1
Date:
17
March
2004
To:
Chris
Miller,
EPA
From:
Maureen
Kaplan
and
Ian
Cadillac,
ERG
Subject:
Observations
on
Engle
et
alia,
"
The
Economic
Impact
of
Proposed
Effluent
Treatment
Options
for
Production
of
Trout
Oncorhynchus
mykiss
in
Flow­
through
Tanks"

Engle
et
alia
is
an
alternative
economic
assessment
of
the
impacts
of
incremental
pollution
control
costs
on
flow­
through
facilities
in
North
Carolina
and
Idaho.
The
authors
appear
to
have
three
different
types
of
analysis:

#
Enterprise
budgets
(
3
sites)

S
North
Carolina
 
68,182
kg
S
Idaho
 
90,909
kg;
and
S
Idaho
 
1,136,363
kg
#
Monte
Carlo
analysis,
and
#
Credit
and
borrowing
analysis
First,
we
examined
the
data
sets
that
form
the
basis
of
the
analysis
and
found
some
discrepancies
regarding
the
number
of
facilities
(
Section
1).
We
found
other
inconsistencies
throughout
the
paper
and
mention
them
where
they
might
lead
to
different
inferences
from
the
study.

Second,
we
investigate
whether
the
data
sets
Engle
et
alia
analyzed
reflect
the
population
that
is
within
the
scope
of
the
rule
(
Section
2).
They
do
not.
They
contain
a
substantial
portion
of
observations
that
are
not
within
the
scope
of
the
rule,
that
is,
they
have
less
than
100,000
lb/
yr
production.

Third,
we
review
the
costs
used
in
the
impact
analysis
(
Section
3).
Even
if
the
EPA
and
Engle
et
alia
methodologies
and
data
sets
are
comparable,
the
estimated
impacts
will
certainly
differ
if
the
pollution
control
costs
differ.
Tetra
Tech
has
the
lead
on
developing
costs
for
EPA
and
are
most
suited
to
review
and
respond
on
this
section.
However,
we
find:

#
an
unreasonable
assumption
that
none
of
the
facilities
have
primary
settling
#
extremely
high
annual
costs
for
land
for
solids
disposal
These
two
factors
drive
the
results
of
the
economic
analyses.
2
The
methodological
discussion
presented
in
Engle
et
alia
steps
from
one
type
of
analysis
to
another
without
clear
breaks.
Our
review
of
the
enterprise
budgets,
Monte
Carlo
analysis,
and
credit/
borrowing
analysis
is
given
in
Sections
4
through
6,
respectively.
In
doing
so,
we
may
have
forced
an
inaccurate
organization
onto
the
discussion.

1.
How
Many
Observations
Are
in
the
Data
Set?

The
author
presents
conflicting
information.
The
introduction
mentions
that
21
of
the
57
commercial
trout
farms
in
North
Carolina
in
2002
were
in
Transylvania
County
(
p.
2).
Under
"
Materials
and
Methods,"
the
authors
state
that
survey
data
were
collected
from
13
of
the
21
farms
in
Transylvania
Country,
NC
for
a
response
rate
of
13/
21
or
62
percent
(
p.
5).
However,
in
the
"
Results
and
Discussion"
section,
the
authors
mention
an
81
percent
response
rate
(
p.
9).
This
implies
17
(
0.81
x
21
=
17.01)
observations
in
the
North
Carolina
data
set.
For
the
rest
of
this
memorandum,
we
assume
the
first
statement
is
the
more
accurate
(
13
farms)
because
it
provides
counts.

The
authors
state
that
they
collected
production
and
financial
information
from
eight
(
8)
trout
farms
in
Idaho
by
interviewing
the
farm
owner
or
manager
(
p.
6).

