
1
Assessing
Prospective
Estimates
of
the
Costs
of
Clean
Air
Act
Compliance:
Are
Costs
Overestimated?

Cynthia
J.
Manson
and
James
E.
Neumann
Industrial
Economics,
Incorporated
Cambridge,
MA
June
29,
2001
INTRODUCTION
Estimates
of
the
cost
of
regulatory
compliance
are
an
important
element
in
regulatory
decision­
making
within
EPA's
clean
air
program.
Although
recently
the
U.
S.
Supreme
Court
reaffirmed
that
the
Clean
Air
Act's
National
Ambient
Air
Quality
Standards
(
NAAQS)
should
be
set
without
regard
to
costs,
the
Court
also
recognized
that
the
Clean
Air
Act
calls
on
states
and
EPA
to
consider
costs
in
implementing
those
standards.
In
addition,
analysis
of
the
cost
of
major
regulatory
actions
is
required
by
several
Executive
Orders.
The
EPA
Office
of
Air
and
Radiation
relies
on
cost
estimates
to
design
economically
efficient
strategies
to
implement
all
aspects
of
the
Clean
Air
Act
and
its
1990
Amendments,
including
federally
mandated
stationary
and
mobile
source
controls
that
support
NAAQS
compliance;
controls
on
hazardous
air
pollutants;
measures
to
reduce
emissions
of
stratospheric
ozone
depleting
substances;
and
requirements
to
limit
acid
rain
precursor
emissions.

Cost
estimates
are
often
a
critical
factor
in
the
choice
of
a
regulatory
option.
The
process
of
estimating
future
compliance
costs,
however,
is
demanding
and
involves
uncertainties.
To
accurately
estimate
costs,
analysts
must,
at
a
minimum:
estimate
the
size
and
character
of
the
existing
and
projected
future
population
of
affected
emission
sources;
determine
the
potential
suite
of
pollutant
control
technologies
or
process
changes
that
can
meet
a
range
of
future
standards;
assess
the
control
efficiency
and
individual
cost
of
relevant
pollutant
control
technologies
(
both
now
and
in
the
future),
with
attention
to
source­
by­
source
variability
in
capital
requirements
and
maintenance
costs;
and
apply
this
information
to
generate
an
aggregate
cost
estimate
for
the
regulation.
The
better
analyses
also
acknowledge
that
technological
diffusion
and
innovation
can
play
a
factor
in
future
cost
estimates,
though
projecting
future
technologies
and
their
costs
is
very
difficult.

In
this
paper,
we
argue
that
there
are
three
key
factors
inherent
in
the
process
of
estimating
regulatory
costs
that
suggest
direct
compliance
cost
projections
may
more
often
be
overestimated
than
underestimated.
These
include:

$
Learning
by
doing
­
As
with
many
other
products,
the
costs
of
producing
pollution
control
equipment
and
the
costs
of
operating
modified
production
systems
may
fall
over
time
or
as
markets
expand
because
of
economies
of
scale
or
other
realized
efficiencies
in
production.
1
The
estimates
were
presented
in
a
September
21,
1990
letter
reply
by
the
then­
Chairman
of
the
CEA,
Michael
J.
Boskin,
to
a
request
for
cost
estimates
for
various
bills
from
John
D.
Dingell,
then­
Chairman
of
the
House
Committee
on
Energy
and
Commerce
(
sometimes
referred
to
as
"
The
Boskin
Letter").
We
present
the
estimate
for
the
Senate
Compromise
bill,
which
most
closely
approximates
the
CAAA
as
passed.
The
estimates
were
developed
when
the
amendments
were
in
conference.

2
$
Innovations
and
technological
change
­
The
presence
of
a
new
regulatory
requirement
can
spur
a
faster
rate
of
technological
diffusion
and
innovation,
as
incentives
to
comply
in
the
most
cost­
effective
manner
take
hold,
and
innovation
that
occurs
independent
of
the
regulation
can
also
decrease
costs.
Regulatory
analyses
that
assume
constant
technology
miss
both
effects.

$
Cost­
reducing
features
of
regulatory
design
­
Flexible
regulatory
design
provides
more
opportunities
for
cost
saving,
and
may
include:
designs
that
take
advantage
of
market
efficiencies,
such
as
emission
trading;
designs
that
are
responsive
to
technological
limitations
by
tailoring
delayed
implementation
periods;
and
processes
that
involve
significant
stakeholder
involvement
to
build
in
provisions
that
achieve
the
same
environmental
goals
at
lower
cost.
These
features
tend
to
reinforce
and
perpetuate
the
effects
of
learning
by
doing
and
innovations
and
technological
change,
and
also
provide
flexibility
to
respond
to
unanticipated
changes
that
can
decrease
compliance
costs
relative
to
projected
costs.

One
of
the
better
studied
examples
of
the
difficulty
in
predicting
compliance
costs
ex
ante
is
the
Clean
Air
Act
Title
IV
program
for
SO
2
emissions
trading;
this
example
illustrates
the
impact
of
each
of
the
three
factors
listed
above.
Prior
to
the
passage
of
the
Clean
Air
Act
Amendments
of
1990,
the
President's
Council
of
Economic
Advisors
estimated
that
the
annual
costs
for
this
program
would
reach
$
4.2
billion
by
2005.1
By
1997,
it
was
clear
that
the
costs
of
compliance
were
much
lower
than
expected.
In
November
of
1997,
EPA's
Report
to
Congress,
Benefits
and
Costs
of
the
Clean
Air
Act:
1990­
2010,
estimated
annual
compliance
costs
for
this
program
of
$
2.3
and
$
2.0
billion
for
years
2000
and
2010,
respectively.
The
reasons
behind
this
dramatic
reduction
in
cost
estimates
included
the
following:

$
Learning
by
doing
­
The
installed
costs
for
sulfur
scrubbers
proved
to
be
more
costeffective
than
anticipated,
due
in
part
to
better
removal
efficiencies
than
anticipated.

$
Innovations
­
Some
plants
found
that,
as
a
result
of
minor
modifications
in
their
processes,
they
could
achieve
dramatically
greater
abilities
to
burn
low
sulfur
cost
than
anticipated
at
lower
cost.
2
Winston
Harrington,
Robert
D.
Morgenstern,
and
Peter
Nelson,
"
On
the
Accuracy
of
Regulatory
Cost
Estimates,"
Resources
for
the
Future
Discussion
Paper
99­
18,
January
1999.

3
See
Harrington
et
al.
(
1999),
page
23.

4
Some
of
these
factors
also
affect
the
specification
of
benefits.
Without
further
information
on
the
case­
specific
marginal
benefit
and
cost
curves,
it
is
difficult
to
assess
the
overall
effect
of
such
mis­
estimations
on
benefit/
cost
ratios.
In
this
paper,
we
focus
on
factors
that
affect
only
costs.

3
$
Cost­
reducing
features
of
regulatory
design
­
The
sulfur
emissions
trading
program
provided
a
continuing
incentive
to
reduce
costs,
which
rewarded
reductions
achieved
through
learning
by
doing.
In
addition,
the
trading
program
also
provided
incentives
to
take
maximum
advantage
of
two
unanticipated
developments:
(
1)
The
discovery
of
larger
reserves
of
low­
sulfur
coal
in
the
Powder
River
Basin
area
of
Wyoming
and
Montana;
and
(
2)
A
drop
in
costs
for
rail
transportation
of
coal
that
resulted
from
deregulation
of
railroad
rates.

While
the
sulfur
emissions
trading
program
is
one
of
the
richest
examples
of
these
factors
at
work,
it
is
certainly
not
the
only
example.
In
the
pages
that
follow,
we
provide
additional
examples
to
illustrate
the
potential
impact
of
each
of
these
three
factors.

