RFQ Auctions With Supplier Qualification Screening

This article was downloaded by: [14.139.128.
19] On: 28 January 2016, At: 20:50

Publisher: Institute for Operations Research and the Management Sciences (INFORMS)
INFORMS is located in Maryland, USA
Operations Research
Publication details, including instructions for authors and subscription information:
http://pubsonline.informs.org
RFQ Auctions with Supplier Qualification Screening

Zhixi Wan, Damian R. Beil,
To cite this article:

Zhixi Wan, Damian R. Beil, (2009) RFQ Auctions with Supplier Qualification Screening. Operations Research 57(4):934-949.
http://dx.doi.org/10.1287/opre.1080.0657
Full terms and conditions of use: http://pubsonline.informs.org/page/terms-and-conditions
This article may be used only for the purposes of research, teaching, and/or private study. Commercial use
or systematic downloading (by robots or other automatic processes) is prohibited without explicit Publisher
approval, unless otherwise noted. For more information, contact permissions@informs.org.
The Publisher does not warrant or guarantee the articles accuracy, completeness, merchantability, fitness
for a particular purpose, or non-infringement. Descriptions of, or references to, products or publications, or
inclusion of an advertisement in this article, neither constitutes nor implies a guarantee, endorsement, or
support of claims made of that product, publication, or service.
Copyright 2009, INFORMS
Please scroll down for articleit is on subsequent pages
INFORMS is the largest professional society in the world for professionals in the fields of operations research, management
science, and analytics.
For more information on INFORMS, its publications, membership, or meetings visit http://www.informs.org
OPERATIONS RESEARCH
informs
Vol. 57, No. 4, JulyAugust 2009, pp. 934949

issn 0030-364X eissn 1526-5463 09 5704 0934
doi 10.1287/opre.1080.0657
2009 INFORMS
Downloaded from informs.org by [14.139.128.19] on 28 January 2016, at 20:50 . For personal use only, all rights reserved.
RFQ Auctions with Supplier Qualication Screening

Zhixi Wan, Damian R. Beil
Stephen M. Ross School of Business, University of Michigan, Ann Arbor, Michigan 48109
{wanzhixi@umich.edu, dbeil@umich.edu}
We consider a manufacturer using a request-for-quotes (RFQ) reverse auction in combination with supplier qualication
screening to determine which qualied supplier will be awarded a contract. Supplier qualication screening is costly for
the manufacturerfor example, involving reference checks, nancial audits, and on-site visits. The manufacturer seeks to
minimize its total procurement costs, i.e., the contract payment plus qualication costs. Although suppliers can be qualied
prior to the auction (prequalication), we allow the manufacturer to delay all or part of the qualication until after the
auction (postqualication). Using an optimal mechanism analysis, we analytically explore the trade-offs between varying
levels of pre- and postqualication. Although using postqualication causes the expected contract payment to increase (bids
from unqualied suppliers are discarded), we nd that standard industrial practices of prequalication can be improved upon
by judicious use of postqualication, particularly when supplier qualication screening is moderately expensive relative to
the value of the contract to the manufacturer.
Subject classications: bidding/auctions; procurement; supplier qualication; supplier screening; mechanism design.
Area of review: Manufacturing, Service, and Supply Chain Operations.
History: Received June 2006; revisions received November 2007, May 2008, July 2008; accepted August 2008.
Published online in Articles in Advance March 16, 2009.
1. Introduction
for lengthy global supply chains, while price pressures and