2.
The
Engle
et
alia
Data
Sets
and
the
EPA
"
In­
scope"
Community
We
compared
the
data
sets
used
by
Engle
et
alia
to
the
EPA
in­
scope
population
(
i.
e.,
>
100,000
lb/
yr).
EPA
chose
the
100,000
lb/
yr
threshold
to
mitigate
impacts
on
smaller
facilities.
If
there
is
a
basic
difference
between
the
populations
studied,
the
results
will
not
be
comparable.
Table
1
provides
an
alternative
representation
of
one
variable
 
farm
size
 
from
Engle
et
alia,
Table
2.
We
assume
there
are
13
observations
in
North
Carolina
and
eight
in
Idaho.
When
a
population
has
an
even
number
of
observations,
the
median
is
conventionally
calculated
as
the
arithmetic
average
of
the
middle
two
observations.

For
North
Carolina,
the
smallest
farm
in
the
survey
produces
1,000
kg/
yr.
This
farm
is
so
small
that
it
does
not
qualify
to
be
a
CAAP,
never
mind
being
within
the
scope
of
the
rule.
By
the
median
value,
we
have
a
farm
that
is
large
enough
to
be
within
the
scope
of
the
rule.
At
a
minimum,
one
of
13
or
7.7
percent
of
the
NC
farms
are
out
of
scope
of
the
rule.
At
most,
six
of
13
or
nearly
half
the
sample
might
be
outside
the
scope
of
the
rule.
EPA
decided
to
set
the
threshold
at
100,000
lb/
yr
because
of
the
potential
impacts
on
the
smaller
facilities.
When
the
scope
of
the
rule
was
shifted
from
20,000
lb/
yr
to
100,000
lb/
yr,
all
but
two
(
unweighted)
facilities
that
reported
unpaid
labor
and
management
dropped
out
of
scope
(
Cadillac
and
Kaplan
2004).

The
Idaho
data
set
has
eight
observations.
The
smallest
farm
produces
27,000
kg/
yr
so,
although
it
is
a
CAAP,
it
is
not
within
the
scope
of
the
rule.
The
median
value
is
reported
as
1.2
million
kg/
yr
and
the
maximum
value
is
reported
as
1.1
million
kg/
yr
[
sic].
The
average
of
observation
#
4
and
observation
#
5
is
1.2
million
kg/
yr,
so
the
most
likely
situation
is
that
both
observations
are
close
to
1.2
million
kg/
yr.
So
at
least
one
and
possibly
three
observations
are
outside
the
scope
of
the
rule
(
i.
e.,
between
12.5
and
37.5
percent
of
the
population).
This
also
implies
that
the
medium­
sized
Idaho
enterprise
budget
is
based
on,
at
most,
three
observations
and
at
least
one
of
those
is
outside
the
scope
of
the
rule.
3
In
sum,
the
data
sets
used
in
the
Engle
et
alia
analysis
contain
a
substantial
portion
of
observations
that
are
outside
the
scope
of
the
rule.
As
we
discuss
later
in
this
memorandum,
the
inclusions
of
these
outof
scope
facilities
colors
the
results
of
the
Monte
Carlo
sensitivity
analysis
and
the
treatment
of
unpaid
labor.
With
a
basic
difference
in
the
populations
being
studied,
it
is
not
surprising
that
the
findings
will
also
differ.
EPA
had
reasons
to
choose
a
100,000
lb/
yr
threshold.

Table
1
Farm
Sizes
in
Engle
et
alia,
2004
Farm
Size
(
kg)

Observation
North
Carolina
Idaho
1
1,000
27,000
2
3
4
1,200,000
5
6
7
66,000
8
1,136,364
9
10
11
12
13
204,500
Source:
Engle,
et
alia,
Table
2.

