ECONOMIC
ANALYSIS
OF
ENVIRONMENTAL
REGULATIONS:

APPLICATION
OF
THE
RANDOM
UTILITY
MODEL
TO
RECREATIONAL
BENEFIT
ASSESSMENT
FOR
THE
MP&
M
EFFLUENT
GUIDELINE
Final
Report
Lynne
G.
Tudor,
Ph.
D.

Engineering
and
Analysis
Division
Office
of
Water
US
EPA
401
M
Street,
SW
Washington,
DC
20460
Telephone:
(
202)
566­
1043
Tudor.
Lynne@
EPA.
gov
Elena
Besedin,
Ph.
D
.

Abt
Associates,
Inc.

55
Wheeler
Street
Cambridge,
MA
02138
Telephone:
(
617)
349­
2770
Elena_
Besedin@
abtasssoc.
com
Stuart
Smith
Abt
Associates,
Inc.

4800
M
ontgomery
Lane
Bethesda,
MD
20814­
5341
Telephone:
(
301)
347­
5619
Stuart_
Smith@
abtassoc.
com
December
2002
Ohio
Case
Study
2
TABLE
OF
CONTENTS
Introduction
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3
1
Methodology
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3
1.1
Overview
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3
1.2
Modeling
the
Site
Choice
Decision
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
5
1.3
Modeling
Trip
Participation
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
7
1.4
Calculating
Welfare
Changes
from
Water
Quality
Improvements
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
10
1.5
Extrapolating
Results
to
the
State
Level
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
11
2
Data
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
11
2.1
The
Ohio
Data
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
12
2.2
Estimating
the
Price
of
Visits
to
Sites
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
15
2.3
Site
Characteristics
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
15
3
Site
Choice
Model
Estimates
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
18
3.1
Fishing
M
odel
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
19
3.2
Boating
Model
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
20
3.3
Swimming
Model
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
21
3.4
Viewing
(
Near­
water
Activity)
Model
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
21
4
Trip
Participation
Model
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
21
5
Estimating
Benefits
from
Reduced
MP&
M
Discharges
in
Ohio
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
24
5.1
B
enefiting
Reaches
in
Ohio
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
24
5.2
E
stimating
Recreational
Benefits
in
Ohio
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
25
6
Limitations
and
Uncertainty
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
26
6.1
One­
State
Approach
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
26
6.2
Including
One­
Day
Trips
Only
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
27
6.3
Nonuse
Benefits
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
27
6.4
Potential
Sources
of
Survey
Bias
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
27
Glossary
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
29
Acronyms
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
31
References
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
32
Ohio
Case
Study
1
See
Chapter
15
of
 
Economic,
Environmental,
and
Benefits
Analysis
of
the
Final
Metal
Products
&
Machinery
Rule 
(
hereafter
EEBA)
for
detail.

3
INTRODUCTION
The
recreational
benefits
analysis
outlined
in
this
report
focuses
on
Ohio
as
a
case
study
of
the
MP&
M
regulation's
expected
benefits.
EPA
combined
water
quality
modeling
and
a
random
utility
model
of
consumer
behavior
(
RUM)
to
assess
how
changes
in
water
quality
from
the
MP&
M
regulation
will
affect
consumer
valuation
of
water
resources
for
recreational
uses.

The
RUM
analysis
provides
a
framework
for
estimating
the
effect
of
ambient
water
quality
and
other
site
characteristics
on
the
total
number
of
trips
taken
for
different
water­
based
recreation
activities
and
the
allocation
of
these
trips
among
particular
sites.

The
Agency
used
this
case
study
to
address
limitations
inherent
in
the
benefits
transfer
method
used
in
the
analysis
of
recreational
benefits
at
the
national
level.
1
The
RUM
model
assesses
water
quality
characteristics
directly
affected
by
the
MP&
M
regulation,
such
as
presence
of
ambient
water
quality
criteria
(
AWQC)
exceedances
and
nonconventional
nutrient
Total
Kjeldahl
Nitrogen
(
TKN)
concentrations
and
their
effect
on
recreation
behavior.
The
direct
link
between
the
water
quality
measures
included
in
the
RUM
model
and
the
water
quality
measures
affected
by
the
regulation,
as
well
as
the
site
specific
nature
of
the
analysis
reduce
uncertainty
in
benefit
estimates.
In
general,
RUM
models
are
well­
regarded
in
the
economic
literature
and
when
these
models
are
appropriately
applied,
the
results
are
thought
to
be
quite
reliable.

Benefits
transfer
results
are
subject
to
uncertainty
because
water
quality
changes
evaluated
in
available
recreation
demand
studies
are
only
roughly
comparable
with
water
quality
measures
considered
in
regulatory
development.
This
case
study
analysis
improves
upon
previous
recreation
demand
studies
that
focused
mainly
on
directly
observable
water
quality
effects,

e.
g.,
designated
use
support
(
i.
e.,
whether
a
water
body
supports
fishing),
the
presence
of
fish
advisories,
an
oil
sheen,
or
eutrophication.
The
O
hio
case
study
includes
unobservable
water
quality
effects
as
well.
The
MP&
M
regulation
affects
a
broad
range
of
pollutants,
many
of
which
are
toxic
to
human
and
aquatic
life
but
are
not
directly
observable
(
i.
e.,
priority
and
nonconventional
pollutants.
These
unobservable
toxic
pollutants
degrade
aquatic
habitats,
decrease
the
size
and
abundance
of
fish
and
other
aquatic
species,
increase
fish
deformities,
and
change
watershed
species
composition.
Water
quality
changes
(
i.
e.,
changes
in
toxic
pollutant
concentrations)
affect
consumers 
water
resource
valuation
for
recreation,

even
if
consumers
are
unaware
of
changes
in
ambient
pollutant
concentrations.

This
study
allows
for
a
more
complete
estimate
of
recreational
benefits
from
reduced
discharges
of
MP&
M
pollutants.
In
addition
to
estimates
of
recreational
benefits
from
reduced
frequency
of
AWQC
exceedances,
the
Ohio
case
study
evaluated
changes
in
the
water
resource
values
from
reduced
discharges
of
TKN.
The
analysis
also
values
additional
recreational
uses
not
addressed
in
the
national
analysis,
such
as
swimming.

The
study
used
data
from
the
National
Demand
Survey
for
Water­
Based
Recreation
(
NDS),
conducted
by
U.
S.
EPA
and
the
National
Forest
Service,
to
examine
the
effects
o
f
in­
stream
pollutant
concentrations
on
consumer
decisions
to
visit
a
particular
water
body
(
U.
S.
EPA,
1994).

1
METHODOLOGY
1.1
Overview
The
Ohio
study
combines
direct
simulation
and
inferential
analyses
to
assess
how
changes
in
water
quality
will
affect
consumers 
valuation
of
water
resources.

The
direct
simulation
analysis
component
estimates
baseline
and
post­
compliance
water
quality
at
recre
ation
sites
actually
visited
by
the
surveyed
consumers
and
all
other
sites
within
the
consumers'
choice
set,
visited
or
not.

The
inferential
analysis
component,
a
RUM
analysis
of
consumer
behavior,
estimates
the
effect
of
ambient
water
quality
and
other
site
characteristics
on
the
total
number
of
trips
taken
for
different
water­
based
recreation
activities
and
the
allocation
of
these
trips
among
particular
recreational
sites.
The
RUM
analysis
is
a
travel
cost
model
(
TCM),
in
which
the
cost
to
travel
to
a
particular
recreational
site
represents
the
 
price 
of
a
visit.
Ohio
Case
Study
2
The
Agency
also
attempted
a
model
structure
that
allows
for
interaction
among
the
choice
of
recreational
activities.
In
this
model,
a
person
first
chooses
a
recreational
activity
and
then
chooses
a
site.
This
model
did
not
perform
very
well
because
less
than
ten
percent
of
recreational
users
included
in
the
dataset
participate
in
all
four
activities.

4
The
main
advantage
of
the
RUM
model
is
inclusion
of
the
effect
of
substitute
sites
on
site
values.
For
any
particular
site,

assuming
that
it
is
not
totally
unique
in
nature,
the
availability
of
substitutes
makes
the
value
for
that
site
lower
than
it
would
be
without
available
substitutes.

EPA
modeled
two
consumer
decisions:


how
many
water­
based
recreational
trips
to
take
during
the
recreational
season
(
the
trip
participation
model);
and

conditional
on
the
first
decision,
which
recreation
site
to
choose
(
the
site
choice
model).

The
econometric
estimation
proceeded
in
two
steps,
each
correspo
nding
to
the
above
decisions.
The
Agency
estimated
these
decisions
in
reverse
order
(
i.
e.,
EPA
modeled
the
second
decision,
site
choice,
first).


Modeling
the
Site
Choice
Decision
.
Assuming
that
a
consumer
decides
to
take
a
water­
based
recreation
trip,
EPA
estimated
the
likelihood
that
the
consumer
will
choose
a
particular
site
as
a
function
of
site
characteristics,
the
price
paid
per
site
visit,
and
household
income.
A
consumer
weighs
the
attributes
for
various
"
choice
set"
sites
against
the
travel
costs
to
each
site.
These
travel
costs
include
both
the
cost
of
operating
a
vehicle
and
the
opportunity
costs
of
time
spent
traveling.
The
consumer
then
weighs
the
value
given
to
the
site's
attributes
against
the
co
st
of
getting
to
the
site
when
making
a
site
selection.
The
site
choice
model
estimates
how
recreational
users
value
access
to
sp
ecific
sites,
and
estimates
per
trip
economic
values
for
changes
in
water
quality
at
recreational
sites
in
the
study
area.

EPA
estimated
the
site
choice
model
using
a
two­
level
nested
multinomial
logit
(
NMNL)
model,
which
groups
sites
with
similar
characteristics.
The
nested
logit
model
assumes
that
individuals
first
choose
the
group
of
sites
and
then
a
site
within
that
group.
This
study
assumes
that
individuals
first
choose
a
water
body
type
(
Lake
Erie,
rivers,
or
small
lakes)
and
then
a
sp
ecific
site.
EPA
used
the
estimated
site­
choice
model
co
efficients
to
estimate
the
value
to
the
consumer
of
being
able
to
choose
among
Ohio
recreation
sites
on
a
given
day.
This
measure
is
referred
to
as
the
 
inclusive
value. 


Modeling
Trip
Frequency.
The
site
choice
models
estimated
in
the
previous
step
treat
the
total
number
of
recreational
trips
taken
each
season
as
exogenous
to
the
site
selection.
The
Agency
estimated
the
expected
number
of
trips
taken
during
the
recreation
season
using
a
Negative
Binomial
Poisson
model
(
Hausman
et
al.,

1995;
Feather
et
al.,
1995;
and
Creel
and
Loomis,
1992),
which
treats
trip
frequency
as
a
pre­
season
decision
regarding
total
participation
in
a
given
recreation
activity.

EPA
estimated
the
total
number
of
trips
during
the
recreation
season
as
a
function
of
the
expected
maximum
utility
(
inclusive
value)
from
recreational
activity
participation
on
a
trip,
and
socioeconomic
characteristics
affecting
demand
for
recreation
trips
(
e.
g.,
number
of
children
in
the
household).
The
coefficient
of
the
individual s
expected
maximum
utility
of
taking
a
trip)
provided
a
means
of
estimating
the
seasonal
welfare
effect
of
water
quality
improvements,
because
changes
in
water
quality
change
the
value
of
available
recreation
sites.

Estimating
the
site
choice
and
total
trip
p
articipation
models
jointly
is
theo
retically
possible,
but
computational
requirements
make
an
integrated
utility­
theoretic
model
infeasible.
EPA
estimated
separate
site
choice
and
trip
frequency
models
for
the
four
recreational
activities:
boating,
swimming,
fishing,
and
near­
water
recreation
(
e.
g.,
viewing
wildlife).
2
The
Agency
used
estimated
coefficients
of
the
indirect
utility
function
with
estimated
changes
in
water
quality
to
calculate
per­
trip
changes
in
consumer
welfare
from
improved
water
quality
at
recreation
sites
within
each
consumer
choice
set.
Trip
frequency
per
season
increases
if
site
water
quality
changes
are
substantial.
A
sample
consumer s
expected
seasonal
welfare
gain
is
therefore
a
function
of
both
welfare
gain
per
trip
and
the
estimated
change
in
number
of
trips
per
season.

Combining
the
trip
frequency
model s
prediction
of
trips
under
the
baseline
and
post­
compliance
and
the
site
choice
model s
corresponding
per­
trip
welfare
measure
yields
the
total
seasonal
welfare
measure.
Ohio
Case
Study
5
EPA
calculated
each
individual s
seasonal
welfare
gain
for
each
recreatio
n
activity
from
post­
compliance
water
quality
changes,
and
then
used
Census
population
data
to
aggregate
the
estimated
welfare
change
to
the
state
level.
The
sum
of
estimated
welfare
changes
over
the
four
recreation
activities
yielded
estimates
of
total
welfare
gain.