a multitude of global supply options compounds the procurement managers challenge of discerning who is an able
supplier and who is a charlatan.
To verify a suppliers qualication and thereby reduce
the likelihood of nonperformance, the procurement function must spend time and money vetting suppliers with
qualication screening. Screening often involves references
checks, nancial status checks, surge capacity verication,
and even site visits to supplier production facilities. Supplier employees might be interviewed and assessed, for
example, to ensure that they understand all the engineering and safety requirements of the product in their charge.
Checks can even reach back into the suppliers supply
chain; to avoid mistakes like those uncovered in the recent
pet food contamination scandal, Procter & Gamble now
vets its suppliers raw ingredients sources to ensure that
they are reputable (Myers 2007).
Typically, qualication screening precedes price negotiations with suppliers. This is particularly common when
buyers use qualication screening before a request-forquotes (RFQ) auction for a well-dened good or service,
a prototypical procurement setting on which we focus this
paper. (For details on procurement auctions in practice, see,
for example, Jap 2003.) To convince the suppliers to bid
aggressively, the buyer touts the fact that only the lowest
price will win the contract when all participants in the auction are absolutely qualied to win the contract.
However, committing to award business to the lowest
bidder requires the buyer to spend signicant time and
resources screening all suppliers entering the auction. This
The average U.S. manufacturer spends 40%60% of its revenue income to purchase goods and services (U.S. Department of Commerce 2005). Vital for most companies, the
procurement function must negotiate reasonable prices with
suppliers andequally importantit must make reasonable efforts to ensure that contracts are made with qualied
suppliers who are indeed able to fulll the contract.
Contracting with unqualied suppliers can have dire consequences for the buyer. Numerous media reports recently
exposing product safety issues in the United States have
traced problems to suppliers failing to meet a buyers
requirements, resulting in dangerous lead paint in toys
(Spencer and Casey 2007), tires lacking proper safety features (Welch 2007), and pet food containing noxious chemicals (Myers 2007). Recalls of faulty products produced by
noncompliant suppliers have cost downstream rms millions of dollars and inicted untold damage on their reputations. Menu Foods market capitalization was slashed
in half soon after the rm recalled over 60 million packages of dog and cat food in March 2007; contaminants
were found to have originated in a suppliers raw ingredients (Myers 2007). Welch (2007) reports that New Jerseybased tire importer Foreign Tire Sales traced its consumers
complaints of faulty tires to an unauthorized design change
made by its supplier, whose design engineer decided to omit
gum strips, apparently unaware of their role in preventing
tread separation. A surprised Foreign Tire Sales was forced
to recall half a million tires, and may go bankrupt as a
result. Vulnerabilities to supplier nonperformance deepen
934
Wan and Beil: RFQ Auctions with Supplier Qualication Screening
Operations Research 57(4), pp. 934949, 2009 INFORMS
paper analyzes how much costly supplier qualication