3.
Pollution
Control
Costs
The
second
factor
driving
the
impact
results
presented
in
Engle
et
alia
are
the
costs
which
are
substantially
higher
than
those
estimated
by
Tetra
Tech.
Engle
et
alia
present
pollution
control
costs
in
Table
5
in
their
report.
The
title
lists
these
as
annual
costs
with
capital
costs
charged
as
annual
depreciation
(
p.
11).
The
major
features
are:

#
All
sites
are
assumed
to
have
no
treatment
in
place.

#
Annual
costs
for
land
for
field
application
drive
the
impact
results.
1In
one
case,
total
labor
time
is
given
in
units
of
$/
yr
when
the
other
labor
entries
are
specified
as
h/
yr.

4
#
Depreciation
period
and
schedule
are
not
specified.
This
overestimates
the
impact
of
capital
costs
on
net
returns.

#
Labor
costs
($/
yr)
are
not
specified.
1
The
number
of
hours
is
provided
but
not
the
rate,
so
we
cannot
calculate
labor
costs.

#
Compliance
monitoring
has
an
entry
for
"
other
costs,"
generally
$
4,008.
This
wouldn't
be
of
concern
except
weekly
monitoring
costs
(
which
we
interpret
as
total
monitoring
costs)
is
listed
as
$
5298.
So
the
unspecified
"
other"
costs
appear
to
be
76
percent
of
total
costs.

Each
point
is
discussed
in
more
detail
in
the
paragraphs
below.

The
assumption
that
all
sites
have
no
treatment
in
place
implies
that
all
sites
incur
costs
for
primary
settling.
This
overstates
costs
for
facilities
that
already
have
primary
treatment
 
which
is
all
the
in­
scope
facilities.
In
contrast,
EPA
includes
costs
for
primary
treatment
where
there
is
none
at
a
site
and
does
not
include
primary
treatment
costs
where
it
already
exists.
EPA's
cost
estimates,
then,
are
more
accurate.
The
Engle
et
alia
costs
are
overstated.

In
Engle
et
alia,
Table
5,
Primary
settling,
costs
for
offline
settling
pond
(
either
with
a
front
end
loader
or
a
vacuum
tank)
contains
a
line
item
for
land
for
field
application
of
the
collected
solids.
These
are
listed
as
annual
costs
and
are
so
large
that
they
drive
the
impacts
analysis.
Table
2
lists
the
total
revenues,
net
revenues,
and
annual
land
application
costs
for
the
three
farms.
Given
that
the
NC
farm
has
a
total
estimated
revenue
of
$
165,000,
it
is
not
surprising
that
an
annual
land
cost
of
$
150,000
renders
the
facility
unprofitable.
Given
that
the
ID
medium
farm
clears
less
than
$
6,000
per
year,
it
is
understandable
that
a
$
22,000
annual
cost
for
solids
disposal
will
render
the
facility
unprofitable.

Table
2
Comparison
of
Total
Revenues,
Net
Revenues,
and
Land
Application
Costs
Enterprises
Parameter
NC
Medium
ID
Medium
ID
Large
Total
Revenues1
$
165,000
$
160,000
$
2,000,000
Net
Revenues/
Returns1
$
8,644
$
5,647
$
284,281
Annual
land
costs
for
field
application
of
solids2
$
150,000
$
22,000
$
275,000
1
Engle
et
alia,
Table
3.
2
Engle
et
alia,
Table
5.
5
Engle
et
alia
presents
capital
costs
as
depreciation,
however,
the
authors
do
not
specify
the
depreciation
method
used
(
e.
g.,
MACRS
or
straight­
line)
or
the
number
of
years
over
which
depreciation
is
taken.
We
specify
that
we
use
MACRS
with
a
half­
year
convention
and
a
10­
year
recovery
period.
But
we
use
it
only
to
calculate
the
tax
shield
in
any
year,
not
the
cost
to
the
owner.
We
first
calculate
the
present
value
of
all
costs
over
the
period
and
then
calculate
a
constant
annual
payment
corresponding
to
the
present
value.
For
example,
the
depreciation
rate
for
Year
1
with
a
mid­
year
convention
and
a
10­
year
recovery
period
is
10
percent.
If
the
equipment
is
assumed
to
be
put
in
place
at
the
beginning
of
the
year,
the
depreciation
rate
is
20
percent.
If
the
equipment
is
assumed
to
be
put
in
place
at
the
beginning
of
the
year
and
the
recovery
period
is
7
years,
the
depreciation
rate
is
28.6
percent.
So
we
do
not
know
how
much
of
the
capital
cost
the
authors
are
assigning
to
the
first
year
revenues.

