Вы находитесь на странице: 1из 12

Allocating Vendor Risks in the Hanford Waste Cleanup

Author(s): J. M. Keisler, W. A. Buehring, P. D. McLaughlin, M. A. Robershotte and R. G.


Whitfield
Source: Interfaces, Vol. 34, No. 3 (May - Jun., 2004), pp. 180-190
Published by: INFORMS
Stable URL: http://www.jstor.org/stable/25062899
Accessed: 23-05-2015 10:27 UTC

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.

INFORMS is collaborating with JSTOR to digitize, preserve and extend access to Interfaces.

http://www.jstor.org

This content downloaded from 80.87.90.164 on Sat, 23 May 2015 10:27:58 UTC
All use subject to JSTOR Terms and Conditions

Interfaces
Vol.

34, No.

iirfJH.

3, May-June

2004,

pp.

180-190

DOI

i0.1287/inte.l040.0078

ISSN0092-21021 EissN 1526-551X1041340310180 ? 2004 INFORMS

Vendor
Waste

Allocating

in the Hanford

Risks
Cleanup
J.M. Keisler

of Management
Sciences
and
Department
100 Morrissey
of Massachusetts
Boston,
University

Information

Systems,
M/5-230,

Boulevard,

of Management,
College
Boston, Massachusetts

02125,

jeff.keisler@umb.edu

and

Decision

Information

Sciences

900, Argonne,

Building

Pacific

W. A. Buehring
Division,

National
9700
Argonne
Laboratory,
Illinois 60439-4832,
wabuehring@anl.gov

P. D. McLaughlin,
Northwest

National

South

Cass

Avenue,

Cass

Avenue,

M. A. Robershotte
Richland,

Laboratory,

{peter.mclaughlin@pnl.gov,

99352

Washington

mark.robershotte@pnl.gov}

R. G. Whitfield
and

Decision

Information

Sciences

900, Argonne,

Building

risks

allocating

as

outsourcing
could

incorrectly

increase

the request for proposals

develop

tank waste

the Hanford
vendors

a way

National
Argonne
Illinois 60439-4832,

to manage

remediation

react

would

to

proposed

costs

9700
Laboratory,
rgwhitfield@anl.gov

risk. We

risks the buyer can share or shift to vendors

determine which

how

view

may

Organizations

Division,

(RFP) for the US Department

we
their

decision-analytic

ones
and

of Energy's

In the model,
in terms
of

(TWRS).
system
risk allocations

1995

Between

dramatically.

developed

and which

used

South

1998,

we

used

this

(DOE's) privatization
an

actions

assessment

and

their

to

approach

it should bear. We

found that
to

approach

initiative

for

to
predict
considered

protocol
We

pricing.

the impact of allocating each major risk to potential vendors, to the DOE, or to both and identified the risk
to vendors plus the costs of any
the DOE's total cost?its direct payments
allocation that would minimize
residual

risks

it

have

costs

increased

Key

words:

decision
This

Allocating

because

risks

inappropriate

the vendor

paper

risks;
analysis:
was
refereed.

to the vendor

to its bids, while


would

not

the RFPs resulted in bids that were

risk allocation,

History:

accepted.

add a large risk premium

vendor would

government:

take

acceptable

have

inappropriate

allocating
adequate

would

risk-reduction

increased

costs

because

the

risks to the DOE also would


measures.

With

the

improved

to the DOE.

agencies.

of Energy
the mid
1990s, the US Department
was
its
nuclear
of
the cleanup
In (DOE)
planning
in
State.
at
site
waste
tanks
the Hanford
Washington
of
the
because
It faced well-documented
problems,

National
Northwest
(PNNL), established
Laboratory
a large privatization
team near the Hanford
site to
was
con
initiative.
to
this
Its
first
task
support
major
tract with one or more vendors
for the development

and the poor characterization


the makeup,
volume,
held in the underground
of the waste
storage tanks.
task was
The DOE's
many
legal
by
complicated
and
the
and
constraints,
political
by
obligations
from a history
of
of the problem
arising
sensitivity
at
Var
the
site.
and
disclosures
delays, disagreements,

of the cleanup.
The DOE
phase
a
issue
had
(RFP),
request for proposals
contract terms.
evaluate and select bids, and negotiate
on a 1995 draft
Comments
from potential
vendors
RFP (TWRS Privatization
that risk allo
1995) showed
a major
concern:
initia
the privatization
cation was
on the contractor's
tive depended
being able to obtain

were
about spending,
concerned
The antici
issues, and radioactivity.
in the tens of billions of dollars.
costs and risks, in 1994 the DOE
a privatization
initiative
it called
what
undertook
as
have
would
contractors
under which
(vendors)
ious

stakeholders

and

the financial market


probably would
risks. Other
risks were potentially
but would
require a high risk premium.
acceptable
at
vendors were
under no obligation
The potential
this point to tell the DOE how high the risk premium
to
and the DOE wanted
really had to be, however,

environmental
was
pated cost
To minimize

much

decision-making
The DOE,
possible.

demonstration
to draft and

financing,
not accept

as
authority and responsibility
with
the Pacific
in partnership

avoid

being

and

certain

locked

into bad

180

This content downloaded from 80.87.90.164 on Sat, 23 May 2015 10:27:58 UTC
All use subject to JSTOR Terms and Conditions

contract

terms.

et al.: Allocating
Keisler
Vendor
Interfaces 34(3), pp. 180-190, ?2004

Risks in the Hanford


INFORMS 181

Waste

Cleanup

Price

RFP

From late 1995 to early 1998, the PNNL contracted


all
with
coauthors
and Whitfield)
(Keisler, Buehring,
at that time from Argonne
National
Laboratory's
and Information
Sciences Division
(ANL) Decision
to support
the risk-management
(headed by
group
and Robershotte)
of its priva
coauthors McLaughlin
tization team. Thus, the PNNL was
the direct client
to
it
of the RFP
for this work,
write
portions
using
that the DOE ultimately
issued (TWRS Privatization
the basic structure for this work
1996). ANL proposed
most
and conducted
and analy
of the assessments
the PNNL advised on the approach, partic
sis, while

Mitigation

Exogenous Amelioration

(DOE)(Vendor)

(DOE)(Vendor)

Mai

interactions with
ipated in the assessments,
managed
the privatization
the results with
team, and integrated
a
the rest of the privatization
We
developed
plan.
to
the
deter
DOE
decision-analytic
approach
help
mine how it should allocate risks within
the RFP, that
is, to what extent should the DOE accept risks and to
what extent should it pass risks on to the vendor?
and assessed
the
We defined
alternative
allocations
risks resulting from them to quantify
their cost impli
for the different
incentives
cations, while
accounting
and risk-bearing
and
the
DOE
its poten
of
capacity
a
to
tial vendors,
RFP
and
successful
further
leading
our
efforts to institutionalize
approach.
in
The problem of how to allocate risks is prevalent
all sorts of contracting
and using RFPs that
situations,
can cause inefficient
outcomes.
allocate
risks poorly
in
Our conceptual
of those
many
approach
applies
situations.

