
what
is
the
largest
single
public
works
project
ever
authorized
by
Congress.
Municipal
taxes
pay
to
maintain
and
operate
the
waste
facilities.
The
expanse
of
controlling
the
non­
municipal
effluents
are
borne
by
industry
(
and
ultimately
the
consumer)
and
other
operators.
The
reach
of
the
law
extends
beyond
effluent
pipes
to
the
many
"
non­
point"
sources
of
water
pollution
such
as
fertilizer
runoff
from
farmers'
fields.

After
a
careful
consideration
of
the
alternatives,
we
decided
to
adopt
a
macro
approach
in
our
study
of
the
intrinsic
benefits
of
water
pollution
control.
A
primary
impetus
for
this
decision
was
the
national
character
of
control
programs.
In
addition
we
were
influenced
by
the
following
considerations
1.
The
results
of
the
various
micro
experiments
suggested
some
of
the
biases
involved
in
the
use
of
surveys
would
be
manageable
at
the
macro
level.

Factors
mitigating
against
a
micro
design:

2.
The
fact
that
unlike
air
pollution,
water
pollution
does
not
lend
itself
to
the
efficient
use
of
site­
specific
visual
aids.

This
is
because:
a)
perception
of
water
quality
is
mediated
strongly
by
individual
settings;
b)
the
diverse
visual
values
of
water
include
everything
from
clarity
to
surface
debris;
and
c)
not
all
visual
degradation
is
due
to
pollution,
making
it
difficult
to
distinguish
between
natural
and
the
human­
produced.
2­
17
3.
The
diversity
of
local
water
bodies
in
many
parts
of
the
country.

Lakes,
streams
and
rivers
each
have
different
characteristics
and
even
within
a
particular
geographical
location
they
may
take
many
different
forms.
This
diversity
poses
great
problems
for
micro
studies
which
seek
to
do
more
than
measure
the
water
quality
benefits
for
a
single
body
of
water.
Air,
in
contrast,
is
a
far
more
homogeneous
medium.

Factors
favoring
a
macro
design:

4.
Both
the
use
of
fresh
water
(
for
recreation,
aesthetic
pleasure,

etc.)
and
the
payment
for
the
cost
of
improving
its
quality
occur
at
the
national
level.
Of
course,
individual
use
takes
place
at
the
local
level,
but
such
use
occurs
all
over
the
country.
Moreover,

some
people
use
water
in
areas
far
distant
from
their
homes.
As
noted
above,
every
person
pays
the
cost
of
improved
water
quality
through
a
combination
of
taxes
and
higher
prices
and
the
cost
is
imposed
as
a
result
of
national
decisions
by
Congress
and
EPA.

5.
The
terms
used
in
the
national
law
mandating
the
water
cleanup
to
describe
the
several
levels
of
water
quality
­­
"
fishable,"

"
swimmable"
­­
are
readily
understood
by
individual
citizens
and
do
not
require
location
specific
visual
aids.

6.
That
a
national
survey
is
particularly
suited,
for
reasons
described
earlier,
for
the
measurement
of
the
intrinsic
value
of
improved
water
quality
for
our
special
task.
2­
18
Figure
2
summarizes
the
major
aspects
of
WTP
benefits
study
design
and
locates
the
RFF
approach
in
relation
to
the
other
types
of
approaches
which
have
been
used
in
the
past.
In
contrast
to
the
earlier
macro
studies,

the
description
of
water
quality
in
our
instrument
is
detailed.
In
contrast
to
the
air
pollution
bidding
games,
we
use
a
national
sample
and
measure
the
benefit
for
the
nation
as
a
whole.
Figure
2
2­
19
Local
Sample
Location
of
benefit
Local
National
Location
of
Benefit
Local
National
TYPOLOGY
OF
WILLINGNESS
TO
PAY
STUDIES
Description
of
Benefits
Description
of
Benefits
1
See
footnote
1,
page
2­
4.

micro
macro
Chapter
3
RESEARCH
PROCEDURES
The
data
for
our
test
of
a
macro
approach
to
estimating
intrinsic
water
quality
benefits
was
gathered
in
1576
personal
interviews
of
a
national
probability
sample
of
persons
18
years
of
age
and
older.
The
sample
was
designed
and
the
interviews
were
conducted
by
the
Roper
Organization
Interviewing
took
place
in
two
waves:
1289
people
were
intervviewed
in
late
January
­
early
February
1980
and
287
in
March
1980.
1
The
sampling
plan
was
a
multistage
probability
sample.
Once
an
eligible
person
was
identified,
as
many
as
four
attempts
were
made
to
arrange
an
interview.

Seventy­
three
percent
of
the
individuals
selected
were
ultimately
interviewed.

A
description
of
the
sampling
design
is
contained
in
Appendix
V.

For
the
entire
sample,
the
chances
are
95
out
of
100
that
the
results
on
a
particular
question
are
within
2
to
3
percentage
points
of
the
results
that
would
have
been
obtained
from
a
very
large
sample
selected
and
interviewed
in
a
similar
manner.

National
surveys
are
very
expensive
to
conduct.
We
were
able
to
minimize
the
costs
of
this
experiment
by
taking
advantage
of
an
ongoing
survey.
After
the
interview
for
the
original
survey
was
completed,
the
interviewers
administered
our
sequence
of
benefits
questions.
From
the
respondents'
perspective,
the
two
interviews
appeared
as
one
long
interview.

1
It
was
originally
intended
that
all
the
interviewing
would
be
done
in
the
initial
period,
but
the
survey
contractor
had
an
unanticipated
shortfall
in
interviews
which
went
unrecognized
for
a
month.
This
necessitated
further
interviewing
to
bring
the
sample
up
to
1500.
3­
2
While
this
procedure
allowed
us
to
have
our
instrument
field
tested
in
a
way
that
was
completely
satisfactory,
budgetary
constraints
limited
the
number
of
questions
we
could
ask
and
prevented
us
from
preparing
a
set
of
briefing
materials
for
the
interviewers.
Consequently,
as
will
be
discussed
at
length
in
later
chapters,
the
percent
of
respondents
who
failed
to
give
the
interviewers
the
amount
they
were
willing
to
pay
for
the
levels
of
water
quality
was
high,
as
was
the
percent
who
gave
zero
bids.
In
this
chapter
we
describe
the
context
of
the
survey
and
the
instrument.

Subsequent
chapters
discuss
the
reliability
and
validity
of
the
responses
and
the
values
people
have
for
water
quality.
The
final
chapter
presents
a
plan
for
revising
the
procedures
to
improve
the
measures
and
increase
the
response
rate
to
the
wtp
questions.

Context
The
RFF
water
benefits
questions
took
about
10­
15
minutes
to
administer.
They
were
preceded
by
a
separate
half­
hour
survey
on
environmental
issues
which
was
conducted
for
another
study.

Since
the
questions
for
this
other
study
set
the
context
for
the
water
benefit
questions
it
is
important
to
outline
briefly
their
content
and
results.
3­
3
We
will
discuss
the
possible
biasing
effect
they
may
have
had
at
a
later
point
in
this
report.

The
environmental
survey
consisted
of
some
100
items
which
probed
the
respondent's
views
about
national
priorities,
environmental
protection,
the
regulation
of
risks,
energy
issues,
values,
and
views
about
government
and
the
environmental
movement.
A
number
of
these
items
were
repeated
from
earlier
surveys
for
trend
purposes.
This
survey
sought
to
probe
beneath
the
respondent's
presumed
predisposition
towards
environmental
protection
(
as
consistently
shown
by
other
national
surveys)
by
asking
questions
which:
a)
forced
the
respondent
to
rank
order
the
environment
among
other
national
priorities,
b)
measured
concern
about
economic
issues
and
energy
shortages,
and
c)
which
forced
the
respondent
to
choose
between
tradeoffs
(
e.
g.
environment
vs.
growth
or
environmental
quality
vs.
lower
cost
of
regulation).
The
questionnaire
for
the
environmental
survey
which
preceded
the
benefits
questions,
including
the
background
questions
used
for
both
studies,
is
in
Appendix
IV.

When
the
respondents
were
forced
to
rank
order
problems
in
terms
of
which
should
have
the
most
government
priority,
"
reducing
pollution
of
air
and
water"
fell
to
sixth
place
(
out
of
10
problems)
from
the
second
place
position
it
held
at
the
time
of
the
original
Earth
Day
in
1970.
Responses
to
other
questions
in
the
environmental
survey
showed
the
respondents
were
extremely
concerned
about
inflation,
energy
problems,
and
defense.

Nevertheless
while
the
environment
is
apparently
no
longer
viewed
as
a
crisis
issue,
overall
support
for
environmental
protection
showed
continued
strength
2
in
the
trend
and
tradeoff
questions,
a
finding
confirmed
by
subsequent
surveys.

For
a
description
of
the
findings
of
the
environmental
survey
see
Public
Opinion
on
Environmental
Issues
(
Council
on
Environmental
Quality,
1980).
3­
4
The
data
from
the
environmental
survey
are
part
of
our
benefits
data
file
and
were
used
in
our
analysis
of
the
benefits
data.
The
environmental
survey
included
several
questions
about
water
quality
issues.
The
respondents
were
asked:

1.
How
worried
or
concerned
they
are
with
"
cleaning
up
our
waterways
and
reducing
water
pollution."
Thirty­
nine
percent
said
they
were
concerned
"
a
great
deal,"
and
at
the
opposite
extreme
16
percent
said
they
were
concerned
not
much
or
not
at
all
about
water
pollution.
(
See
Q.
11c,
Appendix
IV
for
the
marginals
and
comparisons
across
other
areas
of
concern
in
1980).

2.
Their
judgment
about
the
quality
of
the
water
in
the
"
lakes
and
streams
in
this
area"
on
a
self­
anchored
11
step
ladder
for
the
present,
past
(
five
years
ago)
and
the
future
(
five
years
from
now).
Q.
18­
20.
From
this
set
of
questions
it
is
possible
to
calculate
their
optimism
or
pessimism
about
change
in
local
water
quality
over
time.

3.
How
far
in
miles
the
nearest
freshwater
lake
and
river
large
enough
for
boating
are
from
their
home
(
Qs.
33a
and
b).

4.
A
series
of
questions
on
use
of
water
(
Qs.
58­
66)
For
boating,

swimming
and
fishing
in
a
freshwater
lake
or
stream,
respondents
were
asked
whether
they
had
engaged
in
each
activity
in
the
past
two
years,
if
so
whether
they
did
it
within
fifty
miles
of
their
home,
and
how
many
times
they
did
it
during
this
time
period.

We
used
these
questions
for
our
measures
of
recreational
water
use.
3­
5
Water
Pollution
Ladder
and
Value
Levels
The
levels
of
water
quality
for
which
we
sought
WTP
estimates
are
"
boatable,"
"
fishable,"
and
"
swimmable."
We
described
these
levels
in
words
and
depicted
them
graphically
by
means
of
a
water
quality
ladder.

Use
of
these
categories,
two
of
which
are
embodied
in
the
law
mandating
the
national
water
pollution
control
program,
allowed
us
to
avoid
the
methodological
problems
we
would
have
faced
had
we
chosen
to
describe
water
in
terms
of
the
numerous
abstract
technical
measures
of
pollution.
Although
the
boatable­
fishable­
swimmable
categories
are
widely
understood
by
the
public,
they
did
require
further
specification
on
our
part
to
ensure
that
people
perceived
them
in
a
similar
fashion.

We
defined
boatable
water
in
the
text
of
the
question
as
an
intermediate
level
between
water
which
"
has
oil,
raw
sewage
and
other
things
in
it,
has
no
plant
or
animal
life
and
smells
bad"
on
the
one
hand
and
water
which
is
of
fishable
quality
on
the
other.
Fishable
water
covers
a
fairly
large
range
of
water
quality.
Game
fish
like
bass
and
trout
cannot
tolerate
water
that
certain
types
of
fish
such
as
carp
and
catfish
flourish
in.

In
our
pretests
we
initially
experimented
with
two
levels
of
fishable
water
­­
one
for
"
rough"
fish
like
carp
or
catfish
and
the
other
for
game
fish
like
bass
­­
but
we
were
forced
to
abandon
this
distinction
because
people
were
confused
by
it.
We
adopted
a
single
definition
of
"
fishable"
as
water
"
clean
enough
so
that
game
fish
like
bass
can
live
in
it"
under
the
assumption
that
the
words
"
game
fish"

and
"
bass"
had
wide
recognition
and
connoted
water
of
the
quality
level
Congress
had
in
mind.
Swimmable
water
appeared
to
present
less
difficulty
3­
6
for
popular
understanding
since
the
enforcement
of
water
quality
for
swimming
by
health
authorities
has
led
to
widespread
awareness
that
swimming
in
polluted
water
can
cause
sickeness
to
humans.

Because
WTP
questions
have
to
describe
in
some
detail
the
conditions
of
the
"
market"
for
the
good
they
are
inevitably
longer
than
the
usual
survey
research
questions.
Respondents
quickly
become
bored
and
restless
if
material
is
read
to
them
without
giving
them
frequent
opportunities
to
express
judgments
or
to
look
at
visual
aids.
We
designed
the
RFF
instrument
to
be
as
interactive
as
possible
by
interspersing
the
text
with
questions
which
required
the
respondents
to
use
the
newly
described
water
quality
categories.
We
also
handed
them
a
water
quality
ladder
card
which
was
referred
to
constantly
during
the
sequence
of
benefits
questions.

Figure
3.1
shows
the
card.
The
ladder
is
similar
to
the
self­
anchoring
ladder
used
earlier
in
the
interview.
The
top,
step
10,
was
called
the
"
best
possible
water
quality"
and
the
bottom,
step
0,
was
the
"
worst
possible
water
quality."
This
time,
however,
we
anchored
it
by
designating
five
levels
of
water
quality
at
different
steps
on
the
ladder.
Level
E,

at
.8,
was
specified
as
a
point
on
the
ladder
where
the
water
was
even
unfit
for
boating
although
the
active
range
below
2.5
was
described
as
being
of
this
quality.
Level
D,
2.5,
was
where
it
became
okay
for
boating;

C
at
5
was
fishable,
B
at
7
was
swimmable
and
9.5
was
identified
as
A
where
the
water
is
safe
to
drink.
These
numerical
positions
were
estimated
by
indexing
a
set
of
five
objective
scientific
water
quality
parameters
using
a
variant
of
the
National
Sanitation
Foundatin's
Water
Quality
Index
(
Booth
et
al.,
B
Figure
3.1
3­
7
(
WATER
QUALITY
LADDER
CARD)
#
684
BEST
POSSIBLE
WATER
QUALITY
10
9
8
7
6
5
4
3
2
1
0
SAFE
TO
DRINK
SAFE
FOR
SWIMMING
WORST
POSSIBLE
WATER
QUALITY
GAME
FISH
LIKE
BASS
CAN
LIVE
IN
IT
OKAY
FOR
BOATING
3­
8
1976;
McClelland,
1974).
The
method
is
described
in
Appendix
II.

Although
this
is
necessarily
a
tenuous
scaling
procedure,
it
yielded
a
set
of
positions
which
appear
reasonable.
Our
pretests
showed
that
respondents
did
not
seem
to
be
sensitive
to
changes
of
one
or
two
rungs
in
the
location
of
the
water
quality
levels
along
the
scale.

We
introduced
the
market
and
the
ladder
in
the
following
manner:

This
last
group
of
questions
is
about
the
quality
of
water
in
the
nation's
lakes
and
streams.
Comgress
passed
strict
water
pollution
control
laws
in
1972
and
1977.
As
a
result
many
communities
have
to
build
and
run
new
modern
sewage
treatment
plants
and
many
industries
have
to
install
water
pollution
control
equipment.

Here
is
a
picture
of
a
ladder
that
shows
various
levels
of
the
quality
of
water.
(
HAND
RESPONDENT
WATER
QUALITY
LADDER
CARD)
Please
keep
in
mind
that
we
are
not
talking
about
the
drinking
water
in
your
home.
Nor
are
we
talking
about
the
ocean.
We
are
talking
only
about
freshwater
lakes,
rivers
and
streams
that
people
look
at
and
in
which
they
go
boating,
fishing
and
swimming.

The
top
of
the
ladder
stands
for
the
best
possible
quality
of
water,
that
is,
the
purest
spring
water.
The
bottom
stands
for
the
worst
possible
quality
of
water.
Unlike
the
other
ladders
we
have
used
in
this
survey,
on
this
ladder
we
have
marked
different
levels
of
the
quality
of
water.
For
example
.
.
.
.
(
POINT
TO
EACH
LEVEL:
E,
D,
C,
AND
SO
ON,
AS
YOU
READ
STATEMENTS
BELOW)

Level
E
(
POINTING)
is
so
polluted
that
it
has
oil,
raw
sewage
and
other
things
in
it,
has
no
plant
or
animal
life
and
smells
bad
Water
at
level
D
is
okay
for
boating
but
not
for
fishing
or
swimming
Level
C
shows
where
rivers,
lakes
and
streams
are
clean
enough
so
that
game
fish
like
bass
can
live
in
them
Level
B
shows
where
the
water
is
clean
enough
so
that
people
can
swim
in
it
safely
And
at
level
A,
the
quality
of
the
water
is
so
good
that
it
would
be
possible
to
drink
it
directly
from
a
lake
or
stream
if
you
wanted
to
3­
9
We
thus
defined
the
environmental
good
as
freshwater
lakes,
rivers
and
streams
and
distinguished
it
from
drinking
water
and
salt
water.
We
specifically
invoked
visual
values
as
well
as
the
active
use
values
of
boating,
fishing
and
swimming.

Our
intention
was
to
obtain
a
WTP
estimate
for
national
water
quality.

In
order
to
get
the
respondent
to
think
about
the
national
situation
the
interviewer
next
asked:

Now
let's
think
about
all
of
the
nation's
rivers,
lakes
and
streams.
Some
of
them
are
quite
clean
and
others
are
more
or
less
polluted.
Looking
at
this
ladder,
would
you
say
that
all
but
a
tiny
fraction
of
the
nation's
rivers,
lakes
and
streams
are
at
least
at
level
D
in
the
quality
of
their
water
today
or
not?

Strictly
speaking,
the
law
mandates
water
cleanup
for
all
freshwater
bodies.

We
substituted
"
all
but
a
tiny
fraction"
for
"
all"
in
this
and
the
following
questions
because
we
did
not
want
to
unnecessarily
complicate
the
issue
by
having
respondents
speculate
about
the
impossibility
of
every
portion
of
every
water
body
in
the
nation
being
at
a
certain
water
quality
level
at
all
times.
Six
out
of
ten
respondents
agreed
that
today
all
but
a
fraction
of
the
nation's
freshwater
bodies
are
at
level
D
while
17
percent
were
not
sure
and
20
percent
felt
that
level
had
not
yet
been
reached.

The
next
section
of
the
instrument
was
meant
to
introduce
the
respondent
to
two
things:
1)
the
fact
that
water
pollution
control
costs
money
and
2)
that
the
level
of
cleanup
is
a
matter
of
preference.
We
did
this
by
asking
the
following
question:
3­
10
81.
As
you
know
it
takes
money
to
clean
up
our
nation's
lakes
and
rivers.
Taking
that
into
account,
and
thinking
of
overall
water
quality
where
all
but
a
tiny
fraction
of
the
nation's
lakes
and
rivers
are
at
a
particular
level,
which
level
of
overall
water
quality
do
you
think
the
nation
should
plan
to
reach
within
the
next
five
years
or
so
­­
level
E,
D,
C,
B,
or
A?

Eighty­
five
percent
chose
a
goal
of
fishable
or
better
(
C,
B,
or
A)
while
57
percent
chose
swimmable
or
better
(
B
or
A).

Payment
Vehicle
We
used
two
principal
criteria
to
choose
our
payment
vehicle.
The
first
is
realism
­­
the
vehicle
should
match
the
way
people
actually
pay
for
higher
water
quality
as
closely
as
possible.
The
second
criteria
is
conservativism
­­
every
effort
should
be
made
to
avoid
a
false
overstatement
of
willingness
to
pay.
Conservativism
in
question
design
is
important
because
unless
respondents
are
made
to
pay
the
amounts
they
offer,
WTP
studies
are
inevitably
hypothetical
in
character.
The
bias
associated
with
hypothetical
situations
is
towards
overstating
the
amount
the
person
is
willing
to
pay
3
although
the
amount
of
overstatement
is
not
necessarily
large
(
Bohm,
1972)
and
is
sometimes
nonexistent
(
Davis,
1980).
Given
many
economists'
fear
that
the
WTP
methodology
is
biased
upward,
the
findings
of
WTP
questions
will
be
credible
only
if
every
effort
is
made
to
avoid
this
bias.
Our
procedure
was
to
design
our
instrument
so
that,
whenever
possible,
any
bias
present
is
toward
lowering
rather
than
raising
the
WTP
amount.

We
selected
annual
household
payment
in
higher
prices
and
taxes
as
our
payment
vehicle
because
this
is
the
way
people
pay
for
water
pollution
control
programs.
A
portion
of
each
household's
annual
federal
tax
payment
See
Chapter
4.
3­
11
goes
towards
the
expense
of
regulating
water
pollution
and
providing
construction
grants
for
sewage
treatment
plants.
Local
sewage
taxes
pay
for
the
maintenance
of
three
plants.
Those
private
users
who
incur
pollution
control
expenses,
such
as
manufacturing
plants,
ultimately
pass
much
or
all
of
the
cost
along
to
consumers
in
higher
prices.
This
payment
vehicle
is
conservative
because:

Ever
since
the
passage
of
Proposition
13
in
California
in
1977,

opposition
to
the
current
level
of
taxes
is
a
commonly
expressed
attitude
which
is
socially
acceptable
(
even
normative).
Concern
about
inflation
was
the
nation's
"
most
important
problem"
according
to
polls
taken
at
the
time
of
the
RFF
survey.
Thus
we
can
assume
the
words
"
taxes
and
higher
prices"
will
not
be
taken
lightly
by
our
respondents
and
may,
for
some,
have
a
highly
charged
negative
connotation.

By
asking
for
the
annual
amount
a
person
is
willing
to
pay
instead
of
for
a
monthly
amount,
we
avoid
the
possibility
of
an
"
easy
payment
plan"
underestimation.

Starting
Point
Our
review
of
the
literature
on
micro
WTP
studies
and
on
survey
research
more
generally,
identified
starting
point
bias
as
a
particularly
serious
problem
for
our
study.
Because
of
this
we
developed
and
tested
an
alternative
to
the
commonly
used
bidding
game
WTP
method.
In
this
section
we
outline
the
problems
presented
by
the
bidding
game
technique
and
describe
our
alternative
procedure
­­
the
payment
card
method.
3­
12
The
widely
used
bidding
game
format
for
WTP
studies
uses
a
sequence
of
yes/
no
questions
and
normally
requires
the
interviewer
to
begin
the
bidding
process
by
offering
an
initial
amount.
The
subsequent
bids
flow
from
that
point,
albeit
in
either
direction.
If
the
amount
presented
influences
the
respondent's
final
bid
in
some
systematic
way
­­
starting
point
bias
­­
we
have
a
serious
problem.

There
are
a
priori
reasons
for
suspecting
such
a
bias
in
this
type
of
situation.
The
tendency
of
respondents
to
give
a
socially
desirable
answer
(
Edwards,
1957;
Dohrenwend,
1966;
Phillips
and
Clancy,
1970,
1972)

or
to
acquiesce
when
confronted
with
questions
using
a
yes/
no
agree
or
disagree
format
(
Couch
and
Keniston,
1960;
Campbell
et
al.,
1967;
Carr,

1977;
Jackman,
1973;
and
Phillips
and
Clancy,
1970)
is
well
documented.

Accordingly,
when
valuing
a
public
good
like
water
quality,
a
respondent
may
be
reluctant
to
reject
a
starting
bid
even
when
it
is
higher
than
he
is
willing
to
pay
for
fear
of
appearing
cheap
or
lacking
a
social
conscience
(
social
desirability
effect)
and/
or
because
of
a
tendency
on
the
part
of
the
respondent
to
agree
with
suggestions
offered
by
the
interviewer
(
acquiescence
effect).

In
practice,
strong
starting
point
effects
have
been
found
by
some
researchers
doing
micro
WTP
studies
(
Rowe
et
al.,
1979)
although
other
researchers
have
not
found
them
(
Thayer,
et
al.,
forthcoming;
Brookshire,

et
al.,
1979;
Brookshire
et
al.,
1980).
Where
starting
point
bias
has
been
discovered,
the
effect
of
higher
starting
points
is
to
raise
the
mean
WTP
amount.
3­
13
The
acquiescence
effect
shows
a
strong
relationship
with
education
­­

people
with
less
education
are
much
more
likely
to
acquiesce
than
those
with
more
education
(
Jackman,
1973).
This
introduces
a
further
bias.
If
we
assume,
as
studies
have
shown,
that
WTP
varies
by
income
level
and
that
income
is
correlated
with
education,
then
the
potential
for
an
education/
WTP
interaction
effect
is
strong
when
a
single
starting
point
is
used
for
the
entire
sample.
When
choosing
a
single
starting
point,
the
researcher
needs
one
that
will
be
below
the
expected
mean
for
the
entire
sample,
but
not
too
far
below
or
the
process
of
bidding
upward
to
find
the
maximum
WTP
amount
will
be
too
laborious.
An
initial
bid
which
meets
this
requirement
for
the
entire
sample
can
be
expected
to
be
below
the
mean
for
people
in
the
$
15­
25,000
range,
close
to
the
mean
of
the
real
bid
for
someone
in
the
$
8,000­
14,999
income
range
and
above
the
real
mean
bid
for
those
with
lower
incomes.
Since
many
people
in
the
lower
income
range
will
also
have
low
educations,
in
this
situation
they
are
likely,
by
the
operation
of
the
acquiescence
effect,
to
overbid
for
the
good
in
question.
The
reverse
is
less
likely
to
happen
for
those
with
an
income
above
$
25,000
because
their
educational
level
is
higher
(
on
the
average)
and
therefore
their
propensity
for
acquiescence
in
the
interview
situation
is
lower.
Thus
even
if
the
overall
starting
bias
described
earlier
is
not
present,
overstatement
of
benefits
by
lower
income
people
will
bias
the
WTP
amounts
upwards.

A
further
problem
with
the
bidding
game
technique
is
that
the
process
of
iterating
from
a
starting
point
to
a
final
WTP
amount
can
be
tedious
if
the
starting
point
lies
some
distance
from
the
respondent's
real
WTP
amount.
If
the
range
is
narrow
­­
such
that
most
respondents,
for
example,
3­
14
value
a
certain
good
at
between
$
1
and
$
5
per
month
on
their
utility
bill
­­
and
if
the
increments
are
fairly
large
­­
say
$
1
­­
then
the
process
can
be
accomplished
fairly
efficiently.
When
this
is
not
the
case,
the
length
of
the
iteration
process
can
alienate
respondents
or
cause
them
to
cease
bidding
before
reaching
their
maximum
amount.

The
problems
with
the
bidding
game
approach
enumerated
above
are
exacerbated
for
payment
vehicles
like
ours
which
engender
large
bids
(

because
they
ask
for
an
annual
household
amount
for
national
water
quality)

and
which
are
strongly
income
dependent
(
owing
to
the
income
tax
component
of
the
vehicle).
Moreover,
it
seems
questionable
that
the
bidding
game
technique
can
be
used
reliably
by
professional
interviewers
such
as
ours
who
are
spread
across
the
country
and
cannot
be
personally
instructed
in
its
use.
For
these
reasons
we
developed
our
payment
card
technique
to
elicit
the
respondent's
WTP
amounts.

In
this
technique
the
respondent
is
given
a
card
which
contains
a
menu
of
amounts
which
begin
at
$
0
and
increase
by
a
fixed
interval
until
an
arbitrarily
determined
large
amount
is
reached.
When
the
time
comes
to
elicit
the
WTP
amount,
the
respondent
is
asked
to
pick
a
number
off
the
card
(
or
any
number
in
between)
which
"
is
the
most
you
would
be
willing
to
pay
in
taxes
and
higher
prices
each
year"
(
italics
in
the
original)
for
a
given
level
of
water
quality.
The
question
asks
people
to
give
us
the
highest
amount
they
are
willing
to
pay
and
we
accepted
their
answer
as
representing
such
an
amount.
In
our
pretesting
we
tried
asking
people
if
3­
15
they
would
be
willing
to
pay
a
higher
amount
than
the
one
they
picked
and
found
some
people
resented
being
"
pushed"
once
they
had
settled
on
an
amount.

Others
would
give
us
a
higher
amount
but
in
such
a
way
that
we
suspected
they
were
acquiescing
to
interviewer
pressure
rather
than
revealing
their
true
consumer
surplus.

The
payment
card
has
two
special
features:

1.
It
is
anchored.
In
our
initial
pretests
we
found
the
respondents
had
considerable
difficulty
in
determining
their
willingness
to
pay
when
we
used
a
card
which
only
presented
various
dollar
amounts.
A
number
of
them
expressed
embarrasment,
confusion,
or
resentment
at
the
task
and
some
who
gave
us
amounts
indicated
they
were
very
uncertain
about
them.
We
determined
that
the
problem
lay
with
the
lack
of
benchmarks
for
their
estimates.
People
are
not
normally
aware
of
the
total
amounts
they
pay
for
public
goods
even
when
that
amount
comes
out
of
their
taxes,
nor
do
they
know
how
much
they
cost.
Without
a
way
of
psychologically
anchoring
their
estimate
in
some
manner
they
were
not
able
to
arrive
at
meaningful
estimates.

They
needed
benchmarks
of
some
kind
which
would
convey
sufficient
information
without
biasing
their
WTP
amounts.
We
reasoned
that
the
most
appropriate
benchmarks
for
WTP
for
water
pollution
control
would
be
the
amounts
they
are
already
paying
in
higher
prices
and
taxes
for
other
non­
environmental
public
goods.
We
identified
amounts
on
the
card
for
several
such
goods
and
conducted
further
pretests.
These
showed
the
benchmarks
made
the
task
meaningful
for
most
people.

The
use
of
payment
cards
with
benchmarks
raises
the
possibility
of
information
bias.
Are
the
respondents
who
gave
us
amounts
for
water
pollution
3­
16
control
using
the
benchmarks
for
general
orientation
or
are
they
basing
their
amounts
directly
on
the
benchmarks
themselves
in
some
manner?
In
the
former
case
people
would
be
giving
us
unique
values
for
water
quality;
in
the
latter
case
they
would
be
giving
us
values
for
water
quality
relative
to
what
they
think
they
are
paying
for
a
particular
set
of
other
public
goods.
If
the
latter
case
holds
and
their
water
quality
values
are
sensitive
to
changes
in
the
benchmark
amounts
or
to
changes
in
the
set
of
public
goods
identified
on
the
payment
card,
their
validity
as
estimates
of
consumer
surplus
for
water
quality
are
suspect.

We
designed
our
study
to
test
for
information
bias
due
to
the
benchmarks.

Four
different
versions
of
the
payment
cards
were
prepared
and
administered
to
approximately
equivalent
sub­
samples.
Figures
3.2
shows
the
cards
given
to
the
lower­
medium
income
respondents
($
10,000­
14,999
annual
family
income)

for
the
A,
B,
C,
and
D
versions.
These
versions
varied
as
follows:

A
Benchmarks
are
shown
for
the
amounts
we
estimated
the
average
household
of
that
income
level
contributes
to
the
space
program,
highways,
public
education
and
defense.

B
The
same
four
public
goods
and
amounts
as
on
A
plus
police
and
fire
protection.

C
The
same
four
public
goods
used
in
version
A
were
shown,
but
for
amounts
25
percent
higher
than
on
version
A.

D
The
same
four
public
goods
and
amounts
as
in
Version
A,
plus
the
estimated
amount
for
water
pollution
control.

We
added
the
police
and
fire
good
in
version
B
to
see
if
the
insertion
of
a
new
item
in
the
dollar
range
where
water
pollution
benefits
estimates
were
likely
to
fall
would
affect
those
estimates.
Version
C
seeks
to
test
whether
the
actual
amounts
shown
for
the
benchmarks
affect
the
water
pollution
WTP
amounts.
We
purposely
omitted
environmental
goods
in
each
of
the
3­
17
Figure
3.2
PAYMENT
CARDS
FOR
VERSIONS
A,
B,
C,
D
FOR
PEOPLE
WITH
FAMILY
INCOMES
OF
$
10,000­
14,999
3­
18
first
three
versions
to
avoid
having
people
would
tell
us
what
they
think
they
should
give
rather
than
what
they
actually
want
to
pay.
In
version
D
we
added
our
estimate
of
what
average
households
are
actually
paying
for
water
pollution
control
to
see
whether
this
information
actually
does
bias
the
WTP
amounts.

Deriving
the
dollar
estimates
for
each
of
our
benchmark
public
goods
was
a
difficult
task
particularly
because
we
needed
them
for
four
income
levels
as
well
(
see
below).
A
detailed
description
of
our
procedures
is
given
in
Appendix
III.
We
are
satisfied
that
the
estimates
are
sufficiently
close
approximations
to
suffice
for
this
test.
If
it
turned
out
that
people's
WTP
amounts
are
very
sensitive
to
the
benchmark
amounts,
then
much
more
effort
would
be
required
to
improve
the
accuracy
of
these
estimates.

2.
It
is
income
adjusted.
For
the
reasons
stated
earlier,
the
amounts
people
are
actually
paying
for
water
pollution
control
vary
by
income.
This
is
also
the
case
for
the
other
public
goods
which
we
used
as
benchmarks.

We
corrected
for
this
by
developing
benchmark
goods
estimates
for
four
different
income
categories:
I)
family
income
under
$
10,000;
II)
$
10,000­

14,999;
III)
$
15,000­
24,999;
IV)
$
25,000
and
above.
(
Appendix
I
gives
our
public
goods
estimates
for
each
of
these
income
categories).
Each
interviewer
therefore
had
four
different
payment
cards
for
each
of
the
A,
B,
C,

and
D
forms.
At
the
appropriate
point
in
the
interview
the
interviewer
gave
the
respondent
the
payment
card
for
his
or
her
income
category.
(
A
question
on
income
preceded
the
water
quality
benefits
questions.)
For
the
10
percent
of
respondents
who
refused
to
divulge
their
income
our
procedure
was
to
give
them
the
income
card
for
income
level
IV,
the
highest
income
level
as
people
with
higher
incomes
are
more
likely
to
refuse
to
divulge
their
income.
3­
19
Figure
3.3
gives
the
four
forms
used
for
Version
A.
The
card
for
the
lowest
income
category
(
I)
shows
an
annual
defense
figure
of
$
325
while
those
in
the
highest
income
category
were
told
they
are
spending
between
$
3000
and
$
3075
per
year
on
defense.
In
order
to
make
the
stimuli
shown
on
the
payment
cards
as
similar
as
possible
to
each
of
the
four
income
groups
we
varied
the
range
of
potential
amounts.
Each
card
shows
60
amounts.
Income
category
I's
amounts
ranged
from
$
0
to
$
440
while
those
for
IV
were
$
0
to
$
3285.
These
ranges
and
the
intervals
(
which
are
wider
at
the
higher
levels)
were
chosen
so
that
the
visual
pattern
of
public
goods
amounts
was
approximately
the
same
for
each
income
level.
In
each
case
the
maximum
amount
on
the
card
is
roughly
30
percent
greater
than
the
amount
shown
for
defense.

The
following
is
the
text
of
the
first
WTP
question
in
our
instrument.

The
same
text
was
used
for
versions
A,
B,
and
C
with
the
exception
of
the
additional
mention
of
police
and
fire
in
paragraph
two
for
version
B.

82.
Improving
the
quality
of
the
nation's
water
is
just
one
of
many
things
we
all
have
to
pay
for
as
taxpayers
and
as
consumers.
That
is,
the
costs
of
things
like
improving
water
quality
are
paid
partly
by
government
out
of
what
we
pay
in
taxes
and
partly
by
companies
out
of
what
we
pay
for
the
things
they
sell
us.

This
scale
card
shows
about
how
much
people
in
your
general
income
category
paid
in
1979
in
taxes
and
higher
prices
for
things
like
national
defense,
roads
and
highways,
public
schools
and
the
space
program.
(
HAND
RESPONDENT
APPROPRIATE
SCALE
CARD
A­
I,
A­
II,
A­
III,
OR
A­
IV:
LET
RESPONDENT
KEEP
WATER
QUALITY
LADDER
CARD)

You
will
see
different
amounts
of
money
listed
with
words
like
"
highways"
and
"
public
education"
appearing
by
the
amount
of
money
average
size
households
paid
for
each
one
last
year.
"
Highways"
here
refers
to
the
construction
and
maintenance
of
all
the
nation's
highways
and
roads.
"
Public
education"
refers
to
all
public
elementary
and
secondary
schools
but
does
not
include
the
costs
of
public
universities.
3­
20
Figure
3.3
PAYMENT
CARDS
FOR
INCOME
LEVELS
I­
IV
FOR
VERSION
A
3­
21
I
want
to
ask
you
some
questions
about
what
amounts
of
money,
if
any,
you
would
be
willing
to
pay
for
varying
levels
of
overall
water
quality
in
the
nation's
lakes,
rivers
and
streams.
Please
keep
in
mind
that
the
money
would
go
for
sewage
treatment
plants
in
communities
through
various
kinds
of
taxes
(
such
as
withholding
taxes,
sales
taxes
and
sewage
fees)
and
for
pollution
control
equipment
the
government
would
require
industries
to
install,
thus
raising
the
prices
of
what
they
make.