4.
Enterprise
Budgets
Engle
et
alia
Table
3
presents
the
enterprise
budgets
for
trout
production
in
North
Carolina
and
Idaho,
presumably
based
on
information
from
the
survey.
For
comparison
with
the
EPA
analysis,
we
note
what
Engle
et
alia
call
"
net
returns"
is
what
we
would
call
"
net
income,"
that
is,
depreciation
is
included
as
a
cost.
The
economic
analysis
in
the
paper
is
performed
on
a
net
income,
not
cash
flow,
basis.

The
EPA
economic
analysis
accounts
for
taxes
paid
on
taxable
income.
The
enterprise
budgets
have
a
line
item
labeled
"
licenses/
taxes"
but
this
appears
as
a
fixed
cost
and
thus
cannot
refer
to
income
taxes.
So
the
EPA
analysis,
although
done
on
a
cash
flow
basis,
includes
a
cost
that
the
Engle,
et
alia
study
does
not.

4.1
Unpaid
Labor
and
Management
We
had
difficulties
identifying
how
and
how
often
the
authors
had
to
value
unpaid
labor
and
management.
In
Engle
et
alia
Table
2,
labor
and
management
are
given
units
of
hours,
yet
an
unattached
note
C
says
that
labor
costs
include
both
paid
and
unpaid
(
family)
labor.
Elsewhere
in
the
paper,
the
authors
state
that
"...
when
a
dollar
is
assigned
to
the
number
of
hours
worked
by
the
family
members,
net
returns
from
the
enterprise
are
near
the
breakeven
point."
(
p.
11).

Nowhere
do
the
authors
specify
how
they
estimated
the
value
of
unpaid
labor.
Kaplan,
2001
noted
the
analyst's
unintentional
influence
on
the
profitability
shown
in
an
enterprise
budget.
With
returns
near
the
breakeven
point,
very
little
is
needed
to
shift
an
operation
from
profitable
to
unprofitable.

The
authors
do
not
mention
how
many
observations
needed
to
be
adjusted
for
unpaid
labor
and
management.
The
sample
includes
farms
with
production
as
small
as
1,000
kg
in
North
Carolina
and
27,000
kg
in
Idaho.
Small
farms,
such
as
these,
are
likely
to
have
unpaid
labor
and/
or
management.
As
mentioned
in
the
previous
section,
EPA
removed
most
of
the
observations
reporting
unpaid
labor
and/
or
management
from
the
scope
of
the
rule
when
it
set
the
threshold
at
100,000
lb/
yr
(
Cadillac
and
Kaplan,
2004).
2Because
the
costs
for
primary
treatment/
settling
basins/
quiescent
zones
are
part
of
the
first
option
considered
and
all
subsequent
options
are
considered
as
incremental
to
the
costs
of
the
first
option,
no
option
can
pass
if
the
first
option
fails.
So
the
assumption
of
all
sites
needing
primary
treatment
in
the
first
option
results
in
all
sites
failing
all
options.

6
4.2
Results
The
results
of
the
enterprise
analysis
are
presented
in
Engle
et
alia
Table
6.
The
net
returns
for
the
three
farms
match
those
provided
in
Engle
et
alia
Table
3.
Given
the
magnitude
of
the
costs,
particularly
land
for
solids
disposal,
it
is
not
surprising
that
the
medium
NC
and
ID
farms
are
unprofitable
with
the
first
option.
Even
with
the
high
solids
disposal
costs,
the
large
ID
farm
remains
profitable
through
all
the
options.
Having
not
closed
the
facility,
the
authors
raise
the
low
rate
of
return
as
insufficiently
attractive
to
investors.
This
observation
neglects
the
fact
that
by
using
depreciation,
the
first
year
would
look
worst
financially
and
that
subsequent
years
would
look
better.