Conceptual

Approach

a
Allocating
it motivates
age that risk.
the other will

risk to the party better able to control


to man
that party to find the best way
about how
Each party has expectations
behave given the incentives
implied by
to
and those expectations
the risk allocation,
ought
It
is
also
reasonable
inform contractual
negotiations.
to think that the larger party (the government
in this
to
variances
deal
with
better
would
be
able
case)
large
in financial
outcomes
than would
smaller parties
This implies
that, other things remaining
(vendors).
a larger pre
contractors
the
would
demand
equal,

to bear risks than would


mium
To
the government.
must
the decision
allocate
risks efficiently,
makers
take note of all these factors and balance
them.
this as a decision-analysis
Viewing
problem, with
to choose
the DOE as the decision maker, we wanted
maximize
the DOE's
that would
It
uncertain
faced
environmental
utility.
expected
ties and uncertainties
would
about how
vendors
to
the
game
theory,
Inspired by
respond
proposal.
or at least the asymmetric
prescriptive-descriptive
(1982) outlined, we built a prescrip
approach Raiffa
for the DOE's
and a descrip
tive model
decisions
for the vendor's
tive model
decisions.
From the DOE's
the risk allocation

Vendor,

l
Figure 1: Stylized decision trees were used to represent the cash flows
resulting froma contract.For the DOE,the firstnode is the DOE'sdecision
about how to allocate risks in the RFP.The second node is the potential
vendor's decision (an uncertaintyfromthe DOE'sperspective) aboutwhat
price to bid (to be predicted using a vendor-responsemodel). The third
node represents the DOE'sdecisions taken to mitigate risks. The fourth
node consists of the vendor's decisions taken to mitigate risks (inde
pendent of the previous node). Following this are uncertain exogenous
events, afterwhich the DOEacts to ameliorate the outcomes for the events
that have occurred, afterwhich the vendor does the same (independent
of the previous node). The DOE'send-pointvalues are determined by the
contractualpayments itmakes to the vendor alongwith the costs incurred
due to the risks itbears and the events that occur. For the vendor, the first
node is an uncertaintyabout how the DOEdecided to issue the RFP.The
second node is the vendor's decision aboutwhat price to bid. The third
node is a decision node about itssteps tomitigate risks. The fourthnode
is an uncertaintyabout the DOE'ssteps tomitigate risks (independentof
the vendor's

choices

tain exogenous

about

events.

The

the same), and the fifth node consists


fifth node consists
of the vendor's

of uncer
decisions

to ameliorate negative outcomes, and the sixth node is the DOE'sactions


(aboutwhich the vendor is uncertain) to ameliorate risks (again indepen
dent of the previous node).

as a styl
could be viewed
point of view, the problem
tree with nodes
ized decision
its deci
representing
sions and nodes representing
uncertainties
(Figure 1).
From
the vendor's
of view,
the DOE has
point
chosen
its risk allocation. We simplified
the
already
tree and assumed
vendor's
decision
that the vendor's
its chance of getting
the contract,
goal is to maximize
that is, to bid as low a price as possible,
subject to
the constraint

that it still be possible


to finance the
on
That
it
the risks it faces to its
is,
passes
project.
and
financiers
it is these financiers'
risk attitudes
that
we modeled.
The financiers consist of equity and debt providers.
consider
the predicted
variance
Equity
providers
an expected
of cash flows
and demand
return on
look at the
equity based on that risk. Debt providers
that they will be repaid without
likelihood
problems

This content downloaded from 80.87.90.164 on Sat, 23 May 2015 10:27:58 UTC
All use subject to JSTOR Terms and Conditions

Keisler

not only that an interest rate be based on


of risk but also that the project have a
on the
that depends
of equity investment
percentage
of cash flows
level of risk. The prediction
(that will
and the amount of
be used to make debt payments
is
residual cash flow left to give a return on equity)
fixed
and
based on the vendor's
variable
anticipated
costs, the contracted
treated, and the
price for waste
as
in the RFP.
of
risks
allocated
consequences
possible
and demand
that measure

Assessment

of Individual Risks

a list of 100 risks. We


We
started with
combined
some of the risks and narrowed
the list substantially.
the potential
risks for which
The final list contained
was
the
of
allocations
sizable,
range
impact
possible
was substantial,
and the right allocation was not obvi
ous. We examined
nine major
risks, some requiring
of subsidiary
risks:
separate assessment
rate
rates
Interest
(interest
may change between
(1)
the time of the bid and financing),
to inflation
faces exposure
(2) Inflation (the vendor
risk in the draft RFP),
in law (applicable
laws and regulations
(3) Change
may

change),
may
(the vendor
(4) Permitting
obtain necessary
permits),
C (probably
(5) Waste
envelope
waste
to
difficult
process),
cally

uncontrollable

Vendor Risks in the Hanford Waste Cleanup


INFORMS
Interfaces 34(3), pp. 180-190, ?2004

did our assessments


for a representative
poten
to be pre
tial vendor. We
the definitions
intended
to specific
to translate without
cise enough
ambiguity
within
the RFP. When
the DOE
later used
wording
we conducted
to evaluate
dif
the model
specific bids,
assessments
different
vendors.
for
ferent
outcomes.
After
We
then defined
the possible
verbal
of
the
steps
mitigation
obtaining
descriptions
risk
each party would
likely take under each potential
we
outcome
of
assessed
each
allocation,
probabilities
on the likely mitigating
conditioned
steps
(implicitly
taken by both parties). For this step, we tried to take
as well
as the DOE's
the vendor's
view. We
view
same
that vendors would
the
assumed
assign
prob

we

as would
to each outcome
abilities
the DOE, unless
reasons
not
to
there were
the
(for example,
special
a
cer
DOE might
be far more
trustworthy
regarding
or the vendor
tain risk than the vendor
believed,
more
be
about
risks and
different
concerned
might
events than would
the DOE). We also obtained verbal
be likely
of the steps each party would
descriptions
outcomes
to take to ameliorate
of risks
the negative
under

each allocation.
Finally, we obtained

of what
the
verbal descriptions
out
each
be
for
each
side
for
consequences
come of the risk, assuming
both sides had taken the
we assessed
this
From
information,
steps.
predicted
over the
or
estimates
distributions
point
probability
conse
side.
financial
each
The
for
consequences
were
in terms of the net present
defined
quences
at the current
value
(NPV) of cash flows
computed
as increments
from a
risk-free
rate, and expressed
diffi
base case in which
the project proceeds without
that outcomes
of one
culty. In some cases, we noted
would

not

be

the most

able

to

techni

(6) Appropriation
(Congress may not appropriate
for planned
funds adequate
activities),
and decommissioning
(D and
(7) Decontamination
D may be more costly than expected),
(8) Other

et al.: Allocating
182

circumstances

(lawsuit,

tornado), and
earthquake,
sabotage,
risk (the contract
(9) Not-to-exceed
(NTE) price
could be a literal price limit, as
pricing arrangement
in the draft RFP, or could be a target price that allows
selected adjustments).
a background
risk level for the
We also assessed
risk each
that is, the residual
DOE and the vendor,
bear even if all nine risks resolved to their base
would

risk would
determine
the relevance of other risks; for
if the project stopped early, later risks would
example,
not affect cash flows.
The following
of one such assessment,
summary
assessment
D
and D risk, illustrates the
the
for
namely
alternate
and the
of
the
allocations
process
defining
that we

cases.