To
analyze
water
quality
improvement
benefits
in
the
RUM
framework,
EPA
used
available
discharge,
ambient
concentration,

and
other
relevant
data
to
measure
baseline
and
post­
compliance
water
quality
at
the
impact
sites.
Appendix
H
of
the
EEBA
report
provides
detail
on
water
quality
modeling
used
in
this
analysis.

1.2
Modeling
the
Site
Choice
Decision
EPA
used
the
RUM
framework
to
estimate
the
probability
of
a
consumer
visiting
a
recreation
site.
This
framework
is
based
on
the
assumption
that
a
consumer
derives
utility
from
the
recreational
activity
at
each
recreation
site.
Each
visit
decision
involves
choosing
one
site
and
excluding
others.

The
consumer s
decision
involves
comparing
each
site
and
choosing
the
site
that
produces
the
maximum
utility.
An
observer
cannot
measure
all
potential
determinants
of
consumer
utility,
so
the
indirect
utility
function
will
have
a
non­
random
element
(
V)
and
a
random
error
term
(

)
,
such
that
the
actual
determ
inants
of
consumer
utility
V

=
V
+

.
The
probability
(

jn)
that
site
j
will
be
visited
by
an
individual
n
is
defined
as:

(
1)

where:

V
jn
+

jn
=
utility
of
visiting
site
j,
and
V
sn
+

sn
=
utility
of
visiting
a
substitute
site.

Estimating
the
model
requires
specifying
the
functional
form
of
the
indirect
utility
function,
V,
in
which
site
choice
is
modeled
as
a
function
of
site
characteristics
and
the
 
price 
to
visit
particular
sites.
For
example,
a
set
of
conditional
utility
functions
(
one
for
each
site
alternative
j
in
the
choice
set)
can
be
determined
as
follows:

(
2)

where:

V
jn
=
the
utility
realized
from
a
conventional
budget­
constrained,
utility
maximization
model
conditional
on
choice
of
site
j
by
consumer
n;


M
=
marginal
utility
of
income;

M
jn
=
the
income
of
individual
n
availab
le
to
visit
site
j;

P
jn
=
a
composite
measure
of
travel
and
time
costs
for
consumer
n
on
site
alternative
j;


=
a
vector
of
coefficients
representing
the
marginal
utility
of
a
specified
site
characteristic
to
be
estimated
along
with

M
(
e.
g.,
size
of
the
water
body,
presence
of
boating
ramps);
and
X
jn
=
a
vector
of
site
characteristics
for
site
alternative
j
as
perceived
by
consumer
n.
These
characteristics
include
the
actual
monitored
and/
or
modeled
water
quality
parameters
that
are
hypothesized
to
be
determinants
of
consumer
valuation
of
water­
based
recreation
resources,
and
that
may
also
be
affected
by
the
MP&
M
regulation.

The
magnitude
of
the
coefficients
in
Equation
2
reflects
the
relative
importance
of
site
characteristics
when
consumers
decide
which
site
to
visit.
The
coefficients
(

)
of
water
quality
characteristics
of
recreation
sites
are
expected
to
be
po
sitive;
that
is,

all
else
being
equal,
consumers
of
water­
based
recreation
would
prefer
"
cleaner"
recreation
sites.
The
coefficient
on
travel
cost
is
expected
to
be
negative,
i.
e.,
consumers
prefer
lower
travel
costs.

To
estimate
the
site
choice
probabilities,
EPA
specified
and
estimated
a
nested
multinomial
logit
model
(
NMNL)
for
fishing,

boating,
and
swimming
activities.
The
nested
structure
explicitly
groups
similar
alternatives,
which
allows
for
a
richer
pattern
of
substitution
among
alternative
sites.
The
NMNL
is
based
on
the
assumption
that
an
individual
chooses
first
between
groups
of
alternatives
and
then,
within
the
chosen
group,
between
individ
ual
alternatives.
Fo
r
this
analysis,
EPA
grouped
all
recreational
sites
in
Ohio
by
water
body
type
based
on
site
similarities.
EPA
tested
various
alternative
site
groupings,
but
the
Ohio
Case
Study
3
Three
of
the
four
models
(
fishing,
boating,
and
swimming)
passed
specification
tests
for
appropriateness
of
a
nested
structure
(
see
Section
3
for
detail).
Test
results
showed
that
only
two
site
groups
are
appropriate
for
the
boating
and
swimming
models
inland
sites
(
rivers,
small
lakes,
and
reservoirs)
and
Lake
Erie
sites
in
Ohio.
The
fourth
activity,
wildlife
viewing,
did
not
pass
specification
tests
for
a
nested
structure
and
was
estimated
as
a
flat
multinomial
logit
(
MNL)
model.

6
models
presented
here
were
most
successful
at
explaining
the
probability
of
selecting
a
site.
The
best
model
used
the
following
activity­
specific
site
groupings:
3

Fishing
model:


Group
1:
Lake
Erie
sites;


Group
2:
river
sites;


Group
3:
small
lakes
and
reservoirs;


Boating
model:


Group
1:
Lake
Erie
sites;


Group
2:
inland
sites,
including
rivers,
small
lakes,
and
reservoirs;


Swimming
model:


Group
1:
Lake
Erie
sites;


Group
2:
inland
sites,
including
rivers,
small
lakes,
and
reservoirs;


Viewing
model:
EPA
used
a
non­
nested
model
in
which
an
individual
compares
all
sites
and
chooses
the
one
offering
the
highest
utility
level
for
each
trip
occasion.

First,
the
Agency
attempted
to
estimate
a
nested
model
based
on
the
three
water
body
types
lakes,
rivers,
and
Lake
Erie
for
all
four
recreational
activities
included
in
the
analysis.
This
structure,
however,
performed
well
only
for
fishing.
A
two­
nested
model
that
included
inland
and
Lake
Erie
sites
seemed
to
perform
better
for
the
bo
ating
and
swimming
models.

None
of
the
nested
structures
performed
well
for
participants
in
near­
water/
wildlife
viewing
activities.

This
finding
is
no
t
surprising
because
sites
are
grouped
based
on
their
similarities
within
a
given
nest.
It
is
reasonable
to
assume
that
inland
lakes,
rivers,
and
Lake
Erie
sites
are
dissimilar
from
an
angler's
point
of
view,
because
each
of
the
three
water
body
types
is
likely
to
support
different
fish
species.
Lake
sites
may
therefore
not
be
close
substitutes
for
rivers
sites.

For
other
activities,
differences
in
fishery
resources
across
water
body
types
are
unlikely
to
be
important.
Water
body
size
and
the
presence
of
recreational
amenities
are
likely
to
play
a
more
important
role
than
differences
in
fish
species
and
the
type
of
aquatic
habitat.
Lake
and
river
sites
may
therefore
be
regarded
as
substitutes
for
each
other
by
b
oaters
and
swimmers.

Lake
Erie,
on
the
other
hand,
is
a
unique
water
resource
that
differs
from
inland
water
bodies
because
of
its
physical
characteristics
(
e.
g.,
size
and
water
temperature);
river
and
lake
sites
are
therefore
not
likely
to
be
considered
substitutes
for
Lake
Erie
sites.
Finally,
participants
in
near­
water
rec
reation
use
water
resources
indirectly
and
are
therefore
more
likely
to
regard
recreational
sites
located
on
different
water
body
types
as
close
substitutes
to
each
other.
For
this
reason,
the
viewing
model
is
a
simple
logit
model
without
a
nested
structure.

The
models
assume
that
an
individual
first
decides
to
visit
a
specific
water
body
grouping
(
hereafter,
region),
then
decides
which
site
within
that
group
to
visit.
An
individual
probability
of
visiting
site
j,
given
the
choice
of
region
R,
is
a
simple
multinomial
logit.
If
the
random
terms

nj
for
individual
n
at
site
j
are
independently
and
identically
distributed
and
have
an
extreme
value
Weilbull
distribution,
then

jn
takes
the
form
(
McFadden,
1981):

(
3)

where:

 
jn|
r
=
probability
of
selecting
site
j
in
region
r;

=
the
consumer s
utility
from
visiting
site
j;

r
=
regions
­­
 
Lake
Erie, 
 
rivers, 
etc.
as
specified
above
for
a
given
activity;
and
Ohio
Case
Study
7
=
the
sum
of
the
consumer s
utility
at
each
site
j
for
all
sites
in
the
opportunity
set
for
region
r.

Estimated
parameters
of
the
indirect
utility
function
are
then
used
to
estimate
the
inclusive
value.
For
consumer
n,
the
inclusive
value
measures
the
overall
quality
of
recreational
opportunities
for
each
water­
based
activity
and
represents
the
expected
maximum
utility
of
taking
a
trip.
Note
that,
although
EPA
used
a
random
draw
from
the
opportunity
set
for
the
purpose
of
estimating
the
model
parameters,
the
Agency
calculated
the
inclusive
value
(
i.
e.,
the
expected
maximum
utility)

using
all
recreation
sites
in
the
consumer's
opportunity
set
in
a
given
region.

The
inclusive
value
is
calculated
as
the
log
of
the
denominator
in
Equation
2
(
McFadden,
1981).

(
4)

where:

I
r
=
inclusive
value
for
sites
associated
with
region
R;

=
individual
n s
utility
from
visiting
site
j;
and
W
=
a
vector
of
baseline
water
quality
characteristics.

The
probability
of
choosing
a
particular
region
is:

(
5)

where:

 
r
=
probability
of
selecting
region
r;

I
r
=
the
inclusive
values
for
a
given
region;

 
r
=
the
coefficient
on
the
inclusive
value
for
a
given
region;
and
r
=
activity­
specific
regions
(
e.
g.,
 
Lake
Erie, 
 
rivers, 
and
 
small
lakes 
for
fishing).

To
estimate
the
model
described
by
Equations
2
and
5,
EPA
used
a
standard
statistical
software
package,
LIMDEP.

1.3
Modeling
Trip
Participation
After
modeling
the
site
choice
decision,
the
next
step
modeled
the
determinants
of
the
number
of
water­
based
recreation
trips
a
consumer
takes
during
a
season.
To
link
the
quality
of
available
recreation
sites
with
consumer
demand
for
recreation
trips,

EPA
modeled
the
number
of
recreation
trips
taken
during
the
recreation
season
as
a
function
of
the
inclusive
value
estimated
in
the
previous
step
and
socioeconomic
characteristics
affecting
demand
for
recreation
activities.
The
dependent
variable,
the
number
of
recreatio
n
trips
taken
by
an
individual
during
the
recreation
season,
is
an
integer
value
greater
than
o
r
equal
to
zero.
To
account
for
the
non­
negative
property
of
the
dependent
variable,
EPA
used
count
data
models
based
on
probability
densities
that
have
the
non­
negative
integers
as
their
domain.

One
of
the
simplest
count
data
models
is
a
Poisson
estimation
process,
which
is
commonly
used
with
count
data,
such
as
number
of
recreation
trips
taken
during
the
recreation
season.
Inherent
in
the
model
specification
is
the
assumption
that
each
observation
of
a
number
of
trips
is
drawn
from
a
Poisson
distribution.
Such
a
distribution
favors
a
large
number
of
observations
with
small
values
(
e.
g.,
two
trips,
four
trips)
or
zeros,
resulting
in
its
being
skewed
toward
the
lower
end.
Due
to
the
nature
of
the
observed
number
of
trips,
it
is
quite
reasonable
to
assume
that
the
underlying
distribution
can
be
characterized
as
a
Poisson
distribution.
Figure
1
shows
the
number
of
recreation
trips
taken
per
year
and
the
number
of
respondents
who
reported
taking
that
number
of
trips.
Ohio
Case
Study
8
Figure
21.1:
Number
of
Trips
Per
Year
By
Activity
Type
Source:
U.
S.
EPA
analysis
of
NDS
data
(
U.
S.
EPA,
1994)
Ohio
Case
Study
9
Estimating
the
Poisson
model
is
similar
to
estimating
a
nonlinear
regression.
The
single
parameter
of
the
Poisson
distribution
is

,
which
is
both
the
mean
and
variance
of
y
n.
The
probability
that
the
actual
number
of
trips
taken
is
equal
to
the
estimated
number
of
trips
is
estimated
as
follows
(
Green,
1993):

(
6)

where:

Y
n
=
the
actual
number
of
trips
taken
by
an
individual
in
the
sample;

y
n
=
the
estimated
number
of
trips
taken
by
an
individual
in
the
sample;

n
=
1,
2,...,
N,
the
number
of
individuals
in
the
sample;
and

n
=


X,
expected
number
of
trips
for
an
individual
in
the
sample,
where
X
is
a
vector
of
variables
affecting
the
demand
for
recreational
trips
(
e.
g.,
inclusive
values
and
socioeconomic
characteristics)
and
 
is
the
vector
of
estimated
coefficients.