The
bottom
half
of
Engle
et
alia
Table
6
reports
the
findings
for
"
High
Cost
Scenarios."
A
sentence
on
page
7
mentions
that
the
authors
developed
additional
scenarios
to
reflect
farm
businesses
with
land
financing
costs.
They
do
not
mention
the
size,
period,
or
interest
rate
assumed
for
these
land
financing
costs.
The
only
way
to
estimate
the
magnitude
of
this
effect
is
to
take
the
difference
between
the
net
returns
for
the
baseline
scenario
in
the
"
low
cost"
and
"
high
cost"
sections.

The
authors
also
developed
"
high"
cost
scenarios
which
are
worst
case
situations
where
an
owner
must
destroy
existing
tanks
in
order
to
build
quiescent
zones
and/
or
offline
settling
ponds.
It
appears
that
owners
are
therefore
assumed
to:
(
1)
have
no
primary
settling
areas
prior
to
the
rule,
(
2)
incur
a
cost
for
land
for
a
primary
settling
area
[
see
Table
5],
and
(
3)
reduced
production
in
proportion
to
the
number
of
tanks
destroyed.
Small
wonder
that
all
projects
fail
all
Options.
2
The
EPA
survey
found
no
commercial
facilities
within
the
scope
of
the
rule
that
fall
into
these
categories.

5.
Monte
Carlo
Analysis
5.1
Methodology
The
authors
examined
the
variability
of
the
enterprise
budget
results
by
Monte
Carlo
analysis.
The
authors
specify
they
assumed
triangular
distributions
for
the
parameters
in
the
sensitivity
analysis,
however,
they
did
not
specify
exactly
which
parameters
were
in
the
analysis.
Our
two
difficulties
with
the
sensitivity
analysis
are:

#
Whether
they
generated
the
probability
distributions
properly.

#
A
substantial
portion
of
the
enterprise
descriptions
generated
for
the
sensitivity
analysis
represent
farms
outside
the
scope
of
the
regulation.

As
mentioned
by
the
authors,
triangular
distributions
are
characterized
by
minimum,
maximum,
and
most
likely
values
(
p.
8).
Triangular
distributions
are
frequently
used
when
those
three
pieces
of
information
(
max,
min,
and
most
likely)
are
all
that
is
known
about
the
variable.
The
"
most
likely"
value
should
refer
to
the
"
mode."
A
mode
is
the
most
frequently
occurring
value
in
a
set
of
observations.
The
3It
would
be
difficult
to
envision
using
the
same
distributions
to
generate
descriptions
for
two
different­
sized
facilities.

7
authors,
however,
report
median
values
when
describing
the
parameters
listed
in
Engle,
et
alia,
Table
2.
The
median
is
a
number
that
has
the
property
of
having
the
same
number
of
scores
with
smaller
values
as
there
are
with
larger
values.
If
a
population
has
an
odd
number
of
observations
and
the
population
is
sorted
into
rank
order,
the
median
is
the
value
of
the
middle
observation.
The
mode
and
median,
then,
are
not
necessarily
the
same.
We
have
not
examined
in
depth
the
effect
of
using
the
median
instead
of
the
mode
when
generating
random
samples
based
on
a
triangular
distribution.