discussions
We
each risk through
characterized
with
the PNNL privatization
staff experts. We used a
common
the assessments.
For each
form to organize
we
the risk
three
allocations:
alternative
defined
risk,
was
to
to be borne primarily
be
the
shared,
DOE,
by
con
or to be borne primarily
vendor.
We
the
by
both
and
these allocations
structed
using
qualitative
to
alterna
definitions
match
reasonable
quantitative
include in the RFP. This
the DOE might
tive phrasings
even
that
for
for risks allocated pri
meant,
example,
to
the
the
DOE
could still bear some
vendor,
marily
risk does not neces
risk.
shared
Similarly,
tangible
sitate a completely
arrange
symmetric
risk-bearing
so
ment.
The specific vendor was not yet known,

scenarios

risk-management

eters

approximately
and Buehring

and

the

resulting

param

used in the model


(Table 1). It took us
one half day to assess this risk. Keisler
of all the risk
(1996) give the details

assessments.

Results

of Interviews

to Characterize

One

of the

Risks (D and D)
Risk Name:

Decontamination

and

decommission

ing (D and D).


than
D and D may
be more
Description:
costly
not
is
if
the
vendor
provided
expected,
particularly
an incentive,
of risk to the
allocation
with
through
to keep these costs low.
vendor,
in terms
The consequence
is given
Consequences:
over D and D costs,
distribution
of the probability
starts.
to occur 10 years after production
assumed

This content downloaded from 80.87.90.164 on Sat, 23 May 2015 10:27:58 UTC
All use subject to JSTOR Terms and Conditions

et al.: Allocating
Keisler
Vendor
Interfaces 34(3), pp. 180-190, ?2004

Risks in the Hanford


INFORMS

DOEtakes risk

Risk allocation

DOE is responsible

Description

Waste

for D and D.

$5M:10%
-$50 M :40%
-$150M:50%

Probability
distributionfor

No cost

183

Shared risk
Vendor

sets aside

Vendor

takes risk

Vendor is responsibleforD and D; DOEpays

limited funds,

which may be refundedifD and D


costs

Probability
distributionfor
cost to DOE

Cleanup

excess

if vendor

reaches

the

hypotheticallimitof itsabilityto pay.


0:75%
-$50 M :25%

are less than expected.

$2 M:30%
-$40 M :60%
-$140 M: 10%
$3 M: 30%
-$10 M: 70%

$5 M: 50%
-$50

M.-25%

-$100 M: 25%

cost to vendor

Table 1:We summarized the risk-assessment interviews inthe formof quantitativedescriptions of the risksused
as inputsto the simulationmodel. The description for the D and D riskshown here is typical.

Low cost is approximately


less net D and
$5 million
cost than expected. Medium
cost is approximately
more
net D and D cost than expected.
$50 million
cost
is
$150 million more net D
High
approximately
D

and D cost

than expected.

Risk Allocation
Characteristics)
(Representative
The DOE bears the risk: The DOE states up front that
it is responsible,
and the vendor
allocates no money
for D and D.
is defined,
and the ven
Shared risk: Responsibility
a
dor establishes
medium-sized
sinking fund and gets
D
if
is
back
D
and
under
the DOE
money
budget;
covers

overage.

The vendor bears the risk: The vendor pays for D


a
and D and must
establish
large sinking fund from
back.
the start; it gets the unused money
Mitigation
The DOE
standards

and Prevention

Strategies
strict
bears the risk: The DOE establishes
in planning
the vendor must
and
follow
and have close cooperation
with
environ

operation
mental
regulators.
some
Shared risk: The DOE monitors
the vendor
in planning;
the vendor
is responsible
what
for oper
ation with
and
The DOE,
the vendor,
checkpoints.
some
of
the regulators
coordinate
these
together
activities.
The vendor
the risk: The DOE
bears
intervenes
in
extreme
all
actions
the
vendor's
take
cases;
only
into account
their impact on future D and D costs,
and the vendor works hard to establish
rapport with
regulators.

Likelihood
of Occurrence
cost = 10 percent,
The DOE
the risk: Low
bears
cost = 40 percent, high cost = 50 percent.
medium
cost =
Shared risk: Low cost = 30 percent, medium
=
60 percent, high cost
10 percent.
The vendor bears the risk: Low cost = 25 percent,
cost = 50 percent, high cost = 25 percent.
medium
Costs are lowest when
the risk is shared and high
est when
the DOE bears the risk, because
both the

DOE
have

and
the

can
the vendor
to
incentive
do

influence
so under

outcomes
and
a shared-risk

scenario.

Modeling
We

risk-allocation
defined
four main
for
strategies
further analysis
Each
consisted
of
(Table 2).
strategy
nine
allocations
All
each
of
risks.
for
the
of
specific
the realm of possibility.
these strategies were within
Prior to engaging us to work on the problem,
the DOE
was heading
toward something
close to the vendor
bears

strategy
(Strategy 3). This strategy is quite sim
ilar to the 1995 draft RFP, and even after we started
a senior DOE official declared
that the DOE
work,
ven
would
"shift all risks to the vendor."
Potential
turn
in
had
dors
indicated
that just about every risk
should be borne by the DOE
(Strategy 1), although,
the DOE was skeptical of such a position.
predictably,
As we analyzed
individual
risks, the analysis
(and
other factors) influenced
the RFP. By the time we fin
ished analyzing
the various
strategies,
Strategy 2 had

Strategy 1 2 3
DOE
Name
Risk name
Interestrate
Inflation
Changes in law
Permitting
Waste streamC
Appropriation
DandD/RCRA
Sabotage
Earthquake
Tornado
Not to exceed

DOE
DOE
DOE
DOE
DOE
DOE
DOE
Lawsuit
DOE
DOE
DOE
DOE
DOE

Shared
Vendor
Vendor
DOE
Vendor
Vendor
DOE
Shared
Vendor
Vendor
Vendor
Vendor
Vendor
Shared Vendor
Shared
DOE
Vendor
DOE
Vendor
DOE
Vendor
Shared
Vendor

Vendor

Vendor

Table 2: Foreach risk-allocationstrategy,we definedhow each riskwould


be allocated. Sharing risks did notmean that each risk is shared: some
risks are shared, but other risks could still be allocated entirely to one
party or the other.