This
paper
draws
on
a
wide
range
of
existing
literature
that
comments
on
the
accuracy
of
prospective
regulatory
cost
estimates,
and
draws
heavily
on
a
recent
survey
by
Harrington,
Morgenstern,
and
Nelson
(
1999).
2
These
authors
evaluate
the
accuracy
of
EPA
and
OSHA
estimates
of
25
ex
ante
regulatory
costs
relative
to
ex
post
studies
of
actual
costs.
They
conclude:

"
Our
review
of
over
two
dozen
ex
ante­
ex
post
comparisons
indicates
that
ex
ante
estimates
of
total
costs
have
tended
to
exceed
actuals.
We
find
this
to
be
true
of
12
of
the
25
rules
in
our
data
set,
while
for
only
6
were
the
ex
ante
estimates
too
low.
Since
the
overestimates
occur
more
frequently
in
the
larger
rules,
the
dollar­
weighted
predominance
of
overestimates
would
be
even
higher."
3
In
discussing
EPA
cost
estimates,
the
Harrington
et
al.
(
1999)
paper
identifies
several
reasons
for
cost
overestimation
errors,
including
the
difficulty
of
predicting
technological
innovation;
uncounted
cost
reductions
achieved
during
the
regulatory
review
and
public
comment
periods;
reliance
on
maximum
costs
rather
than
means
(
i.
e.,
a
built­
in
conservative
bias);
and
inaccuracies
in
estimating
the
size
of
the
affected
universe.
4
Other
authors
have
commented
on
the
accuracy
of
regulatory
cost
estimates.
Hahn
(
1998)
and
Jaffe
et
al.
(
1995)
conclude
costs
are
consistently
underestimated
because
of
a
failure
to
characterize
the
negative
effects
of
direct
compliance
expenditures
on
investment,
productivity,
and
5
It
can
also
be
argued
that
there
is
a
counterbalancing
effect
of
gains
in
health
on
labor
and
overall
productivity
associated
with
cleaner
air;
this
effect
is
also
rarely
acknowledged
and
quantified
on
the
benefits
side
of
the
equation.

6
See,
for
example,
the
following
citations
from
our
bibliography:
Kolstad
(
2000),
Kemp
(
1997),
Hahn
(
1998),
Hodges
(
undated).

7
Note
that
we
also
exclude
discussion
of
important
sources
of
uncertainty
that
may
limit
analysts'
ability
to
characterize
the
affected
universe
or
the
compliance
scenario;
while
those
factors
may
affect
total
cost
estimates,
they
also
affect
the
implied
environmental
benefit
received
from
incurring
the
cost,
because
emissions
estimates
may
also
be
affected.

4
market
structure,
although
Hahn
(
1998)
acknowledges
that
these
effects
are
very
difficult
to
reliably
quantify.
5
In
addition,
some
authors
have
suggested
that
strategic
incentives
are
an
important
influence
on
cost
estimates.
6
The
implication
of
strategic
biases
is
that
industry
estimates
are
inflated
to
discourage
additional
regulation,
and
regulators'
estimates
are
deflated
to
encourage
regulation.
While
it
would
be
naive
to
fail
to
acknowledge
that
strategic
influences
exist,
it
is
not
our
purpose
to
evaluate
these
claims.
Rather,
we
hope
to
identify
areas
for
objective
analysis
and
potential
adjustment
factors
to
correct
apparent
biases.

In
this
paper,
we
focus
on
those
factors
that
affect
the
accuracy
of
initial
estimates
of
direct
compliance
cost
(
e.
g.,
for
specific
capital
expenditures,
operating
or
maintenance
costs,
or
overall
compliance
costs)
to
achieve
a
given
environmental
goal
(
e.
g.,
a
stated
emission
reduction
or
overall
cap).
Direct
compliance
costs
as
we
consider
them
do
not
include
social
costs,
secondary
impacts,
or
other
real
effects
of
regulation;
accurate
estimation
of
those
impacts,
however,
often
relies
critically
on
initial
estimates
of
direct
costs.
7
Our
conclusion
is
that
systematic
errors
in
cost
estimation
such
as
those
arising
from
the
four
factors
outlined
above
ought
to
be
studied
further
and,
where
possible,
the
effects
of
these
four
factors
should
be
quantitatively
addressed.
The
presence
of
these
factors
does
not
always
imply
that
costs
are
overestimated;
however,
further
research
could
clarify
the
conditions
under
which
costs
might
be
overestimated
or
underestimated,
as
well
as
the
potential
magnitude
of
the
errors.

In
the
sections
that
follow,
we
discuss
each
of
these
factors
in
more
detail,
explain
the
general
theory
supporting
the
hypothesis
that
costs
are
more
likely
to
be
overestimated
than
underestimated,
and
provide
specific
historical
examples
to
support
the
argument.
Finally,
we
summarize
our
conclusions
and
identify
areas
for
further
empirical
research.
8
For
a
discussion
on
the
quantitative
estimate
of
learning
curve
impacts
in
manufacturing,
see
Horngren,
C.,
G.
Foster,
and
S.
Datar,
Chapter
10:
"
Determining
How
Costs
Behave"
in
Cost
Accounting:
A
Managerial
Emphasis,
Prentice
Hall,
1994.
See
also
Viscusi,
W.
Kip
et
al.
Economics
of
Regulation
and
Antitrust.
(
Cambridge:
MIT
Press)
2000.
Viscusi
describes
an
equation
using
cumulative
past
output
as
a
measure
of
experience
(
i.
e.
learning
curve)
to
illustrate
how
a
cost
function
changes
with
a
learning
curve
effect.

9
See,
for
example,
Kemp,
R.
Environmental
Policy
and
Technical
Change:
A
Comparison
of
the
Technological
Impact
of
Policy
Instruments
(
Cheltenham,
U.
K.
and
Brookfield,
U.
S.;
Edward
Elgar)
1997.

10
Kemp
notes
that
markets
for
pollution
control
devices
are
often
dependent
on
clear
regulatory
signals
for
stability
in
certain
cases;
this
observation
would
tend
to
strengthen
the
argument
that
regulatory
activity
and
increased
production
may
often
be
linked.

5
FACTOR
1:
"
LEARNING
BY
DOING"

One
source
of
cost
overestimates
by
both
industry
and
government
is
the
failure
to
predict
decreases
in
the
costs
of
production
over
time
as
a
result
of
the
"
learning
curve
effect"
and
of
increases
in
economies
of
scale.
Both
of
these
concepts
are
well­
known
in
classical
economic
theory,
and
numerous
empirical
studies
and
modeling
efforts
have
been
developed
that
attempt
to
predict
the
extent
to
which
they
affect
prices.

The
concept
of
"
learning
by
doing"
(
or
the
learning
curve
effect),
predicts
that
the
production
costs
decrease
with
repetition,
as
cumulative
experience
increases
(
e.
g.,
as
a
result
of
employee
training
or
modifications
to
production
and
distribution
processes).
Operations
management
literature
and
practice
regularly
addresses
the
learning
curve
effect
in
cost
projections,
both
as
it
occurs
during
the
introduction
of
new
products
or
production
systems
and
as
part
of
continuing
system
efficiency
estimates.
8
The
learning
curve
impact
can
affect
regulatory
cost
estimates
in
two
ways:
first,
the
cost
of
producing
(
and
purchasing)
pollution
control
equipment
will
likely
decrease
over
time;
and
second,
upgraded
production
systems
at
polluting
facilities
can
be
expected
to
become
more
efficient.