In
Section
2,
we
showed
that
at
least
7.7
percent
and
possibly
as
much
as
46
percent
of
the
observations
on
which
the
triangular
distributions
for
the
NC
farm
are
based
on
facilities
outside
the
scope
of
the
analysis.
Presumably,
the
authors
split
the
eight
ID
assumptions
into
two
groups
and
then
used
the
minimum,
median,
and
mean
values
for
the
distributions
for
the
two
Idaho
facilities.
3
This
means
that
descriptions
of
the
medium
Idaho
facility
are
based
on,
at
most,
three
observations
of
which
at
least
one
is
out
of
scope.
The
values
for
the
large
Idaho
facility
might
be
fine
if
the
authors
specified
what
was
used
as
the
lower
bound.
Since
the
median
and
upper
bound
values
are
the
same
(
1.2
million
kg/
yr),
there
may
be
very
little
variation
in
facility
size.

5.2
Results
The
results
of
the
Monte
Carlo
analysis
are
given
in
Engle
et
alia
Table
6.
Because
of
the
option
costs
(
see
Section
3),
the
results
are
not
unexpected.
The
baseline
results,
however,
provide
an
interesting
comparison
to
our
baseline
closure
analysis.
According
to
the
Monte
Carlo
analysis,
prior
to
incurring
additional
costs:

#
Less
than
half
the
NC
farms
are
profitable
#
Only
two
percent
of
the
medium
ID
farms
are
profitable
#
84
percent
of
the
large
ID
farms
are
profitable.

In
contrast,
when
the
EPA
analysis
is
done
on
a
net
income
basis,
60
percent
of
the
in­
scope
facilities
that
incur
costs
(
39
of
66
facilities)
are
profitable
in
the
EPA
analysis.
Thus,
AETF
should
have
no
difficulties
with
the
result
of
the
baseline
closure
analysis.

In
Section
4.1,
we
mention
that
the
medium
NC
and
ID
farms
are
on
the
knife
edge
of
profitability.
This
consistent
with
the
results
of
the
baseline
Monte
Carlo
analysis.
Normal
variation
in
the
parameters
is
sufficient
to
push
98
percent
of
the
medium
ID
farms
and
64
percent
of
the
NC
farms
into
unprofitability
before
the
addition
of
incremental
pollution
control
costs.
Between
the
marginality
of
the
baseline
enterprise
descriptions
and
the
magnitude
of
the
costs,
the
large
impacts
are
an
expected
result.

The
large
ID
farm
remains
profitable
under
the
low
cost
scenario.
However,
the
high
cost
scenario
seems
to
involve
nearly
$
500,000
in
annual
costs
(
i.
e.,
the
change
from
a
net
return
of
$
271,688
in
the
baseline
to
­$
250,000
or
so
after
the
options).
Half
of
the
cost
might
be
the
land
disposal
costs.
We
suspect
the
rest
of
the
large
cost
is
due
primarily
to
the
assumption
that
the
farm
has
to
destroy
existing
tanks
(
and
associated
production)
in
order
to
put
in
primary
settling.
Again,
Tetra
Tech
needs
to
review
the
4
Presumably,
the
model
prevents
the
owner
from
switching
more
production
to
recreational
fish.
Othwerwise,
the
higher
value
harvest
would
alleviate
the
borrowing
constraint.
Engle
et
alia
notes
that
operating
loans
are
frequently
set
as
a
percentage
of
crop
value
(
p.
14).

8
appropriateness
of
the
absence
of
primary
settling,
the
unavailability
of
land,
and
the
need
to
tear
down
productive
assets
for
the
settling
basin.

6.
Credit
and
Borrowing
Analysis
The
three
enterprise
budgets
were
used
to
develop
"
mixed
integer
programming
models."
An
inconsistency
is
that
the
authors
state
the
need
to
consider
sunk
costs
in
an
economic
impact
analysis
(
p.
20),
the
mixed
integer
programming
models
do
not
even
include
fixed
costs,
never
mind
sunk
costs.
Another
inconsistency
is
that
a
cap
on
investment
borrowing
was
set
at
$
90,000
but
this
cap
was
not
used
in
the
base
scenario.