At
the
present
time
the
average
quality
of
water
in
the
nation's
lakes,
rivers
and
streams
is
at
about
level
D
on
the
ladder.
(
POINT
TO
LEVEL
D
ON
WATER
QUALITY
LADDER
CARD)
If
no
more
money
were
spent
at
all
tomorrow
on
water
quality,
the
overall
quality
of
the
nation's
lakes
and
rivers
would
fall
back
to
about
level
E.
(
POINT
TO
LEVEL
E)
People
have
different
ideas
about
how
important
the
quality
of
lakes,
rivers
and
streams
is
to
them
personally.
Thinking
about
your
household's
annual
income
and
the
fact
that
money
spent
for
one
thing
can't
be
spent
for
another,
how
much
do
you
think
it
is
worth
to
you
to
keep
the
water
quality
in
the
nation
from
slipping
from
level
D
back
to
level
E?
That
is,
which
amount
on
this
scale
card,
or
any
amount
in
between,
is
the
most
you
would
be
willing
to
pay
in
taxes
and
higher
prices
each
year
to
keep
the
nation's
overall
water
quality
at
level
D
where
virtually
all
of
it
is
at
least
clean
enough
for
boating?
If
it
is
not
worth
anything
to
you,
please
do
not
hesitate
to
say
so.

Several
aspects
of
question
82
bear
comment.
For
the
purpose
of
convenience
we
started
the
process
of
demand
revelation
with
the
present
level
of
national
water
quality
(
boatable)
and
asked
respondents
to
value
a
reduction
in
this
quality
to
level
E,
non­
boatable.
(
In
subsequent
questions
we
had
them
value
hypothetical
increases
from
boatable
to
fishable
and
then
swimmable.)
In
this
question
we
expanded
the
account
given
in
the
previous
questions
about
how
their
money
would
be
used
and
reinforced
the
ideas
that
the
WTP
amount
would
be
coming
out
of
their
annual
income
and
its
use
for
this
purpose
would
preclude
other
uses
of
the
money.
At
two
points
in
this
question
we
legitimated
a
low
or
zero
WTP
amount
in
an
effort
to
minimize
the
social
desirability
effect.
We
noted
that
"
people
have
different
ideas"
about
the
importance
of
water
quality
to
them
personally
3­
22
and
at
the
conclusion
of
the
question
we
stated:
"
If
it
is
not
worth
anything
to
you,
please
don't
hestiate
to
say
so."

The
response
categories
which
were
supplied
to
the
interviewers
for
this
question
were:

Write
in
amount:
$

Depends
(
voluntary)

Not
sure
Not
worth
anything
Through
a
misunderstanding
the
survey
contractor
did
two
things
which
may
have
biased
the
results.
First
in
this
and
the
next
question,
those
who
responded
"
not
worth
anything"
­­
in
effect
a
$
0
bid
­­
were
not
asked
how
much
they
were
willing
to
pay
for
water
of
higher
quality.
Instead,
the
interviewers
skipped
directly
to
the
last
question.

Presumably
most
of
the
people
who
valued
boatable
water
at
$
0
were
generally
unwilling
to
pay
for
water
pollution
control
of
any
kind
and
would
also
have
valued
fishable
and
swimmable
quality
water
at
$
0.
Our
analysis
of
the
views
of
these
people
about
water
pollution
and
environmental
quality
suggests
that
this
conjecture
is
probably
true
for
most
of
them.
But
some
of
them
may
indeed
only
value
water
nationwide
when
it
reaches
the
fishable
and/
or
swimmable
quality
levels.
If
so,
they
would
have
given
a
WTP
amount
greater
than
$
0
for
the
higher
levels,
if
they
had
the
opportunity,
despite
their
$
0
bid
for
the
lower
level.
Second,
when
the
data
were
keypunched,

the
contractor
restricted
the
WTP
amounts
to
three
colums,
thereby
limiting
the
maximum
WTP
amount
to
$
999.
For
versions
A,
B,
C
combined,
43
People
3­
23
were
recorded
as
WTP
this
maximum
amount
for
level
B.
We
have
no
way
of
knowing
how
many
of
these
people
actually
valued
water
quality
at
an
amount
higher
than
this.
It
is
our
judgment
that
both
these
errors
have
had
only
a
minor
effect
on
our
estimates.
The
direction
of
the
bias
is,
of
course,
conservative.

The
next
question
sought
the
respondents'
WTP
for
fishable
level
C.
resulting
water,

83.
As
I
mentioned
earlier,
almost
all
of
the
rivers
and
lakes
in
the
United
States
are
at
least
at
level
D
in
water
quality.
What
do
you
think
it
is
worth
to
you
not
only
to
keep
them
from
becoming
more
polluted
but
also
to
raise
their
overall
quality
to
level
C?
That
is,
including
the
amount
you
just
gave
me,
which
amount
on
the
scale
card
is
the
most
you
would
be
willing
to
pay
in
taxes
and
higher
prices
each
year
to
raise
the
overall
level
of
water
quality
from
level
D
to
level
C
where
virtually
all
of
it
would
at
least
be
clean
enough
for
fish
like
bass
to
live
in?

The
final
WTP
question
used
the
same
format
for
swimmable
water,

level
B.

84.
What
about
getting
virtually
all
of
the
nation's
lakes
and
rivers
up
to
level
B
on
the
ladder?
Including
the
amounts
of
money
you
have
already
given
me,
which
amount
on
the
scale
card
is
the
most
you
would
be
willing
to
pay
in
taxes
and
higher
prices
each
year
to
make
almost
all
the
nation's
lakes,
rivers
and
streams
clean
enough
so
that
people
could
swim
in
them?

In
two
of
the
versions,
A,
and
C,
we
asked
the
respondents
to
evaluate
the
amount
of
information
we
provided
them
about
the
WTP
exercise.
We
were
precluded
from
asking
this
of
all
the
respondents
because
of
severe
constraints
on
the
length
of
the
questionnaire.
3­
24
85.
Finally,
in
terms
of
your
being
able
to
decide
exactly
how
much
you,
yourself,
would
be
willing
to
pay
as
a
taxpayer
and
consumer
for
better
water
quality,
would
you
say
in
the
last
few
questions
we
gave
you
more
than
enough
information,
about
enough
information,
not
quite
enough,
or
not
enough
information
at
all?
CHAPTER
4
CONTROL
FOR
BIASES
Prior
to
discussing
our
findings
it
is
necessary
to
examine
the
character
of
the
data
we
have
gathered.
To
what
extent
are
they
free
from
bias?
The
micro
willingness­
to­
pay
literature
has
devoted
considerable
attention
to
the
potential
biases,
their
effect
and
how
they
may
be
overcome
(
Schulze,
et
al.,
1980).
Table
4.1
lists
these
potential
biases
and
several
others
which
we
believe
to
be
important.

Table
4.1
POTENTIAL
BIASES
IN
WILLINGNESS
TO
PAY
STUDIES
General
Sampling
Strategic
Sample
Hypothetic
Response
Rate
Instrument
Interview
Starting
Point
Item
non­
response
Payment
Vehicle
Interview
Procedure
Information
Interviewer
Order
4­
2
GENERAL
BIASES
Strategic
and
hypothetic
are
the
two
sources
of
bias
of
greatest
fundamental
concern
to
economists
who
wish
to
evaluate
the
validity
of
willingness
to
pay
surveys.

Strategic
Bias
Its
Nature
Strategic
bias
is
the
attempt
by
respondents
to
influence
the
outcome
of
a
study
in
a
direction
which
favors
the
respondents'
interests
by
deliberately
misrepresenting
their
demand
for
a
good.
In
1954,
Paul
Samuelson
argued
on
free­
rider
grounds
that
a
person
would
be
motivated
to
"
pretend
to
have
less
interest
in
a
given
collective
consumption
activity
than
he
really
has"
and
despaired
of
finding
a
way
of
overcoming
this
problem
(
1954).
Samuelson
assumes
that
the
individual
would
believe
he
or
she
would
have
to
pay
the
amount
he
or
she
declares
as
being
willing
to
pay.
If
this
assumption
is
relaxed,

as
seems
reasonable,
many
economists
believe
an
incentive
to
overestimate
consumption
would
be
prevalent
(
Freeman,
19796:
88).
For
example,
take
a
survey
whose
respondents
believe
the
mean
WTP
amount
for
all
respondents
will
influence
the
government's
provision
of
a
public
good
and
that
they
will
not
be
obligated
to
pay
their
WTP
amount.
If
they
value
the
good,

the
respondents
may
attempt
to
raise
the
mean
(
and
impose
their
preference)

by
overstating
their
willingness
to
pay.
Robert
Crandall
seems
to
have
this
kind
of
situation
in
mind
when
he
wrote:
"
Such
surveys
(
consumer
1
See
Kutz
(
1975)
for
the
the
theoretical
conditions
necessary
for
successful
strategic
behavior.
4­
3
surveys)
are
always
biased
when
the
respondent
knows
that
he
or
she
does
not
have
to
write
a
check
to
confirm
the
answer"
(
Crandall,
1979).
Conversely,

those
who
do
not
value
the
good
very
highly
but
assume
that
many
others
do,

may
underestimate
their
willingness
to
pay
in
order
to
lower
the
mean
and
bring
it
closer
to
their
actual
willingness
to
pay.

Empirical
attempts
to
test
for
strategic
bias
in
willingness
to
pay
studies
and
laboratory
experiments
have
consistently
failed
to
find
it
(
Brookshire,
et
al.,
1979:
22­
23;
V.
L.
Smith,
1977).
A
much
cited
challenge
to
the
notion
that
strategic
bias
can
be
overcome
in
WTP
studies
is
an
experiment
conducted
by
Peter
Bohm.
In
one
of
the
few
attempts
to
compare
hypothetical
WTP
questions
with
the
results
from
identical
non­
hypothetical
situations,
Bohm
(
1972)
conducted
an
experiment
where
participants
bid
for
the
opportunity
to
see
a
closed
circuit
television
program.
He
ran
six
different
versions
of
the
experiment
most
of
which
systematically
introduced
incentives
to
act
strategically
in
a
situation
where
the
respodent
actually
had
to
pay
their
bids.
Only
one
version,
Group
VI,
gave
bids
which
were
significantly
different
from
any
of
the
others.
Since
this
group
was
told
that
they
would
not
actually
have
to
pay
what
they
bid,

Bohm
draws
the
conclusion
that
"
when
no
payments
and/
or
forced
decisions
are
involved
people
will
act
in
an
irresponsible
manner"
(
Bohm,
1972:
125).

In
other
words,
when
the
consequences
for
respondents
are
hypothetical
they
will
overbid.
Careful
examination
of
Bohm's
study
shows
that
this
conclusion
is
unwarranted:
4­
4
1.
Out
of
five
comparisons,
Group
VI's
mean
bid
was
significantly
higher
in
only
one
case
(
Group
III).

2.
Group
VI
was
higher
in
income
than
the
other
groups
which
may
account
for
the
size
of
its
mean
payment.

3.
Group
V
also
did
not
have
to
pay
its
bid.
If
strategic
bias
was
operative,
there
are
reasons
to
think
that
this
group
should
have
had
the
highest
bid
of
all,
but
it
did
not.

4.
Unlike
the
other
groups,
Group
VI
had
one
high
outlier
(
at
50
where
the
median
bid
was
10)
which
raised
its
mean
bid
considerably.

When
the
outlier
is
removed,
its
mean
payment
is
reduced
from
10.19
to
9.45
Kroner
and
the
difference
between
Group
VI
and
Group
III
drops
below
the
.05
level
of
significance.
It
would
appear
that
only
one
person
2
of
54
may
have
acted
"
irresponsibly."

The
incentives
to
misrepresent
preferences
are
minimal
in
most
WTP
surveys
because
respondents
lack
either
the
information
necessary
to
act
strategically
or
the
incentive
to
do
so
because
respondents
do
not
believe
they
will
be
directly
affected
by
the
study's
outcome.
Although
respondents
take
valuation
questions
seriously,
most
do
not
think
their
responses
will
have
an
immediate
effect
on
policy
nor
should
they
since
policy
has
rarely,
if
ever,

been
set
in
this
manner.
The
now
conventional
wisdom
on
strategic
bias
in
WTP
surveys
was
recently
summarized
by
Feenberg
and
Mills
in
their
recent
review
of
water
benefit
analysis.
They
concluded,
"
It
is
unlikely
that
the
problem
is
serious"
(
Feenberg
and
Mills,
1980).

2
We
do
not
believe
the
one
person
acted
strategically
since
an
incentive
to
overbid
in
this
situation
was
not
apparent
although
our
colleague,
Clifford
Russell,
believes
this
to
be
an
example
of
strategic
bias.
4­
5
Our
instrument
was
designed
to
minimize
possible
incentives
to
engage
in
strategic
behavior.
No
policy
outcome
was
mentioned
in
the
instrument
nor
were
respondents
told
how
their
WTP
amounts
would
be
used.
Even
if
respondents
inferred
that
the
study's
findings
are
intended
for
government
guidance
in
some
way,
most
would
be
aware
of
the
indirect
connection
between
such
a
study
and
the
actual
process
by
which
tax
rates
and
prices
are
determined.
_
On
a
priori
grounds,
therefore,
we
would
not
expect
strategic
bias
to
affect
our
results.

(
continue)
4­
6
Distribution
Tests
for
Strategic
Bias
Apart
from
specific
experimental
tests,
two
possible
indicators
of
strategic
bias,
neither
of
them
formalized,
have
been
suggested,
A
distribution
test
was
first
proposed
by
Brookshire,
Ives
and
Schulze
(
1976).

They
hypothesized
that
the
distribution
of
the
WTP
amounts
(
in
their
case,

bids)
will
be
normal
when
strategic
bias
is
absent.
If
it
is
present,
they
predict
a
"
flattened"
distribution.
They
examined
the
distribution
of
responses
for
their
study,
which
involved
the
aesthetic
benefits
of
foregoing
the
siting
of
a
power
plant
near
Lake
Powell,
and
concluded
on
the
basis
of
observation
that
since
the
distribution
was
"
not
flat,"

strategic
behavior
was
unlikely.

This
distribution
test
has
several
weaknesses.

1.
Even
if
we
accept
the
notion
that
non­
strategically
biased
distributions
should
be
normal
it
is
impossible
for
most
WTP
distributions
to
pass
the
standard
statistical
tests
for
normality
such
as
the
Komogorov­
Smirnov
test.
These
tests
assume
that
each
data
point
has
an
equal
probability
of
being
chosen,
but
since
respondents
tend
to
choose
favorite
numbers
(
e.
g.,
5,
10,
20,
25
rather
than
6,
11,
22,
etc.),
the
resulting
distribution
is
always
too
lumpy
to
pass
the
test
even
though
the
distribution
may
appear
to
approximate
a
normal
distribution.

2Clifford
Russell
has
recently
called
our
intention
to
a
grouped
data
normality
test
(
Burlington
and
May,
1958:
180­
181)
which
may
be
an
appropriate
normality
test
for
these
kinds
of
data.
4­
7
2.
The
expectation
that
strategic
behavior
will
flatten
an
otherwise
normal
(
or
approximately
normal)
distribution
is
well
founded,
but
only
if
the
distribution
of
those
who
value
the
public
good
in
question
is
normally
distributed.
In
certain
situations
there
is
reason
to
doubt
that
non­
biased
WTP
amount
distributions
will
be
normal.
Imagine
a
population,
most
of
whom
are
either
environmental
enthusiasts
or
enthusiasts
for
industrial
growth
at
the
lowest
possible
cost.
If
they
all
act
strategically,

flat
distribution
with
lating
at
the
high
end
end.
we
will
get
a
bi­
modal
rather
than
a
the
environmentalists'
amounts
accumuand
the
industrial
enthusiasts'
at
the
other
3.
Since
income
is
the
primary
deterrent
of
willingness
to
pay
and
since
the
distribution
of
income
more
clearly
approximates
3
a
log
normal
curve
than
the
normal
curve.
In
the
absence
of
strategic
bias,
the
distribution
one
would
expect
in
this
situation
would
be
closer
to
a
log­
normal
than
a
normal
distribution.

Figure
4.1
gives
the
distribution
of
the
WTP
amounts
for
fishable
(
level
C)

water
for
questionnaire
versions
A,
B,
and
C
combined.
4
the
distribution
is
3
According
to
O'Brien
(
1979:
855)
the
log­
normal
distribution
is
somewhat
more
skewed
than
the
distribution
of
income
in
the
United
States.

4
Unless
otherwise
specified,
we
will
normally
combine
the
results
for
three
versions,
for
reasons
to
be
explained
below.
Whenever
we
report
the
results
for
one
level,
we
will
use
C,
fishable
water.
Unless
otherwise
specified,
the
results
for
the
other
levels
(
boatable,
swimmable)
parallel
those
for
fishable.
4­
8
Figure
4.1
Frequency
300
270
240
210
180
150
120
80
30
DISTRIBUTION
OF
WTP
AMOUNTS
FOR
FISHABLE
WATER
FOR
VERSIONS
A,
B,
C
COMBINED
INCLUDING
ZERO
AMOUNTS
$
0­
60
61­
180
181­
300
301­
420
421­
540
541­
660
661­
750
781­
900
901­
999
4­
9
dominated
by
the
WTP
amounts
in
the
lowest
category,
$
0­
60.
Of
these,

more
than
half
are
zero
bids.
The
high
occurrence
of
zero
bids
is
one
of
the
two
major
problems
with
our
method
revealed
by
our
experiment
(
the
other
being
the
relatively
high
percent
of
people
who
failed
to
give
any
WTP
amount).
It
is
a
problem
because
it
seems
likely
that
most
of
those
who
gave
zero
bids
actually
have
a
greater
than
zero
value
for
water
quality
and
would
be
willing
to
pay
some
amount,
however
small,
for
water
pollution
control
if
we
had
an
improved
way
of
eliciting
their
true
preferences.

By
probing
zero
responses,
other
studies
have
found
that
some
of
those
who
give
zero
WTP
amounts
do
so
to
protest
some
aspect
of
the
interview
situation.
This
is
undoubtedly
the
case
in
our
situation,
but
we
were
(
continue
)
4­
10
unable,
for
the
reasons
discussed
in
Chapter
3,
to
probe
our
zero
bidders
to
learn
the
reasoning
behind
their
amounts.
(
We
discuss
the
problem
of
zero
bidders
in
detail
later
in
this
chapter
under
item
non­
response
bias.)
Since
we
are
unable
to
separate
the
"
real"
zero
payers
from
the
protest
zero
payers,
our
subsequent
analysi
includes
all
those
who
gave
zero
amounts.
By
doing
this
we
bias
our
findings
downwa
by
some
indeterminate
factor.
However,
for
the
sole
purpose
of
examining
the
distribution
of
the
WTP
amounts,
we
recalculated
the
distribution
leaving
out
all
the
zero
amounts.
The
revised
distribution
is
given
in
Figure
4.2.

1.
At
the
upper
end
the
distribution
falls
off
until
the
highest
category
where
it
increases.
This
is
caused
in
large
part
by
the
arbitrary
$
999
upper
limit
to
our
WTP
amounts.
Since
most
of
those
who
gave
this
amount
are
in
our
highest
income
category,
we
believe
that
if
the
$
999
constraint
had
not
been
introduced
at
the
keypunching
stage,
the
distribution
would
have
tailed
off
gradually.

2.
The
overall
shape
of
the
distribution
is
not
flat.
It
approximates
a
log
normal
distribution,
a
distribution
similar
to
that
reported
by
Brookshire,
et
al.
(
1976)
in
their
Lake
Powell
study,
and
to
the
distribution
of
income
in
the
United
States.
Since
income
is
a
strong
predictor
of
people's
willingness
to
pay
for
water
quality,
as
we
will
see
in
Chapter
5,

we
conclude
that
the
distribution
does
not
suggest
strategic
bias.
4­
11
Figure
4.2
Frequency
160
140
120
100
80
60
40
20
DISTRIBUTION
OF
WTP
AMOUNTS
FOR
FISHABLE
WATER
FOR
VERSIONS
A,
B,
C
COMBINED
EXCLUDING
ZERO
AMOUNTS
$
1­
60
61­
180
181­
300
301­
420
421­
540
541­
660
661­
780
781­
900
901­
999
4­
12
A
second
method
of
testing
the
hypothesis
that
the
distribtuion
of
WTP
amounts
will
be
"
flatter"
than
normal
when
strategic
bias
is
present
is
implied
by
Brookshire,
et
al.
(
1976)
in
their
Lake
Powell
study
when
they
make
the
following
statements:
.
.
.
false
bids
will
be
very
large
relative
to
the
mean
for
environmentalists
and
zero
for
non­
environmentalists
where
bids
are
constrained
to
be
non­
negative
(
1976:
328).

.
.
.
if
strategic
behavior
had
been
prevalent
one
would
expect
a
significant
number
of
high
bids
relative
to
the
mean
bid
(
1976:
340).

This
test
also
has
its
problems.
First,
and
most
important,
we
have
no
objective
way
of
identifying
"
false"
values
since
the
essence
of
the
problem
of
preference
revelation
is
that
"
true
value
is
subjective
and
typically
cannot
be
observed
independently"
(
Freeman,
19796:
97).
Second,

the
simple
fact
that
environmentalists
are
willing
to
pay
more
than
other
people
for
environmental
goods
(
and
non­
environmentalists
less)
does
not
necessarily
imply
strategic
behavior
on
their
part,
especially
when
the
environmental
good
being
valued
is
a
broad
one
like
the
nation's
water
quality.
If
environmentalists
are
true
to
their
professed
ideals,
we
would
expect
them
to
be
willing
to
pay
more
for
water
quality
than
those
of
comparable
income
who
are
less
committed
to
environmentalist
ideals.

Bearing
these
problems
in
mind,
the
best
we
can
do
is
to
arbitrarily
define
certain
WTP
amounts
as
inappropriately
"
high"
or
"
low,"
relative
to
the
respondents'
income
level,
and
see
if
a)
the
percentage
of
people
who
give
bids
of
this
kind
is
large
enough
to
be
troublesome
and
4­
13
b)
if
environmentalists
and
anti­
environmentalists
are
disproportionately
represented
among
those
who
give
such
bids
in
such
a
way
that
the
results
will
be
biased
one
way
or
the
other.

Table
4.
2
divides
those
who
gave
us
amounts
for
fishable
water
into
four
groups:

1.
Those
who
gave
zero.

2.
Those
who
gave
"
low"
amounts
which
we
define
as
any
amount
above
zero
but
equal
to
or
lower
than
half
the
amount
shown
on
the
respondent's
payment
card
as
the
amount
contributed
to
the
space
program.
For
those
in
the
lowest
income
group
this
is
1­
6
dollars;

for
those
in
the
highest
this
is
1­
53
dollars.

3.
Those
who
gave
"
high"
amounts
which
we
arbitrarily
define
as
any
amount
equal
to
or
greater
than
the
amount
shown
for
public
education
on
their
card.
This
amount
was
$
204
for
the
low
income
group
and
S1695
for
the
high
income
group.

4.
Those
who
gave
an
amount
between
the
low
and
high
extremes,
who
we
label
"
normal."

Eighty­
three
percent
of
those
who
gave
amounts
greater
than
Zero5
fall
into
our
"
normal"
category.
Those
in
the
extreme
categories
are
divided,
with
10
percent
giving
"
high"
amounts
and
7
percent
willing
to
pay
low
amounts.
We
conclude
that
those
at
the
extremes
are
relatively
few
in
number
and
rather
evenly
balanced.

The
table
also
shows
some
of
the
characteristics
of
the
people
in
each
of
these
groups.
Comparing
those
in
the
low
category
with
the
normals,
the
lows
have
a
larger
percentage
of
people
in
the
highest
income
category
5
Coding
did
not
distinguish
between
zero
and
one
dollar
responses,
which
were
both
coded
as
zero
(
or,
in
log
responses,
as
one).
4­
14
T
a
b
l
e
4
.
2
PERCENT
OF
THOSE
GIVING
VARIOUS
LEVELS
OF
PAYMENT
WHO
BELONG
TO
CERTAIN
DEMOGRAPHIC
AND
ATTITUDINAL
CATEGORIES
Amount
Willing
to
Pay
for
Fishable
Water
(
level
C)'

$
0
"
LOW"
"
Normal"
"
High"
Cave
No
Amount
Maximum
N
=
2
A
B
C
D
E
F
High
Tucome3
Low
Education:
High
School
and
Below
Age
65
and
Older
High
on
Environmental
Scale
(
2­
4)

Very
Concerned
About
Water
Pollution
Use
Water
for
Recreation
(
183)
4
13%
(
20)

78
(
143)

25
(
46)

6
(
10)

30
(
42)

34
(
62)
(
40)
(
447)

40%
(
16)
23%(
101)

65
(
26)
68
(
275)

13
(
5)
8
(
38)

30
(
11)
30
(
144)

43
(
40)
41
(
196)

62
(
25)
71
(
334)
(
52)
(
445)

48%
(
25)
16%
(
57)

43
(
22)
73
(
328)

0
(
0)
20
(
92)

62
(
35)
20
(
88)

65
(
34)
38
(
168)

83
(
43)
49
(
220)

1
"
Low"
amounts
are
defined
as
any
amount
equal
to
or
lower
than
half
the
amount
people
of
the
respondents'
income
category
were
said
to
spend
on
space.
"
High"
are
amounts
equal
to
or
greater
than
the
education
amount
given
on
the
payment
card.
"
Normal"
are
all
amounts
in
between
the
low
and
high
amounts.

2
Total
N
varies
for
each
of
the
demographic
and
attitudinal
categories.

3
Definitions
of
variables
arc
as
follows:
high
income
=
25t
+
/
low
education
=
high
school
or
below/
high
on
environmental
scale
=
score
of
2­
5
on
a
scale
constructed
from
seven
questions
which
varies
From
­
5
to
+
5
;
See
Appendix
for
a
Full
description
of
the
scale
/
water
user
=
someone
who
has
fished,
boated
or
swam
in
last
two
years.
4
Note
that
these
percents
are
each
independent
of
the
rows
and
colums.
Here,
13
percent
of
those
who
are
willing
to
pay
$
0
have
a
"
high"
income.
4­
15
($
25,000
and
above),
and
a
lower
percentage
of
users
of
freshwater
for
recreation.
Overall,
they
are
as
environmentally
concerned
as
the
normals
but
are
older,
wealthier
and
somewhat
less
likely
to
use
water
for
recreation.
This
combination
of
characteristics
does
not
suggest
upward­
biased
strategic
behavior,
although
it
is
not
inconsistent
with
free
riding.

The
highs
are
also
higher
in
income
than
the
normals.
They
are
much
more
likely
to
be
high
on
our
environmental
scale
­­
and
in
their
concern
about
water
pollution
as
a
problem
­­
and
somewhat
higher
in
recreational
water
use
(
See
Chapter
5
for
a
description
of
these
measures).
Although
we
would
expect
those
who
use
and
value
water
to
place
a
higher
value
on
it
through
their
willingness
to
pay,
and
while
half
of
the
highs
are
in
the
highest
income
category
and
presumably
can
afford
the
amounts
they
said
they
are
willing
to
pay,
these
data
are
consistent
with
the
idea
that
some
of
these
52
people
are
overestimating
their
real
willingness
to
pay.
Whether
this
is
the
result
of
deliberate
calculation
(
strategic
bias)
or
unrealistic
enthusiasm
(
hypothetical
bias)
cannot
be
determined.

We
do
know
they
are
more
than
balanced
by
the
183
zero
bidders.
4­
16
Hypothetic
Bias
Hypothetic
bias
is
the
"
potential
error
induced
by
not
confronting
the
individual
with
the
actual
situation"
(
Schulze,
et
al.,
1980).
In
a
situation
influenced
by
hypothetic
bias
people
are
so
far
removed
from
the
actual
situation
that
they
do
not
have
"
genuine"
opinions.
Perhaps
they
are
being
asked
about
something
which
is
so
far
removed
from
their
experience
and
interests
that
they
are
indifferent
to
the
public
good.
Alternatively,
they
may
have
sufficient
interest
or
potential
interest
in
the
topic
but
the
subject
of
inquiry
is
not
specified
in
sufficient
relevant
detail
in
the
instrument
for
them
to
have
anything
but
superficial
opinions.
This
is
why
social
surveys
sometimes
find
opinions
about
controversial
topics
shift
dramatically
according
to
the
way
contingencies
associated
with
the
issue
are
spelled
out
or
specified.
For
example,
attitudes
towards
nuclear
power
can
be
made
to
shift
by
40
percentage
points
by
varying
the
degree
of
assurance
about
nuclear
safety
in
the
working
of
the
question
(
Mitchell.
1980:
12).

Hypothetic
bias
may
produce
a
variety
of
effects.
One
is
greater
uncertainty
and
ambivalence
on
the
part
of
the
repsondent
compared
with
his
or
her
response
to
a
"
more
realistic"
situation.
The
empirical
consequence
of
this
is
increased
variability
in
responses
and/
or
a
larger
than
normal
number
of
refusals
and
don't
knows.
This
uncertainty
and
ambivalence
means
that
a
respondent's
WTP
amounts
are
much
more
susceptible
to
the
pressures
of
social
desirability.

In
many
cases
(
especially
those
involving
substantial
amounts)
the
direction
of
social
desirability
will
be
ambiguous
or
nonexistent.
Below
we
explore
the
direction
of
hypothetic
bias
for
this
case.
4­
16a
The
other
primary
effect
is
the
rejection
of
some
aspect
of
the
hypothetical
market
in
WTP
surveys,
The
payment
vehicle
is
usually
the
cause
of
this
rejection
which
takes
the
form
of
refusals
or
protest
zero
amounts.
This
effect
is
more
properly
a
separate
component
of
the
larger
context
correspondence
problem
we
discuss
later.
Since
this
response
is
not
due
to
availability
to
visualize
the
market.

Since
WTP
studies
are
by
definition
hypothetical,
the
avoidence
of
hypothetic
bias
requires
ingenuity
on
the
part
of
the
researcher.
It
is
the
burden
of
our
argument
in
this
section
that
hypothetical
or
contingent
markets
can
be
described
in
such
a
way
as
to
minimize
hypothetic
bias.
We
first
discuss
two
preliminary
topics
which
have
not
been
much
discussed
in
the
literature:
the
direction
of
hypothetic
bias
and
the
relationship
between
strategic
and
hypothetic
bias.
We
then
treat
the
question
of
whether
and
under
what
circumstances
survey
research
can
realistically
simulate
markets
for
public
goods,
In
the
final
part
of
this
section
we
consider
the
extent
to
which
our
instrument
suffers
from
context
correspondence
problems.
4­
17
The
Direction
of
the
Bias
The
WTP
literature
habitually
refers
to
hypothetic
"
bias,"
but
does
not
show
what
bias
or
systematic
distortion
of
the
WTP
amounts
is
to
be
expected
from
unrealistic
research
instruments.
Where
people
lack
"
genuine"

opinions
about
a
particular
issue
we
would
expect
their
responses
to
be
more
random
than
would
be
the
case
for
an
issue
on
which
they
held
genuine
opinions.
In
the
former,
more
people
will
"
guess"
rather
than
"
estimate."

Such
guesses
are
vulnerable
to
extraneous
matters
such
as
fatigue,
personal
attraction
to
the
interviewer,
exposure
to
the
evening's
news
on
television,

etc.
For
this
reason,
WTP
amounts
affected
by
hypothetic
bias
will
show
greater
statistical
variance
and
less
reliability
than
those
not
so
affected.
Combined
with
the
constrained
nature
of
WTP
distributions,
this
greater
variance
will
bias
the
WTP
amounts
upwards.

Let
us
consider
this
argument
in
greater
detail.
Given
an
initial
(
in
our
case
the
true)
probability
distribution
with
a
known
mean
and
variance,
increasing
the
variance
of
that
distribution
may
necessarily
result
in
an
increase
in
the
mean
(
or
expected)
value
of
that
probability
function.
This
increase
in
E(
x)
can
be
shown
to
hold
for
many
common
probability
distributions
(
the
common
characteristics
of
which
appear
to
be
a
constraint
on
the
ranges
of
values
which
the
function
can
take).
This
constraint
may
be
definitional
or
artificially
imposed;
in
our
case
this
constraint
is
the
impossibility
of
negative
values.
5a
Two
probability
5a
It
should
be
noted
that
protest
zeros
must
be
removed
before
the
distributional
phenomenon
described
here
can
be
observed.
4­
18
distributions
have
been
proposed
for
WTP
distributions
of
our
type:

lognormal
(
Gramlich,
1977)
and
normal
(
Brookshire,
et
al.,
1976).
6
The
log­
normal
distribution
can
be
defined
for
x
as
x
=
exp(
y)
where
y
=
N(
UJ2).
The
expected
value
of
x
is
E(
x)
=
exp(
U
+
(
l/
2)
32)
and
the
variance
of
x
is
VAR(
x)
=
exp(
29
+
02)
(
e
02
­
1).
It
can
be
straightforwardly
observed
that
an
increase
in
VAR(
x)
causes
an
increase
in
E(
x).

The
normal
distribution
is
the
other
distribution
which
has
been
suggested
as
the
appropriate
distribution
for
WTP
amounts.
Because
the
mean
and
variance
are
independent
from
each
other
in
the
normal
distribution,
increasing
the
variance
of
the
probability
distribution
does
not
change
the
mean.
However
in
the
case
of
WTP
distributions
we
are
not
dealing
with
a
true
normal
distribution,
but
a
normal
distribution
6a
which
is
artificially
constrained
to
be
non­
negative.
We
shall
call
this
distribution
a
constrained
normal.
Through
a
series
of
heuristic
graphs
we
will
show
why
the
mean
WTP
value
increases
for
this
distribution
when
the
variance
of
the
initial
probability
distribution
is
increased.

6
The
increase
in
the
E(
x)
for
an
increase
in
the
variance
of
the
original
chi
square
or
F
distribution
follows
directly
from
the
interdependence
of
the
mean
and
variance
of
a
chi
square
or
F
variable.
See
Hogg
&
Craig
(
1978)
or
Freund
and
Walpole
(
1980)
for
a
detailed
discussion
6a
In
theory,
nothing
prevents
a
legitimate
negative
bid.
Two
examples
of
rational
negative
bids
would
be
a
person
who
feared
clean
water
would
bring
hordes
of
tourists
to
his
or
her
doorstep
or
the
person
who
disliked
environmentalists
so
much
that
the
pleasure
which
clean
water
brought
environmentalists
caused
him
displeasure.
In
practice,
however,
no
governmental
authority
would
pay
a
citizen
in
order
to
provide
him
with
clean
water.
We
believe
that
the
number
of
consumers
whose
true
value
for
water
quality
is
negative
is
sufficiently
small
so
that
we
may
consider
the
constraint
of
non­
negative
values
to
be
inoperable.
This
is
not
necessarily
true
where
the
nature
of
hypothetical
markets
encourages
a
large
increase
in
G2
relative
to
the
true
distribution.
4­
19
First
consider
the
following
graph
of
a
true
probability
distribution:

Figure
A
In
Figure
B
below,
we
increase
the
variance
of
the
original
distribution.

The
mean
of
the
new
distribution
is
the
same
as
the
original
and
is
indicated
as
E(
x).
The
area
shaded
in
to
the
left
of
zero
is
the
area
which
will
be
truncated
if
the
constraint
is
operable.

Figure
B
4­
20
Now
suppose
that
the
distribution
is
constrained
at
zero
so
that
if
x
<
0
then
x
=
0.
The
truncated
area
of
Figure
2
is
rotated
upward
to
the
right
side
of
the
zero
axis
and
the
resulting
distribution
is
shown
in
Figure
C.
In
this
Figure
E(
x)
is
the
expected
value
of
the
original
distribution
and
E(
x')
is
the
expected
value
of
the
constrained
normal
distribution.
In
terms
of
the
definition
of
the
sample
mean
of
a
normal
variable?
=
(
Cxi/
n>
some
of
the
x
'
s
are
greater
than
they
would
have
i
been
in
the
unconstrained
distribution
causing
x1
>?
i.'

Figure
C
7
In
a
more
severe
case
than
our
constrained
normal
distribution
­­
that
of
a
truncated
normal
distribution
where
the
truncated
observations
are
discarded
­­
Cohen
(
1950,
1967)
has
shown
that
the
sample
mean
of
the
resulting
distribution
is
dependent
upon
the
variance.
As
an
example,
if
a
normal
distribution
with
mean
zero
and
variance
3
­
is
truncated
at
zero
and
all
negative
observations
are
discarded
the
resulting
sample
mean
is
which
must
be
greater
than
zero
unless
J
=
0.
4­
21
The
Relationship
Between
Strategic
and
Hypothetical
Bias
A
second
important
aspect
of
hypothetical
bias
which
is
unresolved
in
the
literature
is
the
nature
of
its
relationship
with
strategic
bias.