screening should be performed before the auction. At one
extreme, the buyer uses prequalication only, in which she
fully screens all suppliers entering the auction and commits
to awarding the contract directly to the lowest bidder, so
suppliers in an open-descending auction (were this the format used) would bid down to their true cost (but they may
not have to), and the buyer directly awards the contract
to the lowest bidder. At the other extreme, the buyer uses
postqualication only, in which she screens suppliers only
after seeing their bids in the auction; instead of screening
all suppliers, the buyer homes in on the most promising
bidders and screens them in sequence until nding one who
is qualied. However, this comes with a trade-off: during
the postauction screening process, a bid will have to be
discarded if the bidder is found to be unqualied for the
contract; hence, the winner will be the lowest qualied bidder, if any. In this paper, we consider these two extremes,
and mixtures of the two in which suppliers are partially
screened before the auction.
Our study appears to be the rst to model and analyze
an auction setting in which contracting is contingent upon
passing costly supplier qualication screening, a common
feature of RFQs in practice. The central trade-off of the
research problem is this: although delaying some or all qualication screening until after the auction saves the buyer
qualication screening costs, doing so increases contract
payments and also risks nontransaction (turning to a costly
option outside of the auction, such as internal production)
if all suppliers in the auction turn out to be unqualied. By
optimally balancing these trade-offs, we nd that the buyer
can signicantly reduce total procurement cost (qualication cost plus contract payment) by judiciously delaying all
or part of supplier qualication screening until after the auction, provided the qualication screening is not too costly
(prompting defection to the outside option) nor too cheap
(making total prequalication the best option). Our analysis
endogenizes the level of pre- and postqualication as well
as the negotiation mechanism chosen by the buyer.
Section 2 provides a literature review, and 3 introduces
the model and assumptions. Through a mechanism design
approach, the optimal auction and postqualication structure is derived in 4.2. Behavior of the optimal balance
between pre- and postqualication is characterized in 4.3.
Section 5 provides numerical illustrations, and 6 discusses
practical considerations and extensions. Section 7 contains
concluding remarks. Proofs are provided in the electronic
companion for this paper, which is available as part of
the online version that can be found at http://or.journal.
informs.org/.
2. Literature Review
Many practical decisions faced by procurement managers
have been addressed by academic research: contract type,
such as xed price versus cost plus; negotiation framework
935
such as auction versus face-to-face; and competition
type such as sole versus dual sourcing. See Elmaghraby
(2000) for a detailed survey on procurement studies in
economics and operations management. Our paper focuses
on a procurement situation in which a buyer holds an
auction to award a contract to one of several potential
suppliers. There is a sizeable literature on auctionsbooks
by Krishna (2002) and Milgrom (2004) provide excellent
treatments and detailed referencesbut only a handful of
such studies include the processes that occur before and
after an auction. Typically, these processes seek to mitigate the risk of consummating a transaction in which one
or more parties does not obtain what it expected, what we
will call nonperformance. Note that nonperformance could
describe, for example, an item falling short of the winners expectations, or a winner failing his obligations to
the auctioneer. The supplier qualication process central to
this paperand to our knowledge novel in the literature
is one such process meant to mitigate the risk of supplier
nonperformance in a procurement auction.
One type of nonperformance in forward auctions is the
winners failure to pay. Papers dealing with this issue have
looked at the ability of bidders to borrow money and the
ensuing possibility of broke winners (Zheng 2001), and
the use of deposits or fees forfeited to the auctioneer in
the event a winning bid is reneged upon (Rothkopf 1991,
Waehrer 1995). In a procurement auction context, surety
bonds (analogous to a bid deposit) to partially offset nonperformance costs are examined by Calveras et al. (2004),
whereas Braynov and Sandholm (2003) study a buyer who
is unable to directly verify the trustworthiness of suppliers, but knowing the form of the suppliers cost functions can design bidding options that cause each supplier to
reveal themselves as either a high or low trustworthy type,
allowing the buyer to estimate the expected utility of contracting with that supplier. In contrast, we assume that the
buyer veries (at a cost) that a supplier is qualied up to
some threshold prior to contracting (she will not contract
with unqualied types). Practitioners we have spoken with
use qualication and surety bonds in tandem, the former to
proactively avoid problems (the focus and main contribution area of our paper), the latter to partially recoup costs
if problems arise.
A second type of nonperformance is misevaluation of the
item. For example, costly bid preparation (or costly entry,
or due diligence) plays a central role in forward auctions
for nonstandard, complex items such as an entire company or its assets. To encourage participation in auctions
where bidders trade off their bid preparation costs (possibly millions of dollars) against their anticipated likelihood
of winning the item, Ye (2007) suggests inviting only bidders whose bid in an initial, assumed costless, round of
bidding signals that they stand a good chance of winning
the item. Whereas Ye nds that screening out low-value
bidders can promote competition in a complex item auction
by limiting bidders unnecessary bid preparation costs, we
936
examine screening costs borne by the auctioneer (buyer)
and nd that screening out unqualied suppliers promotes
competition by increasing the likelihood that each bid in
an RFQ auction will be qualied and therefore eligible for
contracting.
In our context of relatively well-specied RFQ auctions,
we assume that a suppliers bid indicates the value offered
to the buyer should that supplier be deemed qualied. This
allows the buyer to delay all or part of the qualication
process until after the auction, at which pointwith bids in
handshe can home in on the suppliers offering the highest
value (who may or may not turn out to be qualied). Other
auction-theoretic papers have focused on situations where
it is costly for the auctioneer to estimate even the value
offered in the suppliers bids. In such situations, the buyer
could employ a sequential search model whereby suppliers
are communicated with and their bids evaluated until either
nding a supplier whose cost is sufciently low or exhausting the supply pool (McAfee and McMillan 1988). In an
effort to explain unconsummated request-for-proposals auctions documented by Snir and Hitt (2003), Carr (2003)
models an auction for professional services where proposals (bids) are difcult to compare; in his model, faced with
high evaluation costs after the auction, the auctioneer might
simply forgo evaluating any proposals in favor of an outside
option. The buyer sometimes turns to an outside option in
our model, but for different reasonseither due to reserve
prices, or after disqualifying all suppliers invited to the
auction.
Methodologically, our study is also related to the screening literature, in particular, studies such as Feinberg and
Huber (1996), which assume that some form of screening
can be performed cheaply (e.g., bids can be observed) relative to more costly forms of screening (e.g., qualication).
In our context, partial qualication screening impacts the
extent to which bidders compete in the auction by creating randomness in the number of qualied bidders in the
auction. For auctions with an uncertain number of bidders,
optimal mechanisms and equilibrium bid functions have
been respectively derived by McAfee and McMillan (1987)
and Harstad et al. (1990), but neither studies qualication
processes or the attendant possibility of having to turn to
an outside option.
Our study is related in spirit to multiple-dimensional auctions (Che 1993, Beil and Wein 2003, Chen 2007) in that
both seek to take nonprice factors into account. Although
the goals of a multiple-dimension auction and qualication processes are related, they are distinct: multiple dimension auctions serve as tools to better express the value of
nonprice abilities of suppliers, such as quality. Qualication
processes seek to verify the ability of a supplier to deliver
on the promises expressed by his bid, be they promises on
price or any other dimensions.
3. Model
We model a risk-neutral, cost-minimizing buyer seeking
to award an indivisible contract to a qualied supplier.