6.1
Operating
Loans
Table
7
appears
to
be
the
result
of
an
analysis
of
the
effects
of
constraining
borrowing
capacity
on
operating
loans.
This
analysis
does
not
include
any
consideration
of
incremental
pollution
control
costs.
The
results
are
intuitive.
As
the
owner
is
permitted
to
borrow
less
and
less
money
to
operate,
the
model
takes
more
and
more
tanks
out
of
operation.
The
reductions
take
place
in
the
food
fish
tanks
because
the
recreational
fish
are
likely
to
have
a
higher
profit
margin.
4
The
analysis
show
how
reductions
in
operating
loans
lead
to
reductions
in
the
amount
of
fish
raised
with
those
costs,
but
analysis
does
not
address
pollution
control
costs
in
any
way.

The
authors
re­
examine
the
situation
where
a
farm
is
required
to
take
tanks
out
of
production
in
order
to
install
a
quiescent
zone
or
settling
pond.
This
leads
to
less
fish
production
which
leads
to
lower
crop
values
which
leads
to
lower
operating
loans
which
leads
to
less
fish
production,
and
so
forth.
This
downward
spiral,
however,
rest
on
the
assumption
that
all
sites
need
to
install
primary
settling
and
that
tanks
will
need
to
be
taken
out
of
operation
to
do
so.
EPA,
however,
found
no
commercial
in­
scope
facilities
that
would
need
to
do
this
to
comply
with
the
rule.

6.2
Capital
or
Investment
Loans
The
analysis
examines
land
costs
and
the
price
to
which
land
costs
would
have
to
drop
in
order
to
expand
production
(
Table
8).
While
an
informative
academic
enterprise,
the
analysis
does
not
relate
to
the
effect
of
effluent
guidelines.

The
analysis
then
changes
to
examining
the
"
full
costs
of
quiescent
zones
and
offline
settling
basins."
(
p.
16).
First,
the
authors
do
not
state
what
they
consider
to
be
the
full
costs
of
primary
settling;
Engle
et
al
Table
5
presents
only
depreciated
costs,
not
total
costs.
The
scenario
examined
in
the
investment
loan
sensitivity
analysis
starts
with
an
investment
cap
of
$
90,000.
Given
that
total
land
or
capital
costs
are
not
specified
in
the
cost
table,
we
can't
say
whether
or
not
this
cap
is
exceeded
by
9
definition.
Given
the
annual
land
cost
for
solids
disposal
exceed
the
$
90,000
cap
for
the
NC
facility,
if
this
is
considered
a
land
cost
then,
by
definition,
all
options
are
infeasible.
The
results
shown
in
Tables
9
and
10
depend
completely
on
the
cost
of
the
primary
settling
structure;
an
aggregate
capital
cost
which
they
do
not
list
and
which
represents
treatment
that
is
already
in
place
for
all
the
commercial
in­
scope
facilities
in
the
EPA
survey.

The
results
for
the
Idaho
farms
are
presented
in
Engle
et
alia
Figure
1.
The
measure
used
is
still
net
returns
above
variable
costs.
As
expected,
the
medium
Idaho
farm
showed
impacts
(
although
still
positive)
while
the
large
farm
remained
profitable.

7.
Paper's
Conclusions
and
Rebuttals
The
authors
see
impacts
on
the
medium
NC
and
Idaho
farms.
Their
sample,
however,
includes
a
substantial
portion
of
observations
that
are
not
within
the
scope
of
the
rule.
EPA
also
saw
potential
impacts
on
small
(<
100,000
lb/
yr)
facilities
and
set
the
threshold
to
remove
them
from
the
scope
of
the
rule.
The
higher
cost
scenarios
are
worst­
case
scenarios
and
so
seeing
impacts
is
not
surprising.

The
authors
state
that
the
risk
analysis
showed
very
low
probabilities
of
generating
positive
net
returns
after
imposing
treatment
options.
What
they
don't
say
is
that
the
medium
farms
in
NC
and
ID
show
low
probabilities
of
generating
positive
net
returns
before
imposing
treatment
options
as
well.