From
Equation
6,
the
expected
number
of
water­
based
recreatio
n
trips
per
recreation
activity
season
taken
by
an
individ
ual
is
given
by:

(
7)

where:

E[
y
n|
x
n]
=
the
expected
number
of
trips,
y
n,
given
x
n;
Var[
y
n|
x
n]
=
the
variance
of
the
number
trips,
y
n,
given
x
n;

=
a
vector
of
coefficients
on
x;
and
x
=
a
matrix
of
socioeconomic
variables
and
inclusive
values.

An
empirical
drawback
of
the
Poisson
model
is
that
the
variance
of
the
number
of
trips
taken
must
be
equal
to
the
mean
number
of
trips,
and
this
equality
is
not
always
supported
by
actual
data.
In
particular,
the
NDS
survey
data
exhibit
overdispersion,
a
condition
where
variance
exceeds
the
mean.
The
estimated
variance­
to­
mean
ratios
of
the
number
of
trips
in
the
NDS
data
sample
are
31,
27.9,
3
5.6,
and
10.5
for
fishing,
swimming,
viewing,
and
boating
trips,
resp
ectively.

Overdispersion
is
therefore
present
in
the
data
set.

To
address
the
problem
of
overdispersion,
EPA
used
the
negative
binomial
regression
model,
an
extension
of
the
Poisson
regression
model,
which
allows
the
variance
of
the
number
of
trips
to
differ
from
the
mean.
In
the
negative
binomial
model,

is
respecified
so
that
(
Green,
1993):

(
8)

where
the
error
term
(

)
has
a
gamma
distribution,
E[
exp(

i
)]
is
equal
to
1.0,
and
the
variance
of

is

.

The
resulting
probability
distribution
is:

(
9)

where:

y
n
=
0,1,2...
number
of
trips
taken
by
individual
n
in
the
sample;

n
=
1,2,...,
N
number
of
individuals
in
the
sample;
and

n
=
expected
number
of
trips
for
an
individual
in
the
sample.

Integrating

from
Equation
9
produces
the
unconditional
distribution
of
y
n.
The
negative
binomial
model
has
an
additional
parameter,
 ,
which
is
the
overdispersion
parameter,
such
that:
Ohio
Case
Study
10
(
10)

The
overdispersion
rate
is
then
given
by
the
following
equation:

(
11)

EPA
used
the
negative
binomial
model
to
predict
the
seasonal
number
of
recreation
trips
for
each
recreation
activity
based
on
the
inclusive
value,
individ
ual
socioeconomic
characteristics,
and
the
overdispersion
parameter,

.
If
the
inclusive
value
has
the
anticipated
positive
sign,
then
increases
in
the
inclusive
value
stemming
from
improved
ambient
water
quality
at
recreation
sites
will
lead
to
an
increase
in
the
number
of
trips.
The
combined
MNL
model
site
choice
and
count
data
trip
participation
models
allowed
the
Agency
to
account
for
changes
in
per­
trip
welfare
values,
and
for
increased
trip
participation
in
response
to
improved
ambient
water
quality
at
recreation
sites.

1.4
Calculating
Welfare
Changes
from
Water
Quality
Improvements
EPA
estimated
the
welfare
change
associated
with
water
quality
improvements
from
the
baseline
to
post­
compliance
conditions
as
a
compensating
variation
(
CV),
which
equates
the
expected
value
of
realized
utility
under
the
baseline
and
post­
compliance
conditions.
The
expected
seasonal
change
in
welfare
attributed
to
the
quality
improvements
for
an
individual
n
in
the
sample
consists
of
two
components:


per
trip
welfare
gain,
and

increased
number
of
trips
under
the
post­
compliance
water
quality
condition.

The
Agency
first
calculated
the
welfare
gain
from
water
quality
improvement
for
each
consumer
on
a
given
day
by
using
a
CV
measure
for
consumer
n
(
Kling
and
Thompson,
1996):

(
12)

where:

CV
n
=
the
compensating
variation
for
individual
n
at
site
j
on
a
given
day;

r
=
 
Lake
Erie, 
 
inland, 
etc.

j
=
1,...
J
r
represents
a
set
of
alternative
sites
for
a
given
recreational
activity
in
region
r;

=
the
inclusive
value
index
(
I);

W
0
=
a
vector
of
information
describing
baseline
water
quality;

W
1
=
a
vector
of
information
describing
post­
compliance
water
quality;
and

M
=
the
implicit
coefficient
on
income
that
influences
recreation
behavior.

In
deriving
Equation
12
,
EPA
assumed
that
the
marginal
utility
of
incom
e,

M
,
is
constant
across
alternatives
(
as
well
as
across
quality
changes).
If
this
assumption
does
not
ap
ply,
the
derivation
of
Eq.
12
is
more
complicated
(
Hausman
et
al.,

1995).
Ohio
Case
Study
4
EPA
selected
this
approach
for
calculating
seasonal
welfare
gain
per
individual
based
on
Dr.
Parsons 
recommendation
(
G.
R.
Parsons,
1999).

5
Section
2.1
provides
a
detailed
description
of
the
data
sample
used
in
the
analysis.

11
EPA
then
estimated
the
low
and
high
values
of
the
seasonal
welfare
gain
for
individual
n
in
the
sam
ple
as
follows:
4
(
13)

(
14)

where:

W
low,
n
=
lower
bound
estimate
of
the
seaso
nal
welfare
gain
for
individual
n;

W
high,
n
=
upper
bound
estimate
of
the
seasonal
welfare
gain
for
individual
n;

I
1
=
the
post­
policy
inclusive
value;

Y
1
=
the
estimated
number
of
trips
after
water
quality
improvement;

I
0
=
the
baseline
inclusive
value;

Y
0
=
the
estimated
number
of
trips
in
the
baseline;
and

µ
=
the
implicit
coefficient
on
income
that
influences
recreation
behavior.

These
estimates
are
per
individual
in
the
population
for
those
individuals
meeting
qualifications
for
inclusion
in
the
NDS
response
set
(
i.
e.,
respondents
whose
home
state
is
Ohio
and
respondents
from
the
neighboring
states
whose
last
trip
was
to
Ohio s
sites).
5
EPA
extrapolated
the
estimates
of
value
per
individual
to
the
Ohio
state
level
based
on
Census
data
(
U.
S.

Bureau
of
the
Census,
2000).
The
following
section
details
the
extrapolation
method
used
in
the
analysis.

1.5
Extrapolating
Results
to
the
State
Level
EPA
used
a
simplified
extrapolation
technique
to
estimate
the
state­
level
benefits.
EPA
first
estimated
the
number
of
participants
in
fishing,
swimming,
boating,
and
wildlife
viewing
in
Ohio,
based
on
the
estimated
percentage
of
the
NDS
survey
respondents
residing
in
Ohio
who
participate
in
a
given
activity
and
the
state
adult
po
pulation.
The
2000
Census
data
provide
information
on
the
number
of
Ohio
residents
aged
16
and
older.
EPA
then
multiplied
the
estimated
average
seasonal
welfare
gain
per
participant
in
a
given
recreational
activity
by
the
corresponding
number
of
recreational
users.
The
total
welfare
gain
to
the
users
of
water­
based
recreation
in
Ohio
is
the
sum
of
fishing,
swimming,
boating,
and
wildlife
viewing
benefits.

2
DATA
This
section
describes
the
data
and
supporting
analyses
required
to
implement
the
RUM
analysis.
The
following
general
categories
of
data
and
supporting
analyses
are
required:


information
on
the
consumers
of
water­
based
recreation
responding
to
the
NDS
in
Ohio;


recreation
sites
identified
for
the
water
quality
and
RUM
analyses,
including
the
sites
visited
by
consumers
of
waterbased
recreational
activity
and
supplemental
sites
in
their
choice
sets;


estimated
price
of
visiting
the
sites.
The
 
visit
price 
is
estimated
as
a
function
of
travel
distance
(
and
travel
time)

between
each
consumer s
hometown
and
each
site
in
the
choice
set;
and

information
on
site
characteristics
likely
to
be
important
determinants
of
consumer
behavior.
Of
particular
importance
to
this
analysis
are
the
water
quality
and
related
characteristics
of
sites
in
the
choice
set,
and
how
those
characteristics
may
be
expected
to
change
as
a
result
of
regulation.
Ohio
Case
Study
6
These
additional
observations
total
11
across
the
four
activities
and
thus
represent
only
a
small
fraction
of
total
observations.

Including
only
Ohio
respondents
in
the
trip
participation
models
underestimates
the
benefits
associated
with
water
quality
improvements,

because
the
welfare
gains
to
recreators
from
neighboring
states
are
ignored.

12
The
following
sections
discuss
each
category
of
data
and/
or
supporting
analysis
below.

2.1
The
Ohio
Data
EPA
obtained
information
on
survey
respondent
socioeconomic
characteristics
and
recreation
behavior,
including
last
trip
profile
and
the
annual
number
of
trips
associated
with
each
water­
based
activity,
from
the
NDS
(
U.
S.
EPA,
1994).
The
1994
survey
collected
data
on
demographic
characteristics
and
water­
based
recreation
behavior
using
a
nationwide
stratified
random
sample
of
13,059
individuals
aged
16
and
over.
Respondents
reported
on
water­
based
recreation
trips
taken
within
the
past
12
months,
including
the
primary
purpose
of
their
trips
(
e.
g.,
fishing,
boating,
swimming,
and
viewing),
total
number
of
trips,
trip
length,
distance
to
the
recreatio
n
site(
s),
and
number
of
participants.
W
here
fishing
was
the
primary
purpose
of
a
trip,
respondents
were
also
asked
to
state
the
number
of
fish
caught.
Table
1
shows
the
number
of
trips
taken
per
year
by
primary
recreation
activity,
as
reported
in
the
NDS.

EPA
selected
case
study
observations
for
Ohio
residents
who
took
trips
within
or
outside
of
the
sta
te.
Trips
to
O
hio
recreation
sites
by
residents
of
neighboring
states
were
also
included
in
the
site
choice
models,
but
not
in
the
trip
participation
models.
6
All
four
activity
models
included
single­
day
trips
only.
EPA
included
only
activity
participants
with
valid
hometown
ZIP
codes,
whose
destination
site
was
uniquely
identified.
The
Agency
used
data
on
both
Ohio
participants
and
Ohio
non­
participants
to
estimate
total
seasonal
trips,
but
included
only
Ohio
participants
and
several
residents
of
nearby
states
in
the
site
choice
models.
Although
they
could
not
be
used
in
the
site
choice
model,
participant
observations
from
Ohio
with
missing
location
information
were
used
to
analyze
the
number
of
trips.
Tables
1
and
2
list
valid
observations
by
activity,

residence,
and
model
type.
Figure
2
illustrates
the
distribution
of
the
sample
observations
in
relation
to
the
location
of
MP&
M
facilities
affected
by
the
rule
in
Ohio.

Table
1:
Classification
of
Sample
Observations
for
Estimation
of
the
Site
Choice
Models
Total
Ohio
Residents
Ohio
Residents
with
Last
Trip
In­
State
Valid
Ohio
Residents
with
Last
Trip
In­

State
Valid
Ohio
Residents
with
Last
Trip
Outside
State
Valid
Nonresidents
with
Last
Trip
in
Ohio
Valid
for
Site
Choice
Model
Participants
(
Total)
609
408
237
35
11
297
Fishing
122
103
66
9
0
84
Swimming
147
100
58
14
2
76
Viewing
231
126
64
2
7
73
Boating
109
79
49
10
2
64
Source:
U.
S.
EPA
analysis.
Ohio
Case
Study
13
Table
2:
Classification
of
Sample
Observations
for
Estimation
of
the
Trip
Participation
Models
Ohio
Residents
Total
Residents
with
Last
Trip
In­
State
Residents
with
Last
Trip
Outside
State
Valid
for
Trip
Participation
Model
Non­
Participants
300
291
Participants
(
Total)
609
408
34
322
Fishing
122
103
4
84
Swimming
147
100
9
78
Viewing
231
126
7
75
Boating
109
79
14
85
Total
Observations
909
408
34
613
Source:
U.
S.
EPA
analysis.
Ohio
Case
Study
14
Figure
2:
Location
of
MP&
M
Facilities
in
Relation
to
the
Visited
Sites
Source:
U.
S.
EPA
analysis.
Ohio
Case
Study
7
The
program
was
created
by
Daniel
Hellerstein
and
is
available
through
the
USDA
at
http://
usda.
maunlib.
cornell.
edu/
datasets/
general/
93014.

8
Note
that
all
expenditures
are
in
1993
dollars
because
the
NDS
trip
choices
and
the
associated
expenditure
occurred
in
1993.

9
The
estimate
of
motor
vehicle
cost
per
mile
was
based
on
estimates
compiled
by
the
Insurance
Information
Institute.

10
The
average
travel
time
to
the
visited
site
was
fairly
uniform
across
the
activities.
Average
one­
way
time
to
the
visited
site
was
51.38
minutes,
71.64
minutes,
43.76
minutes,
and
58.57
minutes
for
fishing,
boating,
swimming,
and
viewing,
respectively.