This content downloaded from 80.87.90.164 on Sat, 23 May 2015 10:27:58 UTC
All use subject to JSTOR Terms and Conditions

Keisler

Strategy
Vendormean cost increment
Vendor standarddeviation
DOEmean cost increment
DOEstandarddeviation

1 ($)

2 ($)

157
55,876
336,731
133,409

32,660
61,780
107,625
82,517

3 ($)
106,152
94,655
60,400
70,675

Table 3:We generated key summary statistics from the risk simulations
(figures inthousands).These results described the impactof risksharing,
butwe still needed to incorporatethemain cost, direct payments fromthe
DOEto the vendor.

at some levels but this support was neither


request we also
complete nor final. At management's
differed only
included Strategy 4 (not shown), which
2
the DOE
from
and
under
which
slightly
Strategy
fared slightly worse.
We used a Monte Carlo simulation with
1,000 iter
support

to produce
for each risk. For each
profiles
we
the
increments
from the base
summed
iteration,
case due to each risk. The result, after accounting
for
a frequency
was
over
distribution
risk interactions,

ations

costs in the
the total increments
(in which
higher
to reduced profits) to the base case,
chart correspond
for both the vendor and the DOE,
including payments
for each of the strategies
considered
(Table 3).
As a measure
of the reliability
of the simulation
we used the stan
results under repeated simulations,
of the average cost for a given strategy,
dard deviation
was
which
$2 million.
approximately
Vendor

Response
the price the vendor would
then had to predict
under
each
strategy, which we could combine
charge
with
the DOE's risk profile to obtain a net-cost distri
bution for the DOE.
conversations
with
the privati
Based on extensive
zation
financial
task leader for the TWRS and with
we developed
external
financial
consultants,
equa
set a price in
tions to predict how the vendor would
to any given set of risks and risk-allocation
response
these equations using experts'
decisions. We validated
specific changes to the risk pro
predictions
regarding
estimates were
file for which
fairly easy
subjective
that
to provide. We then had other experts confirm
realistic. Our
the parameter
values we used were
was
to codify
rather than to
purpose
judgments

We

firm
describe exactly and in detail what any particular
would
do.
in the financial-risk
used
One
variable
important
for which we
the probability
of default,
model was
as a proxy the probability
that the vendor would
stake.
loss larger than its equity
after-tax
? i<
we
h
that
the
calculated
probability
Specifically,
-(j + k), where
h is the baseline NPV
of vendor
cash flows dis
counted at the risk-free rate,

used
face

an

et al.: Allocating
184

Vendor Risks in the Hanford Waste Cleanup


INFORMS
Interfaces 34(3), pp. 180-190, ?2004

from
i is the net impact on vendor
cash flows
at the
all risks, also in terms of dollars discounted
the probability
dis
risk-free rate (we approximated
to the
tribution on i by fitting a normal distribution
simulation

results),
; is the amount of equity the vendor puts up, and
tax benefit
k is the relative
in
of reduced
profits
cases compared
to the baseline.
unfavorable
We determined
cash flows by entering
the pro
a
terms
into
and
financial
price
posed
financing
on
stan
and
based
Paananen
model
1995)
(Weimar
and
the
dard accounting
pri
practice
developed
by
to predict
the DOE's
vatization
finance task manager
costs in a deterministic
and the vendor's
base case
and to clarify the literal implications
of contract terms
(Figure 2).
We assumed
that the vendor would
raise prices
to being
to take on additional
in response
forced
that are pre
risk in the following
ways:
(1) Risks
would
downside
decrease
the expected
dominantly
the
vendor
raise
would
the
return;
price to bring the
to
return
baseline
back
its
value.
(2) Risks
expected
it harder
that increase the probability
of default make
so we assume
to obtain debt financing,
the interest
rate for debt financing
to the degree
of
corresponds
risk and the percentage
of equity
we
assume would
increase

which

risk increases.

required, both of
when
downside
are concerned with

(3) Equity providers


on both the
and the downside,
predictability
upside
so they would
return on equity
raise the required
the overall risk level increases.
when
The risk profiles
and the vendor-response
mod
rate
the vendor's
internal
els determine
of
required
return (IRR); we must
finan
link this IRR to the main
a target price for the vendor
to determine
cial model
over cash flows.
and the corresponding
distribution
in turn determine
These cash flows
debt terms (frac
tion of debt and debt interest rate) and costs, from
we
which
the probability
of default.
The
calculate
an
below
define
equations
upward
implicitly
slop
ing curve with
capital costs on the y-axis and prob
on the x-axis. For given
ability of default
financing
terms, there is also an implicit curve for price versus
is downward
of default, which
probability
sloping
a cushion
because
the
price gives the vendor
raising
that lowers the probability
of default. We then use a
to identify the min
modified
search
binary
algorithm
imum price at which
the vendor
could still finance
that
the project, subject to the following
constraints:
the IRR is no less than that given by Equation
(1),
that the debt fraction is no greater than that given by
(2) and (3), and that the debt interest rate is
Equations
no less than that given by Equation
(4). The solution
the curve representing
lies where
price versus proba
curve representing
the
of
intersects
the
default
bility
versus
cost
financier's
of
default.
capital
probability

This content downloaded from 80.87.90.164 on Sat, 23 May 2015 10:27:58 UTC
All use subject to JSTOR Terms and Conditions

et al.: Allocating
Vendor
Keisler
Interfaces 34(3), pp. 180-190, ?2004

Risks in the Hanford


INFORMS

Simulation

Strategy

output

Waste

Cleanup

Summary
statistics

185
Financing
terms

Pro forma
financi?is

Objectives

DOE RFP
risk
allocation

Figure2: This stylized influencediagram represents the financial relationships in the model. The model pre
dicts price from risk allocation and other assumptions. The influencediagram has two features thatmake it
unconventional?the circular relationship involvingprice and probabilityof default and the fact thatmean, vari
ance, and probabilityof default are summarystatistics derived from the probabilitydistributionover the vendor
cost increment(derived fromend points of the decision tree inFigure 1).

In other words,
the project's
the solution
lies where
pro forma NPV including debt costs is exactly 0 when
at the predicted
IRR requirement.
discounted
to Represent
Used
Vendor
Financial
Equations
to
Risks
Response
Programmatic
a few key consid
After obtaining
advice concerning
the following
erations, we postulated
relationships.
The financial experts based their estimates
of param
eter values on their knowledge
of other debt offerings,
in which
situations
debt
(which
including
ratings
to default
have a known
relation
rates) and corre
terms were available.
sponding
The required after-tax
internal rate of return (IRR)
is given by
IRR = k1+k2x

A + fc3xB,

where

(1)

=
kx 0.03 (the risk-free rate),
= 0.0275
k2
(the increase in IRR required to accom
and
modate
of average downside),
$10 million
= 0.025
to
rate
variance
interest
of
(the
k3
sensitivity
in NPV of cash flows).
In this equation
and in the ones that follow,
A denotes
in the vendor
cost distribu
the variance
in dollars2 x 109), and
tion (expressed
B denotes
cost increment
the vendor mean
(in
dollars xlO5) from the simulation
results.