A supplier is called qualied if the buyer is willing to transact with the supplier without performing additional due
diligence to verify that this supplier satises all preaward
requirements. These requirements vary widely in practice
depending on the buyers needs and the contract type (see
Leenders et al. 2006 for a discussion of purchasing processes). The constituent requirements themselves exhibit
varying degrees of standardization. Among the more standard requirements are the need to verify the suppliers
reputability (e.g., through published ratings) and ability to
ramp up production. Not all requirements are so transparent, because qualication can encompass relational aspects
that are difcult to codify; e.g., a just-in-time manufacturer
we spoke with visits supplier management to ensure they
and the supplier see eye to eye on lean principles before
awarding contracts. These verication processes take time
and can be costly, particularly if involving visits to distant
supplier facilities. As is common in industry, we will refer
to this verication as the qualication process, or the act of
qualifying a supplier. A supplier is described as qualied
once he successfully passes the qualication process.1
To begin formalizing the model, we imagine a continuum of qualication requirements, with zero representing
requirements that every supplier satises and one representing requirements that virtually no supplier satises. We let
q0 be the buyers qualication threshold, a scalar between
zero and one representing the preaward requirements. For
each supplier i, we will dene his qualication level qi
as the maximum qualication threshold that supplier i can
pass. Due to opaque requirements set by the buyer, such as
needing to see eye to eye on lean principles, or to create
rapport with the buyers internal customer (e.g., the engineering department), supplier i does not precisely know its
true qualication level qi , but the buyer and supplier share
the common belief that qi is distributed according to distribution H on domain 0 1. This setup could model, for
example, a buyer deciding to outsource a portion of its
production currently done in-house, facing new suppliers
she knows little about, and who in turn know little about
her (possibly idiosyncratic) qualication requirements. The
strictness of the buyers preaward requirements is captured
by 1 H q0 , the probability that qi q0 .
The buyer and each supplier responding to the RFQ are
equally unsure of the suppliers qualication until costly
qualication verication is undertaken by the buyer. If
qi q0 , the buyers qualication process on supplier i
would reveal that supplier i is qualied. The qualication
cost the buyer would incur to do so is denoted by K, the
total cost to the buyer of verifying that an individual supplier meets all requirements to be deemed qualied. For
example, K may include the cost of purchasing and testing
supplier products, travel to supplier facilities abroad, etc.
For simplicity, we assume that K is the same for all suppliers. This assumption is most appropriate when suppliers
are similar, at least in terms of the cost drivers of qualication, such as distance from the buyer or number of units
937
that the buyer must purchase to run a test sample. A more