A
separate
but
related
force
affecting
cost
forecasting
is
the
impact
of
changes
in
economies
of
scale.
The
cost
implications
of
economies
of
scale
are
well
documented
and
can
be
expected
if
regulations
create
demand
for
certain
products
or
technologies.
However,
it
is
difficult
to
isolate
and
predict
the
impact
on
costs
that
would
result
from
a
regulation.
9
Even
when
demand
for
these
technologies
can
be
predicted
to
increase
substantially,
it
is
difficult
to
predict
the
extent
to
which
polluting
companies
will
enjoy
cost
reductions
associated
with
increased
production.
Market
characteristics
such
as
the
existence
of
patents,
the
competitive
position
and
market
power
of
technology
producers,
and
the
development
of
alternative
technologies
also
influence
the
extent
to
which
producers
can
reduce
costs
and
pass
on
cost
reductions
to
regulated
facilities.
10
11
Skip
Laitner,
of
EPA's
Office
of
Air
Programs,
has
also
completed
a
number
of
publications
on
the
impact
of
the
learning
curve
in
agency
cost
estimations.
Time
constraints
did
not
allow
us
to
collect
and
use
these
sources,
but
they
may
provide
additional
theoretical
detail
and
empirical
examples
that
would
support
the
above
discussion.

12
Another
example
relevant
to
a
discussion
of
learning
curve
impacts
and
economies
of
scale
relates
to
stationary
source
NOx
Control.
A
1991
study
by
the
Electric
Power
Research
Institute
(
EPRI)
estimated
costs
of
installing
selective
catalytic
reduction
(
SCR)
at
$
96­$
105
per
kilowatt
for
dry
bottom
coal­
fired
boilers,
and
$
125­$
140
for
coal­
fired
cyclone
boilers
(
1989
dollars).
A
1994
installation
of
a
cyclone
boiler
at
a
plant
in
New
Hampshire
cost
$
70
per
kilowatt
(
1995
dollars),
and
a
1996
engineering
estimate
suggested
that
the
total
costs
were
$
55
to
$
84
per
kilowatt
for
implementing
six
systems
on
dry
bottom
boilers.
Further
research
is
necessary
to
verify
the
estimates
provided
and
to
determine
to
what
extent
the
reported
differences
are
the
result
of
the
learning
curve
effect
or
economies
of
scale.
See:
attachment
to
Testimony
of
the
Institute
of
Clean
Air
Companies,
Inc.
before
the
Subcommittee
on
Manufacturing
and
Competitiveness
on
EPA's
Revised
NAAQS
for
Particulate
Matter
and
Ozone,
September
24,
1997.

13
Note
also
that
a
January
23,
1995
EPA
fact
sheet
"
Existing
Medical
Waste
Incinerators
C
Proposed
Subpart
Cc
Emission
Guidelines"
identifies
annual
costs
of
$
351
million
per
year
for
a
universe
of
5,000
facilities,
implying
an
annual
compliance
cost
of
$
70,000
per
unit.
This
figure
is
likely
to
include
both
annualized
capital
costs
and
annually
occurring
compliance­
related
costs.

6
Both
the
learning
curve
and
changes
in
economies
of
scale
are
often
associated
with
the
broader
phenomenon
of
technological
innovation,
in
part
because
they
are
often
most
dramatic
when
products
and
technologies
are
relatively
new.
11
However,
we
consider
them
separately
from
innovation
(
which
is
discussed
below)
because
the
influence
is
not
specifically
restricted
to
the
development
of
new
technologies
and
products.
Moreover,
while
broadly
defined
innovation
may
occur
in
response
to
a
wide
range
of
regulatory
programs,
significant
learning
curve
and
scale
economy
impacts
are
most
likely
to
be
associated
with
regulations
that
anticipate
(
or
require)
the
use
of
specific
identified
technologies.
Below
we
provide
examples
of
air­
related
technology
cost
estimates
that
failed
to
account
for
decreasing
costs.
12
Medical
Waste
Incinerator
Continuous
Emission
Monitoring
Systems
The
Institute
of
Clean
Air
Companies
(
ICAC)
produced
a
study
in
1997
responding
to
EPA's
proposed
maximum
achievable
control
technology
(
MACT)
standards
for
medical
waste
incinerators
(
MWIs).
13
ICAC's
study
forecasts
the
costs
of
continuous
emissions
monitoring
systems
(
CEMS)
for
carbon
monoxide
and
oxygen
emissions
from
MWIs.
CO
and
O
2
emissions
are
indicators
of
incinerator
efficiency;
inefficient
incineration
can
result
in
release
of
partially
combusted
materials,
including
air
toxics,
if
temperatures
are
too
low.
14
"
Testing
and
Monitoring
Options
and
Costs
for
MWIs",
Memorandum
from
Thomas
Holloway
of
MRI
to
Rick
Copland,
Dated
May
20,
1996;
cited
in
Testimony
of
the
Institute
of
Clean
Air
Companies,
Inc.
before
the
Subcommittee
on
Manufacturing
and
Competitiveness
on
EPA's
Revised
NAAQS
for
Particulate
Matter
and
Ozone,
September
24,
1997.

15
Air
Emissions
Monitoring
for
Safe
and
Efficient
Medical
Waste
Incinerator
Operation,
ICAC,
September
1997.

7
A
1996
study
by
EPA
estimated
that
the
cost
of
purchasing
a
CO/
O
2/
opacity
monitoring
system
would
be
$
160,000;
annual
operating
costs
were
estimated
at
$
96,000.14
The
ICAC
study
uses
1997
current
vendor
quotes
to
compare
to
Agency
cost
model
results
for
actual
installation
costs
for
existing
CEMS
with
current
market
quotes
for
the
same
processes.
Examples
include:

°
A
CO/
O
2
/
opacity
monitoring
system
that
cost
$
141,000
in
1994
would
cost
$
131,000
in
1997
because
of
technical
advances
such
as
the
use
of
a
combined
CO/
O
2
analyzer.
The
document
further
notes
that
recent
actual
bids
for
installation,
start­
up,
certification,
and
other
services
are
also
lower
than
1994
estimates.

°
A
second
CO/
O
2/
opacity
monitoring
system
sold
in
1991
cost
$
139,745;
including
equipment,
installation,
training
and
certification
testing;
by
1996
costs
for
a
more
complex
set
of
two
colocated
systems
was
estimated
at
roughly
$
180,000;
or
just
under
$
90,000
per
system.

In
explaining
the
change
in
CEMS
capital
costs,
ICAC
identifies
both
technical
innovation
and
"
increased
competition"
as
factors.
In
addition,
the
discussion
notes
that
the
technologies
are
easier
to
install
and
maintain,
indicating
a
learning
curve
in
installation.
15
Low
Emission
Vehicle
(
LEV)
Technology
In
1998
Tom
Cackette
of
California
Air
Resources
Board
(
CARB)
presented
a
case
study
comparing
projections
by
CARB
and
by
auto
industry
consultants
of
the
incremental
costs
for
new
or
revised
emission
control
devices
meeting
the
Low
Emission
Vehicles
(
LEV)
standards
for
the
state
of
California.
The
comparison
is
as
follows:
8
Exhibit
1
COMPARISON
OF
COST
ESTIMATES
FOR
LEV
EMISSION
CONTROL
DEVICES
Engine
Type
1994
Industry
Estimate
1994$/(
1998$)
a
1994
CARB
Estimate
1994$/(
1998$)
a
1998
Actual
Retail
Cost
1998$
b
Average
Fleet
Estimate
$
788
($
867)
c
$
114
($
125)
$
83
4­
Cylinder
Honda
Civic
­
$
85
($
94)
$
75
6­
Cylinder
Toyota
Camry
­
$
137
($
151)
$
79
Ford
V8
­
$
140
($
154)
$
152
Source:
EPA
Center
on
Airborne
Organics
1998
Summery
Symposium
Report:
Summary
of
Tom
Cackette
presentation.
Available
at
http://
web.
mit.
edu/
afs/
athena.
mit.
edu/
org/
a/
airquality/
www/
rep1998.
html.
Notes:
a
For
the
purposes
of
comparison
with
1998
actual
retail
costs,
we
have
included
in
italics
adjusted
estimate
of
1994
CARB
and
industry
estimate
reflecting
inflation
according
to
the
Consumer
Price
Index
from
1994
to
1998.
Source:
Council
of
Economic
Advisors
Report
of
the
President
Table
B­
63
(
January
2001).
b
Note
that
retail
cost
may
reflect
some
absorption
of
costs
by
manufacturers
to
ensure
competitive
product
pricing,
but
this
effect
is
unlikely
to
explain
the
entire
discrepancies
across
estimates.
c
The
summary
of
Mr.
Cackette's
work
provided
only
a
general
industry
cost
estimate;
industry
estimates
may
have
included
more
detailed
estimates
for
specific
engines.