If the probability
the maximum

cent,

of default
(P) is less than five per
debt fraction (DF) is given by

DF = fc4- k5 x P - 0.03 x (A/k6)m,


fc4

0.95

(the maximum
case with
zero

debt

where
fraction

(2)
for

of
percent
hypothetical
probability
default),
= 1.5
of the debt fraction to the
(the sensitivity
k5
of default),
probability
= 3.23
to the mini
(the variance
k6
corresponding
mum
vendor
risk case, including
the assumed
level
of background
risk), and
= 3.0
of the debt
(the exponential
k10
sensitivity
to variance). We used
fraction
the last part of the
in particular
to calibrate predicted
financ
equation
were
aware
We
ing terms for experts'
judgments.
that it lacked a further theoretical
basis, but given
our time constraints, we used it, albeit
In
cautiously.
we
later applications,
this
element
of
the
implemented
model more elegantly.
If the probability
is greater than or equal
of default
to five percent,
the debt fraction is given by
DF = fc4- k5 x P - 0.03 x (A/ke)m
-fc7x(P-T),

This content downloaded from 80.87.90.164 on Sat, 23 May 2015 10:27:58 UTC
All use subject to JSTOR Terms and Conditions

where

(3)

Keisler

= 16
of debt fraction
to
k7
(marginal
sensitivity
rate for risky projects),
default
and
T = 0.05 (the probability
of default
corresponding
to financiers'
threshold of great concern). This thresh
to a downgrading
old is analogous
of the debt on the
as the
project, which we assessed
point
subjectively
at which
as
the project would
from
business
change
into what
The minimum

usual

given

financiers
required

would
debt

view
interest

as risky.
rate (DR)

by
DR = M + k8 x P? Y + k9 x A/k6,

is

= 0.0737

(the T-bond

yield

at the

where
time

(4)
of

the

study),
y =

to be near
the maximum
0.07, considered
allowable
of
default,
probability
P = the probability
of default
(computed
by the
model),
= 0.138
of the sensitivity
of the
(the coefficient
k8
debt interest rate to the probability
of default),
and
=
of the debt interest rate to
k9 0.015 (the sensitivity
in cash flows).
the variance
Where
the financial
advisors
identified
possible,
in
the
for
market;
analogous
financing
packages
a
a
with
bonds
have
histori
example,
given
rating
cal annual default
rate and a known
risk premium
over T-bonds.
a
bond
By finding
rating correspond
= 0.07 and
a linear
ing to y
assuming
relationship
between
rate, one can estimate
yield and default
k8.
some parameters
We derived
directly from such mar
some using
ket data, estimated
subjective
judgment
informed
such
and estimated
such
data,
others,
by
as k10,
what
about
by obtaining
expert
judgments
terms ought
to be for a small number
of
financing
artificial
scenarios
and
the
values
specified
fitting
to mimic
those expert judgments. We compared
the
a
of
the
for
wide
of
cases,
model,
range
implications

et al.: Allocating
186

Vendor Risks in the Hanford Waste Cleanup


Interfaces 34(3), pp. 180-190, ?2004
INFORMS

with

the intuitions
and expectations
of the financial
and of the privatization
team management.
experts
We concluded
that the financing
terms and the cor
case were
reason
for
baseline
the
price
responding
able. The model's
predicted
impact of risk allocations
on
same order as the finan
terms
is
of
the
financing
cial advisors'
estimates
of those impacts
for several
were
reference
that
to
risks
particular
fairly easy
these parameters
would
judge. Our experts
thought
over
reasonable
results
the
of
range
provide
strategies
we examined
for considering
the
and, in particular,
incremental
of
to
the
risk-allocation
impact
changes
strategy.

Numerical

Results

our model, we obtained vendor


finan
By optimizing
terms
cial
to
the DOE's payment
(Table 4). By adding
to the DOE's
the vendor
simulation
statis
summary
obtained
the financial measures
of direct
tics, we
interest to the DOE (Table 4).
over vendor
The distributions
cost (Figure 3) and
net costs for the entire contract period
the DOE's
some expected
show
(Figure 4) for each strategy
such as the low risk to the vendor
characteristics,
and the wider
the DOE
spread for the DOE when
bears the risks. The vendor's
risk distribution
showed
much
loss
than
for
for
greater potential
gain, which
in
initial
the
of the draft
discussions
explained why,
vendors
indicated
they would
require higher
than
the
DOE
had expected.
risks
Most
payments
were
to be assigned
to one party or the other; how
case
reduced
the
ever, the middle
(RFP) actually
DOE's
total downside
risk. Some of the distributions
to cases in which
had spikes near $0, corresponding
the vendor would
be made whole
(that is, the DOE
would
reimburse
the vendor
for all expenses
includ
after
the
ended.
costs)
ing financing
project
RFP,

Strategy
Vendor

financial measures

Probabilityof default (P)


Required internalrateof return(IRR)
Debt fraction(%) (DF)
Debt interestrate (DR)
Price paid to the vendor by DOE
DOEpaymentto vendor ($K)

3.4%

5.17%

8.23%

11.6%

14.3%

30.3%

86.6%

82.7%

30.6%

9.3%

9.8%

11.6%

$602,255

$636,530

$1,397,921

$336,731
$133,409
$938,986

$107,625
$82,517
$744,155

$60,400
$70,675
$1,458,321

DOEfinancialmeasures
DOE mean

cost

increment

DOEstandarddeviation
DOEtotalcost

Table 4: Key financialmeasures of interestto the vendorwere trackedforeach strategy.When the vendor bears
all risks (Strategy3), the difficultyof financingthe project results in the DOEhaving tomake much higher pay
ments to the vendor.Although the DOEcouldminimize its direct payments to the vendor by accepting all risks,
thiswould usually lead to higher total costs.