general model would allow all the qualication verication
cost to differ for different suppliers. We leave this ex ante
asymmetric suppliers extension for future research.
On the other hand, if qi < q0 , supplier i would be rejected
during the buyers qualication process after failing to meet
a requirement. In this latter case, how much cost would
the buyer incur? Assuming that the requirements are nested
(passing a larger threshold implies passing a smaller threshold, but not vice versa), the cost strictly increases with qi
and approaches K as qi approaches q0 . To streamline the
exposition, we further assume that the cost is linear and
normalized such that a threshold of zero costs zero to verify, implying that weeding out an unqualied supplier i
costs the buyer qi /q0 K. This is without loss of generality because any nonlinear and strictly increasing cost
function of qi over 0 q0 can be renormalized to be linear by redening qi and renormalizing the distribution H .
Note that in this paper we use the terms increase and
decrease in their weak sense.
Each supplier i privately knows his cost to fulll the contract, xi , which he observes perfectly prior to the auction.
Cost xi is distributed according to a commonly known distribution F with density f on domain 0 1, which for simplicity is assumed to be statistically independent of other
suppliers costs (6.1 relaxes this assumption) and is own
qualication level. In reality, a supplier who is likely to be
qualied might be expected to have relatively high costs
if qualication requires that costly spare capacity be kept
on hand for an ability to ramp up production in the face
of surge orders from the buyer; on the other hand, lower
costs might prevail if qualication includes lean principles that impart efciency. We leave such issues for future
research. For simplicity, we will assume that all distributions in this paper have positive and continuous densities
over their domains.
The buyer utilizes an auction to extract private cost information from the risk-neutral suppliers, whoas is standard
in the auction literatureare assumed to be fully rational
players following a Bayesian Nash bidding equilibrium in
which bidding is assumed to be costless. To capture practical demands on the buyers resources, such as auction
participant training and technical support, we assume that
the buyer will have a nite number of bidders in the auction, denoted by n, which is treated as xed and exogenous
in our analysis.
We now describe the three-stage procurement process the
buyer uses for supplier qualication screening (qualication
checks) and supplier cost screening (competitive auction).
Prequalication Stage. The buyer announces n prequalication thresholds qi q0 (i = 1 n), where qi represents the requirements the ith auction participant must
meet to participate in the auction. Each prequalication
threshold qi represents part ( qi < q0 ) or all ( qi = q0 ) of
the requirements that a supplier must satisfy in order for
the buyer to consider them qualied for contract award.

The prequalication thresholds are not assumed to be identical, and can be different for different auction participants
(the less general but perhaps more equitable case of symmetric prequalication thresholds is analyzed in 4.4). The
buyer faces a pool of ex ante symmetric suppliers. To nd
its ith auction participant, the buyer veries requirements
of supply pool members one at a time until nding one
who passes prequalication threshold qi . (For simplicity
we assume an innite supplier pool.) After doing so for
each i = 1 n, the buyer has n prequalied suppliers,
where the ith participant (we call this participant bidder i)
has qualication level qi qi .
Auction Stage. The n prequalied bidders participate
in a price-only auction. When bidding, bidder i, who passed
prequalication level qi , is estimated to be qualied with
probability
i =
1 H q0

1 H qi
This probability, referred to as bidder is qualication probability, is strictly increasing in qi . In particular, i equals
its upper bound 1 when qi = q0 , and equals its lower bound
1 H q0 when qi = 0. In the sequel, the i s gure prominently in the analysis and are, for convenience,
directly used as the buyers prequalication threshold decisions. This is without loss of generality: because H is
strictly increasing, there is a one-to-one correspondence
between qi and i , and we dene function q 1
0 q0 such that