The
results
in
Exhibit
1
indicate
that
CARB's
estimates
were
generally
accurate
or
conservative
(
note
that
CARB's
1994
estimate
of
$
140
is
equivalent
to
$
154
in
1998),
and
the
industry
estimate
was
very
high.
Cackette
attributes
the
overestimate
to
several
factors,
including:
projected
use
of
more
expensive
equipment,
overestimates
of
warranty
and
component
costs,
and
an
assumption
that
redesign
of
models
for
catalysts
would
require
unscheduled
production
delays.
In
fact,
"
sheet
metal
changes"
to
relocate
catalysts
were
achieved
during
normal
model
revisions.
While
it
is
unclear
what
part
of
the
total
estimate
is
attributable
to
component
costs
and
sheet
metal
changes,
it
is
likely
that
the
learning
curve
effect
influenced
both
elements,
and
economies
of
scale
may
also
have
reduced
component
costs
as
demand
for
technology
increased.

SO2
Emissions
Trading
Program
­
Scrubber
Technology
Prices
"
Scrubbers"
(
i.
e.,
Flue
Gas
Desulfurization
(
FGD)
technology)
are
an
essential
tool
in
the
effort
to
reduce
sulfur
dioxide
(
SO
2)
emissions
as
part
of
Title
IV
"
acid
rain"
program.
The
cost
of
scrubber
technology
has
been
widely
debated
and
documented,
and
shows
consistent
downward
trends
over
time.
Multiple
analyses
of
scrubbers
identify
market
shifts
that
reflect
both
learning
curve
and
scale
economy
impacts.
Several
examples
are:
16
See
Smith,
J.
and
S.
Dalton,
"
FGD
Markets
&
Business
in
an
Age
of
Retail
Wheeling"
presented
at
EPA/
EPRI/
DOE
SO
2
Symposium,
1995,
cited
in
Testimony
of
the
Institute
of
Clean
Air
Companies,
Inc.
before
the
Subcommittee
on
Manufacturing
and
Competitiveness
on
EPA's
Revised
NAAQS
for
Particulate
Matter
and
Ozone,
September
24,
1997.

17
Institute
of
Clean
Air
Companies,
Scrubber
Myths
&
Realities,
White
Paper,
May
1995.

18
Burtraw,
Dallas.
"
Cost
Savings
Sans
Allowance
Trades?
Evaluating
the
SO2
Emission
Trading
Program
to
Date",
Resources
for
the
Future,
February
1996.

9
Testimony
in
1997
by
ICAC
noted
that
overall
scrubber
purchase
and
installation
was
less
than
predicted.
In
1989,
the
utility
industry
estimated
FGD
retrofit
costs
as
$
382
per
kilowatt
at
an
"
average
sized
coal­
fired
boiler."
However,
as
of
1997
the
average
FGD
retrofit
cost
had
been
$
227
per
kilowatt
and
was
still
declining.
16
A
separate
1995
white
paper
by
ICAC
identified
some
of
the
specific
causes
of
the
disparity:
17
°
Alterations
to
the
scrubbers,
including
larger
absorber
modules
and
the
recognition
that
spare
absorbers
were
unnecessary
decreased
capital
costs
by
over
30%.

°
Other
cost­
savings
resulted
from
elimination
of
flue
gas
reheat,
incorporation
of
additives
in
the
process
design,
simplification
of
design,
by­
product
management,
higher
velocity
absorbers,
alternative
duct
work
designs,
and
reduced
reagent
preparation
costs.

°
Market
pressures
such
as
vendor
competition
and
diversity
of
FGD
systems
had
lowered
costs.

These
observations
were
supported
by
a
1996
study
by
Dallas
Burtraw
which
noted
the
GAO
had
found
that
prices
for
scrubbers
had
decreased
by
over
50
percent
since
1989,
and
that
the
elimination
of
the
spare
absorber
modules
was
specifically
important.
In
addition,
the
paper
noted
that
reduced
maintenance
costs
and
increased
utilization
rates
had
contributed
to
lower
compliance
costs.
18
While
some
of
the
cost
decreases
are
the
result
of
innovative
alterations
to
the
scrubber
units,
the
general
technology
has
not
been
replaced;
cost
savings
were
attributable
in
part
to
more
gradual
adjustments
in
production
and
installation
as
the
technology
became
more
common.

FACTOR
2:
INNOVATIONS
AND
TECHNOLOGICAL
CHANGE
Economists
generally
agree
that
technological
innovation
is
a
key
force
in
decreasing
the
costs
of
compliance
with
regulations
over
time,
and
that
failure
of
government
and
industry
to
account
for
19
Logic
dictates
that
innovations
which
increase
costs
will
generally
not
be
implemented;
as
a
result,
innovation
tends
to
reduce
costs.

20
Certain
regulatory
assessment
efforts
have
attempted
to
address
technological
change.
For
example,
for
EPA's
retrospective
assessment
of
the
Clean
Air
Act,
Benefits
and
Costs
of
the
Clean
Air
Act;
1970
to
1990,
the
Agency
estimated
the
total
social
costs
of
the
regulatory
program
using
the
Jorgenson/
Wilcoxen
model.
This
model
incorporates
an
internal
function
that
considers
the
impact
of
innovation
on
production
and
demand
for
capital.
However,
the
general
equilibrium
impacts
of
a
single
rulemaking
are
likely
to
be
too
subtle
in
most
cases
to
be
measured
with
large
scale
models,
and
are
not
generally
included
in
compliance
cost
analyses.

21
For
one
well­
known
discussion
of
the
positive
link
between
regulation
and
innovation,
see
Michael
Porter
and
Claas
Var
der
Linde,
"
Green
and
Competitive,"
Harvard
Business
Review,
Sep­
Oct.
1995,
p.
120.

22
Kemp,
R.
Environmental
Policy
and
Technical
Change:
A
Comparison
of
the
Technological
Impact
of
Policy
Instruments
(
Cheltenham,
U.
K.
and
Brookfield,
U.
S.;
Edward
Elgar)
1997.

10
innovation
is
often
an
important
element
of
compliance
cost
overestimates.
19
Harrington
et
al.
(
1999)
identify
the
failure
to
account
for
innovation
as
a
chief
reason
for
the
inaccuracies
of
cost
predictions.
As
Harrington
notes,
however,
regulatory
cost
estimates
by
both
industry
and
government
have
historically
avoided
making
ex
ante
predictions
about
technological
change
that
would
occur
either
as
a
result
of
the
proposed
regulation,
or
as
a
logical
extension
of
the
learning
curve
effect
and
other
general
economic
realities.
20
Technological
change
is
notoriously
difficult
to
predict
using
neo­
classical
economic
theory
because
by
definition
it
alters
the
economics
of
production.
Even
more
sensitive
is
any
attempt
to
reflect
a
causal
relationship
between
regulation
and
innovation
in
cost
estimates.
While
there
is
theoretical
and
empirical
support
for
the
notion
that
regulatory
challenges
may
encourage
innovation
in
many
cases
(
i.
e.,
by
requiring
new
levels
of
performance
or
creating
incentives
for
reduction
of
certain
outputs),
regulations
which
require
specific
technologies
may
in
fact
limit
innovation
once
implementation
is
achieved
(
though
cost
reductions
through
learning
curve
effects
would
still
be
relevant).
21
Finally,
the
economic
implications
of
innovation
frequently
include
secondary
impacts
such
as
productivity
(
i.
e.,
if
innovation
includes
a
significant
production
process
alteration),
and
it
can
be
difficult
to
isolate
the
narrower
compliance
cost
implications
of
shifts
in
technology.
However,
in
some
cases
regulations
can
be
designed
to
provide
continuing
incentives
to
innovate;
in
these
instances
there
can
be
a
clear
positive
correlation
between
regulatory
activities
and
the
pace
of
innovation.