When
statements
are
made
that:
"
The
hypothetical
nature
of
such
(
WTP)
surveys
may
then,
in
actuality,
aid
in
eliciting
bids
which
are
not
strategically
biased"
(
Schulze,
et
al.,
1980:
11)
the
implication
is
that
hypothetical
bias
is
the
opposite
of
strategic
bias.
According
to
this
logic,
strategic
bias
occurs
because
people
believe
the
situation
is
"
real"
and
cover
up
their
"
genuine"
opinions
to
suit
their
perceived
interests
whereas
it
is
the
unreality
of
the
situation
which
promotes
hypothetical
bias.
We
believe
it
is
more
correct
to
distinguish
strategic
from
hypothetical
bias
in
terms
of
the
types
of
realism
involved,
however.
Strategic
bias
is
promoted
when
the
consequences
of
the
WTP
questions
are
perceived
by
the
respondent
as
real.
Hypothetical
bias,
in
contrast,
is
induced
when
the
market
described
to
the
respondent
is
not
realistic
enough.
These
two
factors
may
vary
independently
as
shown
in
Table
4.3.
Respondents
may
perceive
that
they
either
will
have
to
pay
the
amount
they
state
for
(
continue)
4­
22
Table
4.3
TYPES
OF
REALISM
AND
STRATEGIC
AND
HYPOTHETIC
BIAS
Perceived
Consequence
for
Respondent
4­
23
the
public
good
or
that
their
responses
will
directly
influence
public
policy.
On
the
table
this
is
described
as
a
direct
consequence
and
promotes
strategic
bias.
Alternatively
this
consequence
may
not
seem
likely
to
them,
a
perception
which
appears
to
be
the
general
rule
among
respondents
in
WTP
studies
including
this
one.
Turning
to
the
other
dimension,
hypothetic
bias
is
minimized
when
the
hypothetical
market
is
credible
or
plausible
to
respondents
in
that
it
accords
sufficiently
with
their
understanding
of
how
the
world
works
and
imposes
realistic
(
albeit
hypothetical)
constraints
on
preferences
(
by
introducing
cost,

for
example).
It
is
the
absence
of
this
market
realism
which
promotes
hypothetical
bias.
Both
biases
are
minimized,
therefore,
when
consequence
realism
is
low
and
market
realism
is
high
(
cell
2
in
the
Table
4.3).

Schulze,
et
al.,
in
a
discussion
of
hypothetic
bias
argue
that
both
consequence
and
market
realism
are
necessary
for
WTP
surveys
(
cell
1):

"
The
contingent
valuation
approach
requires
postulating
a
change
in
environmental
attributes
such
that
it
is
believable
to
the
individual
and
accurately
depicts
a
potential
change.
The
change
must
be
fully
understandable
to
him,
i.
e.,
he
must
be
able
to
understand
most,
if
not
all,
of
its
ramifications.
The
individual
also
must
believe
that
the
change
might
occur
and
that
his
contingent
valuation
or
behavioral
changes
will
affect
both
the
possibility
and
magnitude
of
change
in
the
environmental
attribute
or
quality.
If
these
conditions
are
not
fulfilled,
the
hypothetical
nature
of
contingent
valuation
approaches
will
make
their
application
utterly
useless."
(
Schulze,
et
al,,
1980:
14).
4­
24
We
agree
with
the
first
part
of
their
statement,
but
not
the
second
part.

We
do
not
believe,
as
they
apparently
do,
that
consequence
realism
is
necessary
for
a
credible
survey.
Certainly
none
of
the
WTP
surveys
reported
in
the
literature
on
air
and
water
pollution
have
achieved
it,
a
judgment
in
which
Schulze
and
his
colleagues
concur;
and
if
they
had,
strategic
bias
would
become
a
genuine
problem
for
WTP
surveys.
In
what
follows
we
argue
that
properly
designed
surveys
can
describe
situations
with
sufficient
realism
to
elicit
meaningful
responses
and
discuss
the
adequacy
of
our
questionnaire
in
this
regard.
We
then
propose
theoretically
based
regression
estimations
as
an
appropriate
test
for
hypothetical
bias.

Survey
Research
and
Market
Simulation
According
to
Randall,
et
al.
(
1974:
135)
the
validity
of
WTP
surveys
"
depends
on
the
reliability
with
which
stated
hypothetical
behavior
is
converted
to
action,
should
the
hypothetical
situation
posted
in
the
game
arise
in
actuality."
The
challenge
is
to
create
a
believable
and
meaningful
set
of
questions
which
will
simulate
a
market
for
the
public
good
in
question,

Some
would
argue
that
this
is
an
impossible
task,
that
survey
research
is
too
removed
from
reality
to
be
able
to
predict
behavior.
This
view
seems
to
lie
behind
the
remarks
of
Gary
Fromm
that
"
It
is
well
known
that
surveys
that
ask
hypothetical
questions
rarely
enjoy
accurate
responses"

(
Fromm,
:
172).
4­
25
In
fact,
as
Howard
Schuman
and
Michael
Johnson
(
1976)
show
in
their
major
literature
review
of
the
relationship
between
attitudes
and
behavior,

most
studies
which
measure
people's
attitudes
and
their
subsequent
behavior
show
positive
results.
At
the
individual
level,
for
example,
those
Army
trainees
who
say
they
are
eager
for
combat
are
significantly
more
likely
to
perform
well
in
combat
several
months
later
(
Stouffer,
et
al.,
1949)
and
persons
who
say
they
support
open
housing
are
far
more
likely
(
70%)
to
sign
an
open
housing
petition
three
months
later
than
those
who
expressed
opposition
to
open
housing
(
22%)
(
Brannon,
et
al.,
1973).
One
study
of
four
elections
showed
behavioral
intention
predicted
correctly
to
actual
vote
for
83
percent
of
the
respondents
who
voted
(
Kelley
and
Mirer,
1974).

Schuman
and
Johnson
cite
numerous
other
examples
of
attitude
behavior
correlations
and
conclude
that
the
attitude­
subsequent
behavior
correlations
which
occur
"
are
large
enough
to
indicate
that
important
causal
forces
are
involved"
(
Schuman
and
Johnson,
1976:
199)
although
the
variance
explained
by
attitudinal
intention
is
usually
fairly
modest.

The
most
impressive
demonstrations
of
attitude­
behavior
correlations
occur
at
the
aggregate
level.
Modern
election
polls
predict
election
results
with
great
accuracy.
The
1980
presidential
election
was
no
exception
to
this
generalization
because
the
polls
which
took
place
immediately
before
the
vote
caught
the
last
minute
shift
which
brought
President
Reagan
to
power
(
Ladd
and
Ferree,
1981).
For
many
years
the
Institute
for
Social
Research
at
the
University
of
Michigan
has
used
4­
26
survey
research
to
measure
consumer
sentiments
and
probe
the
psychology
of
economic
behavior.
Their
Index
of
Consumer
Sentiment
represents
a
macro
measure
reflect&
g
the
changes
in
attitudes
and
expectations
of
all
Americans.
For
the
past
25
years
it
has
declined
substantially
prior
to
the
onset
of
every
recession
and
it
advanced
prior
to
the
beginnings
of
periods
of
economic
recovery
(
Katona
with
Morgan,
1980).
These
correlations
occur
despite
the
fact
that
the
University
of
Michigan
economists
are
unable
to
predict
an
individual's
spending
or
saving
on
the
basis
of
changes
in
his
or
her
attitudes
and
expectations.
They
attribute
this
paradox
to
fact
that
individual
consumer
behavior
is
influenced
by
a
large
number
of
factors
including
situational,
attitudinal,
and
physical
(
fatigue)
which
make
accurate
predictions
of
individual
behavior
difficult
to
make.
The
volatility
of
individual
behavior
is
smoothed
out
for
aggregations
of
people;
mood,

individual
differences
in
how
people
react
to
the
particular
stage
in
the
business
cycle,
individual
reactions
to
whether
or
not
they
have
recently
purchased
large
consumer
durables
and
the
like
are
averaged
across
the
sample
(
Katona
with
Morgan,
1980:
60).
This
is
a
strong
argument
for
the
validity
of
surveys
(
provided
the
questions
are
well
worded
and
the
sampling
is
adequate)
as
measures
of
aggregate
benefits.
4­
27
We
conclude
that
properly
designed
survey
questions
do
have
the
potential
to
ap
proximate
real
situations
sufficiently
to
elicit
"
responsible"
responses
which
can
be
predictive
of
behavior
under
the
defined
circumstances
contained
in
the
questions
(
Brookshire,
et
al.,
1979:
30­
31).
Schuman
and
Johnson
analyze
the
design
factors
which
improve
behavioral
predictions,

One
of
the
most
important
is
the
degree
of
congruence
between
the
expressed
attitude
and
behavior.
Heberlein
and
Black
(
1976),
for
example,
found
(
continue)
4­
28
attitude­
behavior
correlations
increased
from
.12
to
.59
for
the
use
of
leadfree
gasoline
when
the
predictive
attitudes
shifted
from
general
interest
in
environmental
issues
to
a
question
about
the
degree
of
personal
obligation
the
respondent
felt
to
buy
lead­
free
gasoline.
In
a
similar
vein,
Brookshire,
d'Arge
and
Schulze
cite
the
psychologists'
Ajzen
and
Fishbein's
well
known
dictum
that
behavioral
intention
and
the
actual
behavior
"
should
correspond,
in
terms
of
the
action,
its
context,
its
target
and
its
time
frame"
(
Brookshire,
et
al.,
1979:
25).

A
second
important
design
factor
is
the
degree
of
information
presented
about
the
consequences
of
an
attitude,
particularly
its
financial
implications.

The
more
fully
these
consequences
are
specified,
the
more
realistic
the
response.
In
the
1960s
Gallup
consistently
found
a
majority
of
people
favored
foreign
aid
when
they
were
asked:
"
In
general,
how
do
you
feel
about
foreign
aid
­­
are
you
for
it,
or
against
it?"
In
a
national
survey
during
the
same
time
period,
Lloyd
Free
and
Hadley
Cantril
introduced
the
pocketbook
aspect
of
the
issue
in
a
question
which
asked
whether
"
government
spending
for
this
purpose
(
foreign
aid)
should
be
kept
at
least
at
the
present
level,
or
reduced
or
ended
altogether?"
When
costs
were
raised
in
this
manner
the
majority
position
shifted
from
favoring
foreign
aid
to
wanting
it
reduced
or
see
also
Mueller,
1963).
ended
(
Free
and
Cantril,
1967:
72;/
A
similar
shift
occurred
in
a
poll
conducted
in
the
Swedish
city
of
Malmo.
In
this
case
a
sample
was
asked
whether
they
would
like
the
Swedish
government
to
increase
aid
to
less­
developed
nations.

Later,
in
the
same
questionnaire,
the
respondents
were
asked
whether
they
would
like
this
to
take
place
"
even
if
taxes
would
be
raised
in
proportion."

Half
the
supporters
of
increased
aid
vanished
when
the
question
was
phrased
this
way,
leaving
only
20
percent
who
were
willing
to
pay
for
increased
aid
(
Bohm,
1979:
146).
4­
29
The
shifts
in
opinion
evoked
by
the
changes
in
question
wording
are
understandable
because
we
would
expect
higher
demand
for
free
goods
according
to
economic
theory,
The
Swedes
who
favor
foreign
aid
in
the
first
question
consist
of
two
types
of
people:
1)
those
who
favor
it
in
the
abstract
but
who
are
not
willing
to
pay
for
it
when
reminded
of
that
contingency
and
2)
those
who
favor
it
in
the
abstract
and
who
are
also
willing
to
pay
for
it,
The
second
question
induces
those
in
category
2)

above
to
relinquish
their
support
by
introducing
the
contingency
of
cost.

WTP
studies
go
one
step
further,
of
course,
and
ask
respondents
to
specify
the
amount
of
money
they
personally
are
willing
to
pay,
This
and
the
fact
that
many
other
contingencies
are
spelled
out
in
the
questionnaire
makes
them
a
far
more
realistic
measure
of
attitudes
than
ordinary
survey
research
items.
4­
30
Context
Correspondence
As
we
noted
in
Chapter
2,
there
are
special
challenges
in
devising
a
macro
WTP
instrument
which
is
sufficiently
realistic
to
avoid
hypothetical
bias,
We
made
special
efforts,
as
described
in
Chapter
3,
to
present
the
market
for
national
water
quality
in
terms
that
are
understandable
to
the
respondent
and
which
related
as
closely
as
possible
to
the
way
the
respondent
actually
contributes
to
the
provision
of
water
quality.
We
will
not
repeat
that
discussion
here,
but
will
amplify
it
by
discussing
the
degree
to
which
our
instrument
is
threatened
by
context
correspondence
problems,
a
particular
7
form
of
hypothetic
bias.

As
described
by
Brookshire,
et
al.
(
1979,
26ff),
these
problems
occur
"
where
the
initial
rights
and
endowments
as
well
as
the
terminal
rights
and
endowments
are
far
removed
from
the
actual
situation."
The
primary
example
of
the
context
correspondence
problem
is
the
failure
of
questions
using
the
willingness
to
accept
compensation
format
to
elicit
meaningful
answers.

The
notion
of
being
"
bribed"
to
tolerate
pollution
is
so
far
out
of
people's
ordinary
comprehension
that
many
people
apparently
consider
it
immoral
and
refuse
to
value
the
environmental
good
at
anything
less
than
infinity
(
Randall,
et
al.,
1974;
Blank,
et
al.,
1977:
Brookshire,

and
above
in
Chapter
1).
Is
is
possible
that
the
high
percent
of
noplays
and
zero
bidders
we
found
is
an
indicator
that
our
instrument
suffers
from
context
correspondence
problems?

7
Brookshire,
et
al.,
say
a
high
percentage
of
protest
votes
is
an
indicator
of
context
correspondence
problems
(
1979:
28)
4­
31
On
an
a
priori
basis
we
do
not
believe
this
to
be
the
case.
The
initial
endowment
of
boatable
water
nationally
and
the
notion
that
people
are
paying
for
water
quality
of
this
level
in
taxes
and
higher
prices
seems
well
within
people's
understanding,
particularly
since
they
are
already
paying
for
water
quality
in
this
manner
(
although
they
may
not
have
thought
about
it),
Our
instrument
assumes
a
structure
of
rights
in
which
fresh
water
is
a
common
property
resource
which
can
be
used
for
various
purposes,

The
simulated
market
provides
a
situation
in
which
the
individual
buy
improved
water
quality
situations
by
paying
higher
taxes
and
prices.
It
assumes
that
these
cannot
be
provided
free
of
charge.
It
is
possible
that
some
people
may
feel
that
businesses
should
pay
the
costs
of
treating
pollution
out
of
profits
instead
of
passing
the
costs
on
to
consumers,
but
surveys
suggest
that
a
large
majority
of
the
public
are
aware
of
the
fact
9
that
these
costs
do
get
passed
on
to
consumers
(
Cambridge
Reports,
1978:
167).

Finally,
the
improved
situations
we
propose,
fishable
and
swimmable
water,

do
not
appear
to
be
so
far
from
the
initial
position
(
boatable
water
nationally)

to
cause
problems
nor
to
deviate
dramatically
from
the
person's
previous
experience
and
preferences.
Most
people
will
have
had
first
hand
contact
with
freshwater
of
those
quality
levels.

However,
when
we
ask
people
to
put
a
dollar
value
on
water
quality
levels
we
are
asking
them
to
do
something
that
is
not
part
of
their
normal
8
In
the
case
of
going
from
boatable
to
non­
boatable
the
respondents
were
buying
the
continuance
of
the
status
quo.
See
the
more
detailed
discussion
of
property
rights
in
Chapter
1
where
we
specify
the
types
of
consumer
surplus
measures
we
employ
in
this
study.
9
Cambridge
Reports
in
a
report
for
the
Shell
Oil
Company
asked
a
national
sample:
"
When
the
government
imposes
new
health
or
safety
standards
on
an
industry
which
single
group
do
you
think
usually
pays
the
cost
of
implementing
those
standards:
the
industry
out
of
its
profits,
workers
in
the
industry
through
lower
wages,
consumers
through
higher
prices
or
the
government
using
tax
money?
Sixty­
two
percent
said
consumers
through
higher
prices
(
Cambridge
Reports
1978:
167)
and
12
percent
"
the
government
using
tax
money."
Only
7%
4­
32
behavioral
repertoire;
both
the
valuing
and
the
contemplation
of
national
water
quality
are
novel
experiences
for
most
people,
By
way
of
contrast,

those
WTP
studies
which
ask
people
to
place
a
value
on
certain
characteristics
of
a
particular
recreational
site
in
terms
of
an
entrance
fee
ask
people
to
perform
a
much
less
novel
act
since
people
are
familiar
with
entrance
fees
and
regularly
make
decisions
about
whether
or
not
they
are
worth
the
price.
Does
this
mean
that
such
a
study
is
necessarily
more
valid
than
ours?
We
think
not,
because
familiarity
may
present
problems
of
its
own.

When
respondents
are
asked
to
express
WTP
amounts
by
the
entrance
fee
vehicle
(
e.
g.
Thayer,
forthcoming)
the
amount
they
give
may
represent
not
what
they
personally
consider
the
benefit
to
be
worth
but
what
they
consider
to
be
a
"
fair"
entrance
fee
based
on
their
experience
with
entrance
fees,
Thus,

novelty
as
such
need
not
be
an
impediment.
What
matters
most
is
whether
respondents
are
made
sufficiently
familiar
with
the
new
situation
in
the
interview.

Where
context
correspondence
is
present
we
will
expect
two
outcomes.
The
first
is
a
greater
incidence
of
item
nonresponse
for
the
WTP
items.
More
people
will
be
unable
to
find
the
situation
meaningful
enough
to
offer
WTP
amounts
or
in
protest
they
will
bid
$
0.
WTP
surveys
test
for
context
correspondence
by
examining
(
and
reporting)
the
rates
of
these
responses.
As
noted
earlier
we
had
large
numbers
of
people
who
failed
to
give
amounts
or
who
gave
$
0
amounts.
In
our
discussion
of
this
problem
below,
under
item
nonresponse
bias,
we
conclude
that
it
is
probably
caused
by
problems
other
than
context
correspondence.
4­
33
Secondly,
if
the
situation
which
respondents
are
valuing
is
too
removed
from
the
experience
or
interests,
their
answers
to
the
WTP
questions
will
be
more
whimsical
than
purposeful
and
should
vary
randomly.
Conversely,
if
the
task
is
meaningful
to
the
respondent,
his
or
her
answers
will
be
constrained
by
the
factors
which
influence
decisions
about
such
expenditures
in
everyday
life:
income
variability.
and
value.
The
context
correspondence
problem
in
this
instance
is
increased/

An
appropriate
test
for
randomness
of
responses
is
the
size
of
R2
in
a
regression
of
WTP
amount
on
theoretically­
based
constraints
1;
(
in
our
case:
recreational
use
of
freshwater,
concern
about
water
pollution,
income,
etc.).
We
report
the
results
of
our
predictive
test
in
Chapter
5.
Our
findings
in
this
respect
are
very
reassuring.

INSTRUMENT
BIASES
The
willingness
to
pay
literature
has
identified
four
instrument
characteristics
which
are
potential
sources
of
bias,
These
are
the
payment
vehicle,
information,
order
and
starting­
point
biases.
A
number
of
studies
have
varied
these
dimensions
systematically
in
an
effort
to
see
whether
or
not
a
particular
instrument
bias
is
present.
Our
effort
in
this
regard
was
limited
to
the
most
innovative
aspect
of
our
instrument;
the
use
of
the
payment
card
to
elicit
the
respondents
WTP
amount.
The
results
of
this
experiment
are
discussed
in
detail
under
starting
point
bias.
The
instrument
was
designed
to
minimize
the
effect
of
each
of
the
other
potential
biases.

Starting
Point
Bias
In
Chapter
3
we
discuss
why
we
believe
starting
point
bias
is
a
serious
problem
for
bidding
game
studies
which
use
payment
vehicles
other
10
For
an
excellent
example,
see
Brookshire,
et
al.,
1980.
4­
34
than
admission
fees
to
measure
people's
willingness
to
pay
for
public
goods.
We
developed
the
anchored
payment
card
as
a
substitute
for
the
opening
bid
on
the
assumption
that
presentation
of
a
large
menu
of
potential
bids
would
minimize
any
tendency
on
the
respondent's
part
to
acquiescece
to
the
interviewer's
suggested
bid.
There
is
the
possibility
of
course,

that
the
payment
card
itself
might
bias
the
WTP
amounts.
To
examine
this
possibility
we
manipulated
the
two
aspects
of
our
payment
cards
which
seemed
to
present
the
greatest
possibility
of
influencing
respondent
WTP
amounts
and
tested
several
different
versions
of
the
payment
card
on
comparable
sub­
samples.

These
variations
and
the
rationale
behind
them
are
as
follows:

1.
The
payment
card
is
anchored
with
estimates
for
non­
environmental
goods.
We
varied
the
number
of
goods
presented
from
four
in
11
versions
A
and
C
to
five
in
Version
B.
The
extra
good
in
Version
B
was
police
and
fire
protection,
The
amount
which
we
estimated
households
spent
on
this
good
($
98,
$
125,
$
312
and
$
626
for
the
12
four
income
levels
)
was
such
that
it
placed
police
and
fire
protection
on
the
payment
card
at
a
place
where
we
guessed
people
might
value
water
quality.
Except
for
the
addition
of
the
fifth
11
In
this
discussion
we
will
only
consider
versions
A,
B,
C,
of
our
instrument.
Version
D
was
significantly
different
and
our
findings
for
this
version
will
be
described
elsewhere.
See
Chapter
3
for
a
description
of
the
research
instrument
and
Appendix
I
for
the
complete
wording
of
all
the
questions.

12
See
Appendix
III
for
the
procedures
used
to
derive
the
public
good
expenditures
and
Appendix
I
for
all
the
payment
cards
used
in
the
study.
4­
35
public
good,
the
payment
cards
for
Version
B
are
identical
to
those
for
Version
A.
If
the
number
or
placement
of
the
anchors
affects
the
starting
point
we
would
expect
the
mean
WTP
amounts
for
B
to
differ
from
the
amounts
for
the
other
versions.

2.
In
order
to
see
whether
people
keyed
their
water
benefit
amounts
to
the
amounts
shown
on
their
card
for
the
other
public
goods,

Version
C
displayed
the
same
four
public
goods
as
Version
A,

but
each
amount
was
increased
by
25
percent.
If
the
dollar
level
of
the
anchor
or
benchmark
goods
determines
the
WTP
amounts
for
water
quality
we
would
expect
higher
mean
amounts
for
Version
C
than
for
Version
A.

Table
4.4
summarizes
the
sample
design
for
our
tests
of
starting
point
bias.

We
used
t
tests
to
test
for
the
hypotheses:

Test
I
Test
II
Where
A,
B,
C
refers
to
versions
A,
B,
C.
Table
4.4
4­
36
STUDY
DESIGN
FOR
EPA
WATER
POLLUTION
BENEFITS
STUDY
AND
NUMBER
OF
CASES
(
IN
PARENTHESIS)

Versions
Scale
cards
with
the
estimated
levels
of
payment
for
space,
A
highways,
public
education
and
defense
for
each
of
the
four
income
categories.

(
431)*

Scale
cards
with
correct
payment
levels
B
for
the
four
public
goods
used
for
A
plus
police
and
fire
(
380)

Scale
cards
with
same
four
public
goods
used
for
A
but
the
payment
C
levels
listed
are
25%
higher
than
those
used
for
Version
A
(
410)

Same
as
A
plus
the
D
estimated
amount
for
water
pollution
control
(
355)
Family
Income
Levels
I
$
9,999
or
less
(
117)

II
$
10,000
to
14,999
(
58)

III
$
15,000
to
24,999
(
112)
Water
Quality
Levels
Amount
willing
to
pay
for:
D
Okay
for
boating
(
2.5
on
10
step
ladder)

C
Game
fish
like
bass
can
live
in
it
(
5.0)

B
Safe
for
swimming
(
7.0)

IV
$
25,000
and
above
or
not
sure/
refused
(
92)

Same
as
A
Same
as
A
I
(
170)
II
(
66)
III
(
98)
IV
(
62)

Same
as
A
I
(
116)
II
(
58)
III
(
126)
IV
(
74)

Same
as
A
I
(
82)
II
(
78)
III
(
103)
IV
(
70)
Same
as
A
Asked
whether
willing
to
pay
the
specific
amount
for
level
C
If
not
willing
to
pay,
asked
how
much
willing
to
keep
level
at
D
If
willing
to
pay
for
C,
asked
how
much
willing
to
pay
for
B
"
The
total
number
of
cases
for
each
version
exceeds
the
sum
of
the
number
of
cases
ascribed
to
each
income
level
for
that
version
owing
to
the
absence
of
income
data
for
some
respondents.
4­
37
The
results
of
these
tests
for
each
income
by
water
quality
level
category
are
given
in
Table
4.5.
Of
the
24
paired
comparisons
only
two
are
significantly
different
from
zero
(
less
than
the
number
positive
findings
one
would
expect
by
chance
at
the
.05
level)
and
both
are
in
the
opposite
direction
to
that
predicted
if
starting
point
bias
is
present.
We
conclude
that
for
I
and
II,
the
null
hypothesis
is
supported:
there
is
no
evidence
of
starting
point
bias.

A
second
test
of
starting
point
bias
was
conducted
using
regression
analysis.
We
made
dummy
variables
for
each
of
the
three
versions.
We
then
estimated
two
sets
of
equations
for
pairs
of
versions.
The
first
used
one
of
the
dummy
variables
as
the
sole
predictor
variable,
the
second
is
identical
to
the
first
except
that
we
added
the
set
of
predictor
variables
which
are
the
best
predictors
of
the
WTP
amounts.
If
Ho
in
Test
II
is
incorrect,

the
dummy
variables
for
the
versions
should
enter
the
equations
significantly
(
as
measured
by
the
t
values).
Table
4.6
presents
the
results
of
these
estimations.
None
of
the
version
dummy
variables
are
significant,
confirming
our
finding
above
that
our
instrument
does
not
suffer
from
starting
point
bias.

On
the
basis
of
these
findings,
which
not
only
show
no
version
effect
but
also
reveal
an
impressive
stability
across
the
versions
in
the
multivariate
estimations,
we
combine
the
three
versions
into
one
data
set
for
all
further
analysis.
4­
38
Table
4.5
Income
Level
Low
1
AB
2
AB
3
AB
High
4
AB
t
TESTS
OF
YEAXS~
FORPAIRED
COMPARISONS
BETWEEN
VERSIONS
A,
B,
C
BY
INCOME
AND
LEVEL
OF
WATER
QUALITY
Level
of
Water
Quality
Boatable
Fishable
Swimmable
AC
AC
AC
AC
BC
AB
AC
BC
AS
BC
AB
AC*
BC
AB
AC
BC
BC
AB
AC
BC
AB
BC
AB
AC
BC
AB
AC
BC
AC
BC
AC
BC
%'
wo
tailed
test,
variances
between
samples
were
compared
and
then
the
t
test
was
computed
on
pooled
or
separate
variables
as
appropriate.

The
one
tailed
t­
test
was
insignificant
for
every
pair
of
A
and
C
for
test
I
since
the
two
significant
pairs
of
A
and
C
(*
in
the
table)
under
the
two
tailed
t
tests
are
in
the
opposite
direction
from
that
predicted
by
Hl
of
test
I.

*
Difference
between
the
means
is
significantly
different
from
0
at
the
5%
level.
Table
4.6
4­
39
Level
C
VERA
VERB
VERC
INCOMER
Intercept
VERA
VERB
N
R2
F
TEST
FOR
STARTING
POINT
BIAS
Variables
Amount
willing
to
pay
annually
for
fishable
water
in
dollars
Dummy
variable
for
Version
A
Dummy
variable
for
Version
B
Dummy
variable
for
Version
C
Household
income
in
dollars
in
10
categories
EDUC
Education
in
7
categories
AGECAT
Age
in
11
categories
ENVINDEX
Index
of
environmental
attitudes*

USERD
Dummy
variable
for
water
use
CNPOLD
Dummy
variable
for
concern
over
water
pollution
Regressions
on
Level
C
for
Versions
A,
B,
C
as
Noted:

A
&
B
A
&
C
B
&
C
179.44
190.6
190.6
Intercept
(
10.7)
(
10.8)
(
11.5)

32.4
21.4
INCOMER
(
1.4)
(.
9)

11.1
EDUC
(­.
5)

515
.003
1.9
500
.002
.79
t
values
are
given
in
parenthesis
VERA
481
AGECAT
.001
.24
ENVINDEX
USERD
CNPOLD
VERB
N
472
467
451
R2
.30
.29
.34
F
37.9
27.3
32.4
A
&
B
­
30.4
(­
0.60)

.0072
(
8.95)

16.8
(
1.85)

­
10.5
(­
2.88)

26.06
(
3.81)

54.41
(
2.33)

44.47
(
1.95)

21.58
(
1.03)
A
&
C
­
8.2
(­.
15)

.0069
(
8.4)

13.9
(
1.4)

­
8.7
(­
2.3)

29.8
(
4.3)

40.9
(
1.74)

48.3
(
2.1)

12.22
(.
58)
B
&
C
­
21.4
(­.
44)

.0073
(
9.3)

15.1
(
1.78)

­
8.4
(­
2.5)

30.9
(
5.2)

27.46
(
1.3)

64.8
(
3.2)

­
12.7
(­.
67)

*
Composed
of
7
items
ranging
from
attitudes
towards
the
environmental
movement
to
the
importance
of
environmental
problems
in
the
respondents
hierarchy
of
issues.
4­
40
Payment
Vehicle
Bias
In
Chapter
3
we
describe
why
we
chose
annual
household
payment
in
higher
prices
and
taxes
for
our
payment
vehicle.
There
we
argue:
1)
that
our
vehicle
realistically
accords
with
the
actual
form
of
payment
for
water
quality
and
2)
that
it
is
familiar
to
respondents
yet
lacks
the
drawbacks
posed
by
some
familiar
vehicles
such
as
entrance
fees
which
may
limit
WTP
responses
to
an
accustomed
payment
range
rather
than
to
a
true
WTP
amount.
A
further
criteria
for
payment
vehicles
imposed
by
economic
theory
is
that
they
should
offer
respondents
the
widest
possible
latitude
of
potential
substitution
across
current
commodities
(
Schulze,

et
al.,
1980:
12).
We
believe
our
vehicle
combines
believability
with
the
widest
latitude
for
substitution,
two
characteristics
which
often
must
be
traded
off
in
WTP
surveys
(
Brookshire,
et
al.,
1979:
23­
4).
In
the
administration
of
the
survey
we
encountered
no
problems
with
the
vehicle.

If
the
vehicle
suffers
from
any
bias
it
is
likely
to
be
downward
owing
to
the
current
national
concern
over
taxes
and
prices.

Information
Bias
Information
bias
occurs
when
the
wording
of
the
instrument
affects
values
elicited
in
ways
unintended
by
the
researcher.
The
result
is
the
introduction
of
contingencies
other
than
those
contained
in
the
formal
hypothetical
situation.
Because
the
opportunities
for
information
bias
in
questions
are
legion,
the
evaluation
of
a
WTP
study
must
the
include
a
review
of
the
wording
of
the
entire
instrument
and
an
examination
of
the
quesiton.
In
Chapter
3
we
introduce
and
describe
the
questions
we
4­
41
used
in
this
study.
Needless
to
say,
we
attempted
to
word
the
instrument
in
such
a
way
that
by
spelling
out
the
tradeoffs,
the
cost,
the
fact
that
they
are
already
paying
for
public
goods,
etc.
the
respondents
were
presented
with
a
credible
hypothetical
market
for
water
value.
We
endeavored
to
word
the
instrument
in
as
neutral
a
manner
as
possible
so
that
neither
the
costs
nor
the
benefits
of
water
quality
were
emphasized
at
the
expense
of
the
other.
Readers
can
judge
the
success
of
our
efforts
for
themselves
by
consulting
Appendix
I
which
contains
the
entire
instrument
in
the
form
it
was
given
to
the
interviewers.

Order
Bias
Order
bias
is
closely
related
to
information
bias.
Some
information
may
influence
people's
responses
in
an
unwelcome
manner
simply
because
of
its
location
in
the
questionnaire.
The
little
research
that
has
been
done
on
order
effects
suggests
that
this
is
not
an
important
source
of
bias
in
surveys
(
Alwin,
1977:
141),
but
good
survey
practice
dictates
that
sensitive
or
potentially
biasing
items
should
be
located
later
in
a
questionnaire,

otherwise
the
sensitive
items
might
lead
respondents
to
prematurely
terminate
the
interview
and
the
biasing
items
might
affect
the
answers
to
questions
which
are
sensitive
to
that
type
of
bias.
In
WTP
surveys
it
is
important
to
avoid
preceding
the
WTP
items
with
questions
which
emphasize
the
benefits
of
the
good
being
valued
at
the
expense
of
the
cost
or
vice
versa.
Rowe,
et
al.
(
1979:
6)
specifically
cite
the
possible
influences
of
early
environmental
attitude
questions
in
this
regard.
4­
42
The
RFF
water
benefits
was
preceded
by
a
half
hour
(
or
more)
interview
on
environmental
and
energy
issues.
The
questionnaire
for
this
study
is
contained­
in
Appendix
IV.
What
bias,
if
any
might
result
from
the
respondent
being
subjected
to
a
searching
interview
about
environmental
protection,
environmental
values,
risk,
energy
source
preferences,
and
government
action
on
these
matters?
Yore
particularly,
might
these
questions
stimulate
a
greater
value
for
environmental
quality
than
would
otherwise
have
been
the
case
and
bias
the
WTP
amount
upwards?
We
think
this
is
unlikely
for
the
following
reasons:

1.
The
earlier
questions
were
realistic
and
balanced
because
they
measured
environmental
values
in
the
context
of
the
tradeoffs
associated
with
obtaining
better
environmental
quality.
They
a)
forced
people
to
rank
order
environmental
goals
with
other
goals
(
Qs.
1­
10),
b)
elicited
people's
views
about
economic
and
energy
problems
(
Qs.
11a,
b,
f;
21a,
f;
26;
40­
46)
and
c)
used
questions
whenever
possible
which
described
the
tradeoffs
entailed
in
minimizing
risk
or
protecting
the
environment
(
e.
g.,
Qs.
31,
34­
36,
39,
53c).

2.
A
contributing
factor
to
the
realism
of
the
RFF
environmental
survey
is
the
unique
historical
context
of
the
survey.
Most
of
the
interviewing
occurred
in
late
January
and
early
February
1980,
a
time
when
the
Iranian
hostage
crisis
and
the
Russian
invasion
of
Afghanistan
were
dominating
the
news.
These
concerns,

added
to
the
great
concern
expressed
by
our
respondents
about
4­
43
inflation
and
higher
prices,
suggest
the
historical
context
did
not
bias
the
respondents
towards
taking
an
environmentally
oriented
position.
If
anything,
the
opposite
is
likely
to
be
the
case.

3.
It
is
possible
to
compare
the
degree
of
environmental
support
revealed
in
the
RFF
questionnaire
with
the
findings
of
a
commercial
phone
survey
(
Opinion
Research
Corporation,
1980)
which
took
place
two
months
after
most
of
the
RFF
interviewing
and
which
repeated
several
key
questions
word
for
word.
The
commercial
survey
found
even
stronger
support
for
environmental
values
than
did
the
RFF
survey.
This
suggests
that
the
format
of
the
RFF
survey
did
not
bias
people
towards
viewing
the
environment
with
special
favor,
but
rather
it
seems
to
have
led
people
to
evaluate
the
issues
with
greater
realism.

In
our
judgment
the
earlier
environmental/
energy
questions
add
to
the
validity
of
the
WTP
study
by
requiring
the
respondents
to
consider
a
wide
range
of
environmental
issues
and
their
tradeoffs
prior
to
evaluating
the
worth
of
water
quality.
It
is
possible,
however,
that
the
length
of
the
first
portion
of
the
survey
may
have
induced
respondent
and
interviewer
fatigue.
If
we
had
used
the
bidding
game
format
fatigue,
if
present,
might
have
biased
the
WTP
results
upwards
by
tempting
respondents
to
acquiesce
to
the
starting
point
more
often
than
would
otherwise
have
been
the
case.

(
or
downwards
by
making
their
willingness
to
pay
bid
lower).
Since
the
payment
card
technique
minimizes
starting
point
bias,
we
have
no
reason
to
believe
that
fatigue
biased
our
results
upwards
in
this
manner.
On
the
contrary,
fatigue
may
be
a
cause
of
the
large
number
of
zero
amounts
and
no
answers
which
we
experienced.
4­
44
SAMPLING
BIASES
There
is
a
set
of
potential
biases
associated
with
the
methodology
of
survey
research
which
have
received
less
emphasis
in
the
WTP
literature
than
they
should.
An
instrument
may
be
entirely
free
from
general
and
instrument
biases,
but
if
it
suffers
from
serious
sample
and
non­
response
problems
its
findings
cannot
be
generalized
reliably
to
a
larger
population
of
any
kind
and
should
not
be
used
to
estimate
aggregate
benefits.
In
the
past
some
WTP
studies
have
made
such
aggregate
benefit
estimates
on
the
basis
of
seriously
flawed
samples
or,
worse,
without
even
reporting
the
information
necesary
to
assess
whether
method
biases
are
present
or
not.