1 H q0

qi = qi H 1 1
i
We primarily use notation qi in lieu of qi . We use the
bold notation for the vector of qualication probabilities
1 2 n , and use for the vector having i = for
all i = 1 n. Unlike a traditional auction where bidding is
directly followed by a contract award decision, this auction
yields a postqualication sequence, which species (based
on auction bids and bidder qualication probabilities) an
ordered subset of bidders who will be postqualied in the
next and nal stage of procurement, the postqualication
stage.
Postqualication Stage. After the auction the buyer
postqualies suppliers up to qualication threshold q0
according to the postqualication sequence until nding a
qualied bidder to contract with, or disqualifying all bidders included in the postqualication sequence. In the latter
case, the buyer turns to her outside option at a cost of Co .
Cost Co describes an option outside the auction, but could,
for example, correspond to in-house production.
The buyer faces two key decisions in the above procurement process: what prequalication thresholds to set;
938
and what auction and postqualication mechanism to use
to nd (if possible) a desirable, qualied bidder with which
to contract. Our analysis rst nds the optimal auction and
postqualication mechanism, and then works backwards to
characterize the optimal prequalication thresholds. Before
moving to this analysis, we conclude this section by discussing the main trade-offs involved.
The choice of prequalication thresholds affects the balance between qualication costs on the one hand and contract payment on the other. If = 1 1, the buyer
performs all due diligence prior to the auctionthe situation tacitly assumed in traditional auction theory. In such
a case the buyer incurs qualication costs for at least n
(possibly more if some are rejected en route to being
qualied) suppliers prior to the auction, and awards the
contract directly to the most attractive bidder after the
auction. On the other hand, = models postponement
of all due diligence; after the auction, the buyer only pays
qualication costs until nding the rst qualied bidder
or turning to her outside option. Clearly, the expected
total cost of qualifying suppliers is greater in the case
where = 1 1. However, the buyer also must consider the expected costs of contracting and nontransaction,
for which the cost relationship can be reversedconsider
the following toy example with two bidders showing why
more prequalication reduces the expected costs of contract
payment and nontransaction.
Suppose that bidder 1s and bidder 2s true costs are
$100,000 and $125,000, respectively, and the auction
reserve price is $150,000. For simplicity, suppose that the
auction mechanism induces truthful bidding. If bidder 1
has been fully qualied ( q1 = q0 ), he is the obvious candidate for contract award. However, to ensure truthful bidding, the mechanisms award and payment rules must be
designed taking bidder incentives into account. An interesting dynamic arises regarding payment to bidder 1. Should
the mechanism pay bidder 1 precisely $125,000, the cost
of the losing bidder? If q2 = q0 , the answeraccording to
the incentive compatibility of the Vickrey auctionis yes,
which is based on the buyers next-best alternative being
bidder 2. However, things are different if q2 < q0 . Intuitively, the mechanism must compensate bidder 1 for the
fact that bidder 2 might be unqualied for the contract,
and hence possibly a nonviable alternative for the buyer.
Thus, the fact that the buyer delayed qualication on bidder 2 means that she must pay bidder 1 more than $125,000
for the contract. On the other hand, if bidder 2s true
cost had been less, say $80,000 instead of $125,000, the
buyer might nd it worthwhile to rst postqualify bidder 2,
which may or may not turn out favorably for the buyer
during postqualication she might discover that bidder 2
is unqualied ( q2 q2 < q0 ), at which point she has
wasted postqualication money on bidder 2 and her nextbest option will be bidder 1, whose payment is then the
reserve price. The optimal mechanism analysis of the next
section determines how the buyer should run an auction

and postqualication process when suppliers are possibly

unqualied, where clearly the buyers expected payment
increases with the amount of qualication that is delayed
until after the auction.
4. Analysis
In this section, we rst x the qualication screening
strategy (equivalently, x the qualication probabilities
1 n ) and compute the buyers total (contracting plus
qualication) expected costs. We begin by deriving PRE,
the expected prequalication cost, in 4.1. In 4.2, we
derive PAY, POST, and NT, respectively, the expected payment to the auction winner, the expected postqualication
cost, and the expected nontransaction cost. These costs are
derived via an optimal mechanism analysis. Having derived
the total expected cost as a function of the qualication
screening policy, we then characterize the optimal qualication screening strategy in 4.3. Following this general
analysis, 4.4 discusses the special (but more fair) case
of symmetric prequalication thresholds.
4.1. Expected Prequalication Cost
As explained in 3, we assume an innite supply pool
from which the buyer samples until nding n suppliers that
pass their respective prequalication thresholds. The buyer
begins by sampling suppliers one at a time until nding
a supplier whose qualication level is at least q1 . This
successful supplier is admitted to the auction as bidder 1.
This process is then repeated for i = 2 n. The ordering
of the i s is not important because failures, i.e., suppliers
found to have a qualication level below qi , are permanently discarded for being unqualied for the contract, and
the supply pool population is innite and samples (supplier
qualication levels) are independent. The expected prequalication cost is comprised of n prequalication successes
plus geometrically distributed numbers of prequalication
failures before each success. Because the qualication
cost is linear in the amount of qualication performed, for
each success the buyer pays qi /q0 K to prequalify a
supplier with qualication level y qi up to the prequalication threshold qi . For each failed prequalication on a supplier with qualication level y < qi , the
buyer pays y/q0 K. Because y is a random variable and
meets or exceeds the prequalication level qi with probability 1 H qi , the buyer expects to pay