In
Environmental
Policy
and
Technical
Change,
René
Kemp
makes
an
important
distinction
between
the
first
two
stages
of
technological
change
(
e.
g.,
invention
and
innovation)
and
the
diffusion
of
technologies.
22
Diffusion
of
technology
is
the
phase
which
determines
the
ultimate
economic
impact
of
the
innovation,
and
may
be
more
easily
linked
to
regulatory
requirements,
since
it
is
in
part
23
Two
other
examples
not
cited
below
involved
regulation
of
emissions
from
coke
oven
batteries
and
benzene
from
chemical
production
plants.
Coke
is
used
in
the
blast
furnace
process
for
producing
steel.
The
strict
limits
on
coke
oven
emissions
promulgated
in
the
early
1990'
s
encouraged
a
trend
that
was
already
occurring
in
the
steel
production
industry
toward
the
use
of
mini­
mills
(
i.
e.,
using
scrap
steel
as
input
rather
than
iron
ore
and
coke).
We
are
not
aware
of
any
ex
post
cost
estimates
to
compare
to
ex
ante
cost
estimates,
but
recent
efforts
re­
visit
the
coke
oven
regulations
in
order
to
set
residual
risk
limits
under
Title
III
may
provide
an
opportunity
to
revisit
this
case.
Relevant
to
benzene
emissions
reductions,
in
the
1970s
ex
ante
estimates
by
industry
for
control
of
benzene
emissions
were
$
350,000;
however,
chemical
production
plants
"
virtually
eliminated"
control
costs
by
developing
a
process
substituting
other
chemicals
for
benzene.
See
Keith
Mason,
"
The
Economic
Impact"
EPA
Journal,
January/
February
1991,
45­
7.

24
Note
that
forecast
costs
varied
with
compounds,
and
that
RAND
actually
underestimated
costs
for
CFC­
113,
though
the
costs
for
CFCs­
11
and
­
12
were
overestimated.
The
reason
for
the
underestimate,
Hammitt
suggests,
may
be
that
reductions
in
CFC­
113
consumption
have
relied
more
on
substituting
other
solvents
and
blends
and
more
efficient
production
instead
of
through
substitution
with
new
compounds:
source:
Hammitt,
J.,
"
Are
the
Costs
of
Proposed
Environmental
Regulations
Overestimated?
Evidence
from
the
CFC
Phaseout,"
Environmental
and
Resource
Economics,
16:
281­
301,
2000.

11
a
function
of
demand.
To
the
extent
that
innovations
represent
incremental
changes,
Kemp
notes
that
a
number
of
models
are
available
to
describe
and
support
predictions
of
both
innovation
and
diffusion
under
certain
market
circumstances;
however,
he
notes
that
for
"
radical"
technology
changes
(
e.
g.,
the
change
in
power
sources
from
physical
to
fossil
fuel
in
1900)
prediction
of
impacts
is
extremely
difficult.

Kemp
outlines
three
general
categories
of
innovation
linked
to
the
adoption
of
environmentally
preferable
technologies:
production
process
changes
by
polluters
(
e.
g.,
end
of
pipe
treatment
and
process
integrated
systems
such
as
CFC
recycling);
product
changes
(
e.
g.,
product
reformulation
to
eliminate
CFCs
and
product
substitution)
and
new
technology
systems
(
e.
g.,
new
energy
production
systems).
The
following
are
examples
of
technological
innovations
in
the
first
two
of
these
categories
that
have
had
significant
impacts
on
initial
cost
estimates.
23
CFC
Phaseout
As
the
U.
S.
developed
and
implemented
the
phaseout
of
CFCs
in
the
late
1980s,
a
series
of
cost
estimates
by
RAND
and
by
EPA
illustrate
the
impact
of
rapid
technological
advances
in
the
development
of
suitable
substitutes.
As
EPA
developed
a
rule
limiting
the
U.
S.
consumption
of
CFCs,
a
report
by
RAND
in
1986
predicted
marginal
control
costs
associated
with
control
of
CFCs­
11,
­
12
and
­
113.
RAND
forecast
that
the
phaseout
would
create
immediate
price
increases
to
six
to
eight
dollars
per
kilogram
of
CFCs
(
compared
to
the
1986
price
of
less
than
two
dollars
per
kilogram).
RAND
also
predicted
that
the
total
decrease
in
consumption
from
baseline
would
be
less
than
20
percent,
even
in
the
face
of
cost
increases
above
200
percent.
24
In
contrast,
the
actual
price
25
See
Hammitt,
J.
"
Are
the
Costs
of
Proposed
Environmental
Regulations
Overestimated?
Evidence
from
the
CFC
Phaseout"
Environmental
and
Resource
Economics,
16:
281­
301,
2000.
Note
also
the
$
30
million
"
golden
carrot"
incentive
from
industry/
NRDC
to
Whirlpool
for
efficient
refrigerator,
and
EPA
cooperation
with
military
to
meet
Montreal
Protocol;
source:
Ozone
Protection
in
the
United
States:
Elements
of
Success,
WRI,
1996.

26
For
comparison,
all
dollar
figures
in
this
paragraph
are
in
1997
dollars.
Source:
Congressional
Testimony
of
P.
N.
Gammelgard,
Vice
President,
American
Petroleum
Institute,
March
18,
1975;
1998
estimates
from
Environmental
Protection
Agency,
"
EPA
Staff
Paper
on
Gasoline
Sulfur
Issues"
May
1,
1998.
Both
quoted
in
Industrial
Economics,
Incorporated,
"
Clean
Air
Act
Cost
and
Benefit
Research,"
unpublished
memorandum
to
Keith
Mason,
August
10,
1998.

12
per
pound
of
CFCs
only
reached
RAND's
predicted
range
in
1990,
when
reductions
of
40
to
60
percent
of
1986
consumption
had
been
achieved.
Moreover,
manufacturers
were
able
to
meet
the
accelerated
timetables
that
EPA
developed
throughout
the
early
1990s,
subsequent
to
Montreal
Protocol
amendments.
Timetables
included
a
1990
goal
of
50
percent
CFC
reduction
by
1998,
a
1992
goal
of
elimination
by
2000,
and
the
final
1993
goal
of
elimination
by
1996.

The
impact
of
innovation,
however,
is
most
apparent
in
the
intermediate
estimates
during
the
development
of
the
CFC
phaseout.
In
1988,
less
than
two
years
after
RAND's
1986
estimate
that
CFC
prices
would
spike
to
over
six
dollars
per
kilogram,
EPA
provided
an
estimate
of
$
3.55
per
kilogram.
This
estimate
reflected
the
options
that
had
begun
to
emerge
in
the
16
months
since
the
RAND
study;
during
that
16
month
period
a
concentrated
research
effort
by
government
and
industry
was
underway
to
identify
CFC
substitutes.
The
continuing
development
of
economic
substitutes
resulted
in
a
surprising
reduction
in
total
cost
estimates
by
1993,
even
as
the
schedule
for
implementation
of
CFC
ban
was
accelerated.
The
identification
of
HCF­
134a
as
an
effective
substitute
was
responsible
for
a
large
part
of
the
cost
reductions;
process
changes
reducing
need
for
solvent
have
also
helped.
25
Other
rapid
innovations
included
the
development
of
CFC
recycling.