Sample
Bias
Scientific
sampling
is
a
process
by
which
elements
of
a
population
are
chosen
in
such
a
way
that
information
about
those
elements
can
be
generalized
within
known
error
ranges
to
the
population
from
which
the
elements
are
drawn.
Methods
of
sampling
are
well
grounded
in
statistical
and
probability
theory.
There
are
numerous
sampling
techniques
but
the
distinguishing
characteristics
of
a
properly
designed
sample
are
that
all
the
units
in
the
target
population
have
a
known,
nonzero
chance
of
being
included
in
the
sample,
and
the
sample
design
is
described
in
sufficient
13
detail
to
permit
reasonably
accurate
calculation
of
sampling
errors.

Sampling
bias
occurs
when
samples
are
not
properly
designed
or
reported.

=
For
a
presentation
of
sampling
theory
and
design
for
the
non­
technical
reader
see
Williams
(
1978).
For
a
discussion
of
sampling
for
surveys
see
Babbie
(
1973:
73­
130)
and,
especially
Sudman's
excellent
book,
Applied
Sampling
(
1976).
4­
45
The
sampling
method
used
for
the
RFF
survey
is
a
probability
sample,

the
more
rigorous
of
the
two
sampling
methods
regularly
used
by
commercial
survey
research
firms
(
the
other
being
the
modified
probability
sample).

A
description
of
the
sample,
which
was
designed
by
the
Roper
Organization,

is
presented
in
Appendix
V.
It
ensures
that
all
noninstitutionalized
persons,
18
years
of
age
or
older,
who
live
in
the
lower
48
states
have
a
known
probability
of
being
interviewed.

There
are
many
considerations
which
enter
into
the
decision
about
how
many
people
to
interview
for
a
study,
but
the
basic
tradeoff
is
between
cost
and
accuracy,
Presuming
that
the
respondents
are
selected
according
to
sampling
theory,
the
smaller
the
size
of
a
set
of
respondents
(
which
may
range
from
the
entire
sample
to
a
sub­
sample
of
special
interest
to
the
analyst
such
as
environmental
activists),
the
larger
the
sampling
error.
For
a
simple
random
sample,
the
error
range
at
the
.05
level
of
confidence
is
3
percent
for
1,067
respondents
and
7
percent
for
196
(
Backstrom
and
Hursh,
1963:
33).
For
a
sample
of
50,
the
Opinion
Research
Corporation
estimates
a
14%
sampling
error.
Thus,
if
25
percent
of
a
sample
of
50
say
they
went
boating
at
least
once
in
the
past
two
years,
the
true
value
will
lie
between
11
and
39
percent,
95
percent
of
the
time.
Obviously,

if
these
50
people
were
not
chosen
by
proper
sampling
techniques
the
error
range
is
unknown,
and
it
is
impossible
to
say
anything
about
what
percent
of
any
larger
population
(
such
as
the
people
who
live
in
the
area
where
the
interviewing
took
place)
went
boating
in
the
last
two
years.

14
For
this
reason,
a
true
sample
of
1500
people
allows
Gallup
to
predict
14
We
use
"
true"
here
to
refer
to
a
probability
based
sample,
4­
46
a
national
election
with
great
accuracy
whereas
a
non­
true
sample
of
100,000
is
worthless
for
this
purpose
as
the
Literary
Guild
Magazine
learned
to
its
chagrin
when
it
predicted
Landon
over
Franklin
Roosevelt.

The
RFF
survey
results
are
based
on
a
total
sample
size
of
1576.
Much
of
the
analysis
in
this
report
is
based
on
versions
A,
B,
and
C
for
(
N=
1221)
for
wh
we
have
approximately
700
valid
answers
to
our
WTP
questions.

Response
Rate
The
results
of
a
sample
survey
can
be
biased
if
"
significant"
number
of
people
selected
to
be
part
of
the
sample
refuse
to
be
interviewed
or
are
unavailable
to
the
interviewer
because
of
travel,
sickness
or
work
at
the
time
the
interviewer
calls.
When
this
occurs,
bias
is
introduced
because
those
not
interviewed
are
likely
to
differ
from
those
who
were
interviewed
in
systematic
ways.
For
example,
they
may
be
more
or
less
environmentally
oriented.
The
question
of
what
constitutes
a
significant
number
does
not
have
a
simple
answer
owing
to
variations
in
sampling
design
(
some
call
for
substitutions
on
a
prespecified
basis
where
the
person
sampled
is
not
available
at
the
time
of
the
interview),
in
interview
method
(
rates
differ
for
the
telephone,
mail
and
personal
interview
techniques),
and
in
the
method
of
calculating
the
response
rate
(
since
non­
responses
can
be
due
to
outright
refusals,
to
not
being
at
home,
to
terminating
the
interview
before
it
is
completed,
etc.
the
way
of
calculating
the
rate
varies
according
to
what
is
defined
as
a
non­
response)
(
Dillman,

1978:
49­
52).
4­
47
When
there
are
no
established
criteria
for
determining
the
quality
of
the
response
rate,
as
is
the
case
for
most
surveys
which
are
not
conducted
by
professional
survey
research
organizations,
researchers
should
provide
sufficient
information
to
enable
the
reader
to
evaluate
the
sampling
15
implementation.
In
our
case,
we
used
a
professional
organization
and
well
established
sampling
procedures.
The
response
rate
for
our
survey
is
73
percent,
computed
upon
the
number
of
interviews
completed
in
households
containing
people
eligible
for
an
interview.
Those
not
interviewed
included
people
who
refused
and
those
who
were
not
at
home
even
after
the
interviewers
made
up
to
three
call
backs
to
reach
the
person
in
the
household
designated
to
be
interviewed
by
the
sampling
plan.
This
response
rate
is
well
16
within
current
national
sample
survey
practice
using
this
methodology.

A
comparison
between
the
RFF
sample
and
census
data
for
age,
education,

income,
sex,
race
and
region
shows
the
RFF
sample
to
be
a
close
approximation
of
the
nation
on
all
but
education
and
those
with
the
highest
income
(
Table
4.7).
Those
with
a
less
than
high
school
education
and
the
highest
income
are
somewhat
under
represented
,
a
common
occurrence
in
sample
surveys
as
these
people
are
among
those
most
likely
to
unavailable
(
the
rich
travel
or
are
less
accessible;
those
with
low
educations
are
disproportionately
15
The
Colorado
State
researchers,
for
example,
describe
their
samples
in
admirable
detail
(
Walsh,
et
al.,
1978:
19­
23)
and
include
a
table
which
informs
the
reader
that
of
600
people
originally
selected
for
interview,
48
letters
were
returned,
231
could
not
be
contacted
by
phone,
119
refused
to
be
interviewed
when
contacted
and
202
were
interviewed.

16
Although
it
is
impossible
to
make
a
direct
comparison,
our
73
percent
may
be
compared
to
the
37%
rate
achieved
by
the
Colorado
State
researchers
(
excluding
the
returned
letters,
but
including
in
the
base
those
the
interviewers
could
not
reach
and
those
who
refused?.
4­
48
Table
4.7
DISTRIBUTING
OF
RESOURCES
FOR
THE
FUTURE
SURVEY
ON
KEY
DEMOGRAPHIC
VARIABLES
Census
18
­
24
25
­
34
35
­
44
45
­
54
55
­
64
65
+

Less
than
High
school
High
School
some
college
college
Under
$
9,999
$
10
­
14,999
15
­
24,999
25
+
refused
RFF
16%
18%
26
22
15
16
14
15
15
13
15
I.
6
100
100
(
age
18+)
(
age
25+)

25
32
38
37
20
1.5
17
16
100
100
25
24
16
17
28
31
22
28
1
0
­
­
­
1.01
100
Male
Female
Black
White
New
England
Mid
Atlantic
East
North
Central
West
North
Central
South
Atlantic
East
South
Central
West
South
Central
Mountain
Pacific
RFF
Census
47%
48.7%
53
51.3
100
100.0
12
12
87
88
99
100
7
17
17
9
17
6
9
5
14
101
6
17
19
8
16
6
10
5
14
101
Current
Population
Reports
(
Population
characteristics:
Profile
or
the
United
States:
1979)
Series
P­
20,
No.
350,
U.
S.
Department
of
Commerce,
Bureau
of
the
Census,
May
1980.
4­
49
among
the
very
old).
Other
factors
may
play
a
role
here
too,
The
10
percent
who
refused
to
reveal
their
incomes
may
be
disproportionately
well
off.

The
census
data
are
not
from
the
1980
census
(
which
was
unavailable
when
the
table
was
constructed)
which
presumably
will
show
a
higher
percent
of
people
with
college
educations
than
the
earlier
census
estimates.

INTERVIEW
BIASES
Item
Nonresponse
Bias
Respondents
invariably
fail
to
answer
at
least
one
question
in
an
interview.

This
presents
a
problem
when
the
analyst
wishes
to
generalize
from
a
sample
to
a
population.
Item
nonresponse
bias
is
the
distortion
in
the
estimate
of
the
population
characteristics
for
a
variable
caused
by
people
failing
to
answer
a
question.

As
noted
earlier,
this
type
of
bias
is
the
one
which
presented
the
greatest
problem
in
this
study.
Considering
only
those
who
answered
versions
A,
B,
C
(
as
has
been
our
practice),
38
percent
failed
to
answer
for
our
WTP
questions
and
16
percent
gave
a
$
0
amount.
Strictly
speaking,
the
zero
amounts
are
responses
and
we
treated
them
as
such,

but
they
bear
further
analysis,
Since
other
studies
have
found
that
a
portion
of
the
zero
bids
represent
protest
bids
and
not
true
zero
valuations,
it
is
appropriate
to
treat
them
here
under
the
item
nonresponse
bias
rubric.

Let
us
consider
those
who
failed
to
give
any
amount
first,
In
national
surveys
it
is
common
for
the
don't
knows
to
range
from
5­
10
percent
for
relatively
demanding
questions.
This
was
the
case
with
the
questions
which
immediately
preceded
the
WTP
items
in
our
questionnaire.
4­
50
It
asked
respondents
for
their
water
quality
preference
and
received
an
11
percent
nonresponse
rate.
In
comparison,
the
38
percent
for
the
WTP
items
is
obviously
high.
The
three
most
likely
explanations
for
this
are:
1)
The
general
difficulty
of
WTP
questions;
2)
The
peculiar
difficulty
of
our
questions;
3)
The
interviewing
situation
for
our
study.
We
will
discuss
each
in
turn
before
concluding
that
a
combination
of
the
first
and
last
of
these
factors
is
the
most
likely
explanation
for
our
high
nonresponse
rate.

WTP
surveys
are
very
demanding
of
respondents
and
it
should
not
be
surprising
if,
for
comparable
samples,
they
experience
higher
item
nonresponse
rates
that
surveys
using
more
common
types
of
question.
The
WTP
instrument
asks
the
respondent
to
attend
to
a
description
of
the
hypothetical
market
which
is
necessarily
detailed.
It
requires
the
respondent
to
value
in
dollars
an
amenity
the
respondent
does
not
customarily
view
in
that
manner.
This
is
an
intellectually
demanding
task
and
requires
a
motivational
commitment
which
may
be
lacking
for
people
for
whom
the
public
good
being
valued
is
not
particularly
salient.
We
reviewed
13
WTP
studies
to
compare
their
item
nonresponse
rates
on
their
WTP
questions,
Unfortunately,

less
than
half
of
these
studies
provide
enough
information
about
item
nonresponse
to
enable
us
to
include
them
in
the
comparison.
For
the
six
which
did,
the
rates
ranged
from
1
percent
for
Robert
Davis'
pioneering
study
of
visits
to
the
Maine
woods
(
Knetsch
and
Davis,
1966)
to
32
percent
for
a
sales
tax
vehicle
used
to
study
the
value
of
air
visibility
in
the
4­
51
Four
Corner's
area
(
Randall,
et
al.,
1974).
In
between
were
item
nonresponse
rates
of
2
percent
(
elk
licenses,
Brookshire,
et
al.,
1980),

8
percent
(
damage
from
surface
mining,
Randall,
et
al.,
1978),
11
percent,

17
(
air
visibility,
Brookshire,
et
al.,
1980).

14
percent
(
sales
tax,
Walsh,
et
al.,
1978),
20
percent
(
utility
bill,

Brookshire,
et
al.,
1980),
and
21
percent
(
electric
bill,
Randall,
et
al.,
1974).

These
data
suggest
the
following
conclusions:
1)
on
the
average,
WTP
studies
tend
to
have
somewhat
higher
item
nonresponse
rates
than
regular
survey
questions
and
yet
2)
under
certain
conditions
these
rates
are
very
low.
In
Davis'
case,
he
personally
conducted
all
his
interviews
in
the
Maine
woods
and
reports
very
high
rapport
with
his
respondents.
The
elk
license
payment
vehicle
of
Brookshire,
et
al.
(
1980)
is
specifically
and
traditionally
tied
to
the
good
being
valued.
Because
entrance
fee
vehicles
have
the
same
characteristics,
we
would
also
expect
them
to
have
low
item
nonresponse
rates.
Studies
like
ours
which
use
bidding
vehicles
that
are
less
specific
or
traditionally
tied
to
the
good
may
expect
higher
item
nonresponse
rates.

The
second
hypothesized
cause
of
item
nonresponse
is
our
question
wording.
While
we
have
identified
minor
changes
which
will
make
the
questions
clearer
and
more
interesting
to
the
respondents
we
are
not
aware
of
serious
problems
in
this
area.
In
our
pretest
with
a
specially
trained
interviewer
only
two
people
of
38
failed
to
give
WTP
amounts.

17
Neither
of
the
last
two
studies
specifically
report
item
nonresponse
rates.
We
infer
these
values
from
Randall,
et
al.'
s,
"
unusable"
survey
figure
and
Brookshire,
et
al.'
s
"
deletions"
for
reasons
not
explained
(
presumably
because
the
respondents
gave
no
amount.
4­
52
The
interview
situation
is
another
matter.
We
believe
this
is
a
major
contributor
to
the
high
item
nonresponse
for
several
reasons,
First,
as
noted
earlier,
we
were
able
to
obtain
a
national
sample
at
low
cost
because
we
were
able
to
add
the
benefits
questionnaire
to
an
existing
survey,

Because
of
this,
as
mentioned
previously,
the
WTP
instrument
was
administered
after
the
respondents
(
and
the
interviewer)
had
already
spent
at
least
a
half
hour
on
the
environment/
energy
survey.
For
certain
categories
of
people,
especially
the
aged
and
those
with
low
levels
of
education,
the
preceeding
interview
probably
took
longer
than
a
half
hour
with
correspondingly
greater
fatigue
effects.
Second,
because
our
budget
was
limited,
(
and
our
purposes
experimental)
we
did
not
provide
the
interviewers
with
the
kind
of
detailed
instructions
which
we
would
provide
for
a
full
scale
benefit
estimation
study.
These
instructions
would
include
procedures
for
handling
various
types
of
respondent:
queries
and
instructions
for
encouraging
reluctant
players
to
give
WTP
amounts.
Third,
the
same
budget
constraints
restricted
the
length
of
our
WTP
instrument.
The
addition
of
several
followup
questions
in
the
instrument
itself
which
would
probe
non­
responses
(
and
zero
amounts)
would
enable
us
to
identify
respondents
who
would
give
us
WTP
amounts
after
further
explanations.

To
summarize,
the
most
likely
explanation
for
our
high
item
nonresponse
rate
is
a
combination
of
the
inherent
difficulty
of
WTP
questions,

and
the
limitations
of
our
interview
situation.
Appropriate
changes
in
the
latter,
combined
with
a
fine
tuning
of
the
questions,
should
reduce
the
item
nonresponse
rate
to
a
tolerable
level,
Because
of
the
inherent
4­
53
difficulty
of
these
types
of
questions,
it
will
be
very
difficult
to
bring
item
nonresponse
rates
from
10­
15
percent
for
WTP
surveys
of
the
general
public.
Rates
of
this
level
should
not
unduly
bias
the
final
estimates
if
weighting
procedures
are
used
to
compensate
for
the
nonrespondents.
We
discuss
these
matters
further
in
Chapter
6.

How
will
our
item
nonresponse
rate
of
39
percent
bias
these
data?

Put
another
way,
this
question
becomes:
What
kinds
of
people
failed
to
respond
to
our
WTP
questions?
We
estimated
a
logit
regression
equation
for
a
combination
of
background
variables
and
key
attitude
items
which
is
presented
in
Table
4.8.
Definitions
for
these
variables
are
given
on
Table
4.6,
page
4­
39
The
dependent
variable
is
a
dummy
with
the
nonrespondents
set
at
1
and
all
those
who
gave
WTP
amounts
greater
than
zero
for
fishable
water
at
0.
(
Thus
we
drop
those
who
gave
zero
amounts
from
the
following
analysis).
The
overall
predictive
accuracy
coefficient
of
.27
indicates
a
moderate
fit.
Older
people,
blacks
and
those
who
are
uncertain
about
the
nation's
water
quality
goals
(
0.81
SPRECHLD)
were
especially
likely
(
p.
=
.
OO1>
and
those
respondents
low
in
income
and
education
were
very
likely
(
p.
=
.
O1>
to
be
among
the
nonrespondents.
The
respondent's
sex
and
use
of
water
for
recreation
were
also
significantly
related
to
the
dependent
variable.
This
profile
is
consistent
with
the
hypothesis
that
people
for
whom
the
issue
is
less
salient
(
SPRECHLD,
RACED)
and/
or
for
whom
the
WTP
instrument
is
difficult
to
answer(
AGECAT,
EDUC,
SPRECHLD)
are
more
likely
to
be
among
the
nonrespondents
to
the
WTP
items.
It
is
noteworthy
that
environmental
and
water
quality
attitudes
(
ENVINDEX,
CNPOLD)
are
not
significant
in
this
equation.
Table
4.8
4­
54
LOGITl
REGRESSIONS
RELATING
BACKGROUND
AND
ATTITUDINAL
VARIABLES
TO
CERTAIN
TYPES
OF
WILLINGNESS
TO
PAY
RESPONSES
FOR
FISHABLE
WATER2
Dependent
variable;
Dependent
variable;
Independent
1
=
zero
wtp
amount;
1
=
'
don't
know'
how
much
Variables
0
=
WTP
amount
greater
willing
to
pay
than
zero
0
=
WTP
amount
greater
than
zero
Intercept
INCOMER
­.
0002*
­.
00002**

EDUC
­.
42***
­.
23**

AGECAT
.14**
.09***

RACED
­.
95**
­
1.38***

SEXD
­.
10
.39*

USERD
­
1.11***
­.
44*

ENVINDEX
­.
44***
­.
08
CNPOLD
SPRECHLD3
­.
23
­.
15
­.
96**
­
1.68***

N
695
783
Likelihood
ratio
index
.31
.18
R2
index
(
D)
.25
.19
Percent
correctly
predicted
zero
amounts
84%
other
amounts
86
Predictive
accuracy
coefficient
.47
don't
know
77%
other
amounts
78
.27
*
p
<
.05
/
**
p
<
l
.01
/
***
p
<
.001
1Xaximum
likelihood
estimates
are
computed
by
the
Newton­
Raphson
method.
(
SAS
Institute,
1980).

2For
Versions
A,
B,
and
C
combined.

3Dummy
variable
where
1
=
nation
should
plan
to
achieve
nationwide
water
quality
of
fishable
or
better
within
the
next
five
years
(
Q.
81);
0
=
all
other
responses
of
which
"
not
sure"
comprises
two­
thirds
and
preference
for
nationwide
water
quality
lower
than
fishable
comprises
one­
twelfth.
4­
55
From
what
we
know
about
the
willingness
to
pay
for
water
quality
of
other
respondents,
the
bias
given
our
estimates
by
the
high
item
nonresponse
rate
is
upwards.
The
older,
less
educated
and
lower
income
people
who
expressed
WTP
amounts
gave
lower
amounts,
other
things
being
equal,
than
their
peers,
and
we
would
expect
the
addition
of
a
significant
number
of
the
nonrespondents
to
those
giving
WTP
amounts
to
lower
the
mean
WTP
value
for
water
quality.

Turning
now
to
the
zero
amounts,
sixteen
percent
of
our
sample
gave
WTP
amounts
of
$
0
for
fishable
water.
It
is
very
difficult
to
compare
this
with
the
experience
of
other
WTP
studies
since
only
four
of
the
13
studies
reviewed
report
the
total
percent
of
$
0
bids.
For
these
studies
the
zero
amounts
varied
as
follows:
1
percent,
Maine
Woods
(
Knetsch
and
Davis,
1966);
2
percent
for
sales
tax
vehicle
and
26
percent
for
utility
bill
option,
water
quality
in
the
South
Platte
River
Basin
(
Walsh,
et
al.,
1978);
6
percent
for
non­
reservation
residents,
air
pollution
visibility
in
Four
Corner's
area
(
Randall,
et
al.,
1974);
and
7­
32
percent,
depending
on
WTP
version,
decreased
risk
from
nuclear
plant
accidents
(
Mulligan,
1978).
Our
level
of
zero
amounts
is
somewhere
in
the
middle
of
this
distribution,
but
we
do
not
regard
this
level
of
zero
amounts
as
acceptable,
especially
since
we
already
have
a
high
nonresponse
rate
for
the
WTP
questions.
4­
56
their
true
value
in
any
future
use
of
our
instrument.
The
factors
discussed
above
for
nonresponse
are
also
the
likely
cause
of
the
zero
amounts.
Question
wording
probably
played
a
much
larger
role
in
stimulating
the
zero
responses,
however.
Endeavoring
to
legitimate
low
values
for
respondents
who
might
have
been
hesitant
to
express
their
"
true"
feelings
about
water
quality,
we
ended
the
first
WTP
question
in
the
series
by
saying:
"
If
it
is
not
worth
anything
to
you,

please
do
not
hesitate
to
say
so."
In
retrospect
we
believe
this
was
too
strong
a
statement
which
unnecesarily
promoted
zero
responses
by
some
who
probably
have
valued
water
at
greater
than
zero
but
who
were
reluctant
to
undertake
the
mental
effort
necessary
to
arrive
at
that
value.

We
will
substitute
another
type
of
encouragement
to
respondents
to
give
We
estimated
a
logit
regression
for
a
dummy
variable
with
zero
WTP
set
at
1
and
those
who
gave
amounts
greater
than
zero
at
0.
This
regression
is
also
reported
in
Table
4.7.
This
estimation
has
superior
predictive
power
to
the
parallel
one
for
nonrespondents
(
predictive
accuracy
coefficient
of
.47).
Comparing
the
two
equations
we
find
recreational
use
and
environmentalism
play
a
greater
role
in
predicting
the
zero
bidders,
who
tend
to
use
water
less
and
are
weaker
in
their
support
for
environmentalism.
These
findings
are
consiste
with
the
hypothesis
that
zero
bids
do
represent
low
(
if
not
zero)
value
for
water
quality.
However,
the
importance
of
age,
also
significant
in
the
equation
at
the
.001
level,
and
the
role
of
race
and
education
(.
01),
4­
57
parallel
their
place
in
the
nonresponse
equation
and
suggests
that
zero
bids
may
also
be
partially
due
to
people
protesting
the
WTP
format
or
expressing
an
unwillingness
to
answer
the
question.

The
bias
introduced
by
the
large
number
of
zero
bidders
is
to
make
our
estimates
lower
than
they
would
be
if
we
had
fewer
zero
bidders.

From
the
findings
of
other
WTP
studies
which
have
asked
their
zero
bidders
why
they
bid
zero
(
Rowe,
et
al.,
1979a;
Thayer
forthcoming,

Brookshire,
et
al.,
1980;
Brookshire,
et
al.,
1976)
it
seems
very
likely
that
some
of
our
zero
bidders
are
probably
protesting
the
instrument
rather
than
really
valuing
water
quality
at
$
0.
An
indeterminate
number
of
the
remaining
zero
bidders,
while
not
protesting,
nevertheless
probably
value
water
quality
at
least
somewhat
higher
than
$
0
and
could
be
induced
to
bid
higher
by
the
changes
described
above.

(
continue)
4­
58
Interview
Procedure
and
Interviewer
Biases
Two
other
interview
method
biases
remain
to
be
discussed.
The
interview
procedure­
bias
refers
to
bias
introduced
by
the
manner
of
conducting
the
interview.
Interviewing
takes
place
by
either
personal
interview,

telephone
or
mail.
The
differences
involved
in
choosing
between
these
methods
including
cost,
return
rate,
ease
of
asking
sensitive
questions,
and
ease
of
asking
complex
questions.
Although
it
is
the
most
expensive
method,
the
personal
interview
method
is
superior
to
the
other
methods
on
all
dimensions
(
Dillman,
1978:
74­
76;
on
social
desirability
see
Bradburn
and
Sudman,
1979:
8).
The
personal
interview
method
is
especially
preferable
for
WTP
surveys
because
it
permits
the
researcher
to
use
visual
displays
such
as
our
ladder
and
payment
cards
and
it
is
the
most
successful
of
these
methods
when
the
questions
are
potentially
tedious
and
boring
(
Dillman,
1978:
75).
The
only
viable
alternative
would
be
the
mail
survey,

a
method
used
only
twice
in
a
WTP
study
to
our
knowledge
(
Bishop
and
Heberlein,

1980;
Fish
and
Wildlife
Service,
1975)
as
the
need
to
create
the
hypothetical
market
in
sufficient
detail
is
too
wordy
for
phone
interviews.

Unlike
the
mail
surveys,
personal
interview
surveys
are
open
to
potential
interviewer
bias.
This
type
of
bias
consists
of
differential
effects
introduced
by
the
individual
interviewers.
In
a
bidding
game,

for
example,
some
interviewers
may
be
more
skillful
in
inducing
respondents
to
increase
their
bids
above
the
starting
point
more
than
others.
If
a
study
uses
relatively
few
interviewers
who
conduct
25
interviews
or
more,

it
is
possible
to
test
for
interviewer
effects
by
holding
the
respondents'
4­
59
personal
characteristics
(
such
as
income)
constant
and
comparing
the
mean
WTP
amounts
to
see
if
they
differ
significantly.
Because
Roper
used
100
interviewers
scattered
across
the
country
to
conduct
our
interviews,

the
number
of
interviews
per
interviewer
is
too
few
to
conduct
this
type
of
test.
With
that
many
interviewers
we
would
expect
individual
interviewer
effects,
if
there
are
any,
to
average
out.
There
is
always
the
possibility
that
the
interviewer
training
may
induce
all
the
interviewers
in
a
project
to
obtain
higher
bids
than
interviewers
trained
by
someone
else
might
with
the
same
questionnaire,
but
there
is
no
easy
way
to
test
for
this
other
than
to
conduct
elaborate
methodological
experiments.

One
advantage
of
our
payment
card
technique
is
that
it
minimizes
the
potential
interviewer
effect
on
the
WTP
amount
as
compared
with
the
bidding
game
method.
Chapter
5
EXPERIMENTAL
BENEFITS
ESTIMATES:
OVERALL,
USE
AND,
INTRINSIC
In
this
chapter
we
examine
the
WTP
amounts
given
by
our
respondents.

The
analysis
begins
with
an
examination
of
the
level
of
benefits
for
national
water
quality
revealed
by
our
respondents.
We
then
test
the
predictive
power
of
a
theoretically­
based
estimation
of
the
amounts;

an
important
test
of
our
instrument's
hypothetical
bias.
The
next
section
presents
our
technique
for
separating
intrinsic
from
recreational
benefits
and
illustrates
it
with
our
data.
In
the
final
section
we
consider
the
regional
variation
in
water
benefits
and
discuss
procedures
by
which
the
data
from
a
national
water
benefits
survey
may
be
helpful
to
those
who
wish
to
estimate
water
benefits
for
sub­
national
areas.

Before
proceeding
further
it
is
important
to
emphasize
that
the
benefit
estimates
we
discuss
below
come
from
experimental
data
and
should
not
be
used
for
making
definitive
national
estimates.
Our
study
was
designed
to
develop
a
new
methodology
and
to
test
it
to
see
if
it
shows
sufficient
promise
for
a
full
scale
application
(
after
appropriate
revision).
As
noted
in
the
last
chapter,
our
macro
WTP
instrument
was
very
successful
with
the
exception
of
the
item
non­
response
rate.
The
nonresponse
rate
problem
is
correctable
(
see
the
Conclusion
for
our
proposals),
but
it
means
the
present
set
of
WTP
amounts
represents
a
selective
rather
than
a
random
sample
of
the
U.
S.
population.
Although
our
data
are
not
sufficiently
representative
for
national
estimates,
they
are
sufficiently
free
from
bias
to
warrant
the
analysis
we
undertake
in
this
chapter.
In
this
sense
the
estimates
discussed
in
the
next
section
may
be
taken
as
illustrative,
in
a
rough
way,

of
the
benefit
estimates
which
a
revised
national
survey
might
produce.
5­
2
ILLUSTRATIVE
ESTIMATES
Taking
into
account
the
above
caveat,
we
discuss
here
the
WTP
amounts
given
by
our
respondents.
This
sample
consists
of
all
those
1
who
were
exposed
to
Versions
A,
B,
or
C
of
the
questionnaire
and
who
gave
us
usable
amounts
(
including
zero
bids).
The
number
of
cases
on
which
the
analyses
in
this
chapter
are
based
vary
from
771
to
695
according
to
whether
or
not
we
had
to
drop
cases
because
of
missing
data
on
individual
items.

Amounts
by
Version
As
described
in
Chapter
2,
the
respondents
valued
three
levels
of
water
quality
which
were
described
in
words
and
depicted
on
the
water
quality
ladder.

They
were
first
asking
how
much
they
were
willing
to
pay
to
maintain
national
water
quality
at
the
boatable
level.
Subsequent
questions
asked
them
their
willingness
to
pay
for
overall
water
quality
to
fishable
quality
and
swimmable
quality.
The
mean
WTP
amounts
given
by
the
respondent
for
the
two
higher
levels
consists
of
the
amounts
they
offered
for
the
lower
levels
plus
any
additional
amount
they
offered
for
the
higher
level.
Table
5.1
gives
the
mean
WTP
amounts
for
each
of
the
three
versions.

1With
the
exception
of
a
handful
of
respondents
whose
answers
to
the
questionnaire
were
so
contradictory
that
they
were
judged
to
be
meaningless.

The
removal
of
these
22
cases
presents
no
bias
to
the
WTP
amounts
as
their
mean
WTP
amount
is
the
same
as
the
entire
sample's.
Appendix
VI
describes
our
rationale
for
dropping
these
respondents
and
gives
information
about
each
case.
5­
3
MEAN
AMOUNTS
WILLING
TO
PAY
ANNUALLY
PER
HOUSEHOLD
FOR
BOATABLE,
FISHABLE
AND
SWIMMABLE
WATER
QUALITY
Table
5.1
IN
THE
UNITED
STATES
BY
VERSION
AND
INCOME
LEVELS
1
2
Income
Levels
1
2
3
4
Total
Version
A
Version
B
Version
C
$
61
(
62)
$
47
(
61)
$
71
(
64)

114
(
38)
124
(
48)
87
(
38)

183
(
78)
135
(
79)
174
(
82)

289
(
73)
262
(
48)
308
(
50)

3
3
3
$
168
(
274)
$
133
(
255)
$
161
(
242)

1
$
77
$
60
$
91
2
161
149
111
3
229
201
223
4
363
347
362
Income
Levels
Version
A
Version
B
Version
C
1
995
$
76
$
103
2
195
163
128
3
268
244
267
4
404
394
375
Total
$
247
$
212
$
222
Total
$
214
$
180
$
198
1
In
this
version
of
the
research
instrument
those
who
did
not
give
an
amount
in
answer
to
the
willingness­
topay
questions
received
no
further
encouragement
to
do
so
by
the
interviewers.

As
a
consequence,
32
percent
of
the
respondents
(
for
fishable
water
it
was
32%
for
version
A;
30%
for
version
B;
and
34%
for
version
C)
did
not
give
amounts.
The
32
percent
who
did
not
give
an
amount
is
comprised
of
24
percent
who
said
they
''
don't
know,"
6
percent
"
it
depends"
and
2
percent
who
refused
to
answer.

2
The
percent
who
said
$
0
were
18%,
22%
and
24%
in
version
A
to
C
respectively.

3The
total
N's
are
larger
than
the
sum
of
the
N's
for
the
four
income
levels
because
they
also
include
those
who
answered
the
willingness­
to­
pay
questions
but
were
not
willing
to
give
their
income.
Since
these
people
could
not
be
assigned
to
their
correct
income
group
the
interviewers
were
told
to
treat
them
as
if
they
were
in
income
level
4.
If
we
include
those
who
did
not
give
an
amount,
the
total
N's
for
the
three
versions
are:
A­
431;
B­
380;
and
C­
410.
5­
4
It
shows
the
following:

1.
The
pattern
of
amounts
is
quite
consistent
across
the
three
versions
of
the
instrument.
As
noted
in
Chapter
4
only
two
of
the
36
between­
version
comparisons
show
differences
that
are
statistically
significant
at
the
.05
level.

2.
The
effect
of
respondent's
income
is
uniformly
strong
as
shown
by
the
column
amounts.
This
is
an
expected
effect,
of
course,

since
people
with
higher
incomes
a)
have
more
disposable
income,

and
b)
were
shown
payment
cards
whose
benchmark
amounts
for
non­
environmental
public
goods
were
higher.

3.
The
WTP
amounts
are
substantial.
This
is
in
contrast
with
the
earlier
macro
WTP
studies
described
in
Chapter
2
which
did
not
describe
the
hypothetical
market
for
their
goods
in
detail.

Combined
Amounts
The
WTP
amounts
for
the
combined
sample
are
shown
in
Figure
5.1.
The
most
substantial
benefit
is
for
boatable
water
with
a
range
of
$
136­
168
per
annum
per
household.
The
respondents
were
willing
to
pay
S175­
213
for
2
fishable
water,
an
amount
27
percent
higher
than
the
boatable
estimate.

According
to
these
data,
national
water
of
swimmable
quality
yields
a
diminishing
return
as
the
swimmable
WTP
amount
is
only
16
percent
greater
than
the
fishable
amount.

2The
mean
amount
which
this
sample
of
people
is
willing
to
pay
for
swimmable
water
quality
is
approximately
the
amount
paid
in
taxes
and
higher
prices
in
1979
for
water
pollution
control
by
U.
S.
households
according
to
the
estimates
of
the
President's
Council
on
Environmental
Quality.
The
CEQ
estimate
for
1979
amounts
to
$
159
per
household
for
control
instituted
as
a
result
of
federal
pollution
control
programs
and
S25j
for
all
water
quality
expenditures
,
including
those
which
industry
would
have
undertaken
irrespective
of
the
federal
pollution
control
laws
(
Council
on
Environmental
Quality,
1980:
394,
397).

For
these
experimental
data
the
total
annual
benefits
for
swimmable
water
nationwide
lie
somewhere
between
9
and
22
billion
dollars.
No
point
estimate
should
be
inferred
from
this
range
for
the
reasons
explained
in
detail
in
the
report.
5­
5
WHAT
PEOPLE
ARE
WILLING
TO
PAY
EACH
YEAR
PER
HOUSEHOLD
Figure
5.1
FOR
DIFFERENT
LEVELS
OF
NATIONAL
WATER
QUALITY
Amount
Annual
Amounts
at
the
$
300
$
200
$
100
Water
D
C
B
Quality
Boatable
Fishable
Swimmable
Level
Mean
=
$
152
$
194
$
225
SE
of
mean
=
$
8.03
$
9.55
$
10.57
N
=
748
748
748
5­
6
Effect
of
Knowing
Amount
Being
Paid
Some
of
the
earlier
macro
WTP
studies
(
Viladus,
1973)
show
that
people
are
more
willing
to
pay
higher
amounts
for
public
goods
when
they
are
told
the
amount
it
will
cost
(
or
is
costing)
than
when
they
do
not
have
this
information.
In
order
to
see
if
this
is
the
case
in
our
study,
we
departed
from
our
previous
format
in
Version
D
of
our
research
instrument
and
told
the
respondents
what
they
are
paying
for
water
pollution
control.
3
In
our
case
the
revealed
value
for
water
quality
in
Version
D
is
quite
similar
to
that
for
the
combined
A,
B,
C
versions
where
the
respondents
were
not
told
how
much
they
are
paying.

Forty­
seven
percent
of
the
354
respondents
to
Version
D
said
they
were
willing
to
pay
the
amount
shown
on
their
card
for
water
pollution
control
(
which
they
were
told
would
raise
the
overall
level
of
national
water
quality
to
fishable
in
the
next
few
years)
and
12
percent
volunteered
that
"
it
depends."

Thirty
percent
were
not
willing,
11
percent
were
not
sure
or
didn't
know,

and
less
than
one
percent
did
not
answer
the
question.
Those
who
were
not
willing
to
pay
the
amount
were
asked
how
much
they
were
willing
to
pay
to
keep
the
quality
of
water
at
boatable
quality
whereas
those
who
were
willing
to
pay
the
amount
were
asked
to
value
an
increase
in
quality
from
fishable
to
swimmable
(
level
B).
It
is
possible
to
calculate
values
for
fishable
4
and
swimmable
water
from
these
data.
The
Version
D
range
for
fishable
water
3
They
were
shown
on
the
payment
card
an
estimate
of
what
households
in
the
respondents'
income
range
were
actually
paying
for
water
pollution
control.