n

qi
1
1
PRE =
K+
q0
1 H qi
i=1

qi
y
K dH y
(1)
q0 H qi
y=0
4.2. Expected Costs in Auction and
Postqualication
The auction and postqualication stages together comprise
a mechanism of awarding a contract to one of n bidders
939
or to the buyers outside option. The buyer designs this

mechanism to minimize her expected auction and postqualication cost PAY + POST + NT. The mechanism extracts
private cost information from the suppliers, as is common
in traditional mechanism design, but also involves postqualication screening, which is not traditional in the mechanism design literature. The auction and postqualication
mechanism has three components: a set of possible messages (or bids) for each bidder; a rule that describes
how the buyer sequences the bidders for postqualication;
and a payment rule that maps bids to an amount transferred between the buyer and bidders. The postqualication
sequencing rule is the analogue of the allocation rule in
standard mechanism design problems absent postqualication; it captures the fact that the allocation decision itself is
not entirely in the buyers hands because postqualication
outcomes (the realization of qi s) affect whether or not a
bidder will actually be qualied for contract award.
Both rules (sequencing and payment) are functions of the
messages sent by bidders. Thanks to the revelation principle (e.g., Gibbard 1973, Green and Laffont 1977, Dasgupta
et al. 1979), given any mechanism and an equilibrium for
that mechanism, there exists an outcome equivalent direct
mechanism in which it is an equilibrium for each bidder
to bid his true cost. This allows us to restrict our optimal
mechanism search to direct mechanisms; in what follows,
the set of messages sent is assumed without loss of generality to be just the bidder cost vector x x1 x2 xn .
The remainder of this subsection formalizes the mechanism
design problem.
We will let Qx denote the postqualication sequencing
rule, an ordered subset of 1 n representing the bidders that will be postqualied in a determined sequence.
For example, a postqualication sequence 3 1 5 2 for
n = 5 means that the buyer would postqualify bidder 3, bidder 1, bidder 5, and bidder 2 in sequence until nding the
rst qualied bidder or disqualifying all these four bidders,
without postqualifying bidder 4 at all. For a given vector of
costs x, using Q and we can compute i , the probability
that bidder i wins the contract, as

i x = i
1 j if i Qx
j ranks ahead
of i in Qx
i x = 0
(2)
if i Qx
This assumes that the buyer uses a deterministic sequencing rule, although this is without loss of generality; see
the proof of Proposition 1. The mechanisms payment rule,
which we will denote as Mx n , is the expected monetary transfer from the buyer to the bidders.
Of course, to constitute a viable direct mechanism, Q
and M must satisfy constraints ensuring both incentive
compatibility (truthful bidding in equilibrium) and individual rationality (bidders expect nonnegative prots by
participating). To write down these constraints for bidder i, rst dene xi = x1 x2 xi1 xi+1 xn , and
let
Fi be the joint distribution of xi . Dene mi zi
Mi zi xi dFi xi to be the expected payment to bidxi
der i when i reports cost zi and all others report their true
costs, and likewise dene i zi xi i zi xi dFi xi
to be the probability bidder i is awarded the contract when
i reports zi to the auctioneer and all others report their true
costs. Under truthful bidding, if all bidders j = i report their
true costs, bidder is expected prot from reporting zi is
given by mi zi i zi xi . Thus, incentive compatibility and
individual rationality can be expressed as
Ui xi mi xi i xi xi = max mi zi i zi xi
zi 0 1
(incentive compatibility) (3)

Ui xi 0
(individual rationality)
(4)
Ui xi is the expected prot of bidder i in equilibrium.