Fuel
Desulfurization
In
1975,
the
American
Petroleum
Institute
(
API)
estimated
its
cost
of
converting
facilities
to
produce
low
sulfur
(
100
ppm)
gasoline
at
$
11.8
billion
in
1997
dollars.
26
In
1998,
an
EPA/
DOE
study
estimated
that
the
total
actual
cost
of
the
same
reduction
was
$
2.4
to
$
3.9
billion
(
based
on
1975
production
levels).
API
also
re­
examined
the
retrofit
costs
in
1998,
and
produced
an
estimate
of
$
4.0
billion
overall.
The
discrepancy
between
the
1975
estimate
and
the
1998
estimates
is
due
in
part
to
the
fact
that
the
1975
estimates
were
based
only
on
currently
available
sulfur
removal
technologies,
and
did
not
take
into
account
technological
innovation.
On
the
other
hand,
EPA
and
later
API
estimates
incorporate
the
emergence
of
sulfur
removal
technologies
that
could
achieve
levels
of
40
ppm
at
a
cost
of
only
1­
2
cents
per
gallon.

Similarly,
initial
industry
cost
estimates
of
lowering
sulfur
levels
in
diesel
fuel
were
overestimated
as
a
result
of
failing
to
account
for
innovation.
In
Congressional
testimony
in
1987,
27
Sources:
Congressional
Testimony
of
James
B.
Hermiller,
Vice
President
of
Refining,
Standard
Oil
Company;
June
17,
1987;
U.
S.
Environmental
Protection
Agency,
Regulation
of
Fuels
and
Fuel
Additives:
Fuel
Quality
Regulations
for
Highway
Diesel
Fuel
Sold
in
1993
and
Later
Calendar
Years,
1990;
and
Department
of
Energy
Web
Site
information
obtained
on
August
7,
1998;
both
quoted
in
Industrial
Economics,
Incorporated,
"
Clean
Air
Act
Cost
and
Benefit
Research,"
unpublished
memorandum
to
Keith
Mason,
August
10,
1998.

28
Harrington,
Winston,
"
Predicting
the
costs
of
environmental
regulations:
how
accurate
are
regulators'
estimates?"
Environment,
Sept.
1999.

29
See
Burtraw
(
1996).

13
Standard
Oil
Company
cited
1986
API
estimates
that
reducing
diesel
fuel
sulfur
levels
to
0.05
weight
percent
would
increase
costs
to
refiners
and
translate
into
an
8.0
cent
per
gallon
increase
to
consumers
(
Standard
Oil
also
testified
that
reducing
diesel
fuel
sulfur
levels
would
not
contribute
to
reduction
of
acidic
deposition).
In
contrast,
a
1986
EPA
study
that
predicted
innovation­
related
cost
savings
estimated
costs
to
be
only
1.2
cents
per
gallon.
Additional
studies
included
a
1989
EPA
estimate
of
between
1.8
to
2.3
cents
per
gallon,
and
in
1989
Frontier
Refining,
a
small
refiner,
estimated
its
costs
to
produce
low­
sulfur
fuel
(
assuming
a
complete
switch
to
low­
sulfur)
at
four
cents
per
gallon.
27
SO2
Emissions
Trading
Program
When
original
cost
estimates
were
first
made,
blending
coals
was
thought
to
be
impractical.
Innovations
arose
through
experiments
and
minor
modifications
that
allowed
low­
sulfur
coal
to
be
blended
with
high­
sulfur
coal
up
to
a
40/
60
mixture,
much
higher
than
original
estimates
of
a
5/
95
mixture.
28
This
allowed
many
utilities
to
effectively
use
cheaper
low­
sulfur
coal
for
cost
savings.

With
the
low
transportation
costs
due
to
railroad
deregulation
and
the
ability
to
utilize
a
larger
area
of
the
Powder
River
Basin,
in
1998
western
low­
sulfur
coal
cost
only
$
12
per
ton
in
comparison
to
the
national
average
coal
prices
of
$
22.
Powder
River
Basin
coal
prices
were
even
lower
at
$
5
per
ton.
Fuel­
switching
and
fuel­
blending
were
thus
critical
in
reducing
compliance
costs,
and
to
date,
they
have
been
the
most
popular
compliance
option
for
utilities.
29
Other
innovations
are
expected
to
reduce
system
costs
by
another
30
percent.
For
example,
manufacturers
of
scrubber
technologies
are
making
use
of
higher
gas
velocity
absorbers
and
advanced
nozzle
arrangements.
Also,
technological
improvements
in
scrubbing
resulted
in
more
reliable
removal
than
anticipated
(
95
percent
removal
rather
than
85
percent
predicted
removal).
14
FACTOR
3:
REGULATORY
DESIGN
The
design
of
flexible
regulatory
requirements
that
are
responsive
to
the
particular
circumstances
of
compliance
can
provide
strong
incentives
for
cost­
reduction,
the
effect
of
which
may
be
difficult
to
anticipate
or
conservatively
underestimated.
The
efficiency
enhancing
qualities
of
an
emissions
trading
scheme
may
be
well
understood
for
a
static
scenario
of
plant­
level
marginal
emission
reduction
costs;
the
initial
efficiencies
of
a
trading
program
may
therefore
be
accurately
reflected
in
the
first
estimates
of
compliance
costs,
provided
there
is
good
information
on
marginal
costs.
A
trading
program,
however,
is
also
flexible
enough
to
accommodate
changes
over
time
in
the
distribution
of
marginal
costs
for
additional
pollutant
emission
reductions.
Trading
systems,
if
properly
designed,
provide
continuing
incentives
for
additional
cost­
saving
technological
innovations
and
opportunistic
responses
to
unanticipated
events.
Examples
of
unanticipated
events
that
may
affect
compliance
options
include
discovery
of
natural
resource
supplies
that
reduce
prices
of
raw
materials;
changes
in
markets
(
e.
g.,
due
to
unrelated
regulatory
actions,
international
actions
such
as
OPEC
supply
decisions,
and/
or
consumer
preferences);
and
events
such
as
weather
patterns
that
alter
resource
use
patterns.

Programs
such
as
these
also
provide
compliance
alternatives
(
i.
e.,
incentives
to
achieve
pollution
prevention
goals
that
bring
a
facility
below
regulatory
requirement
cut­
offs
and
therefore
reduce
compliance
costs).
In
addition,
some
U.
S.
programs
make
effective
use
of
technology
phasein
schedules
that
provide
lead
time
for
research
and
development,
where
warranted,
and
for
diffusion
of
newer
technologies
before
compliance
deadlines
trigger
capital
"
lock­
in".
As
Kemp
(
1997)
points
out,
however,
these
lead­
in
periods,
if
they
are
too
long,
can
also
serve
to
diffuse
incentives
for
rapid
technological
change.
Kemp
nonetheless
agrees
that
lead­
in
periods
can
contribute
to
lower
overall
costs,
because
they
prevent
commitment
to
a
potentially
long­
lived
but
more
expensive
technology
that
would
otherwise
have
been
replaced
by
newer,
less­
expensive
innovations.

In
addition,
many
CAA
rules
include
an
iterative
process
of
developing
cost
estimates
that
involves
key
stakeholders,
including
industry.
Through
the
process
of
commenting
on
and
adjusting
both
rule
provisions
and
cost
estimates
during
the
rule­
making
process,
the
rule
can
achieve
the
same
environmental
goals
at
lower
cost.
Many
cost­
saving
refinements
occur
between
rule
proposal
and
promulgation;
these
refinements
can
result
in
cost
savings
for
industry
and
in
considerable
changes
to
both
industry
and
government
estimates
to
achieve
a
specific
environmental
goal.