4
In
making
this
calculation
we
assign
each
person
who
is
willing
to
pay
the
amount
shown
on
the
payment
card
for
water
pollution
control
that
value
as
their
WTP
value
for
fishable
water.
said
"
it
depends"
Under
the
assumption
that
those
who
would
be
willing
to
pay
that
amount
too
if
they
could
be
assured
that
it
would
achieve
the
fishable
water
quality
goal,
we
also
counted
them
as
willing
to
pay
the
amount
shown.
Those
who
gave
amounts
for
boatable
water
but
not
for
fishable,
were
counted
as
also
willing
to
pay
the
boatable
amounts
for
fishable
water
quality.
5­
7
EXPLANATION
OF
WILLINGNESS
TO
PAY
FOR
WATER
QUALITY
Model
Specification
quality
is
$
185­
233
compared
to
the
A,
B,
C
combined
range
of
$
175­
213.

The
WTP
amounts
for
swimmable
water
given
by
the
Version
D
respondents
are
somewhat
higher
than
those
given
by
the
respondents
to
the
other
version.

A
test
of
the
hypotheticality
of
WTP
studies
is
whether
or
not
the
respondent's
values
can
be
explained
by
a
set
of
theoretically
relevant
factors.
If
the
WTP
questions
are
sufficiently
meaningful
to
the
respondent,

his
or
her
answers
should
be
constrained
by
those
factors
which
affect
such
matters
in
everyday
circumstances.
Surprisingly
few
WTP
studies
have
reported
regression
estimations
and
of
these
only
one
or
two
include
the
range
of
factors
which
theory
and
empirical
research
suggest
as
possible
explanatory
factors.
5
We
propose
the
following
as
the
appropriate
determinants
of
willingness
to
pay:

WTP
=
f(
Respondents'
Income,
Education,
Age,
Environmental
Attitudes,
Availability
of
Freshwater,
Attitudes
Towards
Water
Quality)

In
our
original
estimation
several
of
these
factors
did
not
enter
into
the
6
equation
significantly.
Hence
we
removed
these
variables
and
re­
estimated
5For
WTP
studies
which
report
lack
of
success
in
explaining
the
bids
by
regression
equations
see
Eastman,
et
al.
(
1978)
and
Thayer(
forthcoming).
The
only
studies
which
use
a
range
of
variables
comparable
to
ours
include,
interestingly,
the
two
previous
WTP
studies
of
water
quality
(
Gramlich,
1977;
Walsh,
et
al.
1978)
in
the
published
literature.

6These
include
several
dimensions
of
the
respondents'
attitudes
toward
water
quality
(
e.
g.
desired
quality
levels
of
national
freshwater,
perceived
changes
in
local
water
quality)
and
the
availability
of
freshwater
for
recreational
use.
5­
8
the
equations.
The
coefficients
and
the
significance
levels
of
the
remaining
variables
were
not
appreciably
different
from
the
larger
equations.

Because
we
believe
that
major
conceptual
and
definitional
problems
exist
with
some
of
the
nonsignificant
variables
we
will
not
report
the
results
of
these
larger
equations
here.
The
variables
which
remain
and
our
measures
of
them
are
as
follows:

Income
­­
The
higher
the
respondents'
family
income,
the
larger
the
amount
of
disposable
income
the
respondent
has
available
for
water
quality.

We
measured
income
by
the
standard
survey
research
procedure
of
presenting
the
respondent
with
a
card
which
contains
a
list
of
income
categories.
The
respondent
was
asked:
"
Would
you
call
off
the
letter
of
the
category
that
best
describes
the
combined
(
emphasis
in
the
original)
annual
income
of
all
members
of
this
household,
including
wages
or
salary,
pensions,

benefits,
interest
or
dividends,
and
all
other
sources?"
Thus
we
asked
for
household
not
personal
income.
Table
5.2
presents
the
list
of
income
categories
and
the
percent
of
respondents
in
each
category.
Note
that
10
Percent
of
the
respondents
refused
to
reveal
their
household
income.

This
level
of
item
nonresponse
is
within
the
range
found
by
the
major
survey
research
organizations
in
national
samples
of
our
type.
We
decided
not
to
substitute
mean
values
for
these
cases
but
simply
to
drop
them
from
the
regression
part
of
our
analysis.
5­
9
Table
5.2
INCOME
RANGES
FOR
THE
RFF
SURVEY
Income
Range
Under
$
4,000
Percent
of
Sample
1
Levels
used
for
Payment
Cards
7%

$
4,000
to
$
5,999
7
5
7
I
$
6,000
to
$
7,999
$
8,000
to
$
9,999
$
10,000
to
$
11,999
7
II
$
12,000
to
$
14,999
9
$
15,000
to
$
19,999
13
III
$
20,000
to
$
24,999
15
$
25,000
to
$
49,999
19
$
50,000
and
over
3
Not
sure/
refused
10
IV
1
These
data
are
for
the
entire
sample,
all
versions.
5­
10
Following
the
standard
procedure
(
Kemnta,
1971)
for
incorporating
grouped
income
data
in
regression
equations
where
the
actual
income
is
unobtainable,
we
assigned
each
respondent
the
mid
point
for
his
or
her
income
category.
A
value
of
$
60,000
was
used
for
the
$
50,000
and
over
category.

Age
­­
Studies
of
the
determinants
of
environmental
attitudes
identify
age
as
an
important
predictor
(
Dunlap
and
Van
Liere,
1978;
Mitchell,
1980:
44).

Younger
respondents
are
somewhat
more
supportive
of
environmental
protection
than
older
respondents.
The
WTP
studies
which
report
regression
estimations
show
mixed
findings
on
the
relationship
between
age
and
willingness
to
pay
for
environmental
public
goods.
Walsh,
et
al.
(
1978:
66)
found
a
significant
negative
relationship
between
age
and
willingness
to
pay
for
water
quality
in
the
South
Platte
River
Basin.
Age
did
not
enter
significantly
into
the
regressions
estimated
by
Gramlich
in
his
study
of
the
Charles
River
Basin
(
1977:
187)
and
in
Eastman,
et
al.'
s
(
1978:
22)
study
of
air
visibility
in
the
Four
Corners
area
it
showed
no
consistent
pattern.

Our
age
measure
consists
of
a
card
listing
eleven
age
categories
from
which
the
respondent
chose
the
correct
age
group
for
him
or
herself.

The
first
two
age
categories
are
18­
21
and
22­
24.
Beginning
with
age
25­
29,

the
categories
proceed
by
five
year
intervals
until
the
last
group
which
was
defined
as
65
or
older.
If
the
respondent
refused
to
provide
the
age
information,
the
interviewer
was
instructed
to
make
an
estimate.
We
coded
the
age
variable
at
the
mid
points
for
each
age
category.
For
the
65
and
over
category
we
used
70
which
is
the
approximate
mid
point
of
this
age
category
according
to
census
data.
5­
11
Education
­­
Education
is
also
correlated
with
support
for
environmental
protection;
the
higher
the
educational
level,
the
greater
the
level
of
environmental
concern
(
Dunlap
and
Van
Liere,
1978:
9;
Mitchell,
1980:
44).

Two
WTP
studies
also
report
a
similar
relationship
with
willingness
to
pay
for
environmental
public
goods
(
Walsh,
et
al.,
1978:
60;
Gramlich,
1977:
187).

Our
measure
of
education
consists
of
six
categories,
ranging
from
noschool
to­
grade
8
to
post
graduate
education
(
17
years
of
formal
education
or
more).
Each
category
was
designed
to
be
a
qualitatively
equivalent
increase
in
educational
attainment
from
the
next
lower
category
with
special
weight
given
to
the
completion
of
high
school
and
college.
7
For
this
reason
our
variable
consists
of
the
categories
instead
of
the
mid
point
of
the
years
of
education
represented
by
each
category.

Environmental
Attitudes
­­
Numerous
social
surveys
have
measured
people's
attitudes
towards
environmental
issues
(
for
a
review
see
Dunlap
and
Van
Liere,
1978).
The
questions
used
for
this
purpose
measure
a
wide
variety
of
dimensions
such
as
concern,
perceived
seriousness,

tradeoffs,
and
relative
importance.
On
each
of
these
dimensions
7
These
levels
are
as
follows:

Code
Education
Category
(
no.
of
yrs)
Percent
in
Total
Sample
2
No
school,
grade
school
(
l­
8)
9x
3
Some
high
school
(
9­
11)
16
4
High
school
graduate
(
12)
38
5
Some
college
(
13­
15)
20
6
College
graduate
(
16)
11
7
Post
graduate
(
17+)
6
No
response
1
5­
12
people
can
be
arrayed
along
a
continuum
from
those
who
describe
themselves
as
valuing
environmental
amenities
a
great
deal
to
those
for
whom
environ­
.

mental
amenities
have
lesser
value.
It
is
to
be
expected
that
people's
WTP
for
environmental
amenities
should
be
related
to
their
"
environmentalism"

as
revealed
by
these
kinds
of
attitude
questions.
The
only
previous
attempt
to
our
knowledge
to
demonstrate
this
in
WTP
studies
failed
to
find
a
relationship,
however,
The
Colorado
State
study
included
a
question
about
the
respondents'
general
awareness
of
environmental
problems
in
the
study
area
which
did
not
enter
into
any
of
their
regression
estimations
(
Walsh,

1978:
83­
4.
88­
9).

The
portion
of
our
research
instrument
preceeding
the
WTP
instrument
contained
a
large
number
of
environmental
attitude
measures.
From
these
we
constructed
7
item
environmental
index
(
ENVINDEX).
The
items
for
this
index
were
chosen
subjectively.
We
included
items
which
our
previous
analysis
of
these
data
had
shown
to
be
measures
of
the
degree
to
which
the
respondent
valued
environmental
goods.
In
addition
to
an
item
which
posed
tradeoffs
between
environmental
protection
and
cost,
the
index
includes
items
which
measure
the
respondents'
attitude
toward
the
environmental
movement,
the
degree
to
which
they
rank
environmental
concerns
high
or
low
compared
to
other
national
priorities,
and
whether
they
have
lobbied
public
officials
by
letter
or
personal
contact
on
an
environmental
issue.

The
items
contained
in
the
index,
its
manner
of
construction
and
its
distribution
are
described
in
Appendix
VIII.
To
test
its
metric
qualities
5­
13
we
re­
estimated
our
regression
equations
using
several
different
forms
of
the
index
to
see
if
the
parameters
of
the
other
variables
or
the
R
2
of
the
equations
were
affected.
The
results
of
these
tests
suggest
the
8
use
of
the
linear
form.

Concern
About
Water
Pollution
­­
None
of
the
items
in
the
environmental
index
treat
water
pollution
because
we
wanted
to
see
if
concern
about
water
pollution
had
the
separate
effect
on
willingness
to
pay
we
thought
it
should.

The
item
in
our
questionnaire
which
measured
water
pollution
concern
was
one
of
a
series
of
items
about
which
the
respondent
was
asked:

(
Q.
11)
Now
I'd
like
to
find
out
how
worried
or
concerned
you
are
about
a
number
of
problems
I
am
going
to
mention:
a
great
deal,
a
fair
amount,
not
very
much,
or
not
at
all.
If
you
aren't
really
concerned
about
some
of
these
matters,
don't
hesitate
to
say
so.

C.
Cleaning
up
our
waterways
and
reducing
water
pollution.

In
answer
to
thise
question,
thirty­
nine
percent
said
they
were
concerned
a
great
deal,
44
percent
a
fair
amount,
13
percent
not
very
much
and
3
percent
not
at
all.
We
constructed
a
dummy
variable
CWPOLD)
where
1
­
those
who
say
they
are
concerned
a
great
deal
and
0
=
the
remainder.

Recreational
Use
of
Water
­­
We
reasoned
that
the
greater
the
respondent's
recreational
use
of
freshwater,
the
greater
value
water
pollution
control
8
We
estimated
equation
2
(
Table
5.4)
using
squared
and
cubed
forms
of
ENVINDEX
in
addition
to
ENVINDEX.
The
squared
and
cubed
forms
were
insignificant
Equation
2
was
also
estimated
substituting
the
log
ENVINDEX
for
ENVINDEX.
The
R2
of
this
equation
was
lower.
In
both
of
these
cases
we
used
F
tests
to
test
whether
any
of
these
alternative
equations
had
significantly
different
coefficients
for
the
other
parameters
in
the
equation
2.
Each
F
test
of
the
paired
coefficients
was
insignificant.
As
a
result
of
these
tests
we
decided
to
use
the
linear
form
of
the
index.
5­
14
would
have
for
him
or
her.
Previous
WTP
studies
examined
the
relationship
between
recreational
use
and
willingness
to
pay
without
finding
any
correlation.
The
Colorado
State
study
regressed
the
reported
number
of
water­
based
recreation
activity
days
experienced
annually
in
the
South
Platte
River
Basin
and
the
degree
to
which
respondents
liked
outdoor
water­
based
recreation
on
their
WTP
measures
and
found
no
effect
(
Walsh,

et
al.,
1978:
52,
69­
72).
Similar
findings
of
no
or
marginal
significance
for
recreational
use
are
also
reported
for
air
quality
(
Eastman,
et
al.,

1978:
16­
17)
and
water
quality
(
Gramlich,
1977:
187).

We
measured
recreational
freshwater
use
by
a
series
of
questions
(
Qs.
58­
66
in
Appendix
IV)
which
asked
the
respondent
whether
in
the
past
two
years
he
or
she
had
gone:

"
sailing,
canoeing,
power
boating,
water
skiing
and
the
like"

"
swimming
in
a
freshwater
lake
or
stream
as
opposed
to
a
swimming
pool
or
the
ocean"

"
fishing
in
a
freshwater
lake
or
stream"

Each
person
who
said
yes
to
an
item
was
asked
further
whether
he
or
she
did
this
"
within
fifty
miles
of
your
home,
or
farther
away,
or
both?"

and
"
roughly
how
many
times
would
you
say
you
(
did
the
activity)
over
the
past
two
years?"
Personal
use
of
freshwater
for
these
purposes
varied
from
34
percent
who
went
fishing
to
39
percent
who
went
boating,
We
tested
various
forms
of
a
recreational
measure
and
our
tests
showed
that
neither
the
location
of
use
nor
the
amount
of
use
contributed
to
the
estimation,

a
finding
similar
to
the
Colorado
State
study.
We
therefore
created
a
simple
dummy
variable,
USERD,
which
was
set
at
1
for
those
who
reported
freshwater
use
of
any
kind
over
the
past
two
years
(
60
percent
of
the
sample)

and
0
for
those
who
reported
no
personal
use
during
this
time
period.
5­
15
Estimation
Our
final
explanatory
model
for
national
water
quality
values
consists
of
six
variables:
three
are
socioeconomic
characteristics,
two
are
attitudinal
measures
and
one
is
a
self­
reported
behavioral
measure.
Table
5.3
gives
the
Pearson(
r)
correlation
matrix
for
these
variables.
Although
no
correlation
is
.40
or
above,
three
of
the
fifteen
are
above
.30.

Multicollinearity
cannot
be
ruled
out,
but
the
symptom
of
insignificant
coefficient
estimators
in
conjunction
with
large
R
values
was
not
observed.
2
(
continue
)
5­
16
Table
5.3
CORRELATION
MATRIX
FOR
VARIABLES
USED
IN
THE
REGRESSION
EQUATIONS
INCOME
AGE
EDUC
ENVINDEX
CWPOLD
USERD
Variable
INCOME
AGE
EDUC
ENVINDEX
CWPOLD
USERD
INCOME
AGE
1.00000
­
0.07698
0.0000
0.0425
­
0.07698
1.00000
0.0425
0.0000
0.37733
­
0.27897
0.0001
0.0001
0.05241
­
0.25041
0.1675
0.0001
­
0.05756
­
0.05206
0.1295
0.1704
0.16160
­
0.32212
0.0001
0.0001
EDUC
ENVINDENX
CWPOLD
USERD
0.37733
0.05241
­
0.05756
0.16160
0.0001
0.1675
0.1295
0.0001
­
0.27897
­
0.25041
­
0.05206
­
0.32212
0.0001
0.0001
0.1704
0.0001
1.00000
0.
20955
0.02733
0.19735
0.0000
0.0001
0.4719
0.0001
0.20955
1.00000
0.34516
0.23361
0.0001
0.0000
0.0001
0.0001
0.02733
0.34516
1.00000
­
0.00231
0.4719
0.0001
0.0000
0.9516
0.19785
0.23361
­
0.00231
1.00000
0.0001
0.0001
0.9516
0.0000
N
Mean
Std
Dev
Sum
Minimum
Maximum
695
19946.8
13647.8
13863000
2000
60000
695
42.3
16.0
29418
20
70
695
4.3
1.3
2978
2
7
695
6.4
1.8
4439
1
11
695
0.4
0.5
285
0
1
695
0.6
0.5
435
0
1
5­
17
Equations
were
estimated
using
ordinary
least
squares
regression
for
the
three
levels
of
water
quality
as
shown
in
Table
5.4.
The
patterns
for
the
three
levels
are
very
similar
with
the
fit,
as
measured
by
R2,
increasing
slightly
from
.28
for
the
boatable
equation
to
.31
for
the
swimmable
one.
Using
the
swimmable
equation
as
our
example,
each
of
the
independent
variables
is
statistically
significant
at
the
.05
level
or
better.
Income
is
the
major
factor
in
the
equation
followed
by
the
environmental
index.
Despite
its
affinity
with
the
index,
concern
about
water
pollution
enters
separately
at
a
highly
significant
level.
The
recreation
use
variable
also
enters,

although
in
the
boatable
equation
its
t
value
is
slightly
below
the
.05
level.

Alternative
functional
forms
for
these
equations
were
tested.
The
most
obvious
candidate
for
an
alternative
form,
considering
our
strong
income
effect,
is
a
log­
log
estimation
(
Gramlich,
1977).
The
results
for
this
type
of
estimation
were
not
appreciably
different
or
better
than
the
OLS
estimation
except
that
the
significance
of
the
recreational
use
variable
was
increased.
6
6
The
results
of
the
log­
log
estimation
for
fishable
waters
are
as
follows:

Dependent
Variable
=
Log
of
Level
C
Coefficient
t
Intercept
­
4.24
­
4.89
LOG
INCOMER
0.70
7.50
EDUC
.29
4.73
AGECAT
­.
13
­
5.53
ENVINDEX
.32
7.06
USERD
.85
5.39
CWPOLD
N
=
645
.27
R2
=
.39
1.81
F
=
74.33
5­
18
Table
5.4
OLS
REGRESSION
OF
DEMOGRAPHIC
AND
ATTITUDINAL
VARIABLES
ON
WILLINGNESS
TO
PAY
AMOUNTS
FOR
THREE
LEVELS
OF
NATIONAL
WATER
QUALITY1
INTERCEPT
Levels
of
Water
Quality
eq.
1
Boatable
(
D)
eq.
2
Fishable
(
C)

Coefficient
(
t)

­
141.91
(­
3.07)
­
163.83
(­
3.03)

INCOME
AGE
.0058
(
10.36)
.0072
(
10.95)

­
1.34
.­
2.85)
­
1.84
(­
3.25)

EDUC
14.39
(
2.27)
15.15
(
2.04)

ENVINDEX
21.81
(
4.79)
28.74
(
5.40)

CWPOLD
USERD
47.90
(
3.11)
51.18
(
2.84)

27.25
(
1.71)
40.88
(
2.20)

N
695
695
R2
.28
.31
F
44.54
50.61
eq.
3
Swimmable
(
B)

­
143.47
(­
2.41)

.0075
(
10.43)

­
2.60
(­
4.16)

17.35
(
2.12)

31.77
(
5.46)

56.68
(
2.86)

45.52
(
2.23)

695
.31
51.39
1
For
Versions
A,
B,
C
combined
less
a
few
cases
which
were
dropped
for
reasons
described
in
Appendix
VI.
5­
19
Heteroskedasticity
is
to
be
expected
in
regression
equations
which
use
any
kind
of
consumer
expenditure
data
(
Prais
and
Houthacker,
1955)

and
our
estimations
are
no
exception.
Initial
tests
of
heteroskedasticity
showed
we
had
heteroskedasticity
with
respect
to
almost
very
variable.

Since
the
presence
of
heteroskedasticity
indicates
that
the
OLS
assumption
of
a
covariance
matrix
of
the
form
a31
has
been
violated,
a
generalized
least
squares
(
GLS)
procedure
must
be
used
to
obtain
correct
parameter
estimates.

(
Johnson,
1932;
Rao,
1965).
The
GLS
procedure
uses
the
covariance
matrix
2
instead
of
,
z'I.
The
GLS
estimator
of
E
is
and
the
variance
of
the
GLS
estimator
is
When
2
­
1
is
known,
estimation
of
the
GLS
estimator
is
straightfoward.

When
Q­
l
is
not
known,
special
techniques
must
be
used
to
estimate
it.

2
Standard
adjustments
such
as
weighting
by
l/
income
(
Johnson,
1972)
or
11;"
(
Goldberger,
1964)
did
not
correct
the
problem.
Since
the
standard
constructive
tests
for
heteroskedasticicy
are
not
appropriate
for
a
combination
of
dummy
and
continuous
variables
such
as
ours
(
except
for
some
maximum
likelihood
estimators
and
some
sophisticated
grouping
techniques
which
are
almost
impossible
to
implement)
we
devised
our
own
test.
Inspired
by
the
Park
test,
the
Carson­
Vaughan
constructive
test
uses
a
semilog
weight
7
transformation.

7
See
Appendix
VIII
for
an
extended
discussion.
5­
20
Table
5.5
presents
the
estimations
corrected
for
heteroskedasticity.

The
income
coefficients
and
significance
levels
are
now
20
percent
lower
than
in
the
OLS
equations.
Significance
levels
for
education
and
the
two
environmental
attitude
variables
are
also
reduced
while
those
for
age
and
recreational
use
are
increased
somewhat.

To
give
an
indication
of
price
flexibility
we
calculated
the
ranges
shown
in
Table
5.6.
The
range
is
from
moderate
inelastic
to
unitary
elasticity.
They
are
slightly
higher
but
in
the
same
general
range
as
those
found
by
Brookshire,
et
al.
(
1980:
485)
for
elk
hunting
(.
306)
and
Randall,
et
al.
(
1974:
147)
for
air
pollution
(.
39
­
.65).

Given
the
size
of
our
sample,
the
fact
that
our
explanatory
variables
are
chosen
for
their
theoretical
relevance,
and
the
cross­
sectional
character
of
data;
the
variance
explained
by
our
model
is
reasonably
high.
We
regard
this
as
important
evidence
that
the
contingent
market
described
in
our
research
instrument
is
sufficiently
realistic
to
minimize
hypothetic
bias.
5­
21
Table
5.5
1
ADJUSTED
REGRESSION
OF
DEMOGRAPHIC
AND
ATTITUDINAL
VARIABLES
ON
WILLINGNESS
TO
PAY
AMOUNTS
FOR
THREE
LEVELS
OF
NATIONAL
WATER
QUALITY
Levels
of
Water
Quality
eq.
4
Boatable
(
D)
eq.
5
Fishable
(
C)
eq.
6
Swimmable
(
B)

INTERCEPT
­
30.61
(­
1.14)
­
25.63
(.
80)
5.97
(.
17)

INCOMER
.0047
(
8.71)
.0058
(
9.06)
.0062
(
8.75)

AGE
­
1.01
(­
3.71)
­
1.48
EDUC
8.70
(
2.24)
10.37
ENVINDEX
8.42
(
3.28)
11.04
CWPOLD
30.34
(
3.09)
34.30
USERD
24.06
(
2.69)
32.92
N
2
F
695
695
.28
.32
45.02
52.82
(­
4.56)
­
2.15
(­
5.77)

(
2.25)
12.52
(
2.47)

(
3.63)
12.14
(
3.56)

(
2.97)
38.62
(
2.91)

(
3.07)
30.73
(
2.58)

695
.33
55.79
1
Data
are
adjusted
for
heteroskedasticity
by
the
Carson­
Vaughan
Constructive
Test
(
see
Appendix
VIII
for
description).
Table
5.6
5­
22
PRICE
FLEXIBILITY
OF
INCOME
Level
D
.68
­
1.06
Level
C
.70
­
1.12
Level
B
.69
­
1.12
The
high
end
of
the
range
for
the
price
flexibility
of
income
for
the
different
levels
of
water
quality
was
estimated
from
the
equation:

(
1)
Log(
Leve1
X)
=
Intercept
+
BILog(
Income)

The
low
end
of
the
range
was
estimated
from
the
equation:

(
2)
Log(
Leve1
X)
=
Intercept
+
PILog(
Income)
+
S2Educ
+

BjPge
+
Sl~
NV~
3DEx
+
B5usEm
+
S6cxPo~
D
Because
income
is
moderately
correlated
with
some
of
the
variables
in
(
2)
only
a
range
rather
than
a
point
estimate
can
be
given.
5­
23
INTRINSIC
AND
RECREATION
BENEFITS
In
Chapter
1
we
identified
direct
use
recreation
benefits
and
intrinsic
benefits
(
which
include
indirect,
option
and
existence
benefits)
as
the
subject
matter
of
our
research.
Unlike
the
Colorado
State
researchers,

we
did
not
ask
our
respondents
separate
WTP
questions
for
each
type
of
benefit
we
sought
to
measure.
We
believe
it
is
beyond
the
capability
of
many
respondents
to
reliably
determine
the
separate
value
they
have
for
sub­
categories
of
water
benefits
and
the
results
of
the
Colorado
State
study
confirm
us
in
this
belief.
Our
approach
adopts
a
different
technique
which
we
will
describe
and
illustrate
with
our
data.

At
the
heart
of
the
distinction
between
recreational
and
intrinsic
benefits
is
the
direct
use
vs.
other­
than­
direct­
use
distinction.
The
latter,
our
intrinsic
category,
includes
a
wide
array
of
benefits
ranging
from
indirect
benefits
to
duck
hunters
of
"
clean"
water
to
the
pleasure
gained
from
knowing
that
the
nation's
freshwater
bodies
have
attained
a
certain
quality
level.
Since
our
WTP
questions
measure
the
overall
value
respondents
have
for
water
quality,
the
amount
given
by
each
respondent
represents
the
combination
of
recreational
and
intrinsic
values
held
by
that
person.
We
reason
the
values
expressed
by
the
respondents
who
do
not
engage
in
in­
stream
recreation
should
be
almost
purely
intrinsic
in
nature.
In
calculating
the
average
WTP
amount
for
the
non­
recreator's
alone,
therefore,
we
get
an
approximation
of
the
intrinsic
value
o
water
quality.
By
subtracting
the
non­
recreator's
WTP
amount
from
the
total
the
recreators
are
willing
to
pay,
we
can
estimate,
in
a
rough
way,
the
portion
of
the
recreator's
benefits
which
are
attributable
to
intrinsic
values.
5­
24
Of
the
832
respondents
for
whom
we
have
use
and
WTP
data,
323
or
39
percent
reported
that
they
had
not
boated,
fished
or
swum
in
freshwater
in
the
past
two
years.
These
non­
users
gave
a
mean
WTP
amount
for
fishable
water
(
level
C)
of
$
111.
Bearing
in
mind
the
crudity
of
our
use
measurement
(
which
we
will
discuss
later)
$
111
may
be
regarded
as
an
estimate
of
the
mean
intrinsic
value
which
fishable
level
water
quality
nationwide
has
for
our
sample.
The
mean
WTP
amount
given
by
the
users
(
61
percent
of
our
sample)

was
$
237.
By
assuming
that
users
value
the
intrinsic
benefits
of
freshwater
at
the
same
level
as
the
non­
users,
we
can
subtract
$
111
from
$
237
to
arrive
at
a
mean
recreational
benefit
of
$
126
for
the
users.
By
these
calculations,

about
intrinsic
benefits
are
large;
comprising/
45
percent
of
the
benefits
for
each
user
($
111/
237);
100
percent
of
the
benefits
for
the
non­
users
($
111/
S111);
about
55
8
and/
percent
of
the
total
mean
benefit
for
the
sample
as
a
whole
($
111/$
194)
.

An
alternative
way
to
estimate
intrinsic
benefits
is
to
estimate
equation
7.

Eq.
7:
WTP
=
WTP
Total
Intrinsic
+
WTPRecreation
This
may
be
done
by
regressing
USERD
on
the
WTP
amount
for
fishable
water.

Table
5.7
gives
the
results.
Both
the
intercept
and
the
USERD
terms
are
highly
significant.
The
coefficient
of
the
intercept
may
be
interpreted
as
the
intrinsic
value.
This
amount,
$
113,
is
very
close
to
the
$
111
arrived
at
by
the
other
method.

In
an
effort
to
see
whether
it
is
possible
to
gain
insight
into
the
differential
contribution
to
the
equation
of
the
three
types
of
freshwater
use
which
comprise
the
USERD
variable,
we
estimated
equation
8
(
Table
5.8).

8
From
Table
5.1.
Table
5.7
5­
25
USER
OLS
INTRINSIC
BENEFIT
ESTIMATE
FOR
FISHABLE
WATER
QUALITY
Intercept
USERD
Coefficient
112.6
131.7
N
=
794
R2=
.06
t
7.7
7.0
F
=
49.0
USERD
=
Dummy
variable
where
1
=
personal
use
of
freshwater
for
fishing,
boating,
or
swimming
in
the
past
two
years.
Table
5.8
5­
26
BOAT,
SWIM,
FISH
OLS
INTRINSIC
BENEFIT
ESTIMATE
FOR
FISHABLE
WATER
QUALITY
Eq.
8
Coefficient
Intercept
120.1
BOAT
93.8
FISH
22.5
SWIM
75.4
N
=
792
R2=
.08
F
22.1
t
9.3
4.4
1.1
3.6
BOAT
=
Dummy
variable
where
1
=
boated
on
freshwater
in
last
two
years.

FISH
=
Dummy
variable
where
1
=
fished
in
freshwater
in
last
two
years.

SWIM
=
Dummy
variable
where
1
=
swam
in
freshwater
in
last
two
years.
5­
27
Collinearity
between
boating,
fishing,
and
swimming
precludes
making
firm
estimates
of
the
size
and
significance
of
the
coefficients
on
boating,

fishing
and
swimming,
so
we
will
only
highlight
major
differences
between
the
types
of
recreation.
9
The
intrinsic
term
(
intercept)
remains
stable
and
gains
in
significance.
However,
only
two
of
the
three
types
of
uses,

boating
and
swimming,
have
significant
t
values.
Fishing
is
not
a
good
predictor
of
the
respondent's
value
for
fishable
water,
an
anomaly
which
is
not
easy
to
interpret.
On
the
hypothesis
that
there
may
be
an
interaction
between
fishing
and
income
which
depresses
the
effect
of
fishing
use
in
an
equation
which
includes
people
from
all
income
levels,
we
reestimated
equation
8
for
each
of
our
four
income
levels.
According
to
the
t
statistics
for
this
new
estimation,
which
are
shown
in
Table
5.9,
fishing
continues
to
be
non­
significant.
A
more
detailed
analysis
of
this
question,
which
we
have
not
undertaken
at
this
point,
may
provide
clues
to
why
fishing
is
unrelated
to
people's
value
for
national
water
quality
at
the
fishable
level.

Table
5.9
also
shows
some
interesting
findings
with
respect
to
the
other
two
recreation
variables
and
the
USERD
measures.
At
the
lower
income
levels,
boating
and
swimming
have
significant
t
values
whereas
at
the
higher
two
levels
(
with
the
exception
of
swimming
for
the
highest
income
level)
the
values
are
not
significant.
Likewise,
USERD
is
strongly
9
It
may
be
possible
to
use
ridge
regression
to
arrive
at
more
accurate
parameter
estimates.
Table
5.9
5­
28
t
RATIOS
FOR
REGRESSION
OF
USE
ON
WILLINGNESS
TO
PAY
FOR
FISHABLE
WATER
(
C)
l
HOLDING
INCOME
CONSTANT
Recreational
Use
of
Water
in
last
two
years
Income
Level
USERD
/
BOAT
FISH
SWIM
I.
$
0
­
9,999
5.3
2.6
.03
2.7
II.
$
10,000
­
14,999
4.8
2.0
1.5
3.0
III.
$
15,000
­
24,999
1.9
1.4
1.3
.6
IV.
$
25,000
and
over
1.8
.8
.5
2.7
R2
for
BOAT
+
SWIM
+
FISH
(
Eq.
8)

.16
.21
.03
.07
Underlined
t
values
are
significant
at
1
.05.

1
Using
equation
8,
Table
5.8.
5­
29
significant
for
income
levels
I
and
II
and
barely
significant
for
III
and
IV.

This
suggests
that
recreational
use
is
an
important
determinant
of
the
value
lower
income
people
have
for
water
quality,
This
is
confirmed
by
the
R2s
of
.15
and
.21
for
these
regressions
(
equation
8,
for
income
levels
I
and
II
on
WTP
for
fishable
quality
water).
Using
our
regression
estimation
technique
described
earlier,
we
calculated
the
intrinsic
benefits
for
each
of
the
four
income
groups.
Table
5.10
gives
the
results
which
show
the
dominance
of
recreational
benefits
for
the
people
in
the
lower
income
categories.
Only
one­
third
of
the
WTP
amounts
expressed
by
those
in
income
levels
I
and
II
may
be
attributed
to
intrinsic
benefits
by
our
technique.

For
the
two
higher
income
groups
almost
three­
fourths
of
the
benefits
are
shown
to
be
intrinsic.

We
are
encouraged
by
these
results
which
suggest
this
approach
to
estimating
intrinsic
benefits
is
worth
pursuing
further.
In
the
Conclusion
we
propose
refinements
for
the
questionnaire
and
in
our
analytic
techniques
which
will
enable
us
to
make
reliable
intrinsic
estimates.
Table
5.10
PERCENTAGE
OF
FISHABLE
WATER
QUALITY
WTP
BENEFITS
ESTIMATED
AS
INTRINSIC
BY
INCOME
LEVEL
1
5­
30
Benefits
Intrinsic
Benefits
as
Income
Level
Intrinsic
User
Total
Percent
of
Total
Benefits
I.
$
0
­
9,999
$
30
$
172
$
102
29%

II.
$
10,000
­
14,999
47
125
172
38
III.
$
15,000
­
24,999
171
64
235
73
IV.
$
25,000
and
over
296
111
407
73
1
Versions
A,
B,
C
combined.
Estimated
using
equation
7
,
Table
5.7.
5­
31
REGIONAL
ESTIMATIONS
the
As
a
test
of
/
robustness
of
our
estimations
we
used
our
final
(
corrected)

regression
model
(
eq.
5,
p.
5­
21)
to
predict
the
regional
willingness
to
pay
for
national
water
of
fishable
quality.
To
do
this
we
substituted
the
regional
mean
value
for
the
variables
in
equation
(
5)
and
calculated
a
predicted
WT
amount
for
each
of
the
nine
census
regions.
The
actual
WTP
amount
was
calculated
for
the
same
regions.
The
two
values
are
shown
on
the
map
in
Figure
5.2.
For
all
but
two
of
the
regions
the
fit
is
very
close
and
confirms
the
stability
of
our
regression
model.
Only
in
the
Pacific
and
the
East
North
Central,
the
two
regions
with
highest
mean
WTP
amounts,
did
the
predicted
amounts
differ
by
more
than
two
standard
errors
of
the
mean
from
the
actual.
When
we
estimated
equation
(
5)
using
dummy
variables
for
eight
of
the
nine
regions,
the
distinctiveness
of
these
regions
was
confirmed
as
they
were
only
ones
with
significant
t
values.
(
The
coefficients
of
the
model's
other
variables
were
not
significantly
changed
in
the
regional
dummy
estimation.)

Although
the
difference
between
the
actual
and
expected
amounts
is
relatively
modest,
these
results
suggest
that
for
these
two
regions
one
or
more
explanatory
factors
unique
to
these
regions
may
be
at
work
in
addition
to
income,
education,
recreational
use,
concern
about
water
pollution
and
environmentalism.
However,
we
know
from
our
analysis
of
other
data
in
the
survey
that
respondents
in
these
regions
do
not
differ
significantly
from
those
in
other
regions
in
either
their
evaluation
of
the
5­
32
Figure
5.2
ACTUAL
AND
PREDICTED
WTP
AMOUNTS
FOR
NATIONAL
SWIMMABLE
WATER
QUALITY
BY
CENSUS
REGION
5­
33
quality
level
of
the
local
freshwater
or
in
their
perception
of
the
change
in
quality
of
freshwater
in
their
locality
during
the
past
five
years.
Possibly
the
presence
of
the
Great
Lakes
and
the
abundant
freshwater
resources
in
the
Michigan
penninsula
and
Wisconsin
and
the
equally
unique
water
resources
of
the
California
and
the
Pacific
Northwest
give
water
quality
a
greater
salience
for
the
residents
of
these
areas
which
translates
into
these
higher
values.