15
2.2
Estimating
the
Price
of
Visits
to
Sites
EPA
estimated
trip
 
price 
for
each
consumer
of
water­
based
recreation
as
the
sum
of
travel
costs
plus
the
opportunity
cost
of
time,
following
the
procedure
described
in
Haab
et
al.
(
2000).
Based
on
Parsons
and
Kealy
(
1992),
this
study
assumed
that
time
spent
 
on­
site 
is
constant
across
sites
and
can
be
ignored
in
the
p
rice
calculation.

To
estimate
consumers 
travel
costs,
EPA
first
used
ZipFip
software
to
calculate
the
one­
way
distance
to
each
site
for
each
participant.
7
The
average
estimated
one­
way
distance
to
the
site
visited
is
37.56
miles.
EPA
then
multiplied
round­
trip
distance
by
average
motor
vehicle
cost
per
mile
($
0.29,
1993
do
llars).
8,9
The
model
adds
the
opportunity
cost
of
travel
time,

measured
in
terms
of
wages
lost,
to
the
travel
cost
for
those
who
would
have
lost
income
by
taking
the
recreation
trip.
For
these
consumers
the
dummy
variable
LOSEINC
equals
one.
Travel
times
equal
the
round­
trip
distance
divided
by
a
travel
speed
of
40
mph
and
multiplied
by
the
individual s
hourly
wage
as
calculated
below.

The
travel
cost
variable
in
the
mo
del
was
calculated
as
follows:

(
15)

Individuals
not
losing
income
(
e.
g.,
individuals
taking
vacation
or
a
weekend
trip
or
individuals
whose
work
schedule
is
not
flexible)
do
not
face
lost
wages
as
a
result
of
the
trip
and
inclusion
of
the
opportunity
cost
of
time
would
be
inappropriate.

These
consumers
still
have
an
opportunity
cost
for
their
travel
time,
which
could
otherwise
be
spent
doing
something
else,

like
fishing.
In
other
words,
a
shorter
distance
traveled
allows
for
a
longer
time
spent
at
the
recreation
site.
For
these
consumers,
the
analysis
included
an
ad
ditional
round­
trip
travel
time
variable
calculated
as:

(
16)

The
average
one­
way
estimated
travel
time
to
the
visited
site
is
56.34
minutes.
10
2.3
Site
Characteristics
EPA
identified
1,954
recreation
sites
on
1,631
reaches
in
the
universal
opportunity
set.
Of
these,
580
observations
are
known
recreational
sites
(
e.
g.
parks);
1,366
observations
are
Reach
File
1
(
RF1)
reaches
without
a
known
recreational
site;
and
eight
observations
are
neither
located
in
RF1
nor
identified
as
known
recreation
sites
but
were
visited
by
an
NDS
respondent.
Ohio
Case
Study
11
McFadden
(
1981)
has
shown
that
estimating
a
model
using
random
draws
can
give
unbiased
estimates
of
the
model
with
the
full
set
of
alternatives.

12
Travel
distance
from
respondent s
hometown
to
the
Lake
Erie
sites
did
not
exceed
250
miles.

16
Each
consumer
choice
set
theoretically
includes
hundreds
of
substitutable
recreation
sites
in
Ohio
and
in
the
neighboring
states.
To
prevent
the
recreation
site
analysis
from
becoming
unmanageable,
EPA
analyzed
a
sample
of
recreation
sites
for
each
consumer
observation.
The
Agency
then
created
a
randomly­
drawn
reduced
choice
set
for
each
recreational
activity
as
follows:
11

Fishing.
The
reduced
choice
set
consists
of
20
Lake
Erie
sites,
20
river
sites,
and
20
small
lakes/
reservoirs.
Thus,
a
total
individual
choice
set
consists
of
60
alternatives
(
including
the
chosen
site);


Boating.
The
reduced
choice
set
consists
of
20
Lake
Erie
sites
and
20
inland
recreation
sites
(
including
rivers
and
lakes/
reservoirs).
A
total
individual
choice
set
consists
of
40
alternative
sites
(
including
the
chosen
site);


Swimming.
Similar
to
boating,
the
reduced
choice
set
consists
of
20
Lake
Erie
sites
and
20
inland
recreation
sites
(
including
rivers
and
lakes/
reservoirs).
A
total
individual
choice
set
consists
of
40
alternative
sites
(
including
the
chosen
site);


Wildlife
View
ing.
The
reduced
choice
set
consists
of
40
sites,
including
Lake
Erie,
river,
and
small
lake/
reservoir
sites.

Each
participant
choice
set,
by
definition,
includes
the
site
actually
visited
by
the
respondent.
For
each
consumer,
EPA
drew
the
additional
sites
from
a
geographic
area
defined
by
a
distance
constraint
(
and
the
water
body
types
listed
above).
The
Agency
used
a
120­
mile
distance
limit
for
inland
recreation
sites
(
Ohio
rivers,
small
lakes,
or
reservoirs).
All
Lake
Erie
sites
are
eligible
for
inclusion
in
the
choice
sets
for
all
models.
EPA
assumed
that
consumers
of
water­
based
recreation
would
be
willing
to
travel
farther
to
visit
Lake
Erie
sites,
because
this
water
resource
presents
unique
recreational
opportunities.
12
EPA
used
the
resulting
aggregate
choice
set
of
sites
for
all
individuals
participating
in
a
given
recreation
activity
to
model
consumer
decisions
regarding
trip
allocation
across
recreation
sites.

The
Agency
used
two
classes
of
characteristics
to
estimate
site
choice:


those
unaffected
by
the
MP&
M
regulation,
but
likely
to
determine
valuation
of
water­
based
recreational
resources;

and

those
affected
by
the
regulation
and
hypothesized
to
be
significant
in
explaining
recreation
behavior
and
resource
valuation.

Regulation­
independent
site
characteristics
include
water
body
type
and
size,
location
characteristics,
and
the
presence
of
site
amenities
(
e.
g.,
boat
ramps,
swimming
beaches,
picnic
areas).
Regulation­
dependent
site
characteristics
include
regulationaffected
water
quality
variables.

a.
Regulation­
independent
site
characteristics
Site
characteristics
that
are
likely
to
be
important
determinants
of
consumer
valuation
of
water­
based
recreational
resources
but
that
are
independent
of
the
MP&
M
regulation
include
general
site
descriptors.
These
descriptors
include
the
type
and
size
of
the
water
body
and
location
characteristics,
and
the
presence
of
site
amenities.
EPA
obtained
data
on
regulationindependent
site
characteristics
from
two
main
sources,
RF1
and
the
Ohio
Department
of
Natural
Resources
(
ODNR).

RF1
provided
water
body
type
(
i.
e.,
lake,
river,
or
reservoir)
and
physical
dimension
(
i.
e.,
length,
width,
and
depth).
The
dummy
variables,
LAKE
ERIE,
RIVER,
and
LAKE
characterize
water
body
types.
If
a
site
is
located
on
Lake
Erie,
LAKE
ERIE
takes
the
value
of
1;
0
otherwise.
If
a
site
is
located
on
river,
RIVER
takes
the
value
o
f
1;
0
otherwise.
Finally,
if
a
site
is
located
on
a
small
lake
or
reservoir,
LAKE
takes
the
value
1;
0
otherwise.
Water
body
size
was
determined
by
the
length
of
the
reach
segment
in
miles
for
rivers
and
Lake
Erie
sites.
For
small
lakes
and
reservoirs,
the
appropriate
water
body
size
is
the
water
body
area
in
acres.
The
site
choice
models
use
the
logarithm
of
water
body
size
as
a
measure
of
site
importance,
Ohio
Case
Study
13
EPA
uses
the
logarithm
of
water
body
size
because
it
expects
the
effect
of
water
body
size
on
utility
to
diminish
as
that
size
increases.

17
because
people
are
more
likely
to
be
aware
of
large
water
bodies.
13
Water
body
size
data
for
sites
not
located
in
RF1
came
from
the
ODNR.

ODNR,
supplemented
by
the
Ohio
Atlas
and
Gazetteer,
provided
data
on
recreational
amenities
and
site
setting
(
e.
g.,

presence/
absence
of
boat
ramps,
swimming
beaches,
or
picnic
areas;
public
accessibility;
and
size
of
land
available
for
recreation).
EPA
used
land
available
for
recreation,
LN(
LAND),
(
e.
g.,
acreage
of
state
park,
fishing,
hunting,
and
other
recreation
areas)
to
approximate
site
setting
and
attractiveness.
Dummy
variables
represent
the
presence
of
three
recreational
amenities:
BEACH
is
a
swimming
beach;
RAM
P
is
a
boating
ramp;
and
PARK
indicates
a
park.

b.
Regulation­
dependent
site
characteristics
Selecting
regulation­
dependent
site
characteristic
variables
that
are
both
policy­
relevant
and
significant
in
explaining
recreation
behavior
proved
challenging.
MP&
M
facilities
discharge
many
pollutants,
m
ost
of
them
unlikely
to
have
visible
indicators
of
degraded
water
quality
(
e.
g.,
odor,
reduced
turbidity,
etc).
EPA
hypothesized
that
pollutant
loadings
can,

nonetheless,
reduce
the
likelihood
of
selecting
a
recreation
site.
Reduced
pollutant
discharges
improve
water
quality
and
aquatic
habitat,
thereby
increasing
fish
populations
and
enhancing
the
recreational
fishing
experience.
In
addition,
in­
stream
nutrient
concentrations
are
good
predictors
o
f
eutrop
hication,
which
causes
aesthetic
losses
and
may
thus
affect
the
utility
of
a
water
resource
for
all
four
recreational
uses.

The
connection
between
the
policy
variables
(
i.
e.,
the
change
in
concentrations
of
MP&
M
pollutants)
and
the
effects
perceived
by
consumers
(
e.
g.,
increased
catch
rate,
increased
size
of
fish,
greater
diversity
of
sp
ecies,
or
improved
aesthetic
qualities
of
the
water
body)
are
not
modeled
directly,
but
are
captured
implicitly
in
the
differential
valuation
of
water
resources
as
reflected
in
the
RUM
analyses.

EPA
considered
two
types
of
pollutant
effects
in
defining
water
quality
variables
for
model
inclusion:


visible
or
otherwise
directly
perceivable
effects
(
e.
g.,
water
turbidity);
and

unobservable
toxic
effects
likely
to
imp
act
aquatic
habitat
and
species
adversely.

The
Agency
accounted
for
directly
observable
effects
using
the
ambient
concentrations
of
nutrients
(
e.
g.,
TKN)
as
an
explanatory
variable.

Rather
than
include
the
concentrations
of
all
toxic
pollutants
separately,
EPA
constructed
a
variable
to
reflect
the
adverse
impact
potential
of
toxic
pollutants
on
aquatic
habitat.
EPA
identified
recreation
sites
at
which
estimated
concentrations
of
one
or
more
MP&
M
pollutants
exceeds
AWQC
limits
for
aquatic
life
protection,
to
assess
the
likely
adverse
impacts
on
aquatic
organisms.
A
dummy
variable,
AWQC_
EX,
takes
the
value
of
1
if
in­
stream
concentrations
of
at
least
one
MP&
M
pollutant
exceed
AWQC
limits
for
aquatic
life
protection,
0
otherwise.
This
approach
accounts
for
the
fact
that
adverse
effects
on
aquatic
hab
itat
are
no
t
likely
to
occur
below
a
certain
thresho
ld
level.

c.
Biological
factors
Numerous
biological
parameters
(
e.
g.,
abundance
of
sport
fish)
that
are
a
function
of
the
availability
and
quality
of
suitab
le
habitat
for
breeding
and
feeding
are
also
likely
to
affect
recreation
behavior.
To
account
for
biological
parameters
affecting
the
demand
for
water­
based
recreation,
EPA
used
relative
fish
abundance
(
Biomass)
obtained
from
the
Ohio
Water
Resource
Inventory
(
OWRI)
database
(
OH
EPA,
1996).
Relative
fish
abundance
is
measured
as
the
total
fish
weight
(
in
kg)
per
300
meters.
Because
this
variable
reflects
presence
of
both
tolerant
and
intolerant
fish
species,
it
is
less
correlated
with
the
two
regulation­
dependent
water
quality
variables
(
i.
e.,
TKN
and
AWQC)
included
in
the
analysis
compared
to
the
index
of
well­
being
(
IWB2)
used
in
the
proposed
rule
analysis.