This content downloaded from 80.87.90.164 on Sat, 23 May 2015 10:27:58 UTC
All use subject to JSTOR Terms and Conditions

et al.: Allocating
Vendor
Keisler
Interfaces 34(3), pp. 180-190, ?2004

Waste

Risks in the Hanford


INFORMS 187

Cleanup

Strategy 2. By bearing all risk, the DOE


the vendor
the lowest price, because
obtain
cost increment
and vari
have a lower mean
would
ance and thus the best financing
terms. By sharing
est costs with

would

Frequency
in simulation
results

to the vendor
risks, the DOE would make payments
than
that would
be six percent
($34 million)
higher
to
but
the
the
DOE
those under Strategy
1,
savings
lead to a net cost
from impacts of residual risks would
over
20
for
the
DOE
of
($194 million).
percent
savings

-$200M

Frequency
in simulation

in which
to the strategy
the vendor bears
Compared
2 requires the DOE to
all risks (Strategy 3), Strategy
risks that are 78 percent
pay costs due to residual
to the
but
the
total cost savings
($47 million)
higher
it
because
DOE are nearly 50 percent
($714 million)

results
-$200M

results
-$200M

$200M

Vendor Cost Increment

Figure3: The vendor's distributionof cost increments (the deviation from


contracted payments in the base case) depends on the risk allocation.
Under
over

to the vendor. The


pay a smaller risk premium
the project infeasible.
price for Strategy 3 would make
In spite of the cost difference,
the standard deviation
2 was only slightly
of the DOE costs under Strategy

would

Frequency
in simulation

the shared-risk
its potential

costs,

the vendor
strategy,
but its mean costs

has somewhat
do not increase

greater
much.

spread

and
for
In analyzing
the results
the strategies
we
the
risks
demonstrated
each,
that, indeed, sharing
all the risks to the
would
lead to savings. Allocating
not do enough
the vendor would
DOE would mean
all the risks to the
to keep costs low, while
allocating
cause the vendor
to demand
too high
vendor would
a risk premium;
the low
the
DOE
would
obtain
thus,

Frequency
in simulation

than under Strategy


3. Strategy
3, the costli
higher
est approach, was the DOE's preferred
strategy going
into this effort, but the figure given, $714 million,
per
overstates
the
under
potential
savings accruing
haps
2: if bids came in that high, the DOE could
Strategy
a revised RFP.
have issued, with difficulty,
Our presentation
of these results triggered an ani
At that time, the DOE managers
mated
discussion.
from an RFP that
switched
their support
essentially
to one that
allocated most of the risks to the vendor
RFP by
then refined
the proposed
shared risks. We
a one-way
allo
for
sensitivity
performing
analysis
2 as a basis. We
cation of each risk, using Strategy
the allocation
for each risk that minimized
identified
the DOE's cost when other risks were
left unchanged.
From

these

which

resembled

the

vendor

DOENet Cost

Frequency
in simulation

Risk
Shared

the worst

results

Frequency
in simulation

$1.5B

DOENetCost

Vendor Bears Risk

$0.5B

U
$1B

DOENet Cost
$L5B

we generated
for each risk-allocation
strategy
Figure 4: The histograms
dominates
the two extreme risk allo
show that sharing risk stochastically
cost and its risk.
cations and so reduces both the DOE's expected

changes-in-law,
In
risks.
partic

all

risks

and

such risks into their price.


incorporate
we
added
6, in which
comparison,
Strategy
are

shared,

as

well

as

many

other

varia

that we

at the DOE's
tions (not shown)
considered
the DOE
and the vendor
share all
request. When
cost increment
to the DOE
is higher
risks, the mean
to the vendor,
than in Strategy
2, as is the payment
so Strategy
worse
6 is $24 million
than Strat
5 had a mean
cost increment
to the
egy 2. Strategy
than Strategy 2, in exchange
DOE of $5 million more
in payments
to the ven
for a $15 million
reduction
of
million
$10
dor, for an additional
savings
along
a slight reduction
in the standard deviation
with
of
$2 million.
and

results

rate,

appropriation

to the vendors
risks
ular, by completely
allocating
that they believed were under the DOE's
control, for
D
and
and
the
D,
appropriation,
permitting,
example,
assume
increase its cost: vendors would
DOE would
For

$0.5B

interest

the
and

waste-stream-C,

constructed
5,
Strategy
2, except that the DOE and

Strategy

shared

results

$0.5B

we

allocations

This content downloaded from 80.87.90.164 on Sat, 23 May 2015 10:27:58 UTC
All use subject to JSTOR Terms and Conditions

et al.: Allocating
188

Keisler

the allocation
of the other risks, the
By changing
DOE would
obtain
cost
relatively minor
potential
increases or cost savings. This implied that if the ven
dors seemed very concerned
about risks for which
the model
little increase in costs (for exam
predicted
an additional
$50 million
ple, if the vendor demanded
to accept waste-stream-C
then
DOE
the
could
risk),
on those risks (for
make
concessions
accept
example,
the risk rather than pay the premium).
Such risks
or
have
value
for
vendor
the ven
the
might
symbolic
dor might have a more pessimistic
view of the situa
tion than it merited.
if the vendors were
Conversely,
more
a
to
risk
than
the model
pre
accept
willing
to
the
DOE
could
obtain
concessions
from
dicted,
try
to reduce its total cost. The entire portfo
the vendor
as a unit.
lio of risks matters
and must be considered
For example,
of
the
impact
although
shifting the NTE
risk entirely to the vendor
from Strategy 2 (in which
it
is shared) is $41 million,
the cost of shifting the NTE
risk to the vendor when
the vendor
already bears all
be much
other risks would
In
larger ($198 million).
can bear a little risk comfort
the vendors
other words,
a certain threshold, vendors will not
ably, but beyond
assume
risk without
additional
the DOE
penalizing
dramatically.
The final RFP was

very close to our recommenda


tion of Strategy 5. It had some refinements
in
based
on
our
further
of
model.
The
DOE
part
application
received
several bids and selected
two vendors
for
the technology-development
cost
the
stage. Although
as
included
financing charges that the DOE perceived
came
in
the
bids
than
lower
the
DOE's
high,
previ
ous cost estimates
and lower than bids that it had
of risk in the RFP.
expected prior to the reallocation

Decision-Support

System

the RFP, we
it
anticipated
bidders'
and
suggestions
the bid-selection
and negoti
requests quickly during
some parts of the model
ation phases. We improved
to speed the DOE's
and to take advantage
response
of insights
the
then created
way. We
gained
along
a
rather
than a one
(DSS)
system
decision-support

When
would

the DOE
issued
to evaluate
need

off model.

used

neater risk templates


and assess large models.

to characterize
basic templates
common
to many procurement

to make
it easier to
We developed
four
risks that would
be

contracts,
specifically
and ongo
termination,
of
the
contracted
stage
one type of risk tem

permitting,
processing,
early
risks. For each
ing financial
project, users must
complete
for several risks of that type. The DOE
plate, possibly
to successfully
would
have
resolve
risks from each
move
on
to
to
next
the
This
fact led to
stage
stage.
the next feature. We modeled
risk interactions
using
a
a cell entry of 1 or 0
in which
matrix,
precedence
indicates whether
the DOE must
resolve
successfully
the risk in the corresponding
column before going on
row. This
the risk in the corresponding
it easy to add individual
risks without
and allowed us to
the rest of the simulation
in a more compact and compre
interactions

to encounter
matrix

made

affecting
represent
hensive

way.