In
the
next
chapter
we
propose
a
technique
by
which
our
regional
models
may
be
used
to
estimate
water
quality
benefits
for
small
geographical
areas.
Chapter
6
CONCLUSION
AND
RECOMMENDATIONS
In
this
study
we
have
developed
and
tested
a
macro
WTP
method
for
valuing
the
benefits
of
national
water
quality.
The
advantage
of
this
method
is
the
ease
by
which
benefits
can
be
reliably
aggregated
to
the
sampling
frame,
in
our
case
the
nation.
With
one
exception
the
method
was
shown
to
be
resistant
to
the
several
biases
which
threaten
WTP
studies.
In
the
course
of
this
study
we
also
addressed
a
number
of
theoretical
and
methodological
issues
including
the
types
of
water
quality
benefits,
the
role
of
implied
property
rights
in
WTP
surveys,
the
appropriate
consumer
surplus
measures
to
use
in
WTP
studies,
the
relationship
between
strategic
and
hypothetic
bias,
the
appropriate
model
for
estimating
WTP
equations,
how
to
correct
for
heteroskedasticity
where
the
independent
variables
include
both
continuous
and
dummy
variables,
and
how
to
measure
the
intrinsic
values
of
water
quality.

Although
our
WTP
instrument
measures
a
wide
range
of
water
quality
benefits
which
accrue
to
individual
citizens,
it
does
not
measure
all
such
benefits.
Water
pollution
is
not
described
as
irreversible
in
our
contingent
market,
so
possible
long
term
personal
option
or
intergenerational
option
benefits
(
e.
g.
from
the
avoidance
of
contamination
of
water
bodies
by
certain
toxic
chemicals)
are
not
included.
Neither
are
possible
drinking
water
benefits.

One
principle
we
followed
in
designing
our
instrument
was
to
enhance
the
credibility
of
the
estimates
by
adopting
conservative
procedures
whenever
possible.
For
example,
given
a
choice
between
monthly
payments
or
an
annual
payment
we
chose
the
latter
6­
2
Table
6.1
DIRECTION
OF
BIASES
IN
THE
RFF
SURVEY
Type
of
Potential
Bias
Survey
Context
and
Construction
External
Political
Context
Environmental
Trade­
off
Questions
Vehicle
(
Taxes
and
Prices)
Payment
Schedule
(
Yearly)
Implicit
­
No
Permanent
Pollution
Damage
Zero
Encouragement
Different
Payment
Cards
Interviewer
Effects
X
Response
Inclusion
of
Protest
Zero's
Traditional
Biases
Strategic
Hypothetic
Estimation
Techniques
Maximum
Amount
Constrained
at
$
999
Substituting
amount
from
lower
level
if
amount
for
level
being
analyzed
missing
Intrinsic
Estimation
Procedure
Upward
X
?
X
Direction
of
Probable
Bias
Downward
?
X
X
?
X
?
X
X
X
X
x
X
X
Intermediate
?
X
X
None
?
indicates
uncertainty
about
whether
or
not
the
bias
is
present.
If
present,
it
is
in
the
direction
shown.
6­
3
on
the
grounds
that
it
showed
the
respondent
the
full
magnitude
of
his
or
her
value
for
water
quality
whereas
monthly
payments
might
have
induced
an
"
easy
payment
plan"
mentality.
Table
6.
l
summarizes
the
probable
biasing
effect
of
the
present
instrument's
components,
the
response
pattern,
and
our
analytic
procedures.
The
rationale
for
our
judgments
are
contained
in
the
preceeding
chapters,
especially
Chapter
4.

With
the
exception
of
the
item
nonresponse
problem,
our
goal
of
creating
a
WTP
instrument
which
is
reliable
and
credible
was
largely
fulfilled
in
this
study.
Despite
our
conservatism
in
avoiding
instrument
and
procedural
factors
which
might
bias
the
results
upwards,
respondents
express
sizable
value
for
clean
water.
A
large
fraction
of
this
value
comes
from
the
intrinsic
benefits
of
water
quality.
Yet
our
illustrative
estimates
clearly
suggest
that
the
incremental
benefits,
as
measured
by
the
WTP
methodology,
decrease
as
the
level
of
water
quality
being
evaluated
increases.

In
what
follows,
we
outline
the
modifications
in
wording,
procedure
and
analytic
techniques
which
we
have
identified
on
the
basis
of
this
experiment
as
necessary
for
a
successful
use
of
the
instrument
in
a
full
scale
national
water
benefits
survey.
We
are
confident
that
these
modifications
will
overcome.
the
item
nonresponse
problem
and
improve
the
other,
lesser,

weaknesses
in
the
present
form
of
the
instrument.
We
also
discuss
how
the
instrument
can
be
used
to
derive
sub­
national
estimates
and
to
value
other
forms
of
national
water
quality.
6­
4
Overcoming
Item
Nonresponse
Bias
Earlier
in
this
report
we
identified
item
nonresponse
bias
(
including
in
this
discussion
both
nonresponse
and
zero
bids)
as
the
major
problem
with
our
survey.
Some
item
nonresonse
is
inevitable,
of
course.
In
Chapter
4
we
argue
that
WTP
surveys
are
sufficiently
demanding
that
somewhat
higher
item
nonresponse
rates
than
normal
are
to
be
expected
(
e.
g.
10­
20
percent
range)
for
national
probability
surveys
and
that
such
item
nonresponse
(
continue)
6­
5
rates
are
tolerable.
In
our
experimental
test
the
interviewers
did
not
receive
special
instructions
nor
did
they
have
the
opportunity
to
have
their
questions
answered
by
the
researchers.
Moreover,
the
water
benefits
vehicle
was
added
on
to
an
existing
survey
instead
of
comprising
a
survey
in
its
own
right.
We
believe
these
are
the
major
reasons
for
the
high
item
nonresponse
rate.
The
following
measures
are
designed
to
reduce
the
item
nonresponse
bias
to
manageable
proportions:

A.
Field
Work
Procedures
1.
A
pre­
test
should
be
conducted
with
the
revised
instrument
of
the
survey
using
several/
research
organization's
interviewers
to
interview
50­
100
approximately
/
people.
The
interviewers
would
probe
all
item
nonresponses
and
zero
bids
to
ascertain
the
reasons
why
these
were
given.
Following
the
pre­
test
the
interviewers
would
be
debriefed
at
length.

2.
On
the
basis
of
the
pre­
test,
detailed
instructions
for
the
interviewers
would
be
prepared.
These
would
explain
the
study's
procedures
to
the
approximately
100
interviewers
who
will
do
the
final
interviewing.

3.
Since
the
interviewers
for
a
national
survey
are
scattered
across
the
country,
there
is
no
easy
way
to
brief
them
personally.

It
is
possible,
however,
to
call
each
of
them
by
phone
after
they
6­
6
have
received
the
instructional
materials,
but
prior
to
the
interviewing,
to
answer
their
questions.
The
interviewers
can
also
be
encouraged
to
call
the
researchers
collect
if
they
have
substantive
questions
about
the
instrument
which
arise
during
the
course
of
the
interviewing.

B.
Questionnaire
Modifications
1.
At
key
points
in
the
description
of
the
contingent
market,

should
the
questionnaire/
instruct
the
interviewer
to
pause
and
ask
the
respondent
"
Is
that
clear?"
"
DO
you
have
any
questions?"
This
will
encourage
respondents
to
obtain
clarification
and
maintain
an
active
interest
in
the
interview.
The
interviewer
will
be
supplied
with
a
set
of
standard
answers
most
commonly
raised
in
the
pre­
test.

C.
Aggregation
Procedures
to
the
questions
which
were
(
e.
g.
N=
2000)

If
the
national
survey
sample
is
sufficiently
large,
weighting
procedures
can
be
used
to
correct
for
the
biases
introduced
by
item
nonresponse.
Such
procedures
are
routinely
used
by
survey
research
organizations
to
correct
for
sample
nonresponse.
They
involve
the
identification
of
the
relevant
underrepresented
respondent
characteristics
(
e.
g.
old,
black)
and
the
weighting
of
those
who
did
give
responses
so
that
these
respondents
will
more
accurately
represent
the
full
sample
(
e.
g.
old
blacks
would
receive
specified
weight
greater
than
one,
young
whites
would
receive
a
weight
less
than
one,
etc.).
1
1
Holt,
et
al.,
in
a
recent
article
(
1980)
discuss
the
implications
of
using
sample
survey
data
in
regression
analysis
when
the
sample
represents
an
unequal
probability
sample.
They
warn
that
the
bias
in
the
OLS
estimator
b
can
be
large
under
these
circumstances.
On
the
basis
of
simulations
they
recommend
a
procedure
for
most
situations
involving
unequal
probability
sample
data.
Although
our
original
sample
is
an
equal
probability
sample,
because
of
the
item
non­
response
problem
our
effective
sample
for
estimating
the
WTP
amounts
is
of
the
unequal
probability
variety.
We
do
not
use
their
procedure
for
our
data
here
because
we
are
not
trying
to
make
national
estimates
at
this
point.
In
a
subsequent
survey,
however,
we
would
use
their
technique,
if
necessary,
to
correct
for
item
nonresponse.
6­
7
Intrinsic
Benefit
Estimate
We
are
encouraged
by
the
test
of
our
procedure
for
separating
intrinsic
and
recreational
benefits.
Further
refinements
are
necessary,
however,

before
we
can
reliably
estimate
intrinsic
benefits
from
macro
WTP
data
on
water
quality.
1)
Because
of
space
limitations
in
our
questionnaire,
we
limited
our
use
questions
to
the
respondents'
own
experience.
But
our
unit
of
analysis
is
the
household,
not
the
individual
respondent.
Someone
who
does
not
use
freshwater
directly,
but
who
is
married
to
someone
who
does,
may
value
freshwater
quality
for
its
contribution
to
his
or
her
spouse's
enjoyment.

2)
Our
procedure
for
estimating
an
intrinsic
value
for
the
entire
sample
is
oversimplified.
If
non­
users
were
randomly
distributed
among
the
sample
our
device
of
proceeding
directly
from
the
mean
WTP
amount
for
the
non­
users
to
inferring
the
intrinsic
value
of
a
water
quality
level
/
for
the
entire
sample
would
be
defensible.
However,
non­
users
are
not
so
distributed,
but
are
differentially
older
and
black,
for
example,
In
general,
older
people
and
blacks
tend
to
give
lower
WTP
amounts
than
younger
people
and
whites.

It
is
necessary,
therefore,
to
devise
weighting
procedures
based
on
a
comparison
of
the
WTP
smounts
for,
say,
older
users
vs.
older
non­
users,

to
corect
for
this
bias.

3)
Households
who
do
not
currently
use
freshwater
for
recreation
should
be
asked
a
question
about
intended
future
recreational
water
use.
This
will
provide
useful
option
value
information.
6­
8
4)
Questions
need
to
be
asked
about
the
availability
and
use
of
substitute
sources
of
water
for
recreation.
Respondents
who
own
swimming
pools
or
who
belong
to
swimming
clubs
may
value
swimmable
freshwater
less
than
those
who
do
not
have
access
to
such
facilities,

Other
Refinements
and
Techniques
The
strong
correlation
between
the
regional
WTP
estimates
from
our
national
WTP
equation
(
eq.
5­
31­
33
above)
and
the
actual
regional
WTP
amounts
suggest
that
a
scheme
can
be
devised
to
estimate
water
benefits
for
sub­
national
geographic
areas.
Such
a
scheme
would
work
approximately
as
follows:
1)
A
new
(
presumably
more
predictive)
national
benefits
equation
would
be
estimated
from
a
large
national
survey.
2)
Census
data
would
be
used
to
supply
the
area
mean
values
for
the
demographic
variables
of
the
equation
(
e.
g.
income,
education).
3)
A
low
cost
area
telephone
survey
could
measure
the
attitudinal
variables
for
the
equation.

4)
Local
benefits
would
then
be
estimated
using
these
data
and
the
coefficients
from
the
national
equation.
Procedures
would
have
to
be
devised
to
determine
the
correct
apportionment
of
local
and
national
benefits
and
the
appropriate
aggregation
procedure
for
people
and
water
bodies.
One
procedure
for
the
former
is
to
do
a
pilot
regional
or
local
WTP
study
parallel
with
the
national
survey.

In
the
present
study
we
value
a
uniform
level
of
national
water
quality
by
referring
to
the
"
nation's
overall
water
quality
at
level
x
where
virtually
all
of
it
is
at
least
clean
enough
for
x."
Our
method
can
be
adapted
6­
9
to
value
alternative
supply
options
such
as
"
all
the
nation's
waterbodies
except
for
x,
y,
and
z"
with
the
respondent
being
shown
a
map
depciting
the
probable,
location
of
those
waterbodies
which
would
not
meet
a
specified
level.

A
final
refinement,
which
is
applicable
to
WTP
surveys
of
all
kinds,

is
to
ask
a
series
of
questions
to
measure
the
respondents'
firmness
of
opinion
about
his
or
her
WTP
amount.
These
questions
would
show
whether
or
not
the
contingent
market
and
WTP
question
sequence
create
a
sufficiently
meaningful
situation
for
the
respondent.
The
answers
to
these
items
would
provide
an
overall
evaluation
of
the
instrument's
realism
(
and
of
the
danger
of
hypothetic
bias).
They
may
also
be
used
to
identify
individual
respondents
who,
although
they
gave
answers,
really
did
not
have
sufficiently
firm
opinions
to
warrant
the
inclusion
of
their
responses
in
the
analysis.

The
survey
research
form
of
Yankelovich,
Skelly
and
White
have
devised
and
tested
what
they
call
a
"
mushiness
index"
which
can
be
adapted
to
this
purpose.
2
According
to
them:
"
Answers
to
survey
questions
on
such
issues
(
ones
that
are
not
'
thought
through')
are
often
top­
of­
the­
head
and
subject
to
change."
Mushiness
describes
the
volatility
and
changeability
of
the
public's
views.
(
Public
Opinion,
1981:
50).
In
the
RFF
instrument
we
experimented
with
a
single
quality
check
item
which
is
similar
2
We
recommend
including
three
of
the
four
items
in
the
YSW
scale.
These
measure:
1)
the
degree
of
personal
involvement
in
the
issue,
2)
whether
the
person
feels
he
or
she
has
enough
information
about
it
and
3)
the
firmness
with
which
the
person
holds
his
or
her
views.
The
wording
is
contained
in
Public
Opinion
­
(
1981:
50).
6­
10
to
one
of
the
indicators
in
the
Yankelovich,
Skelly
and
White
scale.
(
We
were
only
able
to
include
it
in
two
of
the
four
versions
of
our
questionnaire,

A
and
C).
The
results
of
this
item,
which
asked
people
whether
we
had
supplied
them
with
enough
information
so
that
they
could
decide
how
much
they
would
be
willing
to
pay
for
better
water
quality,
were
encouraging.

Only
12
percent
said
they
did
not
have
"
enough
(
information)
at
all"
while
56
percent
said
they
had
"
about
enough"
or
"
more
than
enough"
(
14
percent).

Twenty­
two
percent
said
they
had
"
not
quite
enough."
R­
1
REFERENCES
Alwin,
Duane
F.
1977.
"
Making
Errors
in
Surveys,
an
Overview,"
Sociological
Methods
&
Research,
vol.
6,
no.
3,
November,
pp.
131­
150.

Amemiya,
Takeshi,
1373.
"
Regression
Analysis
When
the
Dependent
Variable
is
Truncated
Normal,"
Econometrica,
vol.
41,
pp.
997­
1016.

Babbie,
Earl
R.
1973.
Survey
Research
Methods
(
Belmont,
CA.,
Wadsworth
Publishing
Co.).

Backstrom,
Charles
H.
and
Gerald
D.
Hursh.
1963.
Survey
Research
(
Evanston,
Ill.,
Northwestern
University
Press),

Baumol,
William
J.
and
Wallace
E.
Oates,
1979.
Economics,
Environmental
Policy
and
The
Quality
of
Life
Binkley,
Clark
S.
and
W.
Michael
Hanemann.
1978.
"
The
Recreation
and
Benefits
of
Water
Quality
Improvement:
Analysis
of
Day
Trips
in
An
Urban
Setting,"
(
Washington,
D.
C.,
U.
S.
E.
P.
A.).

Bishop,
Richard
C.
and
Thomas
A.
Heberlein,
1980.
"
Simulated
Markets,
Hypothetical
Markets,
and
Travel
Cost
Analysis:
Alternative
Methods
of
Estimating
Outdoor
Recreation
Demand,"
Journal
of
American
Agricultural
Economics,
December
(
No.
187,
Agricultural
Economics
Staff
Paper
Series).

Blank,
Frederick
M.,
et
al.,
1978.
"
Valuation
of
Aesthetic
Preferences:
A
Case
Study
of
the
Economic
Value
of
Visibility,"
(
paper
prepared
by
Resource
and
Environmental
Economics
Laboratory,
University
of
Wyoming,
under
contract
with
Electric
Power
Research
Institute,
Palo
Alot,
CA),
October
6.

Bohm,
Peter,
1979.
"
Estimating
Willingness
to
Pay:
Why
and
How?"
Scandinavian
Journal
of
Economics,
vol.
81,
no.
2.

Bohm,
Peter.
1972.
"
Estimating
Demand
for
Public
Goods:
An
Experiment,"
European
Economic
Review,
vol.
3,
pp.
111­
130.

Bradford,
David
F.
1970.
"
Benefit­
Cost
Analysis
and
Demand
Curves
for
Public
Goods,"
Kyklos,
vol.
23,
pp.
775­
791.

Brannon,
R.,
et
al.,
1973.
"
Attitude
and
Action:
A
Field
Experiment
Joined
to
a
General
Population
Survey,"
American
Sociological
Review,
vol.
38,
pp.
625­
36.

Brookshire,
David
S.,
Ralph
C.
d'Arge,
and
William
D.
Schulze,
1979.
"
Methods
Development
for
Assessing
Tradeoffs
in
Environmental
Management,"
vol.
2,
U.
S.
E.
P.
A.
(
available
from
National
Technical
Information
Service).
R­
2
Berry
C.
Ives,
and
William
D.
Schulze.
1976.
"
The
Valuation
of
Aesthetic
Preferences,"
Journal
of
Environmental
Economics
and
Management,"
vol.
3,
no.
4,
pp.
325­
346.

Alan
Randall
and
John
R.
Stoll.
1980.
"
Valuing
Increments
and
Decrements
in
Natural
Resource
Service
Flows,"
American
Journal
of
Agricultural
Economics,
vol.
62,
no.
3,
pp.
478­
488.

Mark
A.
Thayer,
William
D.
Schulze
and
Ralph
d'Arge,
n.
d.
"
Valuing
Public
Goods:
A
Comparison
of
Survey
and
Hedonic
Approaches,"
(
research
funded
under
grant,
Methods
Development
for
Assessing
Air
Pollution
Control
Benefits)
(
U.
S.
E.
P.
A.).

Burch,
William
R.,
Neil
H.
Cheek,
and
Lee
Taylor,
1972.
Social
Behavior,
Natural
Resources
and
the
Environment
(
New
York,
Harper
&
Rowe).

Burington,
Richard
Stevens
and
Donald
Curtis
May.
1958.
Handbook
of
Probability
and
Statistics
with
Tables
(
Sandusky,
Ohio,
Handbook
Publishers,
Inc.).

Cambridge
Reports.
1978.
"
Public
and
Worker
Attitudes
Toward
Carcinogens
and
Cancer
Risk,"
(
prepared
for
Shell
Oil
Company).

Campbell,
D.
T.,
C.
R.
Seigman
and
M.
B.
Bees,
1967.
"
Direction
of
Wording
Effects
in
the
Relationships
Between
Scales,"
Psychological
Bulletin,
no.
68,
November,
pp.
292­
303.

Carr,
L.
1971.
"
The
Srole
Items
and
Acquiescence,"
American
Sociological
Review,
vol.
35,
April,
pp.
287­
293.

Charbonneau,
J.
John
and
Michael
J.
Hay,
1978.
"
Determinants
and
Economic
Values
of
Hunting
and
Fishing,"
(
paper
presented
at
43rd
North
American
Wildlife
and
Natural
Resources
Conference,
Phoenix,
Arizona,
March
18­
22).

Cicchetti,
Charles
J.,
Anthony
C.
Fisher
and
V.
Kerry
Smith,
1976.
"
An
Econometric
Evaluation
of
a
Generalized
Consumer
Surplus
Measure:
The
Mineral
King
Controversy,"
Econometrica,
vol.
44,
November,
pp.
1269­
1276.

Cicchetti,
Charles
J.,
Robert
K.
Davis,
Steve
H.
Hanke,
Robert
H.
Haveman,
and
Jack
L.
Knetsch.
1972.
"
Benefits
or
Costs?
An
Assessment
of
the
Water
Resources
Council's
Proposed
Principles
and
Standards,"
(
pamphlet
available
from
Department
of
Geography
&
Environmental
Engineering,
Johns
Hopkins
University).

Cicchetti,
Charles
J.
and
A.
Myrick
Freeman
III.
1971.
"
Option
Demand
and
Consumer's
Surplus:
Further
Comment,"
Quarterly
Journal
of
Economics
vol.
85,
no.
3,
August,
pp.
528­
539.
R­
3
Cohen,
A.
C.
Jr.,
1957.
"
On
the
Solution
of
Estimating
Equations
for
Truncated
and
Censored
Samples
from
Normal
Populations,"
Biometrika,
vol.
44,
pp.
225­
236.

1950.
"
Estimating
the
Mean
and
Variance
of
Normal
Populations
from
Singly
and
Doubly
Truncated
Samples,"
Annals
of
Mathematical
Statistics,
vol.
21,
pp.
557­
569.

Couch,
A.
and
K.
Keniston,
1960.
"
Yeasayers
and
Naysayers:
Agreeing
Response
Set
as
a
Personality
Variable,"
Journal
of
Abnormal
and
Social
Psychology
vol.
60,
March,
pp.
151­
174.

Council
on
Environmental
Quality.
1980.
"
Public
Opinion
on
Environmental
Issues,"
(
Washington,
D.
C.,
Council
on
Environmental
Quality).

Crespi,
Irving.
1981.
"
Reporting
the
Polls"
Public
Opinion
vol.
4,
no.
2,
April/
May,
pp.
48­
50.

Curie,
John
Martin,
John
A.
Murphy
and
Andrew
Schmitz.
1971.
"
The
Concept
of
Economic
Surplus
and
Its
Use
in
Economic
Analysis."
The
Economic
Journal,
vol.
81,
no.
324.
December,
pp.
741­
799.

d'Arge,
Ralph,
W.
Schulze,
and
D.
Brookshire,
1980.
"
Benefit­
Cost
Valuation
of
Long
Term
Future
Effects:
The
Case
of
CO
2"
(
paper
prepared
for
RFF/
National
Climate
Program
Office
Workshop,
April
24­
25,
Ft.
Lauderdale,
Florida).

Darling,
Arthur
H.
1973.
"
Measuring
Benefits
Generated
by
Urban
Water
Parks,"
Land
Economics,
vol.
49,
no.
1,
February,
pp.
22­
34.

David,
Elizabeth
L.
1971.
"
Public
Perceptions
of
Water
Quality,"
Water
Resources
Bulletin,
vol.
7,
no.
1,
June,
pp.
453­
457.

Davis,
Robert
K.
1980.
"
Analysis
of
the
Survey
to
Determine
the
Effects
of
Water
Quality
on
Participation
in
Recreation,"
(
memorandum
to
John
Parson,
National
Capital
Region,
National
Park
Service,
(
Washington,
D.
C.,
U.
S.
Department
of
the
Interior)
July
28.

1964.
"
The
Value
of
Big
Game
Hunting
in
a
Private
Forest,"
(
Transactions
of
the
29th
North
American
Wildlife
and
Natural
Resources
Conference,
March),
(
Washington,
D.
C.,
Wildlife
Management
Institute).

1963.
"
Recreation
Planning
as
an
Economic
Problem,"
Natural
Resources
Journal,
vol.
3,
no.
2,
October,
pp.
239­
249.

Dillman,
Don
A.
1978.
Mail
and
Telephone
Surveys
­
The
Total
Design
Method,
(
New
York,
John
Wiley
&
Sons).
R­
4
Dohrenwend,
B.
P.
1966.
"
Social
Status
and
Psychological
Disorder:
An
Issue
of
Substance
and
an
Issue
of
Method,"
American
Sociological
Review,
vol.
31,
February,
pp.
14­
34.

Dunlap,
Riley
E.
and
Kent
D.
Van
Liere.
1978.
"
Environmental
Concern:
A
Bibliography
of
Empirical
Studies
and
Brief
Appraisal
of
the
Literature,"
(
Monticello,
Illinois,
Vance
Bibliographies).

Dunlap,
Thomas
R.,
1981.
DDT:
Scientists,
Citizens
and
Public
Policy
(
Princeton,
N.
J.,
Princeton
University
Press).

Dutta,
M.
1975.
Econometric
Method
(
Cincinnati,
Ohio,
South­
Western
Publishing
Co.).

Eastman,
Clyde,
Alan
Randall,
and
Peggy
L.
Hoffer,
1974,
"
How
Much
To
Abate
Pollution,"
Public
Opinion
Quarterly,
pp.
575­
584.

1978.
"
A
Socioeconomic
Analysis
of
Environmental
Concern:
Case
of
the
Four
Corners
Electric
Power
Complex,"
Bulletin
626
Agricultural
Experiment
Station.

Edwards,
A.
1957.
The
Social
Desirability
Variable
in
Personality
Assessment
and
Research,
(
New
York,
Dryden).

Federal
Energy
Administration,
1977.
"
The
Surveys
of
Public
Attitudes
and
Response
to
Federal
Energy
Policy,"
(
data
prepared
by
Opinion
Research
Corporation;
available
from
Inter­
University
Consortium
for
Political
and
Social
Research,
Ann
Arbor,
Michigan).

Feenberg,
Daniel
and
Edwin
S.
Mills.
1980.
Measuring
the
Benefits
of
Water
Pollution
Abatement
(
New
York,
Academic
Press).

Fischer,
David
W.
1974.
"
Willingness
to
Pay
as
a
Behavioural
Criterion
for
Environmental
Decision­
Making,"
Journal
of
Environmental
Management,
vol.
3,
pp.
29­
41.

Fish
and
Wildlife
Service,
U.
S.
Department
of
Interior.
n.
d.
"
1975
National
Survey
of
Hunting,
Fishing
and
Wildlife
­
Associated
Recreation,"
(
Washington,
D.
C.,
U.
S.
Department
of
Interior).

Free,
Lloyd
A.
and
Hadley
Cantril.
1967.
The
Political
Beliefs
of
Americans
­
A
Study
of
Public
Opinion,
(
New
Brunswick,
N.
J.,
Rutgers
University
Press).
R­
5
Freeman,
A.
Myrick.
1979a.
"
The
Benefits
of
Air
and
Water
Pollution
Control:
A
Review
and
Synthesis
of
Recent
Estimates,"
(
report
prepared
for
Council
on
Environmental
Quality),
December.

1979b.
The
Benefits
of
Environmental
Improvement
­
Theory
and
Practice
(
Baltimore,
Md.,
Johns
Hopkins
University
Press
for
Resources
for
the
Future).

Freund,
John
E.
and
Ronald
E.
Walpole.
1980.
Mathetmatical
Statistics
(
Englewood
Cliffs,
N.
J.,
Prentice­
Hall).

Fuller,
Carol
H.
1974.
"
Weighting
to
Adjust
for
Survey
Nonresponse,"
Public
Opinion
Quarterly
vol.
38,
no.
2,
pp.
239­
246.

Glejser,
H.
1969.
"
A
New
Test
for
Heteroskedasticity,"
Journal
of
the
American
Statistical
Association,
vol.
64,
pp.
316­
323.

Goldberger,
Arthur
S.
1964.
Econometric
Theory
(
New
York,
John
Wiley
&
Sons).

Goldfeld,
S.
M.
and
R.
E.
Quant,
1972.
"
Nonlinear
Methods
in
Econometrics,"
in
D.
W.
Jorgenson
and
J.
Waelbroeck,
eds.
Contributions
to
Economic
Analysis
vol.
77,
(
Amsterdam,
North­
Holland).

and
1965.
"
Some
Tests
for
Homoskedasticity,"
Journal
of
the
American
Statistical
Association,
vol.
60,
pp.
539­
47.

Gramlich,
Frederick
W.
1977.
"
The
Demand
for
Clean
Water:
The
Case
of
the
Charles
River,"
National
Tax
Journal,
vol.
30,
no.
2,
June,
pp.
183­
194.

Greenley,
Douglas
A.,
Richard
G.
Walsh,
and
Robert
A.
Young.
1980.
"
Option
Value:
Empirical
Evidence
from
a
Case
Study
of
Recreation
and
Water
Quality,"
Quarterly
Journal
of
Economics,
forthcoming.

Harris,
Louis
and
Associates.
1969.
"
A
Study
of
the
Attitudes
of
the
American
Public
Toward
Improvement
of
the
Natural
Environment,"
(
prepared
for
the
National
Wildlife
Federation).

Haveman,
Robert,
1975.
"
Discussion
of
Dennis
Tihansky's
Paper,
'
A
Survey
Empirical
Benefit
Studies,"'
in
Henry
M.
Peskin
and
Eugene
P.
Seskin,
eds.
Cost
Benefit
Analysis
and
Water
Pollution
Policy
(
Washington,
D.
C.,
The
Urban
Institute),
pp.
166­
167.

Hay,
Michael
J.
and
Kenneth
McConnell,
1979.
"
An
Analysis
of
Participation
in
Nonconsumptive
Wildlife
Recreation,"
Land
Economics
vol.
55,
November,
pp.
460­
471.
R­
6
Henderson,
James
M.
and
Richard
E.
Quant.
1971.
Microeconomic
Theory:
A
Mathematical
Approach,
(
New
York,
McGraw­
Hill).

Henry,
Claude,
1974.
"
Option
Values
in
the
Economics
of
Irreplaceable
Assets,"
The
Review
of
Economic
Studies
Symposium,
pp.
89­
104.

Hicks,
J.
R.
1956.
A
Revision
of
Demand
Theory
(
Oxford,
Clarendon
Press).

1943.
"
The
Four
Consumer
Surpluses,"
Review
of
Economic
Studies,
vol.
11,
pp.
31­
41.

1941.
"
The
Rehabilitation
of
Consumer's
Surplus,"
Review
of
Economics
Studies
vol.
8.

1939.
"
Foundations
of
Welfare
Economics,
"
Economic
Journal,
vol.
43.

Hogg,
Robert
V.
and
Allen
T.
Craig.
1978.
Introduction
to
Mathematical
Statistics
(
New
York,
MacMillan
Publishing
Co.).

Holt,
D.,
T.
M.
F.
Smith
and
P.
D.
Winter,
1980.
"
Regression
Analysis
of
Data
from
Complex
Surveys,"
J.
R.
Statist.
Soc.
A
143,
part
4,
pp.
474­
487.

Jackman,
M.
R.
1973.
"
Education
and
Prejudice
or
Education
and
Responseset
American
Sociological
Review,
vol.
38,
June,
pp.
327­
339.

Johnston,
J.
1972.
Econometric
Methods,
(
New
York,
McGraw­
Hill).

Katona,
George,
1980.
Essays
on
Behavioral
Economics,
(
Ann
Arbor,
Michigan,
Survey
Research
Center,
University
of
Michigan).

Kelly,
S.
and
T.
W.
Mirer,
1974.
"
The
Simple
Act
of
Voting,"
American
Political
Science
Review,
vol.
68,
pp.
572­
91.

Kmenta,
Jan.
1971.
Elements
of
Econometrics
(
New
York,
MacMillan).

Knetsch,
Jack
L.
and
Robert
K.
Davis.
1966.
"
Comparisons
of
Methods
for
Recreation
Evaluation,"
in
Allen
V.
Kneese
and
Stephen
C.
Smith,
eds,
Water­
Research
(
Baltimore,
Johns
Hopkins
University
Press),
pp.
125­
142,

Krutilla,
John
V.
1967.
"
Conservation
Reconsidered,"
The
American
Economic
Review,
vol.
57,
pp.
787­
786.

and
Anthony
C.
Fisher,
1975.
The
Economics
of
Natural
Environments:
Studies
in
the
Valuation
of
Commodity
and
Amenity
Resources,
(
Baltimore,
Johns
Hopkins
University
Press
for
Resources
for
the
Future).
R­
7
Kurz,
Mordecai.
1974.
"
Experimental
Approach
to
the
Determination
of
the
Determination
of
the
Demand
for
Public
Goods,"
Journal
of
Public
Economics,
vol.
3,
pp.
329­
348.

Ladd,
Everett
C.
and
G.
Donald
Ferree,
1981.
"
Were
the
Pollsters
Really
Wrong?"
Public
Opinion,
vol.
3
no.
6,
December/
January,
pp.
13­
20.

Maddala,
G.
S.
1977.
Econometrics
(
New
York,
McGraw­
Hill).

Maler,
Karl­
Goran.
1974.
Environmental
Economics:
A
Theoretical
Inquiry
(
Baltimore,
Johns
Hopksins
University
Press
for
Resources
for
the
Future).

Malinvaud,
E.
1980.
Statistical
Methods
of
Econometrics
(
Amsterdam,
North­
Holland).

Mishan,
E.
S.
1976a.
Cost
Benefit
Analysis,
2nd
edition
(
New
York,
Praeger).

1976b.
"
The
Use
of
Compensating
and
Equivalence
Variations
in
Cost­
Benefit
Analysis,"
Economica,
vol.
43,
pp.
185­
97.

1971.
Cost­
Benefit
Analysis,
1st
edition
(
New
York,
Praeger).

1960.
"
A
Survey
of
Welfare
Economics,
1939­
1959,"
Economic
Journal,
vol.
70,
June.

1947.
"
Realism
and
Relevance
in
Consumer's
Surplus,"
Review
of
Economics
Studies,
vol.
15.

Mitchell,
Robert
Cameron.
1980.
"
Polling
on
Nuclear
Power:
A
Critique
of
the
Polls
After
Three
Mile
Island,"
in
Albert
H.
Cantril,
ed.
Polling
on
the
Issues
(
Washington,
D.
C.,
Seven
Locks
Press).

Mulligan,
Patricia
J.
1978.
"
Willingness
to
Pay
for
Decreased
Risk
From
Nuclear
Plant
Accidents,"
(
Working
Paper
#
43,
Center
for
the
Study
of
Environmental
Policy,
Pennsylvania
State
University).

Park,
R.
E.
1966.
"
Estimation
with
Heteroskedastic
Error
Terms."
Econometrica,
vol.
34,
no.
4,
October,
p.
888.

Phillips,
D.
L.
and
K.
J.
Clancy,
1970.
"
Response
Bias
in
Field
Studies
of
Mental
Illness,"
American
Sociological
Review,
vol.
35,
June,
pp.
503­
515.

and
1972.
"
Some
Effects
of
Social
Desirability
in
Survey
Studies,"
American
Journal
of
Sociology,
vol.
77,
March,
pp.
921­
940.
R­
8
Prais,
S.
J.
and
H.
S.
Houthakker,
1955.
The
Analysis
of
Family
Budgets
(
New
York,
Cambridge
University
Press).

Rao,
C,
Radhakrishna,
1973.
Linear
Statistical
Inference
and
Its
Applications
(
New
York,
John
Wiley
&
Sons).

Randall,
Alan
and
John
R.
Stoll.
1980.
"
Consumer's
Surplus
in
Commodity
Space,"
American
Economic
Review,
vol.
70,
no.
3,
pp.
449­
455.

Berry
Ives
and
Clyde
Eastman.
1974.
"
Bidding
Games
for
Valuation
of
Aesthetic
Environmental
Improvements,"
Journal
of
Environmental
Economics
and
Management,
vol.
1,
pp.
132­
149.

Orlen
Grunewald,
Angefos
Pagoulatos,
Richard
Ausness,
and
Sue
Johnson,
1978.
"
Estimating
Environmental
Damages
from
Surface
Mining
of
Coal
in
Appalachia:
A
Case
Study,"
(
Washington,
D.
C.,
U.
S.
E.
P.
A.).

and
1978.
"
Reclaiming
Coal
Surface
Mines
in
Central
Appalachia:
A
Case
Study
of
the
Benefits
and
Costs."
Land
Economics,
vol.
54,
no.
4,
November,
pp.
472­
489.

Rosenbaum,
Walter
A.
1977.
"
Slaying
Beautiful
Hypotheses
with
Ugly
Facts:
EPA
and
the
Limits
of
Public
Participation,"
Journal
of
Voluntary
Action
Research,
vol.
6,
nos.
3­
4
(
Summer­
Fall),
pp.
161­
174.

Rowe,
Robert,
Ralph
C.
d'Arge,
and
David
S.
Brookshire,
1979.
"
An
Experiment
on
the
Economic
Value
of
Visibility,"
(
available
from
Department
of
Economics,
University
of
Wyoming,
Laramie).

1979.
"
Environmental
Preferences
and
Effluent
Charges,"
Progress
in
Resource
Management
and
Environmental
Planning,"
2nd
vol.
forthcoming.
and
Samuelson,
Paul.
1955.
"
Diagrammatic
Exposition
of
a
Theory
of
Public
Expenditure,"
Revue
of
Economics
and
Statistics,
vol.
37,
pp.
350­
356.