Chemical
properties
of
the
waters
(
e.
g.,
pollutant
concentrations)
are
likely
to
affect
the
diversity
and
abundance
of
the
fishery
resources.
Biological
parameters
may
also
be
affected
by
numerous
anthropogenic
stressors
unrelated
to
water
quality,
such
as
over­
fishing,
physical
alteration
of
habitat,
invasion
of
exotic
species,
etc.
Although
EPA
used
the
baseline
values
of
relative
fish
abundance
to
estimate
the
site
choice
models,
the
Agency
did
not
estimate
changes
in
biological
parameters
caused
by
the
regulation
analysis
due
to
data
limitations
and
the
challenges
posed
by
modeling
population
impacts
of
a
broad
spectrum
of
pollutants
at
hundreds
of
recreation
sites.
Ohio
Case
Study
18
d.
Presence
of
fish
advisories
Another
important
factor
that
may
affect
a
recreational
consumer s
decision
to
visit
a
pa
rticular
site
is
presence
of
fish
consumption
or
contact
advisories
(
FCAs).
EPA
obtained
information
on
fish
consumption
advisories
and
contact
advisories
at
reaches
in
Ohio
from
the
ODNR
(
Ohio
DNR,
1999).
Fish
consumption
advisories
and
contact
advisories
were
listed
by
the
name
of
the
stream
or
river
with
the
consumption
advisory.
An
advisory
that
applied
to
only
part
of
the
river
included
the
names
of
cities,
towns,
or
highways
to
identify
the
stretch
of
the
reach
for
which
the
advisory
was
relevant.
The
name
of
the
river
and
the
other
geographic
identifiers
were
used
to
assign
reach
numbers
from
RF1
to
the
consumption
advisories.
EPA
created
a
dummy
variable
for
each
type
of
advisory
(
i.
e.,
fish
advisories
and
contact
advisories).
The
variable
takes
the
value
of
1
if
the
relevant
advisories
are
present;
0
otherwise.

3
SITE
CHOICE
MODEL
ESTIMATES
EPA
estimated
four
separate
models
of
recreational
demand:
fishing,
boating,
swimming,
and
viewing.
The
Agency
classified
trips
by
the
primary
activity
listed
by
the
respondent.
All
four
activity
mo
dels
co
ver
single­
day
trips.
EPA
estimated
the
site
choice
model
using
the
site
actually
visited
and
randomly­
drawn
sites
from
the
choice
set
for
each
recreation
activity
as
described
in
Section
2­
3
above.

EPA
estimated
activity
models
for
five
alternative
choice
sets
(
i.
e.,
five
random
draws
from
the
universal
choice
set),

producing
five
sets
of
estimated
coefficients.
Mean
estimates
from
the
five
alternative
draws
represent
EPA s
best
estimate
of
actual
coefficient
values.
Table
3
lists
the
variables
used
as
arguments
in
the
utility
function
and
presents
the
mean
estimation
results
for
the
four
models.
In
estimating
site
choice
models
for
fishing,
boating,
swimming,
and
viewing,
the
Agency
restricted
the
coefficient
on
travel
cost
to
be
equal
across
all
four
models
to
ensure
a
constant
marginal
utility
of
income
across
all
four
activities.

The
following
sections
provide
a
short
description
of
the
results
of
the
site
choice
model
corresponding
to
each
recreation
activity.
Ohio
Case
Study
19
Table
3:
Site
Choice
Model
Estimation
Results
(
Mean
parameter
estimates
from
five
random
draws)
a
Variable
Activity
Fishing
Boating
Swimming
Viewing
TRCOST
b
­
0.044
(­
22.704)
­
0.044
(­
22.704)
­
0.044
(­
22.704)
­
0.044
(­
22.704)

TIME
c
­
1.474
(­
7.482)
­
0.362
(­
4.27)
­
0.436
(­
7.007)
­
0.719
(­
12.647)

RAMP
d
0.878
(
7.509)
N/
A
N/
A
N/
A
LN(
LAND)
e
N/
A
N/
A
0.058
(
2.431)
0.162
(
7.471)

PARK
f
N/
A
N/
A
0.753
(
3.79)
0.787
(
4.638)

BEACH
g
N/
A
N/
A
0.491
(
2.96)
N/
A
LN(
SIZE)
h
All
N/
A
0.502
(
5.777)
­
0.273
(­
6.083)
N/
A
Lake
Erie
0.908
(
6.639)
N/
A
N/
A
0.665
(
10.474)

River
0.171(
1.993)
N/
A
N/
A
­
0.261
(­
4.937)

Lake
0.050
(­
0.348)
N/
A
N/
A
­
0.429
(­
4.329)

Biomass
i
Lake
Erie
N/
A
­
0.130
(­
1.777)
N/
A
N/
A
River
0.068
(
2.328)
0.017
(
0.4432)
N/
A
N/
A
TKN
j
­
0.584
(­
3.763)
­
1.187
(­
6.863)
­
0.660
(­
4.631)
­
0.711
(­
4.401)

AWQC
k
­
0.573
(­
3.698)
­
0.172
(­
1.179)
N/
A
N/
A
Inclusive
Values
ERIE
0.811(
9.895)
0.296
(
6.098)
0.730
(
7.466)
N/
A
Inland
N/
A
0.088
(
2.525)
0.275
(
6.302)
N/
A
RIVER
0.591
(
6.945)
N/
A
N/
A
N/
A
LAKE
0.429
(
2.629)
N/
A
N/
A
N/
A
Adj.
R
2
0.467
0.280
0.408
a
EPA
performed
this
analysis
based
on
five
alternative
draws
to
assess
sensitivity
of
the
estimated
coefficients
with
respect
to
random
draws.

b
Travel
Cost
is
calculated
as
0.29
*
round­
trip
distance.

c
Travel
Time
is
(
round­
trip
distance
/
40)*
Wage.

d
1
if
a
boating
ramp
is
present,
and
0
otherwise.

e
Log
of
the
number
of
land
acres.

f
1
if
the
site
is
a
park,
and
0
otherwise.

g
1
if
a
swimming
beach
is
present,
and
0
otherwise.

h
Log
of
the
size
of
the
water
body.
For
rivers
and
Lake
Erie,
this
is
the
log
of
the
reach
segment
length
or
Lake
Erie
shore
segment
in
miles.
For
lakes,
this
is
log
of
the
lake
circumference.

i
Biomass
is
measured
as
the
total
fish
weight
(
in
kg)
per
300
meters.

j
In­
stream
concentrations
of
TKN
(
mg/
l).

k
1
for
any
reach
if
in­
stream
concentrations
of
at
least
one
MP&
M
pollutant
exceeds
the
AWQC
limits
for
protection
of
aquatic
life,
and
0
otherwise.

Note:
T­
statistic
for
test
that
the
estimated
coefficient
equals
0
is
given
in
parentheses
beside
the
coefficient
estimates.
N/
A
indicates
that
the
variable
was
not
included
in
the
estimation
for
this
activity.

Source:
U.
S.
EPA
analysis.

3.1
Fishing
Model
The
estimated
fishing
model
includes
travel
cost
(
TRCOST),
time
(
TIME)
spent
traveling,
and
site
characteristics.
The
Agency
included
the
following
site
characteristics
in
the
fishing
model:
boat
ramp
(
RAMP),
water
body
size
(
LN(
SIZE)),
Ohio
Case
Study
14
Inclusive
values
equal
to
1
cause
the
model
to
collapse
to
a
flat
multinomial
logit.

20
relative
fish
abundance
(
Biomass),
TKN
concentrations,
and
presence
of
AWQC
exceedances.
Table
3
shows
that
most
coefficients
have
the
expected
sign
and
are
significantly
different
from
zero
at
the
95
th
percentile.
Travel
cost
and
travel
time
have
a
negative
effect
on
the
probability
of
selecting
a
site,
indicating
that
anglers
prefer
to
visit
sites
closer
to
their
homes
(
other
things
being
equal).

Anglers
who
fish
from
a
boat
are
likely
to
view
the
presence
of
a
boat
ramp
as
an
important
factor
that
may
affect
their
site
choice.
However,
the
presence
of
a
boat
ramp
is
unlikely
to
be
important
for
anglers
who
fish
from
shore.
Thus,
the
Agency
used
an
interaction
variable
(
RAMP
x
USE_
BOAT)
such
that
the
ramp
variable
was
turned
on
only
if
the
angler
reported
using
a
b
oat
on
his
last
fishing
trip.
A
positive
sign
on
the
boat
ramp
indicates
that
anglers
owning
a
boat
are
more
likely
to
choose
sites
with
a
bo
at
ramp.

The
water
body
size
has
a
different
effect
on
the
probability
of
selecting
a
site
in
the
Lake
Erie,
river,
and
small
lake/
reservoir
groups.
The
larger
the
river
or
the
Lake
Erie
shore
segment,
the
more
likely
that
anglers
visited
the
reach.
The
size
of
inland
lakes
and
reservoirs
does
not
have
a
significant
effect
on
the
probability
of
visiting
the
site.

The
Agency
used
the
square
roo
t
of
the
fish
weight
per
300
meters
as
a
measure
of
fish
abundance
(
Biomass).
The
probability
of
a
river
site
visit
increases
as
the
relative
fish
abundance
at
the
site
increases.
However,
inclusion
of
this
variable
in
the
Lake
Erie
nest
was
not
significant,
which
indicates
that
relative
fish
abundance
does
not
have
a
significant
effect
on
choosing
a
Lake
Erie
site.
This
finding
is
counterintuitive
and
is
likely
to
be
due
to
the
lack
of
variation
in
the
relative
fish
abundance
variable
for
the
Lake
Erie
sites.
This
variable
was
excluded
from
the
Lake
Erie
nest
in
the
final
model
presented
here.
Data
on
relative
fish
abundance
were
not
availab
le
for
lakes.

Finally,
higher
ambient
concentrations
of
TKN,
which
indicate
potential
eutrophication
problems,
and
presence
of
AWQC
exceedances
negatively
affect
the
probability
of
site
selection.
In
other
words,
anglers
prefer
cleaner
sites,
all
else
being
equal.

Estimated
inclusive
values
on
Lake
Erie
sites,
rivers,
and
small
lakes
lie
within
a
unit
interval
[
0,1]
and
are
significantly
different
from
0,
indicating
that
the
nested
choice
structure
is
appropriate.
14
EPA
found
other
variables,
tested
as
explanatory
variables,
to
be
insignificant,
including
the
presence
of
FCAs.
It
might
be
expected,
a
priori,
that
the
presence
of
an
FCA
decreases
a
site s
likelihood
as
a
fishing
choice.
In
fact,
the
existence
of
FCAs
did
not
significantly
affect
a
site s
probability
of
being
chosen;
59
percent
of
the
sites
actually
chosen
by
NDS
respondents
had
an
FCA
in
place.
Creel
surveys
provided
by
ODNR
indicated
that,
on
average,
anglers
released
70
percent
of
their
catch
(
ODNR,
1997).
This
finding
suggests
that
recreational
anglers
are
aware
of
FCAs,
and
catch
but
do
not
consume
fish
in
the
affected
areas.

3.2
Boating
Model
The
estimated
boating
model
includes
travel
cost
(
TRCOST),
time
(
TIME)
spent
traveling,
and
site
characteristics.
The
Agency
included
the
following
site
characteristics
in
the
boating
model:
water
body
size
(
LN(
SIZE)),
relative
fish
abundance
(
Biomass),
TKN
concentrations,
and
presence
of
AWQC
exceedances.
Table
3
shows
that
most
coefficients
have
the
expected
sign
and
are
significantly
different
from
zero
at
the
95
th
percentile.

Travel
co
st
and
travel
time
have
a
negative
effect
on
the
probability
of
selecting
a
site,
indicating
that
boaters
prefer
to
visit
sites
closer
to
their
homes
(
other
things
being
equal).
However,
the
magnitude
of
the
travel
time
coefficient
indicates
that
boaters
are
willing
to
travel
farther
than
participants
in
other
recreational
activities.
This
is
not
surprising,
since
motorboating
and
sailing
are
restricted
to
the
sites
where
these
activities
are
allowed.
The
positive
coefficient
on
the
water
body
size
variable
(
LN(
SIZE))
ind
icates
that
the
larger
the
water
body
the
more
likely
the
boaters
visited
it.

The
coefficients
on
water
quality
variables
(
TKN
and
AW
QC)
are
negative,
indicating
that
boaters
prefer
to
visit
cleaner
sites.
The
Biomass
coefficient
is
positive,
but
insignificant
for
inland
sites,
and
negative
for
Lake
Erie
sites.
The
negative
coefficient
on
this
variable
is
likely
to
be
due
to
the
fact
that
88
percent
of
the
sample
trips
used
in
this
model
were
motorboating
trips.
Motorboating
itself
is
likely
to
be
a
significant
environmental
stresso
r
for
biological
communities
due
to
noise
and
turbidity
associated
with
this
activity.
Thus,
lower
fish
abundance
at
popular
boating
sites
may
indicate
that
intensive
motorboating
may
adversely
affect
species
abundance.
As
was
the
case
with
the
fishing
model,
the
estimated
Ohio
Case
Study
21
inclusive
value
is
significantly
different
from
zero
and
lies
within
a
unit
interval
[
0.1],
supporting
the
nested
model
framework.