We refined the vendor-response


function and made
on
it more
with
based
further
transparent
meetings
the project's
external
advisors.
The simulation
(and
in the DSS
noted
templates)
specifically
potential
severe write
events
default
and, separately, more
in which
In
lenders would
lose principal.
concerns
the original model,
financier
about write
in the exponential
off events were
reflected
penalty
function we used
in which
for situations
the prob
above
five percent. Consider
ability of default was
risks explicitly
is more precise,
and by
ing write-off
for high
doing so we replaced the exponential
penalty
default probability with
linear functions of the default
and write-off
The DSS
records
future
probabilities.
as parts of the
events
events
and default
write-off
outcomes
of individual
risks as it simulates
them.
off events

not directly
the debt fraction but
calculate
it from a debt-service-coverage-ratio
derives
found in practice.
This structure
facili
requirement
was
more
tated assessments
which
(with
everyone
of the judgmental model
comfortable)
relating project
to financing
statistics
terms. We ended by
summary
and automating
the optimization
sub
streamlining
routine for finding the vendor's minimum
price.
It does
instead

The

answered
system
negotia
resulting
an hour, instead
in
half
about
questions
a day or more
the original model
could
on
the
of
the
take, depending
complexity
question. To
some of the improvements,
we relied on stan
make
to make models
dard methods
faster, more flexible,
In particular, we restructured
and more user-friendly.
to use modular
the model
and automated
components
tors' what-if
of the half

most
tasks. We also made
improvements
conceptual
more
in
that made
the model
than it was
efficient
our earlier efforts; we recommend
such improvements
as a starting point
for similar efforts others might
undertake.

We

structure

Vendor Risks in the Hanford Waste Cleanup


Interfaces 34(3), pp. 180-190, ?2004
INFORMS

Organizational

Impact

The insights
from our analysis
within
team
the privatization
The DOE managers
presented
of
mary
diagram
(Department

produced
and within

consensus

the DOE.
sum
the conceptual
Privatization
Energy
1998, ?5, Figure 4.6) on risk shar
Group
Working
of energy, Hazel O'Leary.
ing to the then secretary
to provide
One purpose
of our effort was
the DOE
with
evidence
about
the potential
of
consequences
that
vendors
alone
should
bear
the
the
risk,
insisting
and we
that doing
this would
showed
successfully

This content downloaded from 80.87.90.164 on Sat, 23 May 2015 10:27:58 UTC
All use subject to JSTOR Terms and Conditions

et al.: Allocating
Keisler
Vendor
Interfaces 34(3), pp. 180-190, ?2004

Risks in the Hanford


INFORMS 189

Waste

Cleanup

a high cost. The DOE included our results in its


on the privatization
to the US Congress
project
of Energy Privatization
(Department
Group
Working
described
the model's
1998, ?3.4), which
impact:

have

attract

report

ing

The DOE also concluded that the level of uncertainty


with respect to design, financing, and regulation at
the end of Part A was such that fixing prices would
an

require

excessive

to compensate

price

for

the

risk

faced by the vendor. Thus, a design phase (referred


to as Part B-l in the contract) was defined to reduce
this uncertainty and to provide the DOE with vari
ous

reviews

with

decision

and

construction

will

and

wastes

Hanford-specific

and

arrangements

technology
and

to

prior
The

operations.

time to verify

allow

equity

points

to

proceeding

design

on
and

debt

optimize

technical

phase

performance

included

the following

lessons

learned:

Need for an equitable risk allocation. In the early stages of


the Phase I contracting approach, the DOE
developing
that
recognized
privatization is effective in shifting sig
risk to the vendor,
performance
to remain
have
with
the DOE.

nificant
would

some

but
For

risks

example,

the DOE recognized that the private sector would not


accept the risk of potential fluctuations in yearly bud
In addition, the DOE recognized
get appropriations.
the need to absorb the risks associated with its own
in

performance
and preexisting

areas

such

as waste

characterization

conditions.

The DOE

group (Department
privatization
working
of Energy Privatization
1998, Case
Working
Group
a
on
for the
report
study 6) prepared
privatization
of
Its
list
learned
from
this
of
lessons
energy.
secretary
included
the
project
following:
Risk Allocation. This is the single most
ture

of

the

to be

"deal"

established.

important

There

must

be

fea

that

it can

performance
for risks
that
furnished

control?for
DOE

risks.

The

it can

control?for

items

and minimum

and
technical
example,
to be
needs
responsible
example,
waste

government
quantities.

The overall
lead to estab
report, which
helped
lishment
of the DOE's Office
of Privatization
and
recommen
Contract Reform,
contained
the following
dation
Group

of Energy Privatization
(Department
1998, ?2, recommendation
5):

community
between

by, for example,


risk and
reward.

ensur

draft RFP.
Afterword
often on an
1998, the DOE used the model
Through
in finalizating
ad hoc basis
the RFP and selecting
and negotiations
vendors
vendor
with
the remaining
into what was called
(of two initially selected) going
Phase 1B-1 (the first stage was Phase 1A). In this case,
the DOE

used

the model

to evaluate
was

whether

the ven

and to
reasonable
proposed
an
al.
et
1998). Over
agreement
negotiate
(McLaughlin
to
the rest of the demonstration
the start
phase up
risk allocation

of

full-scale

mal

the DSS forecast


the opti
production,
to have expected
costs of $3 billion
to

strategy
almost all risk
$5 billion less than strategies allocating
to either the DOE or the vendor.
The larger privatization
effort had a disappointing
reasons
were beyond
for
that
the scope
end, however,
of the RFP risk-allocation
decisions. About a year after

the last agreement,


much of the tech
after developing
the
vendor
raised
its price estimates
nology,
suddenly
for the next stage by more
than 100 percent.
The
reasons behind
this move were
controversial
(Welch
costs had been allocated
2000). The risk of increased
so that when
costs became higher
than the DOE con
sidered acceptable,
the vendor
gave up the technol
and lost out on all incentive
ogy it had developed
com
the DOE paid for the work
while
payments,
a
costs
incurred
and
while
pleted
holding
seeking
to
The
DOE
its
abandon
used
replacement.
option
its contract with
the vendor
it, the pri
(and with
at the end of the design phase
vatization
initiative)
a more
and then adopted
conventional
government
owned,

contractor-operated

approach.

an

equitable allocation of risk between the DOE and the


vendor. The vendor must be held responsible for the
risks

balance

a
This represented
turnaround
from the
complete
view
that the vendors
should
all
risks, which
carry
in response
comments
to the
had led to discouraging

dor's

requirements.