1954.
"
The
Pure
Theory
of
Public
Expenditure,"
The
Review
of
Economics
and
Statistics,
vol.
36,
pp.
387­
89.

1947.
Foundations
of
Economic
Analysis
(
New
York,
Atheneum,
1974).

Schulze,
William
D.,
Ralph
C.
d'Arge,
and
David
Brookshire,
1980,
"
Valuing
Environmental
Commodities:
Some
Recent
Experiments,"
(
prepared
for
U.
S.
E.
P.
A.)
(
available
from
Department
of
Economics,
University
of
Wyoming,
Laramie).
R­
9
Schuman,
Howard
and
Michael
P.
Johnson.
1976.
"
Attitudes
and
Behavior,"
in
Alex
Inkeles,
et
al.,
eds.
Annual
Review
of
Sociology,
vol.
2,
(
Palo
Alto,
Ca.,
Annual
Reviews,
Inc.)

Silverburg,
Eugene,
1978.
The
Structure
of
Economics:
A
Mathematical
Analysis
(
New
York,
McGraw
Hill).

Smith,
V.
L.
1977,
"
The
Principle
of
Unanimity
and
Voluntary
Consent
in
Social
Choice,"
Journal
of
Political
Economy,
vol.
85,
no.
6,
December.
pp.
1125­
1139.

Steeh,
Charlotte
G.,
1981.
"
Trends
in
Nonresponse
Rates,
1952­
1979,"
Public
Opinion
Quarterly,
vol.
45,
pp.
40­
57.

Stouffer,
S.
A.,
et
al.,
1949.
The
American
Soldier:
Combat
and
Its
Aftermath
vol.
2,
(
Princeton,
N.
J.,
Princeton
University
Press).

Sudman,
Seymour,
1976.
Applied
Sampling
(
New
York,
Academic
Press).

Thayer,
Mark
A.
forthcoming.
"
Contingent
Valuation
Techniques
for
Assessing
Environmental
Impacts:
Further
Evidence,"
Journal
of
Environmental
Economics
and
Management.

Tihansky,
Dennis.
1975.
"
A
Survey
of
Empirical
Benefit
Studies,"
in
Henry
M.
Peskin
and
Eugene
P.
Seskin,
eds.
Cost­
Benefit
Analysis
&
Water
Pollution
Policy,
(
Washington,
D.
C.,
The
Urban
Institute),
pp.
127­
172.

U.
S.
Congress.
1977.
"
The
Federal
Water
Pollution
Control
Act
(
P.
L.
92­
500)
as
amended
by
the
Clean
Water
Act
of
1977
in
95th
Congress,
Serial
$
95­
12
(
Washington,
D.
C.,
U.
S.
GPO).

Viladus,
Joseph.
1974.
"
Poll
Finds
Americans
Concerned
About
Environment,"
National
Journal
Reports,
vol.
6,
pp.
634­
635.

1973.
"
The
American
People
and
Their
Environment,"
vol.
1
(
Washington,
D.
C.,
U.
S.
E.
P.
A.).

Walsh,
Richard
A.,
D.
A.
Greenley,
R.
A.
Young,
J.
R.
McKean,
and
A.
A.
Prato.
1978.
"
Option
Values,
Preservation
Values
and
Recreational
Benefits
of
Improved
Water
Quality:
A
Case
Study
of
the
South
Platte
River
Basin,
Colorado,"
(
U.
S.
E.
P.
A.)
(
available
from
NTIS).
R­
10
Weisbrod,
Burton
A.
1964.
"
Collective
Consumption
Services
of
Individual­
Consumption
Goods,"
Quarterly
Journal
of
Economics,
vol.
78,
no.
3,
August,
pp.
471­
477,

Williams,
William
H.
1978.
A
Sampler
on
Sampling,
(
New
York,
John
Wiley
and
Sons).

Willig,
Robert
D.
1976.
"
Consumer's
Surplus
Without
Apology,"
American
Economic
Review,
vol.
66,
no.
4,
September,
pp.
587­
597.

Wyckoff,
J.
B.
1971.
"
Measuring
Intangible
Benefits
­
Some
Needed
Research,"
Water
Resources
Bulletin,
vol.
7,
no.
1,
February,
pp.
11­
16.
Appendix
I
THE
RFF
RESEARCH
INSTRUMENT
A­
I­
1
STUDY
#
684
NOTE:
INSERT
THIS
FORM
AFTER
PAGE
14
OF
WHITE
"
X"
Now
Let's
think
about
all
of
the
nation's
QUESTIONNAIRES
ONLY
AND
ASK
FOLLOWING
Q.
79.
rivers
,
lakes
and
streams.
Some
of
them
are
quite
clean
and
others
are
more
or
less
80.
This
last
group
of
questions
is
about
the
quality
of
water
polluted.
Looking
at
this
Ladder,
would
in
the
nation's
lakes
and
streams.
Conqress
passed
strict
you
say
that
all
but
a
tiny
fraction
of
the
water
pollution
control
laws
in
1972
and
1977.
As
a
re­
nation's
rivers,
lakes
and
streams
are
at
sult
many
communities
have
to
build
and
run
new
modern
least
at
level
D
in
the
quality
of
their
sewage
treatment
plants
and
many
industries
have
to
install
water
today
or
not?
water
pollution
control
equipment.

Here
is
a
picture
of
a
ladder
that
shows
various
levels
of
the
quality
of
water.
(
HAND
RESPONDENT
WATER
QUALITY
LADDER
CARD)
Please
keep
in
mind
that
we
are
not
talking
about
the
drinking
water
in
your
home.
Nor
are
we
talking
about
the
ocean.
We
are
talking
only
freshwater
lakes,
rivers
and
streams
that
people
look
at
and
in
which
they
go
boating,
fishing
and
swimming.

The
top
of
top
ladder
stands
for
the
best
possible
quality
of
water,
that
is,
the
purest
spring
water.
The
bottom
stands
for
the
worst
possible
quality
of
water.
Unlike
the
other
ladders
we
have
used
in
this
survey,
on
this
ladder
we
have
marked
different
levels
of
the
quality
of
water.
For
example....
(
POINT
TO
EACH
LEVEL:
E,
D,
C
AND
SO
ON,
AS
YOU
READ
STATEMENTS
BELOW)

Level
E
(
POINTING)
is
so
polluted
that
it
has
oil,
raw
sewage
and
other
things
in
it,
has
no
plant
or
animal
life
and
smells
bad
Level
B
shows
where
the
water
is
clean
enough
so
that
people
can
swim
in
it
safely
And
at
level
A,
the
quality
of
the
water
is
so
good
that
it
would
be
possible
to
drink
it
directly
from
a
lake
or
scream
if
you
wanted
to
Water
at
level
D
is
okay
for
boating
but
not
for
fishing
or
swimming
Level
C
shows
where
rivers,
lakes
and
streams
are
clean
enough
so
that
game
fish
like
bass
can
live
in
them
All
but
a
fraction
at
level
D..
.
1
Not
at
level
D..................
2
Not
sure........................
3
81.
As
you
know
it
takes
money
to
clean
up
our
nation's
lakes
and
rivers.
Taking
that
into
account,
and
thinking
of
overall
water
quality
where
all
but
a
tiny
fraction
of
an
nation's
lakes
and
rivers
are
at
a
particul
level,
which
level
of
overall
water
quality
do
you
think
the
nation
should
plan
to
reach
within
the
next
five
years
or
so­­
level
E,
D,
C,
B
or
A?

A
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1
B
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.2
C
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3
D
..................
4
E
..................
5
Depends
(
vol.).....
6
Other
(
vol.)
.
.
.
.
.
.
.
7
Not
sure...........
8
IF
$
25,000
AND
ABOVE
OR
NOT
SURE/
REFUSED
USE
CARD
A­
IV
IF
$
15,000
TO
$
24,999
USE
CARD
A­
III
NAIRE.
THEN
LOOK
BELOW
TO
SEE
WHICH
SCALE
CARD
RESPONDENT
USES
IN
QUESTIONS
82
­
84.

IF
LESS
THAN
$
9,999
USE
CARD
A­
I
INTERVIEWER:
CHECK
INCOME
IN
Q.
79
ON
PAGE
14
OF
MAIN
QUESTION­

IF
$
10,000
TO
$
14,999
USE
CARD
A­
II
(
OVER)
A­
I­
2
82.
Improving
the
quality
of
the
nation's
water
is
just
one
of
many
things
we
all
have
to
pay
for
as
tax­
payers
and
and
as
consumers.
That
is,
the
costs
of
things
like
improving
water
quality
are
paid
partly
by
government
out
of
what
we
pay
in
taxes
and
partly
by
companies
out
of
what
we
pay
for
the
things
they
sell
us.
83.
As
I
mentioned
earlier,
almost
all
of
the
rivers
This
scale
card
shows
about
how
much
people
in
your
general
income
category
paid
in
1979
in
taxes
and
higher
prices
for
things
like
national
defense,
roads
and
highways,
public
schools
and
the
space
program.
(
HAND
RESPONDENT
APPROPRIATE
SCALE
CARD
A­
I,
A­
II,
A­
III
,
OR
A­
IV;
LET
RESPONDENT
KEEP
WATER
QUALITY
LADDER
CARD)

You
will
see
different
amounts
of
money
listed
with
words
like
"
hiqhways"
and
"
public
education"
appear­
ing
by
the
amount
of
money
average
size
households
paid
for
each
one
last
year.
"
Highways"
here
refers
to
&
he
construction
and
maintenance
of
all
the
nation's
highways
and
roads.
*
Public
education"
refers
to
all
pubic
elementary
and
secondary
schools
but
does
not
include
the
costs
of
public
universities.
(
SKIP
I
want
to
ask
you
some
questions
about
what
amounts
of
money,
if
any,
you
would
be
willing
to
pay
for
varying
levels
of
overall
water
quality
in
the
nation's
lakes,
rivers
and
streams.
Please
keep
in
mind
that
the
money
would
go
for
sewage
treatment
plants
in
communities
through
various
kinds
of
taxes
(
such
as
withholding
taxes,
sales
taxes
and
sewage
fees)
and
for
pollution
control
equipment
the
govern­
ment
would
require
industries
to
install,
thus
raising
the
prices
of
what
they
make.

At
the
present
time
the
average
quality
of
water
in
the
nation's
lakes,
rivers
and
streams
is
at
about
level
D
on
the
ladder.
(
POINT
TO
LEVEL
D
ON
WATER
QUALITY
LADDER
CARD)
If
no
more
money
were
spent
at
all
tomorrow
on
water
quality,
the
overall
quality
of
the
nation's
lakes
and
rivers
would
fall
back
to
about
level
E.
(
POINT
TO
LEVEL
E)
People
have
different
ideas
about
how
important
the
quality
of
lakes,
rivers
and
streams
is
to
them
personally.
Thinking
about
your
household's
annual
income
and
the
fact
that
money
spent
for
one
think
can't
be
spent
for
another,
how
much
do
you
think
it
is
worth
to
you
to
keep
the
water
quality
in
the
nation
from
slipping
from
level
D
Sack
to
level
E?
That
is,
which
amount
on
this
scale
card,
or
any
amount
in
between,
is
the
most
you
would
be
willing
to
pay
in
taxes
and
higher
prices
each
year
to
keep
the
nation's
overall
water
quality
at
level
D
where
virtually
all
of
it
is
at
least
clean
enough
for
boating?
If
it
is
not
worth
anything
to
you,
please
do
hot
hesitate
to
say
so.
Write
in
amount:
$

Depends
(
vol.)......................
OOX
Not
sure
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
OOY
Not
worth
anything
,..................
001
85.
Finally,
in
terms
of
your
being
able
to
decide
exactly
how
much
you,
yourself,
would
be
willing
to
pay
as
a
taxpayer
and
consumer
for
better
water
quality,
would
you
say
in
the
last
few
questions
we
gave
you
more
than
enough
informa­
tion,
about
enough
information,
not
quite
enough,
or
not
enough
information
at
all?

More
than
enough
1
Not
quite
enough
3
About
enough....
2
Not
enough
at
all
4
Don't
know
.
.
.
.
.
.
.
.
5
and
lakes
in
the
United
States
are
at
least
at
level
D
in
water
quality.
What
do
you
think
it
is
worth
to
you
not
only
to
keep
them
from
be­
coming
more
polluted
but
also
to
raise
their
overall
quality
to
level
C?
What
is,
including
the
amount
you
just
gave
me,
which
amount
on
the
scale
card
is
the
most
you
would
be
willing
to
pay
in
taxes
and
higher
prices
each
year
to
raise
the
overall
level
of
water
quality
from
level
D
to
level
C
where
virtually
all
of
it
would
at
least
be
clean
enough
for
fish
like
bass
to
live
in?

Write
in
amount:
$

Depends
(
vol.)....................
OOX
(
ASK
84)

Not
sure
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
OOY
Not
worth
anything
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
001
TO
85)

84.
What
about
getting
virtually
all
of
the
nation's
lakes
and
rivers
up
to
level
B
on
the
ladder?
Including
the
amounts
of
money
you
have
already
given
me,
which
amount
on
the
scale
card
is
the
most
you
would
be
willing
to
pay
in
taxes
and
higher
prices
each
year
to
make
almost
all
the
nation's
lakes,
rivers
and
streams
clean
enough
so
that
people
could
swim
in
them?

Write
in
amount:
$
Name
Address
NOW,
RETURN
TO
PAGE
14
OF
MAIN
QUESTIONNAIRE
AND
COMPLETE
FACTUAL
SECTION.
Depends
(
vol.)....................
OOX
(
ASK
83)

Not
sure
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
OOY
Not
worth
anything
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
001
(
SKIP
to
85)
A­
I­
3
STUDY
#
684
D
NOTE:
INSERT
THIS
FORM
AFTER
PAGE
14
OF
YELLOW
"
Y"
QUESTIONNAIRES
ONLY
AND
ASK
FOLLOWING
Q.
79.

80.
This
last
group
of
questions
is
about
the
quality
of
water
in
the
nation's
lakes
and
streams.
Congress
passed
strict
water
pollution
control
Laws
in
1972
and
1977.
As
a
re­
sult
many
communities
have
to
build
and
run
new
modern
sewage
treatment
plants
and
many
industries
have
to
install
water
pollution
control
equipment.

Here
is
a
picture
of
a
ladder
that
shows
various
levels
of
the
quality
of
water
(
HAND
RESPONDENT
WATER
QUALITY
LADDER
CARD)
Please
keep
in
mind
that
we
are
not
talking
about
the
drinking
water
in
your
home.
Nor
are
we
talking
about
the
ocean.
We
are
talking
only
about
freshwater
lakes,
rivers
and
streams
that
people
look
at
and
in
which
they
go
boating,
fishing
and
swimming.

The
top
of
the
ladder
stands
for
the
best
possible
quality
of
water,
that
is,
the
purest
spring
water.
The
bottom
stands
for
the
worst
possible
quality
of
water.
Unlike
the
other
ladders
have
used
in
this
survey,
on
this
ladder
we
have
marked
different
levels
of
the
quality
of
water.
For
example..
.
(
POINT
TO
EACH
LEVEL:
E,
D,
C,
AND
SO
ON,
AS
YOU
READ
STATEMENTS
BELOW)

LEVEL
E
(
POINTING)
is
so
polluted
that
it
has
oil,
raw
sewage
and
other
things
in
it,
has
no
plant
or
animal
life
and
smells
bad
Water
at
level
D
is
okay
for
boating
but
not
for
fishing
or
swimming
Level
C
shows
where
rivers,
lakes
and
streams
are
clear
enough
so
that
game
fish
like
bass
can
live
in
them
Level
B
shows
where
the
water
is
clean
enough
so
that
people
can
swim
in
it
safely
And
at
level
A,
the
quality
of
the
water
is
so
good
that
it
would
be
possible
to
drink
it
directly
from
a
lake
or
stream
if
you
wanted
to
Now
let's
think
about
all
of
the
nation's
rivers,
lakes
and
streams.
Some
of
them
are
quite
clean
and
others
are
more
or
less
polluted.
Looking
at
this
ladder,
would
you
say
that
all
but
a
tiny
fraction
of
the
nation's
rivers,
lakes
and
streams
are
at
least
at
level
D
in
the
quality
of
their
water
today
or
not?

All
but
a
fraction
at
Level
D..
1
Not
at
level
D.................
2
Not
sure..
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3
81.
As
you
know
it
takes
money
to
clean
up
our
nation's
lakes
and
rivers.
Taking
that
into
account,
and
thinking
of
overall
water
quality
where
all
but
a
tiny
fraction
of
the
nation's
lakes
and
rivers
are
at
a
particular
level,
which
level
of
overall
water
quality
do
you
think
the
nation
should
plan
to
reach
within
the
next
five
years
or
so­­
level
E,
D,
C,
B
or
A?

A
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1
B
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
2
C
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3
D
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
4
E
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
5
Depends
(
vol.)
......
6
Other
(
vol.)
........
7
Not
sure
.
.
.
.
.
.
.
.
.
.
.
.
8
INTERVIEWER:
CHECK
INCOME
IN
Q.
79
ON
PAGE
14
OF
MAIN
QUESTION­

NAIRE.
THEN
LOOK
BELOW
TO
SEE
WHICH
SCALE
CARD
RESPONDENT
USES
IN
QUESTIONS
82
­
84.

IF
LESS
THAN
$
9,999
USE
CARD
D­
I
IF
$
10,000
TO
$
14,999
USE
CARD
D­
II
IF
$
15,000
TO
$
24,999
USE
CARD
D­
III
IF
$
25,000
AND
ABOVE
OR
NOT
SURE/
REFUSED
USE
CARD
D­
IV
(
OVER)
A­
I­
4
82.
Improving
the
quality
of
the
nation's
water
is
just
one
of
many
things
we
all
have
to
pay
for
as
tax­
payers
and
as
consumers.
That
is,
the
costs
of
things
Like
improving
water
quality
are
paid
partly
by
government
out
of
what
we
pay
in
taxes
and
partly
by
companies
out
of
what
we
pay
for
the
things
they
sell
us.

This
scale
card
shows
about
how
much
people
in
your
general
income
category
paid
in
1979
in
taxes
and
higher
prices
for
things
like
national
defense,
roads
and
highways,
public
schools
and
the
space
program.
(
HAND
RESPONDENT
APPROPRIATE
SCALE
CARD
D­
I,
D­
II,
D­
III
OR
D­
IV;
LET
RESPONDENT
KEEP
WATER
QUALITY
LADDER
CARD)

You
will
see
different
amounts
of
money
listed
with
words
like
"
highways"
and
"
public
education"
appearing
by
the
amount
of
money
average
sire
house­
holds
paid
for
each
one
last
year.
"
Highways"
here
refers
to
the
construction
and
maintenance
of
all
the
nation's
highways
and
roads.
"
Public
education"
refers
to
all
public
elementary
and
secondary
schools
but
does
not
include
the
costs
of
public
universities.

I
want
to
ask
you
some
questions
about
what
amounts
of
money,
if
any,
you
would
be
willing
to
pay
for
varying
levels
of
overall
water
quality
in
the
nation's
lakes,
rivers
and
streams.
Please
keep
in
mind
that
the
money
would
go
for
sewage
treatment
plants
in
communities
through
various
kinds
of
taxes
(
such
as
withholding
taxes
,
sales
taxes
and
sewage
fees)
and
for
pollution
control
equipment
the
govern­
ment
would
require
industries
to
install,
thus
raising
the
prices
of
what
they
make.

You
will
also
see
on
the
scale
card
the
amount
of
money
the
average
household
in
your
general
income
category
paid
last
year
in
taxes
and
higher
prices
to
improve
the
water
quality
of
the
nation's
lakes
and
rivers.
This
share
of
the
nation's
expenditures
to
fight
water
pollution
has
meant
that
so
far
the
average
quality
of
these
bodies
of
water
has
been
raised
from
level
E
to
level
D
on
the
ladder.
(
POINT
TO
LEVELS
E,
AND
D
ON
WATER
QUALITY
LADDER
CARD)
If
this
amount
of
money
continues
to
be
spent
each
year,
the
quality
of
the
water
will
be
raised
up
to
level
C
(
POINT
TO
LEVEL
C)
in
the
next
few
years­­
that
is,
where
virtually
all
of
it
would
be
at
least
clean
enough
for
fishing
First,
as
far
as
you
are
concerned,
are
you
willing
to
pay
this
amount
each
year
to
raise
water
quality
to
level
C
or
not?

.
Yes,
willing...........
1
Depends
(
vol.).........
2
­

No,
not
willing
.
.
.
.
.
.
.
.
3­
F
(
ASK
83)

(
SKIP
TO
84)
­

83.
What
about
getting
the
nation's
lakes
and
rivers
up
to
level
B
on
the
ladder?
Including
the
amount
of
money
indicated
on
the
card
to
get
water
quality
up
to
level
C,
how
much
are
you
willing
to
pay
in
taxes
and
higher
prices
each
year
to
raise
the
water
quality
to
level
B­­
that
is
where
virtually
all
the
nation's
lakes,
rivers
and
streams
are
at
least
clean
enough
to
swim
in
safely?

Write
in
amount:
$

Depends
(
vol)
..................
COX
Not
sure.....;
..................
00Y
Not
worth
anything
.............
001
I
(
SKIP
TO
NAME
AND
ADDRESS
RECORDING
BELOW)

84.
What
about
the
amount
of
money
to
keep
the
quality
of
water
at
level
D?
How
much
do
you
think
you
would
be
willing
to
pay
each
year
in
taxes
and
higher
prices,
if
anything,
to
keep
the
nation's
overall
water
quality
from
slipping
below
level
D
to
level
E
where
it
once
was?
If
it
is
not
worth
anything
to
you,
please
do
not
hesitate
to
say
so.

Write
in
amount:
$

Depends
(
vol.)
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
OOX
Not
sure........................
OOY
Not
worth
anything..
.
.
.
.
.
.
.
.
.
.
.
.
001
.­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­
I
I
I
Name:
i
1
!
Address:
I
I
I
NOW,
RETURN
TO
PAGE
14
OF
MAIN
QUESTIONNAIRE
AND
COMPLETE
FACTUAL
SECTION.

Not
sure
4
A­
I­
7
A­
I­
8
A­
I­
9
A­
II­
1
Appendix
II
THE
WATER
QUALITY
LADDER
William
J.
Vaughan
Water
quality
can
either
be
described
in
terms
of
the
uses
for
which
a
particular
body
of
water
is
suitable
or
in
terms
of
the
objective
characteristics
of
the
water
itself.
In
turn,
objective
characteristics
traverse
a
continuum
from
those
that
are
readily
perceptible
to
those
that
can
only
be
detected
by
scientific
measurement.
In
certain
dimensions
(
e.
g.,
visible
phenomena
such
as
the
extent
of
algal
growth,
the
clearness
of
the
water,
and
the
existence
of
suds,
foam
or
debris
(
David,
1971))
people
at
large
find
it
easy
to
preceive
changes
in
water
quality.
However,
some
characteristics
which
delineate
water
quality
levels
more
finely,
such
as
dissolved
oxygen
content,

escape
visual
and
olfactory
perception.
Thus
it
is
not
surprising
that
people's
ratings
of
water
quality
levels
are
likely
to
exhibit
a
less­
than­
perfect
degree
of
association
with
any
one
or
a
combination
of
the
several
scientific
measures
of
quality
conditions
(
Binkley
and
Hanemann,
1978).
This
poses
a
problem
for
benefit
estimation
because
the
existence
of
a
positive
willingness
to
pay
for
water
quality
improvement
depends
upon
the
ability
of
people
to
perceive
water
quality
changes
when
such
changes
do,
in
fact,
occur.

This
problem
has
lead
previous
investigators
either
to
attempt
to
engineer
the
fortunate
marriage
of
an
objective
water
quality
index
(
based
on
some
weighted
combination
of
scientific
quality
parameters)
and
a
subjective
index
of
publicly
perceived
quality
(
Bouwes
and
Schneider,
1979)
or
to
link
subjective
indices
of
public
perception.
and
expert
perception
(
Dornbusch,
1975).
A­
II­
2
We
chose
to
describe
water
quality
primarily
in
terms
of
the
uses
for
which
water
becomes
suitable,
and
secondarily
in
terms
of
a
few
obvious
water
quality
conditions
(
clearness,
odor,
debris,
etc.).
However,
we
located
the
numerical
position
of
the
five
posited
water
quality
levels
(
Boatable,
Fishable­
2
levels,

Swimmable,
Drinkable)
by
indexing
a
set
of
five
objective
scientific
water
quality
parameters
using
a
variant
of
the
National
Sanitation
Foundation's
Water
Quality
Index
(
Booth
et
al.,
1976;
McClelland,
1974)
along
with
informed
judgment.
In
so
doing
we
hope
to
extablish,
ex­
ante,
an
admittedly
tenuous
link
between
scientifcally
measured
quality
characteristics
(
anchors
of
the
rating
scale)
and
perceived
water
quality
characteristics
(
the
use
and
readily
perceivable
objective
characteristic
descriptors
of
these
anchors).

Specifically,
a
number
of
sources
were
consulted
to
ascertain
the
minimally
acceptable
concentration
levels
of
five
measurable
quality
characteristics
associated
with
five
potential
uses
of
natural
water
courses.
These
were
fecal
coliforms
(
organisms/
100
ml),
dissolved
oxygen
(
mg/
1),
maximum
BOD­
5
(
mg/
1),

turbidity
(
JTU)
and
pH.
1
The
five
quality
measures
were
the
only
ones
for
which
numerical
values
could
be
obtained
across
all
use
classifications,
a
requirement
dictated
by
the
index
approach.
Particular
attention
was
given
to
state
water
quality
standards
(
North
Carolina
Environmental
Management
Commission,
Dorfman
1972))
because
they
report
specific
critical
water
quality
parameters
associated
with
a
set
(
usually
four
or
five)
of
descriptive
water
quality
classifications.

The
consensus
results
for
each
quality
level
are
summarized
in
Table
1.

1
Sources
consulted
include
Thomann
(
1971),
U.
S.
G.
S.
(
1978),
Pickle
et
al.
(
1973),
Davis
(
1968)),
Economics
Research
Associates
(
1979),
Katz
(
1969),
Dorfman
et
al.
(
1972),
North
Carolina
Environmental
Management
Commission,
APHA,
AWWA
and
FSIWA
(
1955),
National
Technical
Advisory
Committee
(
1968),
NAS­
NAE
(
1972),
EPA
(
1976),
Davidson,
Adams
and
Seneca
(
1966),
National
Planning
Association
(
1975).
A­
II­
3
Table
1.
Consensus
Water
Quality
Characteristics
of
Five
Water
Quality
Classes
Water
Quality
Classification
Measurable
Water
Quality
Characteristics
Petal
Dissolved
5­
day
Turbidity
Ph
Coliforms
Oxygen
BOD
(#/
100
ml)
hell)
al
(
mg/
1)
(
JTU)

Acceptable
for
drinking
without
treatment
0
7.0
(
90)
0
5
7.25
Acceptable
for
swimming
200
6.5
(
83)
1.5
10
7.25
Acceptable
for
game
fishing
1000
5.0
(
64)
3.0
50
7.25
Acceptable
for
rough
fishing
1000
4.0
(
51)
3.0
50
7.25
Acceptable
for
boating
2000
3.5
(
45)
4.0
100
4.25
Percent
saturation
at
85Ol'
in
parentheses
A­
II­
4
In
order
to
associate
each
of
the
five
possible
sets
of
scientific
measures
with
a
single­
valued
ordinate
or
the
quality
ladder
a
truncated
version
of
the
National
Sanitation
Foundation
Water
Quality
Index
(
WQI)

was
used:

WQI
=

where
the
quality
of
the
qi=

parameter,
a
number
from
0
to
100
obtained
from
the
transformation
functions
for
water
quality
measures
in
McClelland
(
1974).

the
weight
assigned
to
the
parameter.
The
original
weights
(
wi)
reported
in
McClelland
(
1971)

cover
nine
quality
measures
and
Our
adjusted
weights
cover
a
smaller
number
of
measures
which
also
The
resultant
ladder
appears
in
Figure
1.
A­
II­
5
For
example,
the
index
value
for
the
"
Acceptable
for
Rough
Fishing"

classification
was
developed
as
shown
below:
Weighted
Scale
Value
Scaled
Value
(
q
i)
Weight
Value
Characteristic
Fecal
Coliform
1000/
100m1
20
0.242
1.985
Dissolved
Oxygen
Max
5­
Day
BOD
Turbidity
Notes:

Percent
saturation
at
85
°
F.
3
mg/
1
50
JTU
7.25
44
0.274
2.820
74
0.161
2.000
38
0.129
1.599
93
0.194
2.049
Similar
calculations
for
the
remaining
four
classes
yield
the
water
quality
ladder
shown
in
Figure
1.
A­
II­
6
A­
II­
7
SOURCES
APHA,
AWWA,
and
FSIWA.
1955.
Standard
Methods
for
the
Examination
of
Water,
Sewage,
and
Industrial
Wastes
(
10th
ed.,
NY:
American
Public
Health
Association,
Inc.)

Binkley,
Clark
S.
and
W.
Michael
Hanemann.
1978.
The
Recreation
Benefits
of
Water
Quality
Improvement:
Analysis
of
Day
Trips
in
an
Urban
Setting
(
Washington,
D.
C.:
U.
S.
Environmental
Protection
Agency)

Booth,
William
E.,
Paul
C.
Carubia,
and
Francis
C.
Lutz.
1976.
A
Methodology
for
Comparative
Evaluation
of
Water
Quality
Indices
(
Washington,
D.
C.:
Council
on
Environmental
Quality)
NTIS
PB
251­
572
Bouwes,
Nicolaas
W.,
Robert
Schneider.
1979.
"
Procedures
in
Estimating
Benefits
of
Water
Quality
Change,"
American
Journal
of
Agricultural
Economics.
vol.
,
no.

David,
Elizabeth
L.
1971.
"
Public
Perception
of
Water
Quality,"
Water
Resources
Research
vol.
7,
no.
3.

David
M.
Dornbusch
and
Company,
Inc.
1975.
The
Impact
of
Water
Quality
Improvements
on
Residential
Property
Prices.
Report
prepared
for
the
National
Commission
on
Water
Quality
(
San
Francisco:
David
M.
Dornbusch
and
Company,
Inc.)

Davidson,
Paul,
F.
Gerard
Adams,
and
Joseph
Seneca.
1966.
"
The
Social
Value
of
Water
Recreational
Facilities
Resulting
from
an
Improvement
in
Water
Quality:
The
Delaware
Estuary,"
in
A.
V.
Kneese
and
S.
C.
Smith,
eds.,
Water
Research
(
Baltimore:
Johns
Hopkins
University
Press
for
RFF)

Davis,
Robert
K.
1968.
A
Study
of
Dissolved
Oxygen
in
the
Potomac
Estuary
(
Baltimore,
Md,:
Johns
Hopkins
University
Press
for
RFF)

Dorfman,
Robert,
Henry
D.
Jacoby,
and
Harold
A.
Thomas,
Jr.,
eds.
1972.
Models
for
Managing
Regional
Water
Quality
(
Cambridge,
Mass.:
Harvard
University
Press)

Economics
Research
Associates.
1979.
"
Cost
Impact
of
Marine
Pollution
on
Recreation
Travel
Patterns,"
(
Corvallis,
Ore.:
U.
S.
EPA
Environmental
Research
Laboratory)
68­
01­
3197
NTIS
PB­
290655
Katz,
Max.
1969.
Appendix
F
in
Robert
Nathan
Associates,
"
Mine
Draining
Pollution
and
Recreation
in
Appalachia."
(
Washington,
D.
C.:
Robert
Nathan
Associates)

McClelland,
Nina
I.
1974.
Water
Quality
Index
Application
in
the
Kansas
River
Basin
(
Washington,
D.
C.:
U.
S.
Environmental
Protection
Agency)
EPA­
907/
9­
74­
001
National
Academy
of
Sciences,
National
Academy
of
Sciences
Engineering
Committee
on
Water
Quality
Criteria.
1972.
Water
Quality
Criteria:
1972
(
Washington,

D.
C.:
U.
S.
Environmental
Protection
Agency).
A­
II­
8
National
Planning
Association.
1975.
Water
Related
Recreation
Benefits
Resulting
from
P.
L.
92­
500
(
Washington,
D.
C,;
Prepared
for
Natl.
Comm.
on
Water
Quality
National
Technical
Advisory
Committee.
1968.
Water
Quality
Criteria:
A
Report
of
the
National
Technical
Advisory
Committee
to
the
Secretary
of
the
Interior
(
Washington,
D.
C.:
Federal
Water
Pollution
Control
Adm.)

North
Carolina
Environmental
Management
Commission.
North
Carolina
Adminsitrative
Code
(
Raleigh,
N.
C.:
Environmental
Management
Commission
­
current
through
March
1977)

Pickle,
Hal
B.,
Andrew
C.
Rucks,
and
Renee
Sisson.
1973.
The
Economic
Benefits
of
Abating
Water
Pollution
in
the
Steel,
Textile,
and
Paper
Industries
in
Alabama
(
Auburn,
Ala.:
Water
Resoruces
Research
Institue)

Robert
R.
Nathan
Associates,
Inc.
1969.
"
Mine
Drainage
Pollution
and
Recreation
in
Appalachia."
(
Washington,
D.
C.:
Robert
Nathan
Assocs.)

Thomann,
Robert
V.
1971.
System
Analysis
and
Water
Quality
Management
(
N.
Y.:
Environmental
Research
and
Applications,
Inc.)

U.
S.
Environmental
Protection
Agency.
1976.
Quality
Criteria
for
Water
(
Washington,
D.
C.:
U.
S.
Environmental
Protection
Agency)

U.
S.
Geological
Survey.
1978.
"
Water­
Quality
Indices
for
Specific
Water
Uses."
(
U.
S.
Geological
Survey
Circular
770)
A­
III­
1
Appendix
III
DERIVATION
OF
PUBLIC
GOODS
EXPENDITURES
The
estimated
public
goods
expenditures
used
in
this
study
to
"
anchor"
the
amounts
displayed
on
the
payment
cards
are
shown
in
Table
I
below:

Table
1:
Public
Goods
Expenditure
Estimates
for
Versions
A,
B,
C,
D
by
Income
Class
Income
Category
I.
Less
than
$
10,000
II.
$
10­
15,000
III.
$
15,25,000
IV.
$
25
Public
Good
(
Average
Expenditure
per
Household)

Defense
$
322
(
402)

676
(
845)

1337
(
1671)

3013
(
3766)
Education
$
204
(
255)

446
(
557)

882
(
1103)

1988
(
2485)
Highways
Water
Police
&
Roads
Pollution
&
Fire
Space
$
98
(
123)

192
(
240)

312
(
390)

626
(
782)
$
61
$
33
s
13
(
16)

125
70
27
(
34)

245
139
53
(
66)

562
313
120
(
150)

These
amounts
were
used
to
anchor
the
payment
card
amounts
as
follows:

1.
Version
A
used
four
public
goods
(
Defense,
Education,
Highways,
and
Space
Program).

2.
Version
B
used
five
public
goods
(
Defense,
Education,
Highways,
Police
and
Fire
Protection,
and
Space).

3.
Version
C
used
the
four
public
goods
listed
for
A.
The
public
goods
expenditures
used
in
Version
C
were
25%
higher
than
those
used
in
Version
A.
These
amounts
are
shown
in
parenthesis.

4.
Version
D
used
the
four
public
goods
and
amounts
as
in
Version
X
plus
the
amounts
shown
from
Water
Pollution.
A­
III­
2
Methodology
Since
we
desired
to
take
account
of
public
goods
expenditures
that
were
the
result
of
both
direct
taxes
and
indirect
taxes
(
usually
reflected
in
higher
prices)
we
used
a
formula
that
took
into
account
both
direct
and
indirect
taxation.
Using
the
federal
tax
structure
as
our
base,
43%
of
taxes
come
from
income
taxes
(
direct)
while
57%
come
from
other
taxes
and
charges.
Internal
Revenue
Service
figures
are
also
available
on
the
average
amount
of
income
tax
paid
by
income
category.
Aggregating
the
IRS
categories
by
the
weight
of
the
percent
of
the
population
in
that
category,
we
obtained
for
the
federal
budget.
the
average
federal
income
tax
paid
by
our
four
income
classes.
1
The
following
formula
was
used
to
determine
total
household
expenditures
Average
Federal
Total
Federal
Income
Tax
Paid
(
43%)
+
Indirect
Taxes
(
57%)
=
Household
Expenditures
or
Average
Federal
Income
Tax
Paid
=
Total
Federal
43%
Household
Expenditures
It
is
now
possible
to
solve
the
equation
for
total
federal
household
expenditures
since
average
federal
income
tax
paid
is
known
and
.43
is
a
constant
representing
the
ratio
of
income
tax
to
total
federal
revenues.