3.3
Swimming
Model
EPA
included
the
travel
cost
and
time
variables
(
TRCOST,
TIME),
physical
characteristics
of
the
site,
and
ambient
TKN
concentrations
in
the
swimming
model.
This
model
also
includes
the
presence
of
recreational
amenities
that
are
likely
to
be
important
to
swimmers:
presence
of
a
beach,
the
designation
of
the
site
as
a
p
ark,
and
the
natural
log
of
the
land
acres.
All
estimated
coefficients
have
the
expected
sign
and
are
significantly
different
from
zero
at
the
95
th
percentile.

Price,
travel
time,
the
presence
of
a
park
with
a
beach,
and
the
size
of
the
land
area
around
the
site
all
increase
the
probability
of
a
particular
site
being
chosen
for
swimming.
Swimmers
are
less
likely
to
visit
large
sites
(
referring
to
the
size
of
the
water
body)
or
sites
with
visible
water
quality
effects
as
indicated
by
higher
in­
stream
concentrations
of
TKN.
As
for
the
fishing
and
boating
models,
the
estimated
inclusive
value
is
significantly
different
from
zero
and
lies
within
a
unit
interval
[
0,1]

supporting
the
nested
model
framework.

Again,
some
variables
expected
to
be
significant,
such
the
presence
of
contact
ad
visories,
are
no
t.
This
variable s
insignificance
probably
stems
from
its
scarcity.
Of
1,954
sites
included
in
the
universal
opportunity
set,
contact
advisories
are
in
place
for
only
13.
(
None
of
the
sites
actually
visited
had
contact
advisories
in
place.)
The
probability
that
a
chosen
site
has
contact
advisories
in
place
is
very
small,
because
ind
ividual
choice
sets
are
randomly
selected.

The
fish
Biomass
variable
representing
biological
characteristics
of
a
water
body
also
did
not
have
a
significant
influence
on
consumer
decisions
to
visit
a
particular
site
and
was
dropped
from
the
model.
This
outcome
is
not
surprising,
since
abundant
aquatic
life
may,
in
fact,
interfere
with
swimming
activities.

3.4
Viewing
(
Near­
water
Activity)
Model
EPA
included
the
travel
cost
and
time
variables
(
TRCOST,
TIME),
physical
site
characteristics,
and
ambient
TKN
concentrations
in
the
viewing
model.
In
addition,
the
Agency
included
the
natural
log
of
the
land
acres
and
the
designation
of
the
site
as
a
park.
All
estimated
coefficients
have
the
expected
sign
and
are
significantly
different
from
zero
at
the
95
th
percentile.

The
probability
of
choosing
a
site
for
near­
water
activities
is
most
significantly
related
to
visit
price,
travel
time,
land
size,

and
in­
stream
concentrations
of
TKN.
Similarly
to
the
fishing
model,
the
water
body
size
has
a
different
effect
on
the
probability
of
selecting
a
site
in
the
Lake
Erie,
river,
and
small
lake/
reservoir
groups.
The
larger
the
Lake
Erie
shore
segment,

the
more
likely
that
viewers
visited
the
site.
The
negative
coefficients
on
river
and
inland
lake
size
indicate
that
consumers
prefer
smaller
inland
water
bodies
for
near­
water
and
wildlife
viewing
activities.

4
TRIP
PARTICIPATION
MODEL
EPA
estimated
the
determinants
of
individual
choice
concerning
how
many
trips
to
take
during
a
recreation
season
with
a
separate
model
for
each
of
the
four
activities.
These
participation
models
rely
on
socioeconomic
data,
and
on
estimates
of
individual
utility
(
the
inclusive
value)
derived
from
the
site
choice
models.
V
ariables
of
imp
ortance
include
age,
ethnicity,

gender,
education,
and
the
presence
of
young
or
older
children
in
the
household.
W
hether
or
not
the
individual
owns
a
boat
is
particularly
important
in
boating
participation,
and
is
included
in
the
model
for
that
activity
only.
Variable
definitions
for
the
trip
participation
model
are:


IVBASE:
inclusive
value,
estimated
using
the
coefficients
obtained
from
the
site
choice
models;


#
TRIPS:
number
of
trips
taken
by
the
individual;


AGE:
individual s
age.
If
the
individual
did
not
report
age,
their
age
is
set
to
the
sample
mean;


MALE:
equals
1
if
the
individual
is
a
male,
0
otherwise;


NOHS:
equals
1
if
the
individual
did
not
complete
high
school,
0
otherwise;
Ohio
Case
Study
22

COLLEGE:
equals
1
if
the
individual
completed
college,
0
otherwise;


AFAM:
equals
1
if
the
individual
is
African
American,
0
otherwise;


YNGKIDS:
equals
1
if
there
are
kids
6
years
or
younger,
0
otherwise;


OLDKIDS:
equals
1
if
there
are
kids
7
years
or
older,
0
otherwise;


OWNBT:
equals
1
if
individual
owns
a
boat,
0
otherwise;


Constant:
a
constant
term
representing
each
individual s
utility
associated
with
not
taking
a
trip;
and


(
alpha):
overdispersion
parameter
estimated
by
the
Negative
Binomial
Model.

Table
4
presents
explanatory
variables
and
a
mean
value
for
each.

Table
4:
Mean
Values
for
Explanatory
Variables
Used
in
the
Participation
Models
Variables
(
Mean)
Non­
Participant
(
N=
291)
Boating
(
N=
85)
Fishing
(
N=
84)
Swimming
(
N=
78)
Viewing
(
N=
75)

#
TRIPS
0.00
7.71
10.07
9.46
9.59
AGE
43.99
39.06
38.53
34.76
36.91
MALE
0.33
0.49
0.65
0.47
0.47
NOHS
0.17
0.09
0.14
0.13
0.13
COLLEGE
0.15
0.32
0.20
0.32
0.35
AFAM
0.11
0.02
0.05
0.03
0.12
YNGKIDS
0.18
0.26
0.24
0.24
0.27
OLDKIDS
0.38
0.48
0.58
0.56
0.48
OWNBT
N/
A
0.53
N/
A
N/
A
N/
A
Source:
U.
S.
EPA
analysis.
Ohio
Case
Study
23
Table
5
presents
the
results
for
the
participation
models
of
the
four
recreation
activities.

Table
5:
Trip
Participation
Negative
Binomial
Model
Estimates
Variables/
Statistics
Boating
Fishing
Swimming
Viewing
IVBASE
0.12
0.82
0.72
0.47
(
0.71)
(
2.86
)
(
4.57)
(
3.66)

AGE
­
0.07
­
0.04
­
0.06
­
0.05
(­
4.73)
(
­
2.06)
(
­
2.24)
(
­
2.77)

MALE
1.23
2.22
1.15
0.91
(
2.75)
(
3.25)
(
1.52)
(
2.00)

NOHS
1.29
­
1.09
­
0.92
0.1
(
2.37)
(­
1.56
)
(
­
0.96)
(
0.17)

COLLEGE
­
0.19
­
0.40
0.53
1.22
(­
0.29)
(­
0.721
)
(
0.71)
(
2.05)

AFAM
­
3.74
­
1.44
­
4.07
­
1.16
(­
1.81)
(­
1.53
)
(
­
2.68)
(
­
1.34)

YNGKIDS
1.51
­
0.95
0.35
­
0.17
(
2.96)
(
­
1.26)
(
0.42
)
(
­
0.38)

OLDKIDS
­
1.67
1.11
0.4
0.8
(­
3.58)
(
2.78)
(
0.65)
(
1.81)

OWNBT
3.82
N/
A
N/
A
N/
A
(
5.26)

Constant
0.20
­
5.74
­
0.1
­
1.98
(
0.11)
(
­
3.01)
(
­
0.06)
(
­
1.6)

Alpha

5.77
9.03
8.92
8.17
(
5.85)
(
7.16)
(
6.78)
(
6.03)

Note:
T­
statistic
for
test
that
coefficient
equals
0
is
given
in
parentheses
below
the
coefficient
estimates.
N/
A
indicates
that
the
variable
was
not
included
in
the
estimation
for
this
activity.

Source:
U.
S.
EPA
analysis.

Parameter
estimates
of
the
inclusive
value
index
(
IVBASE)
in
the
swimming,
fishing,
and
viewing
models
are
positive
and
differ
significantly
from
zero
at
the
95
th
percentile,
indicating
that
water
quality
improvements
have
a
positive
effect
on
the
number
of
trips
taken
during
a
recreation
season.

The
estimated
coefficient
on
IVBASE
in
the
boating
model,
while
positive,
was
not
statistically
significant.
Taking
a
boating
trip
often
requires
more
preparation
(
e.
g.,
taking
a
boat
to
the
water
body)
than
taking
other
trips.
Therefore,
although
water
quality
improvements
increase
the
value
of
a
boating
day,
factors
other
than
water
quality
are
likely
to
have
a
stronger
impact
on
the
number
of
boating
trips
per
season.

The
AGE
variable
is
negative
and
significant
for
all
four
recreation
activities:
younger
people
are
likely
to
take
more
recreation
trips.
Ethnicity
and
gender
(
the
AFAM
and
MALE
variables)
also
have
a
significant
impact
on
whether
an
individual
participates
in
water­
based
recreation.
African
Americans
living
in
Ohio
are
less
likely
to
participate
in
any
of
the
four
recreation
activities
than
representatives
of
other
ethnic
groups.
Males
are
more
likely
than
females
to
participate
in
any
of
the
recreation
activities.

Education
also
influences
trip
frequency
significantly.
People
who
did
not
complete
high
school
(
NOHS=
1)
tend
to
take
fewer
fishing
or
swimming
trips.
Those
with
a
college
degree
(
COLLEGE=
1)
are
more
likely
to
participate
in
swimming
and
Ohio
Case
Study
24
viewing.
Respondents
who
attended
college
are
less
likely,
however,
to
participate
in
fishing
and
boating
than
those
who
completed
only
a
high
school
education.
For
the
boating
model,
the
COLLEGE
variable
is
not
significantly
different
from
zero.

The
presence
of
older
children
(
OLDKIDS)
in
the
household
is
associated
with
greater
participation
in
swimming,
viewing
(
near­
water
recreation),
and
fishing
activities,
but
is
no
t
a
significant
determ
inant
in
decisions
to
participate
in
boating.

Younger
children
in
the
household
(
YNGKIDS)
tends
to
lead
to
greater
participation
in
boating
and
swimming,
but
leads
to
fewer
fishing
or
viewing
trips.

5
ESTIMATING
BENEFITS
FROM
REDUCED
MP&
M
DISCHARGES
IN
OHIO
5.1
Benefiting
Reaches
in
Ohio
EPA
identified
reaches
where
it
expects
the
MP&
M
rule
to
eliminate
or
reduce
the
number
of
existing
AWQC
exceedances
(
hereafter,
benefiting
reaches).
The
Agency
first
identified
the
reaches
in
which
baseline
discharges
from
industrial
sources,

including
both
MP&
M
and
non­
MP&
M
facilities,
caused
one
or
more
pollutant
concentrations
to
exceed
AWQC
limits
for
aquatic
species.
A
reach
is
considered
to
benefit
from
the
MP&
M
rule
if
at
least
one
AWQC
exceedance
is
eliminated
due
to
reduced
MP&
M
discharges.
Although
the
method
for
identifying
benefiting
reaches
is
similar
to
the
method
used
in
the
national
analysis
(
see
Chapter
15
o
f
the
EEBA
report
for
detail),
there
are
three
notable
differences:


Unlike
the
national
analysis,
the
Ohio
case
study
incorporates
information
on
all
industrial
and
municipal
point
source
discharges
and
non­
point
sources
to
assess
in­
stream
concentrations
of
toxic
and
nonconventional
pollutants
in
the
baseline
and
post­
compliance.
Appendix
H
of
the
EEBA
report
provides
information
on
the
data
sources
and
methods
used
to
assess
ambient
water
quality
conditions
in
Ohio.


The
water
quality
model
used
in
this
analysis
estimates
ambient
pollutant
concentrations
in
the
reaches
receiving
discharges
from
MP&
M
facilities
and
reaches
below
the
initial
discharge
reach.
Appendix
H
of
the
EEBA
report
provides
detail
on
the
water
quality
model
used
in
this
analysis.


The
analysis
of
recreational
benefits
accounts
for
changes
in
TKN
concentrations.

EPA's
analysis
indicates
that
po
llutant
concentrations
at
the
baseline
discharge
levels
fro
m
all
industrial
sources
(
including
all
MP&
M
facilities)
exceed
acute
exposure
criteria
for
aquatic
life
on
15
reaches,
and
exceed
chronic
exposure
criteria
for
protection
of
aquatic
species
on
21
reaches.
EPA
estimates
that
reducing
pollutant
discharges
from
oily
waste
facilities
directly
discharging
to
the
receiving
streams
would
not
eliminate
all
concentrations
in
excess
of
the
acute
aq
uatic
life
exposure
criteria
or
the
chronic
exposure
criteria
on
any
reach
under
the
final
rule;
it
would
reduce
the
number
of
acute
and
chronic
exceedances
on
one
reach.