The flexibility
that the design phase added proved
new
to risk allocation
Our
important.
approach
kept
con
to
that
viable
up
privatization
point. Our work
vinced
the DOE to take a more businesslike
attitude
toward risk. The DOE's
report to Congress
(Depart
ment
of Energy Privatization
1998,
Working
Group
?3.2)

the business

a proper

Working

The DOE should better integrate the perspectives of the


business community into its privatization initiatives. One
of the keys to successful privatization
at the Depart
ment is the ability to structure business deals that will

Summary
Our

new

led to the DOE's


successful
approach
of an RFP that at one time had large expected
of millions
of dollars or more.
savings in the hundreds
In this case, we
cannot
trace good decisions
about
to good ultimate
risk allocation
but this
outcomes,
the privatization
the good
program
approach allowed
intermediate
outcome
it attracted
two
of continuing;
zero response would
vendors
for Phase
IA, when
issuance

have

been likely given


the DOE's
attitudes
original
toward assigning
risk. In retrospect,
the DOE may
have made
too much
of a stretch in attempting
to
switch to a privatization
it
under
which
would
regime
as just another business.
have to behave

This content downloaded from 80.87.90.164 on Sat, 23 May 2015 10:27:58 UTC
All use subject to JSTOR Terms and Conditions

et al.: Allocating
190

Keisler

That

should not deter


in procurement
as well?from
tal agencies

businesses

others?certainly
and probably

involved

McLaughlin,
Whitfield.

governmen

this approach.
considering
Government
should
the Hanford
consider
agencies
a rather extreme
case.
In
sit
many
cleanup
simpler
a
to
could
this
uations,
help
approach
bureaucracy
overcome
to share risks. The DOE's doc
its reluctance
umented
lessons learned apply in such cases. Compa
nies
also

government
doing business with
agencies may
as a way
find our approach
to help the
useful
in nego
understand
the
agency
company's
positions
tiations that include the allocation
of risks.

tization

and

TWRS

cooperation,
Richmond

and

of
standing
Carol
Sohn
bers

of

programmatic
interest
and

complex
for her

the

DiPrinzio
their

of

the

We

improve
risk
topics.
and

support

including
assistance.
We

response

thank

Bill

who
experts
our under
We
other

thank
mem

Kearns

Paul
thank

and
A.

Raymond

Services

of Scully Capital
in

assistance

contractor

his enthusiastic

paper.
matter

team,

and Brian Oakley

invaluable

ment

this

subject
and

interviews

privatization
for their
Straalsund,

Jerry

on

Lerchen,

Megan
to conduct

us

allowed

for his advice,

the

development
We
model.
also

for

and

refine

thank

Mark

J. Jusko of Argonne National Laboratory for his assistance


with the refinement of the model as well as Chris Klaus for
his

assistance

running

scenarios.

We

thank

two

anonymous

reviewers for their helpful suggestions. Finally, we thank


Detlof von Winterfeldt
for reviewing the initial proposal for
this effort. The original work on which this paper is based
was

funded

by

the US

ber DE-AC06-76RL

Department

of Energy,

contract

num

O1830.

References
of
Department
Immobilization
27,
February
www.hanford.
(

to Congress:
Treatment
and
Report
Radioactive
Tank
Waste.
Retrieved
of Hanford
2002 www.hanford.gov/docs/twrs-atp/toc.html
Energy.

gov / docs / twrs-atp / section3


.html).
1998. Report
of Energy
Privatization
Department
Working
Group.
to the Secretary
theMarket:
The Opportu
of Energy. Harnessing
nities and Challenges
Retrieved
27, 2002
of Privatization.
February
DOE/S-0120.
(www.osti.
www.osti.gov/privatization/report/,
gov/privatization/report/chapt-2.htm).
W. A. Buehring.
J.M,

Keisler,

on
report
estimating
allocation
decisions
Energy
Argonne,

1996. Summary
of the technical
risk
of key programmatic
impact
on Phase
1 bids and U.S. Department
of
the

costs. ANL/DIS/TM-41,
IL.

Argonne

National

Laboratory,

1982. The Art

and Science

Cambridge,

Privatization.

TWRS

Privatization.

DE-RP06-96RL13308,

under

of Energy

1995.

contract

of Negotiation.

DE-AC06-76RLO

Harvard

Univer

MA.
Privatization

TWRS

Request

for Proposals

Request

for Proposals.

November.

(draft). DE-RP06-96RL13308,
1996. TWRS

Privatization

February.

market
1995. Financial
of Phase
Weimar,
M.,
J. Paananen.
analysis
on file in TWRS Privatization
1 for TWRS privatization.
Report
June 9.
Backup Documentation
Library,
costs may
nuclear
delay
Skyrocketing
1. Retrieved
Seattle
Times,
May
2002
seattletimes.nwsource.com/news/local/html98/
27,
.html.
hanf01m_20000501
C.

at

comments

his

H.

sity Press,

Welch,

Acknowledgments
We thank Mark Weimar

P. D., M. A. Robershotte,
R. G.
W. A. Buehring,
1998. Analysis
of the risk allocation
for TWRS priva
IB. PNNL-12028
(limited distribution).
phase
Prepared

for US Department
1830, November.
Raiffa,

Vendor Risks in the Hanford Waste Cleanup


Interfaces 34(3), pp. 180-190, ?2004
INFORMS

2000.

Hanford.

cleanup
February

B. Mellinger,
Senior
George
Manager,
Program
Pacific
Northwest
902 Bat
National
Laboratory,
teile Boulevard,
PO Box 999, Richland, Washington
"I was The Deputy
of the
99352, writes:
Manager
Team (WIT) at the Han
Disposal
Integration
of Energy
ford, Washington
(DOE) site
Department
from 1995-1998.
The WIT was
the Pacific North
west National
team (also referred
to as
Laboratory
the Privatization
the
for assisting
Team) responsible
DOE with developing,
and
implementing,
managing
the contracting
for the cleanup
of massive
strategy
amounts of nuclear waste
stor
stored in underground

Waste

the risk and decision management


age tanks. Under
sub-task for the WIT, extensive work was done in risk
to support
allocation modeling
this multi-billion
dol
lar DOE Tank Waste Retrieval
Priva
(TWRS)
System
tization effort.
"I verify that the authors

of paper 54-702-JK (Keisler,


and Wfhitfield)
Robershotte,
described
for the TWRS Priva

Buehring, McLaughlin,
did in fact do the work
tization team in the roles

and that their


they detailed,
efforts did have the impact they describe. The concept
of proper and appropriate
risk allocation was key in
garnering
private
industry support for this huge and
effort.
the Request
for
complex
cleanup
Specifically,
was
on
in
based
modified
their
(RFP)
part
Proposal
recommendations.
The DOE
to the modified
RFP, which
at
that
time."
proceed

This content downloaded from 80.87.90.164 on Sat, 23 May 2015 10:27:58 UTC
All use subject to JSTOR Terms and Conditions

received
enabled

in response
to
the initiative

bids

Вам также может понравиться