The
credit
analysis
either
examines
academic
cases
with
no
consideration
of
effluent
treatment
costs
or
examines
the
effects
of
massive
costs
that
are
highly
unlikely
to
occur.
All
in­
scope
facilities
in
the
EPA
survey
had
primary
settling
features
in
place
prior
to
the
rulemaking.

Sunk
costs
are
first
mentioned
in
the
conclusions
section.
Presumably
the
authors
include
sunk
costs
in
their
analysis
by
the
inclusion
of
depreciation
as
a
cost
but
depreciation
(
as
calculated
for
tax
purposes)
can
overstate
the
replacement
cost
particularly
in
the
initial
years
of
an
accelerated
cost
recovery
schedule.
The
authors
are
inconsistent
when
they
explain
that
some
farms
which
already
adopted
some
of
the
treatment
options
proposed
by
EPA
might
have
done
so
because
"
sunk
costs
on
older
farms
were
not
generating
cash
expenses."
Depreciation
is
not
a
cash
expense.
If
these
farms
were
able
to
install
some
of
the
treatment
options
because
they
examined
cash
flow
rather
than
net
income,
why
shouldn't
the
analysis
be
performed
on
a
cash
flow
basis?

First,
the
Engle
et
alia
analysis
assumes
that
a
facility
would
expand
its
production
to
cover
the
additional
fixed
costs
and,
thus,
maintain
the
same
level
of
profitability.
Second,
the
high
land
prices
in
the
analysis
coupled
with
limits
on
borrowing
capacity
prevent
the
potential
expansion.
The
third
step
in
the
analysis
assumes
the
facility
would
cannibalize
productive
tanks
for
the
land.
In
contrast,
EPA
found
that
all
in­
scope
facilities
had
primary
treatment
in
place
and
would
not
need
to
incur
the
costs
for
installing
that
component.
Engle
et
alia
note
that
some
of
the
farms
in
their
data
set
had
installed
some
of
the
EPA
treatment
components.
EPA
also
did
not
identify
the
need
for
the
massive
annual
land
costs
for
solids
disposal
that
are
included
in
the
Engle
et
alia
study.
The
EPA
analysis
examines
the
change
in
profitability
due
to
the
costs
of
incremental
pollution
control
all
other
things
being
equal.
The
EPA
analysis
examines
the
hit
the
facility
would
take,
not
what
expansion
is
needed
to
keep
it
at
the
same
level
of
profitability.
Without
the
massive
land
costs,
the
EPA
analysis
has
no
need
for
a
facility
to
cannibalize
itself
as
a
result
of
the
rule.
10
After
that,
the
argument
falls
to
pollution
control
being
a
non­
productive
asset;
an
argument
made
with
every
effluent
guideline.
The
imposition
of
pollution
controls
is
to
address
the
market
failure
that
exempts
CAAPs
from
internalizing
the
social
costs
of
pollution.

8.
References
Cadillac
and
Kaplan.
2004.
Ian
Cadillac
and
Maureen
Kaplan,
ERG.
Concentrated
Aquatic
Animal
Production
Industry:
Unpaid
Labor.
Memorandum
to
Chris
Miller,
EPA.
27
January.

Engle
et
alia.
2004.
Carole
R.
Engle,
Steeve
[
sic]
Pomerleau,
Fary
Fornshell,
Jeffery
M.
Hinshaw,
Debra
Sloan,
and
Skip
Thompson.
The
Economic
Impact
of
Proposed
Effluent
Treatment
Options
for
Production
of
Trout
Onchorhynchus
mykiss
in
Flow­
through
Tanks.
Submitted
to
USDA.
March
2004.

Kaplan.
2001.
Maureen
F.
Kaplan.
Lessons
Learned
from
Sensitivity
Analysis
and
Enterprise
Budgets.
Memorandum
to
the
Aquatic
Animal
Production
Industry
Project
File.
31
December.
Water
Docket
OW­
2002­
0026
DCN
20153.