In
addition,
baseline
pollutant
concentrations
exceed
human
health­
based
AWQC
for
consumption
of
water
and
organisms
on
three
reaches
and
exceed
AQWC
for
consumption
of
organisms
only
on
two
reaches.
EPA
estimates
that
reducing
pollutant
discharges
from
oily
waste
facilities
directly
discharging
to
the
receiving
streams
would
red
uce
the
number
of
pollutants
exceeding
the
human
health­
based
AWQC
on
one
reach
under
the
final
rule;
it
would
not
eliminate
all
human
health­
based
AQWC
exceedances
on
any
reach
in
Ohio.
Table
6
summarizes
these
results.
In
addition,
the
final
regulation
is
estim
ated
to
reduce
in­
stream
concentrations
of
TKN
in
the
affected
reaches.
The
estimated
average
reductions
are
0.54
percent
in
lakes
and
0.45
percent
in
rivers
and
streams.
Ohio
Case
Study
15
The
choice
set
of
recreational
sites
available
to
boaters
was
restricted
to
the
sites
where
motorboating
and
sailing
is
permitted
because
the
majority
of
Ohio
boaters
included
in
this
analysis
used
either
motor
or
sail
boats.
Water
quality
improvements
at
the
sites
where
boating
is
not
allowed
does
not
result
in
welfare
gain
to
boaters.

25
Table
6.:
Estimated
MP&
M
Discharge
Reaches
with
MP&
M
Pollutant
Concentrations
in
Excess
of
AWQC
Limits
for
the
Oily
Wastes
Subcategory
for
Protection
of
Aquatic
Species
or
Human
Health
Regulatory
Status
Number
of
Reaches
with
Concentrations
Exceeding
AWQC
Limits
for
Human
Health
Number
of
Reaches
with
Concentrations
Exceeding
AWQC
Limits
for
Aquatic
Species
Number
of
Benefiting
Reaches
All
AWQC
Exceedances
Eliminated
Reaches
with
Some
AWQC
Exceedances
Eliminated
H20
and
Organisms
Org.

Only
Acute
Chronic
Baseline
3
2
15
21
­
­

Final
Regulation
3
2
15
21
0
1
Source:
U.
S.
EPA
analysis.

5.2
Estimating
Recreational
Benefits
in
Ohio
To
estimate
peoples 
willingness­
to­
pay
for
water
quality
improvements,
the
Agency
first
calculated
per­
person
seasonal
welfare
gain
corresponding
to
the
final
regulation.
Table
7
presents,
for
each
recreation
activity,
the
compensating
variation
per
trip
(
the
median
value
over
all
individuals
in
the
sam
ple)
associated
with
the
reduced
MP&
M
discharges.
Because
the
trip
choices
and
the
associated
expenditures
occurred
in
199
3,
the
welfare
gain
was
calculated
in
1993
dollars
and
then
adjusted
to
2001
dollars
based
on
the
Consumer
Price
Index
(
CPI).

The
model
indicates
that
the
reductions
in
MP&
M
discharges
from
the
final
regulatio
n
result
in
a
modest
increase
in
per­
trip
values
for
three
of
the
four
recreation
activities
(
fishing,
viewing,
and
swimming).
There
is
no
welfare
gain
to
boaters
from
improved
water
quality
under
the
post­
compliance
scenario.
15
Table
7
provides
the
mean
estimates
of
welfare
gain
per
recreational
user
in
Ohio.

Table
7:
Welfare
Gain
per
Recreational
User
in
Ohio
(
2001$)

Per
Trip
Welfare
Gain
Average
Number
of
Trips
per
Person
per
Year
Mean
Seasonal
Welfare
Gain
Fishing
$
0.02
13.6
$
0.17
Boating
$
0.00
6.22
$
0.00
Viewing
$
0.01
9.26
$
0.11
Swimming
$
0.01
8.72
$
0.01
Source:
U.
S.
EPA
analysis.

Table
7
also
reports
seasonal
compensating
variation
per
individual.
The
reported
seasonal
welfare
gain
includes
both
the
increase
in
the
utility
from
better
water
quality
at
the
available
recreation
sites
receiving
MP&
M
discharges
and
the
increase
in
utility
from
greater
recreational
trip
participatio
n.

As
noted
above,
the
Ohio
case
study
evaluated
changes
in
the
water
resource
values
from
both
reduced
discharges
of
TKN
and
reduced
frequency
of
AWQC
exceedances.
Changes
in
TKN
concentration
in
the
Ohio
water
bodies
resulting
from
reduced
MP&
M
discharges
from
the
Oily
Wastes
subcategory
account
for
approximately
96
percent
of
the
monetary
value
of
benefits
resulting
from
the
final
rule.
Ohio
Case
Study
16
Note
that
EPA
used
a
few
observation
on
visitors
from
neighboring
states
to
estimate
site
choice
models.
The
analysis
does
not
include
these
observations
in
calculating
state­
level
benefits
from
water
quality
improvements.

26
Both
the
per­
trip
and
seasonal
welfare
estimates
are
much
lower
than
values
reported
in
the
existing
studies
of
water­
based
recreation.
T
his
is
not
surprising,
since
the
water
quality
changes
expected
fro
m
the
final
rule
are
very
modest.

To
calculate
state­
leve
l
recrea
tional
benefits
from
the
final
rule,
EP
A
first
calculated
seasonal
welfare
gain
from
water
quality
improvements
per
individual
in
the
sample.
The
Agency
then
multiplied
the
average
welfare
gain
per
individual
by
the
corresponding
number
of
participants
in
a
given
activity
(
see
Section
1.5
above
for
detail).
The
resulting
product
is
the
annual
benefit
from
the
final
MP&
M
rule
to
consumers
of
a
given
water­
based
recreation
activity
in
Ohio.
Table
8
summarizes
state­
level
results.

Table
8:
Estimated
Recreational
and
Nonuse
Benefits
from
Reduced
MP&
M
Discharges
from
the
Oily
Wastes
Subcategory
in
Ohio
Activity
Percentage
of
Ohio
Residents
Participating
in
Single­
Day
Trips
(
from
NDS)
Number
of
Participants
Aged
16
and
older
a
Total
Annual
Welfare
Gain
to
Recreational
Users
in
Ohio
Fishing
10.2%
892,283
$
153,102
Boating
7.7%
676,026
$
0
Viewing
9.1%
798,220
$
88,047
Swimming
9.1%
798,220
$
9,783
Total
Recreational
Use
Benefit
$
250,933
Nonuse
Benefits
$
125,466
Total
Recreational
Benefits
(
Use
+
Nonuse)
$
375,859
a
EPA
estimated
the
number
of
participants
in
each
recreation
activity
by
multiplying
the
percent
of
NDS
survey
respondents
from
Ohio
participating
in
a
single
day
trip
for
each
activity
by
the
total
adult
population
aged
16
an
older
(
8,790,969).
This
analysis
uses
the
2000
Census
data
to
estimate
current
population
in
Ohio.

Source:
U.
S.
EPA
analysis.

Under
the
final
regulation,
the
extrapolation
from
the
sample
to
the
adult
pop
ulation
in
Ohio
yields
mean
annual
benefits
estimates
of
$
153,102,
$
9,783,
$
88,047,
and
$
0
(
2001$)
for
fishing,
swimming,
viewing,
and
boating,
respectively.
The
total
mean
recreational
use
benefit
is
$
250,932
(
2001$).
The
Agency
used
the
same
approach
as
in
the
national
analysis
to
estimate
nonuse
benefits
(
see
Section
15
.2.3,
Nonuse
Benefits,
for
detail).
EPA
estimated
nonuse
benefits
as
one­
half
of
recreational
use
benefits
for
low,
mid,
and
high
estimates,
respectively.
The
estimated
mean
nonuse
benefit
is
$
125,466
(
2001$).

6
LIMITATIONS
AND
UNCERTAINTY
6.1
One­
State
Approach
Some
benefits
are
likely
to
be
m
issed
by
a
state­
leve
l
case
stud
y.
For
example,
residents
from
neighboring
states
undoubtedly
recreate
in
Ohio
waters,
and
residents
of
Ohio
undoubtedly
recreate
in
neighboring
states.
A
state­
by­
state
approach
that
restricts
its
analysis
to
recreation
activities
within
the
state
misses
these
categories
of
benefits.
16
This
omission
is
likely
to
be
more
significant
for
unique
locations
of
high
quality
(
e.
g.,
Lake
Erie),
where
participants
travel
significant
distances,
and
for
sites
very
close
to
state
boundaries.
Ohio
Case
Study
17
Westat
(
1989)
uses
ten
or
more
activity­
days
per
year
as
an
indicator
of
an
"
avid"
user.

27
6.2
Including
One­
Day
Trips
Only
Use
of
day­
trips
only
tends
to
understate
recreational
benefits
for
swimming,
fishing,
boating,
and
viewing,
since
recreation
as
part
of
multi­
day
trip
s
is
exclud
ed.
Inclusion
of
multi­
day
trips,
however,
can
be
problematic.
M
ulti­
day
trips
are
freq
uently
multi­
activity
trips.
An
individual
might
travel
a
substantial
distance,
participate
in
several
recreation
activities
and
go
shopping
and
sightseeing,
all
as
part
of
one
trip.
Recreational
benefits
from
improved
recreational
opportunities
for
the
primary
activity
are
overstated
if
all
travel
costs
are
treated
as
though
they
are
associated
with
the
one
recreational
activity
of
interest.
The
total
benefits
per
trip
from
water
quality
improvements
are
not
overstated,
however,
if
individuals
participated
only
in
several
water­
based
activities.

6.3
Nonuse
Benefits
Estimating
nonuse
benefits
using
the
50
percent
rule
is
less
precise
than
using
a
more
so
phisticated
benefits
transfer
approach.

However,
limiting
the
b
enefits
of
water
quality
improvements
only
to
recreational
benefits
would
significantly
underestimate
the
benefits
of
the
rule.
The
effects
of
using
the
simpler
approach,
e.
g.
either
overestimation
or
und
erestimation
of
benefits,
is
unknown.
Other
benefits
include
aesthetic
benefits
for
residents
living
near
water
bodies,
habitat
values
for
a
variety
of
species
(
in
addition
to
recreational
fish),
and
nonuse
values.
To
correct
for
this
limitation
of
using
only
a
travel
cost
model,

EPA
quantified
nonuse
values
in
proportion
to
recreation
values.
This
approach
provides
only
a
rough
approximation
of
the
value
of
water
resources
to
nonusers.
For
example,
some
natural
resources
have
high
use
values
but
small
or
negligible
nonuse
values
(
e.
g.,
cows),
while
other
species
have
very
high
nonuse
values
but
small
or
negligible
use
values
(
e.
g.,
blue
whales).

6.4
Potential
Sources
of
Survey
Bias
The
survey
results
could
suffer
from
bias,
such
as
recall
bias
(
e.
g.,
W
estat,
1989),
nonresponse
bias,
and
sam
pling
effects.

a.
Recall
bias
Recall
bias
can
occur
when
respondents
are
asked
the
number
of
days
in
which
they
recreated
over
the
previous
season,
such
as
in
the
NDS
survey.
Some
researchers
believe
that
recall
bias
tends
to
lead
to
an
overstatement
of
the
number
of
recreation
days,
particularly
for
more
avid
participants.
Avid
participants
tend
to
overstate
the
number
of
recreation
days,
since
they
count
days
in
a
"
typical"
week
and
then
multiply
them
by
the
number
of
weeks
in
the
recreation
season.
17
They
often
neglect
to
consider
days
missed
due
to
bad
weather,
illness,
travel,
or
when
fulfilling
"
atypical"
obligations.
Some
studies
also
found
that
the
more
salient
the
activity,
the
more
"
optimistic"
the
respondent
tends
to
be
in
estimating
number
of
recreation
days.

Individuals
also
have
a
tendency
to
overstate
the
number
of
days
they
participate
in
activities
that
they
enjoy
and
value.

Taken
together,
these
sources
of
recall
bias
may
result
in
an
overstatement
of
the
actual
number
of
recreation
days.

b.
Nonresponse
bias
A
problem
with
sampling
bias
may
arise
when
extrapolating
sample
means
to
population
means.
This
could
happen,
for
example,
when
avid
recreation
participants
are
more
likely
to
respond
to
a
survey
than
those
who
are
not
interested
in
the
forms
of
recreation,
are
unable
to
pa
rticipate,
assume
that
the
survey
is
not
meant
for
them,
or
consider
the
survey
not
worth
their
time.

c.
Sampling
effects
Recreational
demand
studies
frequently
face
two
types
of
observations
that
do
not
fit
general
recreation
patterns:

non­
participants
and
avid
participants:

Non­
participants
are
those
individuals
who
would
not
particip
ate
in
the
recreation
activity
under
any
conditions.
This
analysis
assumes
that
an
individual
is
a
non­
participant
in
a
particular
activity
if
he
or
she
did
no
t
particip
ate
in
that
activity
at
any
site.