The
cost­
reducing
effects
of
some
of
the
previous
examples
cited,
as
well
as
in
other
instances
,
were
facilitated
by
elements
of
regulatory
design,
as
noted
below.

$
Sulfur
emissions
trading:
Trading
programs
provide
continuing
incentives
to
reduce
costs.
In
this
case,
the
program
rewarded
reductions
achieved
through
learning
by
doing,
through
induced
innovation,
and
through
a
confluence
of
good
fortune
in
low
sulfur
coal
markets.
The
"
good
fortune"
resulted
from
the
combination
of
two
unpredictable
events:
the
evolution
of
markets
and
mining
systems
that
made
it
realistic
to
recover
large
amounts
of
low­
sulfur
coal
in
the
Powder
River
Basin
in
Wyoming
and
Montana,
and
the
contemporaneous
reduction
in
transport
costs
as
a
30
Schmalensee,
Richard
et
al.
"
An
Interim
Evaluation
of
Sulfur
Dioxide
Emissions
Trading,"
Journal
of
Economic
Perspectives,
Summer
1998,
12(
3):
53­
68.

31
Coal
transportation
prices
in
the
east
were
20­
26
mills
per
ton
mile.
Competition
in
rail
for
Western
coal
led
to
prices
of
only
10­
14
mills
per
ton
mile;
furthermore,
innovations
and
improvements
in
railroads
occurred
following
deregulation,
improving
the
level
of
service
in
the
transport
of
coal
15
result
of
railroad
deregulation
under
the
Staggers
Act
of
1980.
After
railroad
deregulation,
transport
rates
decreased
by
35
percent
and
utilities
were
surprised
by
the
low
transportation
costs
that
allowed
them
to
effectively
switch
to
low­
sulfur
coal
from
the
Powder
River
Basin
in
Wyoming
and
Montana.
30
Because
transportation
costs
are
50
percent
of
the
purchase
cost
of
low
sulfur
coal
cost,
this
reduction
was
significant,
and
gave
an
cost
advantage
to
western
coal.
31
Coupled
with
the
innovation
of
fuel­
switching
and
fuel­
blending,
and
a
separate,
fortunate
shift
in
demand
toward
regions
closer
to
low­
sulfur
coal
in
the
West,
utilities
were
able
to
greatly
reduce
their
compliance
costs.

$
Phaseout
of
leaded
gasoline.
EPA's
1985
leaded
gasoline
phaseout
rule
involved
a
number
of
sophisticated
cost­
benefit
analyses
in
the
early
1980s;
the
Agency's
1985
Final
Regulatory
Impact
Analysis
for
the
lead
phaseout
rule
projected
annual
costs
to
refineries
of
$
608
million
in
1986,
falling
to
$
441
million
by
1992.
The
RIA
also
predicted
that
costs
to
refiners
could
be
offset
by
between
$
173
and
$
226
million
due
to
the
banking
provision
of
the
rule
which
allowed
refineries
making
reductions
ahead
of
schedule
additional
flexibility
in
meeting
longer­
term
targets.
Refiners
took
good
advantage
of
these
banking
provisions,
and
the
predicted
effect
was
largely
realized.

$
LEV
technology.
Delayed
implementation
periods
in
the
California
LEV
program
facilitated
reductions
in
installed
cost
for
LEV
technology
that
were
achieved
from
"
learning
by
doing"
­
see
Exhibit
1
of
this
paper,
for
example.
It
is
plausible
to
conclude
that
compliance
costs
with
immediate
implementation
very
well
might
have
been
closer
to
the
1994
projections
than
the
1998
realized
costs.

SUMMARY
AND
DIRECTIONS
FOR
FURTHER
RESEARCH
The
preceding
discussion
elaborates
on
three
factors
that
we
find
are
likely
to
contribute
to
an
overall
tendency
to
overestimate
costs
in
ex
ante
regulatory
cost
projections
for
Clean
Air
Act
rules.
A
key
remaining
question
is
whether
these
factors
can
be
quantitatively
accounted
for
in
future
cost
estimation
processes.
Exhibit
2
below
summarizes
our
conclusions
from
the
preceding
sections
and
recommendations
for
research
to
quantify
these
effects.
16
Exhibit
2
SUMMARY
OF
EFFECT
OF
KEY
FACTORS
ON
COST
ESTIMATES
Influencing
Factor
Supporting
Theory
Best
Examples
Potential
for
Quantifying
Effect
"
Learning
by
Doing"
Incentives
exist
to
minimize
costs;
firms
respond
by
growing
more
efficient
and/
or
taking
advantage
of
economies
of
scale.
­
Large
capital
expenditures
­
Goods
and
services
in
newly
expanded
or
created
markets
High.
Quantification
is
most
promising
for
items
similar
to
those
where
historical
data
are
available.

Innovation
and
Technological
Change
Firms
and
third­
parties
respond
to
new
requirements
by
reevaluating
production
methods
and
conducting
research.
Innovations
that
increase
costs
are
not
adopted.
­
Processes
or
products
at
the
edge
of
profitability
­
Cross­
industry
innovations
Low.
Difficult
to
predict
innovation;
exceptions
in
those
cases
where
some
other
baseline
trend
may
exist,
and
presence
of
regulation
accelerates
research
and
innovation.

Regulatory
Design
Increased
flexibility
in
regulatory
design
provides
facilitates
cost
saving
opportunities,
including
incentive
to
take
advantage
of
multiple
effects.
­
Emissions
trading
schemes
­
Delayed
implementation
to
facilitate
innovation
Mixed.
Gains
from
emissions
trading
and
delayed
implementation
are
often
quantified,
but
impacts
of
unanticipated
events
are
random
and
unpredictable.

Prospects
for
quantifying
these
effects
vary
widely
depending
on
case­
specific
factors,
but
a
few
general
conclusions
can
be
derived.
First,
the
potential
for
quantifying
the
effects
of
learning
by
doing
can
be
high.
For
investments
in
relatively
high­
cost
capital,
for
example,
there
is
often
a
thirdparty
vendor
market
that
is
willing
to
share
detailed
historical
cost
information
(
e.
g.,
scrubbers,
continuous
emissions
monitors,
etc.);
incentives
for
regulated
firms
to
share
such
data
may
not
be
as
apparent.
Once
data
are
obtained,
they
can
be
used
to
estimate
quantitatively
the
effect
of
process
and
product
efficiencies
over
time.
In
the
broader
perspective,
it
is
standard
practice
in
the
private
sector
to
take
account
of
learning
efficiencies
in
projecting
costs
for
internal
improvements,
based
on
historical
data
on
learning
curves.
Generalizing
from
the
existing
set
of
well­
defined
cases
outlined
above
to
the
development
of
new
goods
and
services
requires
careful
attention
to
the
details
of
the
case
at
hand,
however.

Second,
the
potential
for
predicting,
and
therefore
quantifying,
innovation
and
technological
change
is
likely
to
be
low.
While
some
additional
effort
to
innovate
can
be
linked
to
the
imposition
of
a
new
requirement,
the
outcome
of
that
effort
is
uncertain.
Nonetheless,
there
are
some
cases
17
where
a
new
regulation
may
simply
accelerate
the
pace
of
technological
change
along
a
wellestablished
path.
For
example,
efforts
in
the
last
1980'
s
and
early
1990'
s
to
control
hazardous
air
pollutant
emissions
from
ovens
used
to
produce
coke,
a
necessary
input
to
steel
produced
in
blast
furnaces,
worked
in
concert
with
other
existing
trends
in
steel
production
that
suggested
mini­
mill
technology
might
be
poised
to
replace
some
aging
blast­
furnace
steel
production.
In
cases
such
as
that
one,
reasonable
efforts
can
be
made
to
bracket
the
effect
of
future
technology
diffusion,
if
not
predict
the
innovation
itself.