1
An
exception
to
this
procedure
was
made
in
the
case
of
the
$
0­
5,000
income
categories.
These
categories
are
not
included
in
our
calculations
for
the
under
$
10,000
income
class
because
they
pay
almost
no
income
taxes
and
would
have
distorted
our
estimate
of
the
non­
income
expenditures
on
public
goods
for
the
under
$
10,000
income
class.
Hence,
our
estimates
of
average
federal
income
tax
paid
by
those
in
the
under
$
10,000
category
are
biased
upward.
A­
III­
3
From
the
1980
United
States
Budget,
defense
spending
accounts
for
24%

of
total
federal
expenditures.
To
calculate
a
household's
(
in
a
given
income
category)
expenditures
for
defense
the
following
formula
was
used:

24%
x
(
Total
Household
Federal
Expenditures)
=
Household
Defense
Expenditure
Expenditures
for
other
public
goods
were
calculated
using
defense
spending
as
a
base.
2
(
HED)
x
=
HEPGX
where
HED
=
Household
Expenditure
on
Defense
TEPGX
=
Total
Expenditures
on
Public
Good
X
TFDE
=
Total
Federal
Defense
Expenditures
HEPGX
=
Household
Expenditures
on
Public
Good
X
For
a
household
in
income
level
I
(
under
$
10,000
annual
income),
expenditures
on
highways
and
roads
were
calculated
as
follows:

$
322
x
where
HED
=
$
322
TEPGX
=$
33,700,000,000
HEPGX
=
$
98
Public
Good
X
=
Highways
and
roads
Estimation
Problems
The
estimates
of
the
public
goods
expenditures
by
income
category
are
only
intended
to
be
rough
"
ball
park"
figures.
They
are
plagued
by
a
number
2
The
estimates
of
expenditures
on
highways
and
roads
included
the
folowing
correction
factor
to
take
account
of
the
regressive
nature
of
gasoline
taxes
which
are
largely
responsible
for
financing
this
public
good.
For
income
category
I
(
under
$
10,000)
the
estimated
household
expenditure
on
highways
and
roads
was
multiplied
by
120%.
For
income
categories
II,
III,
and
IV,
the
correction
factor
was
+
10,
and
­
20,
respectively.
A­
III­
4
of
problems
some
of
which
are
not
easily
tractable.
Since
we
are
attempting
to
obtain
estimates
of
willingness
to
pay
for
water
quality
at
the
time
of
the
interview,
it
is
desirable
to
use
as
current
as
possible
estimates
of
expenditures
on
other
public
goods.
This
desire
presents
three
alternatives:

(
1)
using
the
latest
year
for
which
estimates
were
available
for
all
public
goods
used
which
in
our
case
would
have
been
1976,
(
2)
make
the
heroic
assumption
of
determining
the
rate
at
which
expenditures
on
each
public
good
changed
since
the
last
good
estimate
available,
(
3)
use
the
latest
year
available
for
each
public
good.
We
have
chosen
the
third
alternative,
as
the
drawbacks
of
non­
comparable
years
appeared
better
than
old
numbers
in
the
case
of
(
1)
and
the
expansion
and
contraction
of
several
public
goods
such
as
water
pollution
control,
defense,
and
highways
out
of
sinc
with
any
of
the
standard
indexes
precluded
easy
use
of
(
2).

Discrepancies
in
definitions
also
pose
estimation
problems
in
the
case
of
the
Census
Bureau's
household
definition
and
IRS's
definition
of
non­
business
income
tax
returns.
In
our
case,
there
are
77
million
households
and
87
million
individual
and
joint
income
tax
returns.
We
chose
to
consider
households
and
IRS
tax
returns
and
equivalent
for
the
purpose
of
computing
average
federal
income
tax
paid.

The
most
heroic
assumption
we
made
was
that
the
other
57%
of
the
federal
budget
is
collected
in
the
same
proportion
as
income
tax.
These
indirect
taxes
are
largely
consumption
taxes;
hence
this
assumption
is
probably
not
warranted.
If
the
public
goods
expenditures
on
the
payment
card
showed
itself
to
be
sensitive
to
the
exact
amount
given,
then
a
major
effort
would
be
required
to
achieve
more
accurate
estimates
of
these
expenditures.
A­
III­
5
Version
A
and
Version
C
of
this
survey
were
explicitly
designed
to
test
this
sensitivity.

With
the
exception
of
the
purely
federal
expenditures
of
defense
and
space,
our
implicit
assumption
of
uniform
national
expenditures
by
income
category
is
questionable
although
highways
and
roads
and
water
pollution
control
expenditures
violate
this
assumption
to
a
lesser
degree
than
do
police
and
fire
or
public
education
expenditures.
(
I.
e.,
a
resident
of
New
York
City
pays
much
more
for
police
protection
than
does
someone
in
rural
Iowa).
Further,
the
respondent,
if
he
or
she
is
familiar
with
public
goods
expenditures
is
most
likely
to
be
familiar
with
expenditures
on
these
two
highly
local
public
goods.
If
our
estimates
are
significantly
different
from
the
respondent's
perceptions
of
what
they
are,
the
survey
may
lose
credibility
in
the
eyes
of
the
respondent.
The
extent
of
this
problem,
if
any,
was
not
explored.
A­
III­
6
Sources
A.

B.

C.

D.

E.

F.
Tax
figures
­­
1976
IRS
preliminary
estimates
Total­
federal
income,
defense
expenditures,
space
expenditures
­­
Budget
of
the
United
States,
1980.

Education
figures
­­
HEW
preliminary
estimates
for
primary
and
secondary
education
expenditures
during
the
1978­
79
school
year.

Highways
and
roads
­­
American
Highway
and
Transportation
Builder's
Association
for
1978.

Water
Pollution
­­
CEQ
estimates
for
total
expenditures
on
water
pollution
control
(
December
1978).

Police
and
Fire
­­
Facts
and
Figures
on
Government
Finance
(
Tax
Foundation,
Inc.,
1979).
Appendix
IV
FINAL
RESULTS
OF
THE
RESOURCES
FOR
THE
FUTURE
National
Environmental
Survey
for
the
President's
Council
on
Environmental
Quality
These
results
are
based
on
a
probability
sample
of
1576
persons,
age
18
and
over
living
in
the
continental
United
States
excluding
Alaska.

Initially
1286
persons
were
interviewed
in
person
between
January
26
and
February
9,
1980.
An
additional
sample
of
280
persons
were
interviewed
in
person
later
in
March
to
bring
the
sample
size
up
to
1576.

All
the
data
reported
here
have
been
weighted
using
standard
procedures
to
compensate
for
minor
variations
between
the
final
sample
and
the
actual
distribution
of
basic
population
characteristics.

In
order
to
include
as
many
questions
as
possible
in
the
instrument,

the
sample
was
split
into
two
equivalent
samples.
Most
questions
were
asked
of
the
entire
sample
but
some
were
asked
only
of
the
X
or
the
Y
half.

These
questions
are
identified
on
the
questionnaire.
The
sample
size
for
the
X
version
is
840
and
that
of
the
Y
sample
is
736.

Robert
Cameron
Mitchell
Senior
Fellow
A­
IV­
2
STUDY
NO.
684
(
1002)
JANUARY
1980
COUNTY
PLACE
BLK.
#
5­
1
OMB
Clearance
Number:
116F­
79025
Time
Started
Time
Finished
Total
Minutes
6/
7
Hello,
I'm
from
ROPER
AND
CANTRIL
and
we're
conducting
a
study
all
over
the
country
for
the
United
States
Government
getting
people's
views
about
some
of
the
problems
the
nation
faces.
Your
participation
in
this
survey
is
entirely
voluntary.
All
information
will
be
held
in
the
strictest
confidence
and
will
be
used
only
to
produce
overall
statistical
reports.
We
would
very
much
value
your
cooperation.

1.
First,
I
would
like
to
ask
you
which
three
national
problems
you
would
like
to
see
the
government
devote
most
of
its
attention
to
in
the
next
year
or
two?
(
HAND
RESPONDENT
CARD)

a.
Reducing
racial
discrimination
.................

b.
Reducing
the
amount
of
crime
...................

c.
Beautifying
America
............................

d.
Conquering
"
killer"
diseases
...................

e.
Reducing
pollution
of
air
and
water
............

f.
Helping
people
in
poor
areas
...................

g.
Reducing
unemployment
..........................

h.
Improving
highway
safety
.......................

i.
Improving
housing
and
run­
down
neighborhoods
...

j.
Improving
public
education
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

None
...........................................

No
opinion
.....................................
13%

61
5
41
24
29
48
­

20
35
1
1
8/

9/

10/

11/

12/

13/

14/

15/

16/

17/

18/

19/

2.
There
is
a
lot
of
talk
these
days
about
what
the
aims
country
should
be
for
the
next
ten
years.
On
this
card
are
listed
some
of
the
goals
which
different
people
would
give
top
priority.
(
HAND
RESPONDENT
CARD
Would
you
please
say
which
one
of
these
you,
yourself,
consider
the
most
important?

2.
3.
Most
Next
most
important
important
(
Col.
20)
(
Col.
21)

a.
Maintaining
a
high
rate
of
economic
growth
........................
27%
27%

b.
Making
sure
that
this
country
has
strong
defense
forces
..........

c.
Seeing
that
people
have
more
say
in
how
things
get
decided
at
work
and
in
their
communities
.....................

d.
Protect
nature
from
being
spoiled
and
polluted
...................

None
..............................................................
44
26
19
22
9
21
1
4
No
opinion
.......................................................

3.
And
which
would
be
the
next
most
important?
(
RECORD
ABOVE)

4.
If
you
had
to
choose,
which
one
of
the
things
on
this
card
would
you
say
is
most
desirable?
(
HAND
RESPONDENT
CARD)
4.
5.
Most
Second
desirable
choice
(
Col.
22)
(
Col.
22)

a.
Maintaining
order
in
the
nation
16%
28%

b.
Giving
the
people
more
say
in
important
government
decisions
c.
Fighting
rising
prices
18
30
d.
Protecting
freedom
of
speech
56
22
None
9
17
No
opinion
1
3
5.
What
would
be
your
second
choice.
(
RECORD
ABOVE)
Maintaining
a
stable
economy
Progress
toward
a
less
impersonal,
more
humane
society
.....................
6.
7.
Most
Next
most
important
important
(
Col.
24)
(
Col.
25)

55%
24%

8
14
27
37
9
22
1
3
No
opinion
.
.
.
.
.
.
.
.
.
.
..
e.................
A­
IV­
3
Page
2
6.
Here
is
another
list.
(
HAND
RESPONDENT
CARD)
In
your
opinion,
which
one
of
these
is
most
important?

The
fight
against
crime.................

Progress
toward
a
society
in
which
ideas
can
count
more
than
money
.
.
.
.
.
.
.
.
.

None
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
a.

b.

c.

d.

7.
Which
is
next
most
important?
(
RECORD
ABOVE)

8.
Here
is
a
card
that
includes
all
of
the
goals
listed
on
the
three
cards
you
have
just
looked
at.
(
HAND
RESPONDENT
CARD)
Would
you
tell
me
which
one
of
the
goals
on
this
card
you
consider
the
most
desirable
of
all
a.
Maintaining
a
high
rate
of
economic
growth
.
.
.
.
.
.
.
.
.
.
­
.
.
.

b.
Making
sure
that
this
country
has
strong
defense
forces
.
.
.

c.
Seeing
that
people
have
more
say
in
how
things
get
decided
at
work
and
in
their
communities
...........

d.
Protecting
nature
from
being
spoiled
and
polluted
.
.
.
.
.
.

e.
Maintaining
order
in
the
nation
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

f.
Giving
the
people
more
say
in
important
government
decisions
.........................
8
9
10
Most
Next
most
Lease
desirable
desirable
important
(
Cols.
26,
27)
(
Cols.
28,
29)
(
Cols.
30,
31)

8%
8%
8%

24
13
5
5
5
8
2
5
9
6
­
2
6
6
4
25
17
2
2
5
­

11
14
4
2
2
11
5
15
4
3
3
16
19
1
1
8
g.
Fighting
rising
prices
................,.......
.
.
.
.
.
..­..

h.
Protecting
freedom
of
speech
.
.
.
.
.
.
.
.
..
ss...............

i.
Maintaining
a
stable
economy
..
.
_..
.
.
_
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

j.
Progress
toward
a
less
impersonal,
more
humane
society
k.
The
fight
against
crime
................................

l.
Progress
toward
a
society
in
which
ideas
can
count
more
than
money
..
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

None
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

No
opinion
..
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

9.
Which
is
the
next
most
desirable?
(
RECORD
ABOVE)
A­
IV­
4
a.
How
worried
or
concerned
are
you
about
the
rise
in
prices
and
the
cost
of
living
.............
81%
16%
2%
1%
­
32/
A
great
A
fair
Not
very
Not
at
No
deal
amount
much
all
opinion
The
problems
of
the
poor?
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
b.

c.
Cleaning
up
our
waterways
and
reducing
water
pollution?
..........................

d.

e.
OMITTED
Reducing
the
amount
of
unnecessary
noise
in
this
community
44
42
11
2
1
33/

39
44
13
3
1
34/

No
col.
35
11
20
34
34
1
36/

73
18
5
2
1
37/

36
40
16
7
1
38/

42
27
16
13
1
39/

No
col.
40
f
.
Shortages
of
oil,
gasoline,
coal,
natural
gas,
electricity,
or
other
fuels?
..................

g.
Reducing
air
pollution?
............................

h.
The
purity
of
the
drinking
water
in
your
community?
...........................

i.
OMITTED
Question
asked
for
RFF
in
separate
Roper
survey,
March
1980:
9.
Now,
I'd
like
to
find
out
how
worried
or
concerned
you
are
about
a
number
of
problems
I'm
going
to
mention:
a
great
deal,
a
fair
amount,
not
very
much,
or
not
at
all.
If
you
aren't
really
concerned
about
some
of
these
matters,
don't
hesitate
to
say
so.
First,
(
ask
about
each
item)

A
great
A
fair
Not
very
Not
at
No
deal
amount
much
all
opinion
a.
How
worried
or
concerned
are
you
bout
the
rise
in
prices
and
the
cost
of
2%
­
living?
86%
11%
1%

b.
The
presence
of
toxic
chemicals
such
as
pest­
icides
or
PCBs
in
the
46
32
16
4
2
environment?

c.
Cleaning
up
our
waterways
and
reducing
water
54
33
10
2
1
pollution?

d.
The
disposal
of
industrial
chemical
wastes
that
are
64
26
7
2
1
hazardous?
TREND
ON
CONCERN
ABOUT
INFLATION,
ENERGY
SHORTAGES,

WATER
AND
AIR
POLLUTION,
1972­
1980
Q.
11
Now
I'd
like
to
find
out
how
worried
or
concerned
you
are
about
a
number
of
prob­
lems
I
am
going
to
mention:
a
great
deal,
a
fair
amount,
not
very
much,
or
not
at
all.

If
you
aren't
really
concerned
about
some
of
these
matters,
don't
hesitate
to
say
so.

Rise
in
prices
and
the
cost
of
living.

N
composite
score
Shortages
of
oil,
gasoline,
coal,
natural
gas,
electricity
and
other
fuels.

composite
score
June
1972
April
1974
May
1976
Roper
Rff
1980
March
83%
73%
79%
86%

12
21
17
11
3
3
2
2
1
1
2
1
1
0
0
0
1802
1865
1071
1286
c.
1500
90
93
57
24
72
13
18
5
5
1
3
1
78
61
51
57
38
34
29
35
32
45
33
7
10
13
10
1
2
2
3
2
2
2
2
1
1
84
79
82
60
46
55
36
28
36
32
39
9
12
8
17
1
3
3
7
2
3
2
1
83
76
80
Cleaning
up
our
waterways
and
reducing
water
polluting.

composite
score
Reducing
air
pollultion.

composite
score
Data
for
1972­
1976
are
from
the
"
State
of
the
Nation"
studies
done
by
the
Potomac
Institute.
Only
four
of
the
items
used
in
question
11
of
the
CEQ
questionnaire
were
repeated
from
the
Potomac
Institute
questionnaires.
The
composite
index
is
adopted
from
the
Potomac
Institute
studies.
It
is
calculated
by
dropping
the
'
don't
knows'

and
multiplying
the
number
saying
'
great
deal'
by
100,
those
saying
'
fair
amount'
by
67,
'
not
very
much'
by
33
and
'
not
all'
by
0.
12.
Here
is
a
picture
of
a
ladder.(
HAND
RESPONDENT
CARD)
Let's
suppose
the
top
of
the
ladder
represents
the
best
possible
natural
environmental
situation
for
our
country;
that
is,
all
aspects
of
the
environment­­
air
water,

forests,
wildlife,
waste
disposal,
noise
and
the
lake.
Suppose
the
bottom
of
the
ladder
represents
the
worst
possible
environmental
situation
for
the
United
State.
Please
show
me
on
which
step
of
the
ladder
you
think
the
environment
in
the
United
States
is
at
the
present
time.
(
RECORD
NUMBER
CHOSEN
BELOW)

mean
5.24
42/

Don't
know
4%

13.
On
which
step
would
you
say
it
was
about
five
years
ago?

5.32
42/

Don't
know
6%

14.
Just
as
your
best
guess,
if
things
go
pretty
much
as
they
are
now,
where
do
you
think
the
environment
in
the
U.
S.
will
be
on
the
ladder,
let
us
say,
about
five
years
from
now?

5.08
43/

Don't
know
7%

15.
Thinking
now
just
about
the
quality
of
the
air
in
this
area.
Suppose
the
top
of
the
ladder
represents
the
cleanest
air
possible,
and
the
bottom,
the
most
polluted
air
possible,
Please
show
me
on
which
step
of
the
ladder
you
think
the
air
around
here
is
at
the
present
time.

6.58
Don't
know
2%
44/

16.
On
which
step
would
you
say
it
was
about
five
years
ago?

6.04
45
Don't
know
8%

17.
As
your
best
guess,
where
will
it
be
about
five
years
from
now?

6.04
46
Don't
know
6%

18.
Now,
thinking
just
about
the
quality
of
the
water
in
the
lakes
and
streams
in
this
area.
Suppose
the
top
of
the
ladder
represents
the
cleanest
water
possible
and
the
bottom,
the
most
polluted
water
possible.
Please
show
me
on
which
step
of
the
ladder
you
think
the
water
in
the
lakes
and
streams
around
hers
is
at
the
present
time.

5.64
47
Don't
know
9%

19.
On
which
step
would
you
say
it
was
about
five
years
ago?

5.72
48
Don't
know
13%

20..
As
your
best
guess,
where
will
it
be
about
five
years
from
now?

5­
62
49
MEAN
SELF
ANCHORED
LADDER
RANKINGS
OF
NATIONAL
ENVIRONMENTAL
QUALITY,
LOCAL
WATER
QUALITY
AND
LOCAL
AIR
QUALITY
FOR
PAST,
PRESENT
AND
FUTURE
21.
People
are
affected
in
different
ways
by
the
problems
our
country
faces.
For
each
of
the
problems
I
am
going
to
mention,
please
tell
me
how
much
you,
yourself,
are
affected
in
terms
of
the
kind
of
life
you
live
and
your
personal
enjoyment
of
your
surroundings.
First,
the
energy
shortage:
would
you
say
it
affects
you
a
great
deal,
a
fair
amount,
just
a
little,
or
not
at
all?
(
ASK
ABOUT
EACH
ITEM
BELOW)

A
great
Fair
Just
a
Not
at
No
deal
amount
little
all
opinion
a.
The
energy
shortage?
54%
31%
10%
4%
­

b.
Air
pollution?
18
33
24
22
1
1
c.
Unnecessary
noise
in
your
area?
8
15
25
52
d.
OMITTED
No
opinion
e.
OMITTED
83
11
3
1
2
22.
OMITTED
A­
IV­
9
23.
Now,
I'd
like
to
ask
you
about
some
specific
aspects
of
the
environment.
If
you
aren't
sure
about
some
of
the
things
I
am
going
to
mention
please
don't
hesitate
to
say
so.
First,
which
one,
factories,
automobiles,
or
incinerators,
is
the
major
cause
of
air
pollution
in
this
country?

Factories
39%

Automobiles
45
Incinerators
Not
sure
9
24.
Nuclear
power
plants
are
built
near
bodies
of
water.
Do
you
think
that's
because
the
water
is
used
as
another
source
of
power,
as
a
disposal
place
for
waste,
or
is
used
for
cooling
purposes?

Another
source
of
power
11%

A
disposal
place
for
waste
16
For
cooling
purposes
52
Not
sure
21
25.
Do
you
think
cancer
can
be
caused
in
rats
by
every
chemical,
by
most
chemicals,
or
only
some
chemicals
if
they
are
fed
to
rats
in
a
large
enough
dose?

Every
chemical
...........
14%
59/

Most
chemicals
.
.
.
.
.
.
.
.
.
.
.
31
Only
some
chemicals
......
42
Not
sure.................
12
26.
From
what
you
have
heard
or
read,
do
you
think
we
produce
enough
oil
in
this
country
to
meet
our
present
27.
energy
needs
or
do
we
have
to
import
some
oil
from
other
countries?

Produce
enough
oil
..............
29
60/

Have
to
import
some
oil
.........
63
Not
sure
.
.
.
.
.
.
.
.
.
.
.
.
.
.
..­.......

Do
you
think
that
it
is
possible
for
a
nuclear
power
plant
to
explode
and
cause
a
mushroom­
shaped
cloud
like
the
one
at
Hiroshima
or
don't
you
think
that
is
possible?

Possible.......................
61
Not
possible...................

Not
sure.......................
52
31
16
A­
IV­
10
3a
I'm
going
to
read
you
a
short
list
of
topics
and
incidents
that
have
been
mentioned
in
the
news
media
over
the
past
year
or
so.
As
I
mention
each,
if
you
what
it
refers
to?
(
ASK
ABOUT
EACH
ITEM)
happen
to
have
heard
or
read
about
it,
would
you
please
tell
me
a.
Love
Canal,
near
Niagara
Falls,
New
York:
can
you
tell
me
what
happened
there?
(
DO
NOT
READ
LIST)
CORRECT:
abandoned
hazardous
waste
dump;
chemical
or
toxic
waste
dump;
place
where
chemical
wastes
have
harmed
people
or
made
them
move;
where
drums
of
toxic
chemicals
have
leaked
into
the
soil
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

PARTIALLY
CORRECT:
people
moved
out
of
their
homes;
place
where
there
was
a
problem
with
the
soil.
NO
REFERENCE
TO
CHEMICALS
INCORRECT
.
.
.
.
..­................................­.........................
4
8
65
Not
sure
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
..*..............

b.
The
accident
at
Three
Mile
Island
can
you
tell
me
what
happened
there?
(
DO
NOT
LIST)
CORRECT:
nuclear
reactor
accident;
Harrisburg
incident;
accident
that
almost
caused
a
meltdown
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

PARTIALLY
CORRECT:
Presidential
Commission;
in
Pennsylvania
NO
REFERENCE
TO
NUCLEAR
POWER
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

INCORRECT
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
..*.......................

Not
sure
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

c.
Synfuels
or
synthetic
fuels:
can
you
tell
me
what
they
are?
(
DO
NOT
READ
LIST)

CORRECT:
gas
or
oil
made
from
coal
or
oil
shale
or
tar
sands;
coal
gasification;
coal
liquidification
or
coal
liquids
PARTIALLY
CORRECT:
President
Carter's
new
billion
dollar
plan;
new
source
of
energy
recently
proposed
by
President
Carter
and
debated
in
Congress
64/

INCORRECT
.
.
.
.
.
..­..............­...................................

Not
sure
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

d.
Acid
rain:
can
you
tell
me
what
this
is?
(
DO
NOT
READ
LIST)

CORRECT:
polluted
rain
that
harms
lakes
by
disturbing
the
growth
of
plants,
algae
and
fish;
rain
that
is
like
vinegar;
air
pollution
in
the
rain
that
harms
land
and
water;
sulphur
dioxide
in
the
air
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

PARTIALLY
CORRECT:
references
to
"
rain"
that
do
not
include
­
mention
of
air
pollution
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

INCORRECT
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.._..............................................
22%

74
3
4
19
37
5
15
42
26
6
9
58
62/

63/

65/

Not
sure
..........................
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Page
6
29.
Now
I'd
like
to
ask
you
some
questions
about
air
pollution.
How
serious
do
you
feel
air
pollution
is
in
this
country­­
very
serious,
somewhat
serious
or
not
serious
at
all?

Very
serious
......
Not
serious
at
all
35%
8%
66/
35
Somewhat
serious
..
NO
opinion
.
.
.
.
.
.
.
.
.
­
2
30.
How
about
in
this
area:
how
serious
do
you
think
air
pollution
is
here?

11
Very
serious......
Not
serious
at
all.
48
67/

Somewhat
serious...
39
No
opinion
.
.
.
.
.
2
31.
As
you
may
know,
many
of
the
nation's
marsh
and
swamp
areas
have
had
water
drained
out
of
them
so
that
the
land
can
be
used
for
residential
areas,
factories
or
farming.
Some
people
say
we
should
drain
more
of
these
areas
because
land
for
development
is
becoming
harder
to
find.
Other
people
say
that
these
marsh
and
swamp
areas
should
be
kept
as
they
are
because
they
help
maintain
nature's
balance
by
providing
breeding
areas
for
fish
and
feeding
places
for
ducks
among
other
things.
Who
do
you
agree
with
most­­
those
who
feel
these
swamp
and
marsh
areas
should
continue
to
be
drained,
or
those
who
feel
they
should
be
preserved
in
their
natural
state,
or
don't
you
have
a
strong
feeling
one
way
or
the
other?

Continue
to
drain.
10
No
strong
feeling
..
11
68/
8
Preserve
in
Depends
(
vol.)
.....

natural
state
.
.
.
.
.
.
65
No
opinion
........
6
32.
Finding
new
places
to
build
new
industrial
and
power
plants
is
sometimes
difficult
these
days.
I'm
going
to
mention
five
types
of
buildings
or
sites.
Assuming
that
they
would
be
built
and
operated
according
to
govern­
ment
environmental
and
safety
regulations,
you
might
or
might
not
feel
strongly
about
living
close
to
them.
For
each
type
of
plant
please
tell
me
the
closest
such
a
plant
could
be
built
from
your
home
before
you
would
want
to
move
to
another
place
or
to
actively
protest,
or
whether
it
wouldn't
matter
to
you
one
way
or
another
how
close
it
was?
(
IF
DISTANCE
GIVEN,
GET
DISTANCE
IN
MILES)

a.
First,
what
about
a
ten­
story
office
building?
mean
=
5.8
miles
Less
than
1
mile
69/
70
1
or
more
miles
(
write
in
number)

Wouldn't
matter
one
way
or
other.

Don't
want
it
at
any
distance
(
vol.)

No
opinion
b.
A
power
plant
that
uses
coal
for
fuel?
mean=
20.5
miles
Less
than
1
mile
71/
72
1
or
more
miles
(
write
in
number)

Wouldn't
matter
one
way
or
other.

Don't
want
it
at
any
distance
(
vol.)

No
opinion
c.
A
nuclear
power
plant
mean=
91.0
miles
Less
than
1
mile
73/
74
1
or
more
miles
(
write
in
number)
d.
A
large
industrial
plant
or
factory?
mean=
13.9
miles
Less
than
1
mile
75/
76
1
or
more
miles
(
write
in
number)

Wouldn't
matter
one
way
or
other.

Don't
want
it
at
any
distance
(
vol.)

No
opinion
e.
How
about
a
disposal
site
for
hazardous
waste
chemicals
if
the
government
said
disposal
could
be
done
safely
and
that
the
site
would
be
inspected
regularly
for
possible
problems?
mean=
81.4
miles
Less
than
1
mile
77/
78
1
or
more
miles
(
write
in
number)

Wouldn't
matter
one
way
or
other.

Don't
want
it
at
any
distance
(
vol.)

No
opinion
A­
IV­
12
Would
you
please
tell
me
approximately
how
far
away
the
following
are
from
your
home?
mean=
18.8
miles
a.
The
nearest
freshwater
lake?

Less
than
1
mile
6/
7
1
or
more
miles
(
write
in
number)

No
opinion
b.
The
nearest
river
large
enough
for
boating?
mean=
23.3
miles
Less
than
1
mile
.
.
.
.
.
.
.
.
.
.
.
.
.
8/
9
1
or
more
miles
(
write
in
number)

No
opinion
.
.
.
.
.
.
.
.
­­­
.
.
.
.
.
.
.

c.
The
nearest
industrial
plant
or
power
plant?
mean=
7.9
miles
Less
than
1
mile
.
.
.
.
.
.
.
­
­.
­.
._

1
or
more
miles
(
write
in
number)
10/
11
No
opinion
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
14%

d.
The
nearest
nuclear
power
plant
that
is
either
operating
now
or
under
construction?
mean=
90.8
miles
Less
than
1
mile
.................
12/
13
4
Depends
(
vol.)

holding
down
costs
rather
than
requiring
stricter
controls
.
.
.
.
34%

OR
Pollution
control
requirements
and
standards
have
gone
too
far;
they
already
cost
more
than
they
are
worth
13%

Depends
vol.)
5%

No
opinion
5%
42%
14/
No
opinion
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
32%

I
am
going
to
read
you
three
points
of
view
regarding
pollution
control.
Please
tell
me
which
one
best
represents
your
opinion.

READ:

Protecting
the
environment
is
so
important
that
requirements
and
standards
cannot
be
too
high,
and
continuing
improvement
must
be
made
regardless
of
cost
OR
We
have
made
enough
progress
on
cleaning
up
the
environment
that
we
should
now
concentrate
on
42%
15/

17
24
Higher
prices
........

Higher
taxes
.........

Neither
(
vol.)
......

Both
(
vol.)
..........

Depends
(
vol.)
.......

No
opinion
..........

36.
I
am
going
to
read
you
three
statements
about
en­
vironmental
protection
and
economic
growth.
Please
listen
carefully
and
tell
me
which
statement
you
agree
with
the
most.

READ:

We
can
achieve
our
current
goals
16/
of
environmental
protection
and
economic
growth
at
the
same
time
39%

OR
We
must
relax
environmental
standards
in
order
to
achieve
economic
growth
OR
6
We
must
accept
a
slower
rate
of
economic
growth
in
order
to
protect
the
environment
......................
27
No
opinion
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
9
No
cols.
17­
23
A­
IV­
13
certain
uses
of
the
chemical.
The
second
is
that
the
government
should
require
clear
warning
I
am
going
to
describe
four
different
kinds
of
chemicals
which
studies
have
shown
to
cause
cancer
in
some
people.
I
would
like
you
to
tell
me
which
one
of
the
approaches
listed
on
this
card
you
think
the
Federal
Government
should
take
for
each
chemical.
(
HAND
RESPONDENT
CARD)
The
first
approach
is
that
the
government
should
ban
labels
on
all
products
using
the
chemical
but
let
them
continue
to
be
sold.
The
third
approach
is
that
the
government
should
not
regulate
the
chemical
at
all.
(
READ
EACH
ITEM)

Warning
NOT
No
Ban
label
regulate
opinion
a.
The
first
kind
of
chemical
that
has
been
shown
to
cause
cancer
is
one
commonly
used
to
preserve
food
like
bacon.
Which
one
approach
should
the
government
take?
(
PROBE:
That
is,
chemicals
like
nitrites)
..................................

b.
The
second
is
a
chemical
used
to
color
food
like
in
some
hair
dyes
...............................

c.
The
third
is
saccharin
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

d.
the
fourth
is
a
chemical
used
to
color
food
like
hot
dogs,
soft
drinks
and
ice
cream.
(
PROBE:
that
is,
Red
Dye
#
2)
..
I.........................
33%

31
16
47
37%
5%
5%

60
3
6
66
12
­

44
4
5
17/

18/

19/

20/

38.
Even
though
human
beings
may
not
be
directly
affected,
there
are
various
risks
that
some
chemicals
pose
to
the
environment.
I
am
going
to
describe
some
of
these
risks
and
ask
you
what
action,
if
any,
you
think
the
government
should
take.
(
HAND
RESPONDENT
CARD)
(
READ
EACH
ITEM)

a.
First,
the
disposal
of
hazardous
chemical
wastes
far
out
in
the
ocean
where
they
harm
the
ocean
environment
where
they
are
disposed.
Do
you
think
the
government
should
prohibit
the
disposal,
in
pose
controls
but
not
prohibit
it,
urge
corporations
to
follow
certain
procedures
but
not
require
it,
or
not
take
any
action
at
all?

38a.
b.
c.
Disposal
chemicals
Disposal
in
ocean
in
pesticides
in
rivers
(
Col.
21)
(
Col.
22)
(
Col.
23)

Should
prohibit
........................
57%
36%
70%

Impose
controls
but
not
prohibit
it
....
29
52
21
Urge
corporations
to
follow
certain
procedures
but
not
require
it
..........
6
6
5
Not
take
any
action
at
all
..
.
.
.
.
.
.
.
.
.
.
.
2
1
1
No
opinion
........................
4
4
3
b.
What
about
the
use
of
chemicals
such
as
certain
pesticides
which
increase
crop
production
but
which
kill
wildlife.
Again,
looking
at
the
card,
what
action,
if
any,
do
you
think
the
government
should
take?
(
RECORD
ABOVE)

c.
And,
the
current
practice
of
releasing
into
rivers
certain
industrial
chemicals
which
have
been
shown
to
harm
fish.
What
action,
if
any,
do
you
think
the
government
should
take?
(
RECORD
ABOVE)
A­
IV­
14
Page
8
39.
Each
year
industry
introduces
thousands
of
new
chemicals
into
the
marketplace.
Some
people
think
that
companies
should
be
required
to
submit
information
to
the
government
before
any
of
the
new
chemicals
can
be
used
and
that
the
government
should
require
tests
of
those
chemicals
which
it
believes
may
be
dangerous
to
the
public.
Other
people
feel
that
government
screening
of
all
new
chemicals
would
be
expensive,
might
keep
potentially
useful
chemicals
off
the
market
and
that
it
is
enough
for
the
government
to
take
action
after
a
chemical
has
been
shown
to
cause
a
problem.
How
do
you
feel­­
should
the
government
(
1)
wait
until
a
chemical
already
in
use
has
been
shown
to
present
a
problem
before
taking
action
or
(
2)
conduct
an
extensive
screening
program
to
try
to
make
sure
all
chemicals
are
safe
before
they
are
used?

Wait
until
there
is
a
problem...
8%
14/

Conduct
screening
programs
before
chemical
is
used..
.
.
.
.
.
.
.
85
Depends
(
vol.)..................
4
No
opinion
...............­......
4
40.
Here
is
a
list
of
several
ways
to
get
energy.
(
HAND
RESPONDENT
CARD)
Looking
ahead
to
the
year
2000,
and
this
nation's
energy
needs,
which
two
or
three
of
these
sources
of
energy
do
you
think
we
should
concentrate
on
the
most?

READ
WHILE
RESPONDENT
LOOKS
AT
CARD:

This
list
includes
coal;
nuclear
energy;
energy
conservation
steps
such
as
more
and
better
home
insulation
and
cars
that
get
good
mileage;
water
power
from
dams
or
waterfalls;
solar
energy
including
energy
from
the
sun
and
the
wind;
oil
and
natural
gas;
and
synfuels
which
are
a
new
kind
of
fuel
made
by
industrial
plants
which
convert
oil
shale
into
oil
or
coal
to
liquid
or
gas.
Which
two
or
three
do
you
think
we
should
concentrate
on
the
most?

40.
Concentrate
on
most
(
Col.
25)

Coal..
.
.
*.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
36%

Nuclear
energy
.
.
.
.
_
.
.
.
.
.
.
.
.
.
.
.
.
.
25
33
Energy
conservation
....................

Water
power
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
31
Solar
energy
.
.
_.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
­.
61
Oil
and
natural
gas
..........
.._.....
28
26
Synfuels
.
.
.
.
.
.
.._.................

None
­

No
opinion
2
41.
Spend
least
effort
(
Col.
16)

9%

33
3
10
6
9
9
6
15
43.
Solar
energy
is
a
very
broad
term
that
includes
many
ways
of
using
the
sun
to
produce
energy.
Here
is
a
list
of
ways
that
can
be
used
by
individual
home
owners
to
produce
energy
from
the
sun.
(
HAN
h.
Passive
solar
desigh
such
as
having
houses
face
towards
the
south,
using
walls,
floors,
ceilings
or
water
containers
for
heat
storage,
using
skyline
for
heat
We
should
continue
to
built
more
nuclear
power
plant­
as
ncedec........................................

No
more
new
plants
should
be
planned
out
we
should
continue
to
use
the
ones
already
in
operation
and
finish
those
now
under
construction.
_.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
4­

We
should
stop
building
nuclear
plants
including
those
under
construction
and
shut
down
the
existing
ones
as
soon
as
possible................................
20
uncertain
9
.
.
.
..­............­.................­..........

2
So
answer
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
..­.

a.
Solar
panels
to
heat
water
for
homes
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

b.
Solar
panels
to
heat
space
in
homes
C
.
Wood
stoves
to
heat
space
in
homes
d.
Solar
swimming
pool,
hot
tub
spa
heaters.....................

e.
Solar
cells
to
produce
electricity.

f.
Small
windmills
to
produce
electricity.......................

g.
Using
farm
wasted
to
produce
gas
for
fuel.
.
.
.
.
..­...................