This
assumption
tends
to
understate
benefits,
since
some
individuals
may
not
have
participated
during
the
sampling
period
simply
by
chance,
or
because
price/
quality
conditions
were
unfavorable
during
the
sampling
period.

Avid
particip
ants
can
also
be
problematic
because
they
claim
to
participate
in
an
activity
an
inordinate
number
of
times.
This
Ohio
Case
Study
28
reported
level
of
activity
is
sometimes
correct,
but
often
overstated,
perhaps
due
to
recall
bias
(
see
Westat,
1989).
Even
where
the
reports
are
correct,
these
observations
tend
to
be
overly
influential.
EPA
dropped
observations
of
participants
who
reported
more
than
100
trips
per
year
when
estimating
trip
participation
mo
dels,
to
correct
for
potential
bias
caused
by
these
observations.
Ohio
Case
Study
29
GLOSSARY
ambient
water
quality
criteria
(
AWQC):
levels
of
water
quality
expected
to
rend
er
a
body
of
water
suitable
for
its
designated
use.
Criteria
are
based
on
specific
levels
of
pollutants
that
would
make
the
water
harmful
if
used
for
drinking,

swimming,
farming,
fish
production,
or
industrial
processes.
(
http://
www.
epa.
gov/
OCEPAterms/
aterms.
html)

compensating
variation
(
CV):
the
amount
of
money
a
person
would
need
to
pay
or
receive
in
order
to
leave
that
person
as
well
off
as
they
were
before
a
change.

consumer
choice
set:
the
set
of
alternatives
(
e.
g.,
alternative
recreation
sites)
from
which
a
consumer
may
choose.

exogenous:
external
to
the
inner
workings
of
a
system
or
model;
variables
are
exogenous
to
the
extent
that
they
are
"
given"

and
not
the
result
of
the
operation
o
f
the
system
or
anything
going
on
in
the
mo
del
itself.

expected
maximum
utility:
see
 
inclusive
value. 

fish
biomass
(
Biomass):
measure
of
biological
factors
in
the
water
body
represented
by
the
total
fish
weight
in
kilograms
per
300
meters.

fish
consumption
advisories
(
FCAs):
an
official
notification
to
the
public
about
specific
areas
where
fish
tissue
samples
have
been
found
to
be
contaminated
by
toxic
chemicals
which
exceed
FDA
action
limits
or
other
accepted
guidelines.

Advisories
may
be
species
specific
or
community
wide.

inclusive
value:
the
value
to
the
consumer
of
being
able
to
choose
among
X
alternatives
(
e.
g.,
among
a
number
of
recreational
sites)
on
a
given
trip
occasion.

indirect
utility
function:
gives
the
maximum
value
of
utility
for
any
given
prices
and
money
income.
The
indirect
utility
function
is
obtained
when
the
quantity
of
goods
that
maximizes
consumer
utility
subject
to
a
budget
constraint
are
substituted
into
a
utility
function.

inferential
analyses:
based
on
interpretation.

multinomial
logit
(
MNL):
a
utility
maximization
model.
In
this
model,
an
individual
is
assumed
to
have
preferences
defined
over
a
set
of
alternatives
(
e.
g.,
recreation
sites).
The
choice
model
takes
the
form
of
comparing
utilities
from
different
alternatives
and
choosing
the
one
that
produces
the
maximum
utility.
In
this
framework,
observed
data
consist
of
attributes
of
the
choices
(
e.
g.,
available
recreational
amenities
at
different
sites)
and
the
choice
actually
made.
Usually
no
characteristics
of
the
individuals
are
observed
beyond
their
actual
choice.

National
Demand
Survey
for
Water­
Based
Recreation
(
NDS):
a
U.
S.
EPA
survey
of
recreational
behavior.
The
1994
survey
collected
data
on
socioeconomic
characteristics
and
water­
based
recreation
behavior
using
a
nationwide
stratified
random
sample
of
13,059
individuals
aged
16
and
over.
(
http://
www.
epa.
gov/
opei)

negative
binomial
regression
model:
an
extension
of
the
Poisson
regression
model
that
allows
the
variance
of
the
process
to
differ
from
the
mean
(
see
also
Poisson
distribution
and
Poisson
estimation
process).

Negative
Binomial
Poisson
model:
(
see
negative
binomial
regression
model).

nested
multinomial
logit
model
(
NMNL):
an
extension
of
MNL
(
see
above).
In
this
model,
an
individual
is
assumed
to
choose
among
different
groups
of
alternatives
first
(
i.
e.,
Great
Lakes
or
inland
recreatio
n
sites)
and
then
to
choose
specific
alternatives
(
e.
g.,
a
particular
river
reach,
lake,
or
Great
Lakes
site)
in
the
choice
set
for
each
group.

nonconventional
pollutants:
a
catch­
all
catego
ry
that
includ
es
all
po
llutants
that
are
not
classified
as
priority
pollutants
or
conventional
pollutants.

Ohio
Water
Resource
Inventory
(
OWRI):
a
biennial
report
to
U.
S.
EPA
and
Congress
required
by
Section
305(
b)
of
the
Clean
Water
Act.
The
report
is
composed
of
four
major
sections:
(
1)
inland
rivers
and
streams,
wetlands,
Lake
Erie,
and
water
program
description;
(
2)
fish
tissue
contaminants;
(
3)
inland
lakes,
ponds,
and
reservoirs;
and
(
4)
groundwater.
Ohio
Case
Study
30
overdispersion:
condition
for
a
distribution
where
the
variance
exceeds
the
mean.
It
usually
signifies
a
nonrandom
dispersion,
for
example
the
case
where
a
small
minority
of
the
population
is
responsible
for
the
majority
of
recreational
trips
taken.

Poisson
distribution:
a
random
variable
X
is
defined
to
have
a
Poisson
distribution
if
the
probability
density
of
X
is
given
by
f
x(
X)=
f
x
(
X;

)
=
e­


x
/
x!
for
x
=
0,1,2
..,
and
0
otherwise.
In
this
model,

is
both
the
mean
and
variance
of
X.

Poisson
estimation
process:
used
to
model
discrete
random
variables.
Typically,
a
Poisson
random
variable
is
a
count
of
the
number
of
events
that
occur
in
a
certain
time
interval
or
spatial
area,
for
example,
the
number
of
recreational
trips
taken
during
a
recreational
season.

priority
pollutants:
126
individual
chemicals
that
EPA
routinely
analyzes
when
assessing
contaminated
surface
water,

sediment,
groundwater,
or
soil
samples.

random
utility
model
(
RUM):
a
model
of
consumer
behavior.
The
model
contains
observable
determinants
of
consumer
behavior
and
a
random
element.

Reach
File
1
(
RF1):
a
database
of
approximately
700,000
miles
of
streams
and
open
waters
in
the
conterminous
United
States.
The
database
contains
information
on
stream
flow,
time
travel
velocity,
reach
length,
width,
depth,
and
other
stream
attributes.

site
choice
model:
used
to
determine
which
recreational
site
is
chosen
by
the
consumer.
EPA
estimated
the
likelihood
that
the
consumer
will
cho
ose
a
particular
site
as
a
function
of
site
characteristics,
the
price
paid
per
site
visit,
and
household
income.

Total
Kjeldahl
Nitrogen
(
TKN):
the
total
of
organic
and
ammonia
nitrogen.
TKN
is
determined
in
the
same
manner
as
organic
nitrogen,
except
that
the
ammonia
is
not
driven
off
before
the
digestion
step
.

travel
cost
model
(
TCM):
method
to
determine
the
value
of
an
event
by
evaluating
expenditures
by
participants.
Travel
costs
are
used
as
a
proxy
for
price
in
deriving
demand
curves
for
recreation
sites.

(
http://
www.
damagevaluation.
com/
glossary.
htm)

total
seasonal
welfare:
see
 
welfare
effect. 

trip
participation
model:
used
to
estimate
the
number
of
water­
based
recreational
trips
taken
during
the
recreation
season.

EPA
estimated
the
total
number
of
trips
during
the
recreation
season
as
a
function
of
the
expected
maximum
utility
(
inclusive
value)
from
recreational
activity
participation
on
a
trip
and
socioeconomic
characteristics
affecting
demand
for
recreation
trips
(
e.
g.,
number
of
children
in
the
household).

utility­
theoretic:
consistent
with
the
behavioral
postulate
that
individuals
act
to
maximize
their
welfare
(
utility)
that
underlines
the
structure
of
models
of
consumer
behavior.

welfare
effect:
gain
or
loss
of
welfare
to
the
group
of
individuals
(
e.
g.,
fishermen)
as
a
whole.
Ohio
Case
Study
31
ACRONYMS
AWQC:
ambient
water
quality
criteria
CV:
compensating
variation
FCAs:
fish
consumption
advisories
IWB2:
index
of
well­
being
LIMDEP:
Limited
Dependent
Variable
MNL:
multinomial
logit
NDS:
National
Demand
Survey
for
Water­
Based
Recreation
NMNL:
nested
multinomial
logit
model
ODNR:
Ohio
Department
of
Natural
Resources
OWRI:
Ohio
W
ater
Resource
Inventory
RUM:
random
utility
model
RF1:
Reach
File
1
TKN:
Total
Kjeldahl
Nitrogen
TCM:
travel
cost
model
Ohio
Case
Study
32
REFERENCES
Bockstael,
N.
E.,
I.
E.
Strand,
and
W.
M.
Hanemann.
1987.
T
ime
and
the
Recreation
Demand
Model.
American
Journal
of
Agricultural
Economics
69:
213­
32.

Creel,
M.
and
J.
Loomis.
1992.
 
Recreation
Value
of
Water
to
W
etlands
in
the
San
Joaquin
V
alley:
Linked
Multinomial
Logit
and
Count
Data
Trip
Frequency
Models. 
Water
Resource
Research.
Vol
28.
No
10,
October:
2597­
2606.

Feather,
Peter
M
.,
Daniel
Hellerstein,
and
Theodore
Tomasi.
1995.
 
A
D
iscrete­
Count
M
odel
of
Recreation
Demand. 

Journal
of
Environmental
Economics
and
Management,
29:
214­
227.

Green,
W
.
H.
19
93.
Econometric
Analysis.
New
York,
NY:
Macmillan
Publishing
Company.

Hausman,
J.,
G.
Leonard,
and
D
.
McFadden.
1995.
 
A
Utility­
Consistent,
Combined
D
iscrete
Choice
and
Count
D
ata
Model:
Assessing
Recreational
Use
Losses
Due
to
Natural
Resource
Damage. 
J.
of
Public
Economics
No
56
pp.
1­
30.

Jones,
C.
A.
and
Y.
D.
Sung.
1993.
Valuation
of
Environmental
Quality
at
Michigan
Recreational
Fishing
Sites:

Methodological
Issues
and
Policy
Application.
Final
Report.
EPA
Contract
No.
CR­
816247­
01­
2.
September.

Kling,
C.
L.
and
C.
J.
Thomson.
1996.
 
The
Implication
o
f
Model
Specification
for
Welfare
E
stimation
in
Nested
Logit
Models. 
American
Journal
of
Agricultural
Economics
Association,
No.
78,
February:
103­
114.

McFadden,
D.
1981.
 
Econometric
Models
of
Probabilistic
Choice. 
In:
C.
F.
Manski
and
D.
L.
McFadden,
eds.,
Structural
Analysis
of
Discrete
Data.
Cambridge,
MA:
MIT
Press.

Ohio
Atlas
and
Gazetteer,
The.
1995.
Freeport,
ME:
Delorme.

Ohio
Department
of
Natural
Resources,
Division
of
Wildlife.
Creel
Survey
Summaries
from
1992
to
1997.

Ohio
Department
of
Natural
Resources,
Division
of
Wildlife.
1999.
Fish
Consumption
Advisories.

Ohio
Environmental
Protection
Agency.
1996.
Ohio
Waste
Resource
Inventory
Volume
1:
Summary
Status,
and
Trends;
and
Volume
3:
Ohio
Public
Lakes,
Ponds,
and
Reservoirs.
www.
Chagrin.
EPA.
State.
OH.
US/
document_
index.

Parsons,
G
.
and
M
.
J.
Kealy.
199
2.
 
Randomly
Drawn
Opportunity
Sets
in
a
Random
Utility
Model
of
Lake
Recreation. 

Land
Economics
68
N
o.
4:
418­
33.

Parsons,
G.
R.
1999.
Comments
on
Assessing
the
Recreational
Benefits
of
the
MP&
M
Regulation:
A
State­
level
Case
Study
Based
on
the
Random
Utility
Model
Approach.
Memo
to
Abt
Associates
Inc.,
August.

U.
S.
Department
of
Commerce,
Bureau
of
the
Census.
2000.
http://
www.
state.
oh.
us/
odhs/
octf/
stats/
gjcs/
ohio.
pdf.

U.
S.
Environmental
Protection
Agency
(
U.
S.
EPA).
19
94.
National
Demand
for
Water­
Based
Recreation
Survey.

Washington,
D.
C.:
Office
of
Policy
Evaluation
and
Information.