Third,
methods
exist
to
evaluate
the
first­
order
cost
reductions
associated
with
regulatory
design,
but
there
is
no
way
to
anticipate
the
effects
of
good
fortune
and
innovation
that
sometimes
characterize
the
benefits
of
a
flexible
regulatory
design.
The
first­
order
effects
include
such
factors
as
allowing
plants
with
high
costs
of
compliance
to
purchase
reductions
from
firms
with
lower
costs
of
compliance
­
what
is
needed
to
quantify
these
effects
in
an
ex
ante
position
is
knowledge
of
plantspecific
marginal
cost
curves.
Flexible
regulatory
design
also
provides
continuous
incentives
to
innovate
and
take
advantage
of
unanticipated
events;
as
stated
above,
innovation
and
unrelated
events
are
difficult
or
impossible
to
predict.
Nonetheless,
more
use
of
sensitivity
tests
of
key
factors
in
cost
estimates
could
be
used
to
better
characterize
the
potential
impact
of
surprises.

Overall,
it
appears
that
additional
effort
devoted
to
data
collection
and
study
of
the
systematic
effects
of
learning
by
doing
could
improve
the
accuracy
of
cost
estimates.
Strategic
incentives
and
competitive
dynamics
within
industries
(
i.
e.,
firms
with
competitive
advantages
in
compliance
are
unlikely
to
want
to
share
it
publicly)
present
barriers
to
an
exhaustive
quantitative
analysis
of
this
effect.
Nonetheless,
the
examples
presented
here,
along
with
the
over
thirty­
years
of
experience
with
Clean
Air
Act
compliance,
suggest
that
the
database
for
quantifying
these
effects
ought
to
be
significantly
enhanced.
18
REFERENCES
Boskin,
M..
September
21,
1990.
Letter
reply
by
the
then­
Chairman
of
the
CEA,
Michael
J.
Boskin,
to
a
request
for
cost
estimates
for
various
bills
from
John
D.
Dingell,
then­
Chairman
of
the
House
Committee
on
Energy
and
Commerce.

Burtraw,
D.,
February,
1996.
"
Cost
Savings
Sans
Allowance
Trades?
Evaluating
the
SO2
Emission
Trading
Program
to
Date,"
Resources
for
the
Future.

Cook,
E.,
editor,
1996.
Ozone
Protection
in
the
United
States:
Elements
of
Success,
World
Resources
Institute.

Hahn,
R.,
1998.
"
Government
Analysis
of
the
Benefits
and
Costs
of
Regulation"
Journal
of
Economic
Perspectives,
12
(
4)
201­
210.

Hammitt,
J.,
2000.
"
Are
the
Costs
of
Proposed
Environmental
Regulations
Overestimated?
Evidence
from
the
CFC
Phaseout,"
Environmental
and
Resource
Economics,
16:
281­
301.

Harrington
W.,
R.
Morgenstern,
and
P.
Nelson,
1999.
"
On
the
Accuracy
of
Regulatory
Cost
Estimates.
White
Paper,
Resources
for
the
Future.

Harrington,
W.,
1999.
"
Predicting
the
costs
of
environmental
regulations:
how
accurate
are
regulators'
estimates?"
Environment,
September.

Hodges,
H.,
undated.
"
Falling
Prices:
Cost
of
Complying
with
Environmental
Regulations
Almost
Always
Less
than
Advertised,"
Briefing
Paper,
Economic
Policy.

Holloway,
T.,
May
20,
1996.
"
Testing
and
Monitoring
Options
and
Costs
for
MWIs,"
Memorandum
to
Rick
Copland,
cited
in
Testimony
of
the
Institute
of
Clean
Air
Companies,
Inc.
before
the
Subcommittee
on
Manufacturing
and
Competitiveness
on
EPA's
Revised
NAAQS
for
Particulate
Matter
and
Ozone,
September
24,
1997.

Horngren,
C.,
G.
Foster,
and
S.
Datar,
1994.
"
Determining
How
Costs
Behave"
in
Cost
Accounting:
A
Managerial
Emphasis,
Prentice
Hall.

Illinois
House
Special
Committee
On
Gas
Pricing,
February
1,
2001.
Final
Report.

Industrial
Economics,
Incorporated,
August
10,
1998.
"
Clean
Air
Act
Cost
and
Benefit
Research,"
unpublished
memorandum
to
Keith
Mason,
U.
S.
Environmental
Protection
Agency.

Institute
of
Clean
Air
Companies,
September
24,
1997.
Testimony
before
the
Subcommittee
on
Manufacturing
and
Competitiveness
on
EPA's
Revised
NAAQS
for
Particulate
Matter
and
Ozone.
19
Institute
of
Clean
Air
Companies,
September
1997.
Air
Emissions
Monitoring
for
Safe
and
Efficient
Medical
Waste
Incinerator
Operation.

Institute
of
Clean
Air
Companies,
May
1995.
Scrubber
Myths
&
Realities,
White
Paper.

Jaffe,
Adam
B.,
Steven
R.
Peterson,
Paul
R.
Portney,
and
Robert
N.
Stavins.
1995.
"
Environmental
Regulation
and
the
Competitiveness
of
U.
S.
Manufacturing:
What
Does
the
Evidence
Tell
Us?"
Journal
of
Economic
Literature,
33
(
March
1995):
132­
163.

Kemp,
R.,
1997.
Environmental
Policy
and
Technical
Change:
A
Comparison
of
the
Technological
Impact
of
Policy
Instruments.
Cheltenham,
U.
K.
and
Brookfield,
U.
S.;
Edward
Elgar.

Kolstad,
C.
2000.
Environmental
Economics.
New
York:
Oxford
University
Press.

Mason,
K.,
January/
February
1991.
"
The
Economic
Impact"
EPA
Journal,
45­
7.

Morgenstern,
R.,
1997.
Economic
Analyses
at
EPA,
Washington,
D.
C.:
Resources
for
the
Future.

Porter,
M.
and
C
Var
der
Linde,
1995.
"
Green
and
Competitive,"
Harvard
Business
Review,
September/
October.

Schmalensee,
R.
et
al.
1998.
"
An
Interim
Evaluation
of
Sulfur
Dioxide
Emissions
Trading,"
Journal
of
Economic
Perspectives,
12(
3):
53­
68.

Smith,
J.
and
S.
Dalton,
1995.
"
FGD
Markets
&
Business
in
an
Age
of
Retail
Wheeling"
presented
at
EPA/
EPRI/
DOE
SO
2
Symposium.

Summary
Symposium
Report,
1998.
Tom
Cackette;
California
Air
Resources
Board
(
CARB)
Case
Study
Presentation
to
EPA
Center
on
Airborne
Organics.
Available
at
http://
web.
mit.
edu/
afs/
athena.
mit.
edu/
org/
a/
airquality/
www/
rep1998.
html.

U.
S.
Environmental
Protection
Agency,
January
23,
1995.
"
Existing
Medical
Waste
Incinerators
C
Proposed
Subpart
Cc
Emission
Guidelines."

U.
S.
Environmental
Protection
Agency,
1997.
Benefits
and
Costs
of
the
Clean
Air
Act:
1990­
2010,
Report
to
Congress.

Valdmanis,
R.,
March
23,
2001,
"
Looser
Green
Rules
Won't
Stop
U.
S.
Gasoline
Spike,"
Reuters.

Viscusi,
W.
Kip
et
al.
2000.
Economics
of
Regulation
and
Antitrust.
Cambridge:
MIT
Press.