For Q and M satisfying (3) and (4), we now compute NT,
POST, and PAY.
Expected Nontransaction Cost. If no bidder wins the
contract, the buyer must turn to the outside option at
cost Co . Letting F denote the joint
distribution
of x, from

(2) it is easy to see that NT = x 1 ni=1 i xCo dFx.
Expected Postqualication Cost. Because i x is the
probability that i wins the contract, i x/i must be the
probability that bidder i undergoes postqualication. Taking
the expected cost of postqualifying bidder i, and dividing
this cost by i and calling the result ai , yields
q0
qi qi
1
ai
K dH qi
1 H q0 qi =qi
q0
+
q0 qi
K
q0
(5)
Thus,
n the expected postqualication cost is POST =
i=1 i xai dFx
x
Expected Payment. Finding the optimal mechanism is
aided by expressing the buyers expected payment (up to
an additive constant) as a function of award probabilities
(as is standard in mechanism design analyses). Applying
the envelope theorem (see Milgrom 2004, p. 67) to (3)
implies that Ui
xi = i xi for xi 0 1. Treating this
as a differential equation and using the expected prot at
xi = 1 as an integration constant yields an equation for
bidder is equilibrium expected prot:
1
mi xi i xi xi = mi 1 i 1 +
i zi dzi
(6)
xi
Solving for mi , the buyers expected payment to bidder i

given the reported cost xi , and then integrating over x and
summing over n, yields
n
n

PAY = mi 1 i 1 +
i xxi dFx
i=1
i=1 x
where xi xi +
F xi

f xi
(7)
940
The value xi , which equals bidder is true cost, xi , plus

F xi /f xi , the informational rent accruing to bidder is
private knowledge of his cost xi , is commonly referred to
as is virtual cost in the mechanism design literature.
Mechanism Design Program. If (6) holds, it is easy
to check that incentive compatibility constraint (3) is guaranteed if the postqualication sequencing rule is such
that i zi is decreasing in zi 0 1. (See, for example,
Myerson 1981, Lemma 2 for a similar result.) Furthermore,
from (6), we have that individual rationality constraint (4)
holds as long as the payment and postqualication sequencing rules satisfy mi 1 i 1 0. Clearly, Equation (7)
implies that a cost-minimizing buyer will set mi 1
i 1 = 0. Thus, to nd the cost-minimizing direct mechanism Q M employed by the buyer in the auction and
postqualication stage, it is sufcient to solve the following
program,
n

min
i xxi +ai CodFx+Co (8a)
Q, M
i=1 x
PAY+POST+NT
subject to mi xi = i xi xi +
xi
i zi dzi
i
(8b)
and verify that at this solution i zi is decreasing in zi

0 1 for all i. Program (8) ignores the prequalication
cost, which is considered sunk by the auction and postqualication mechanism. Note that if the buyer selects the prequalication thresholds to be qi = q0 for all ii.e., she
performs all qualication screening prior to the auction
then i = 1 for all i and the above program devolves into
the standard optimal auction analysis of Myerson (1981).
4.2.1. The Optimal Auction and Postqualication
Mechanism. We now nd an optimal solution Q M
to program (8). To ensure that i zi is decreasing, we will
assume that the virtual cost xi is strictly increasing in
the true cost xi (see the proof of Proposition 1). This standard, technical assumption is satised, for example, if F is
logconcave; see Bagnoli and Bergstrom (2005) for details
about logconcave functions, which include uniform, normal, logistic, and exponential distributions.
Proposition 1. An optimal direct, individually rational,
and incentive-compatible auction and postqualication
mechanism Q M that minimizes PAY + POST + NT is
as follows.
Set Q x such that the buyer only postqualies those
bidders whose xi and i are such that xi + ai Co ,
and sequence all such bidders for postqualication according to ascending value of xi + ai , breaking ties
randomly.
Set M x such that the buyer pays the contract winner, i.e., the rst bidder (if any) deemed qualied by postqualication, a payment of
1 z x
i
i
i
xi +
(9)
dzi
i x
xi
and pays nothing to all other bidders, where i x is calculated via Equation (2) given Q x.
Note that Proposition 1 implies that the buyer should
set bidder-specic reserve prices ri i = 1 n, such that
ri = maxmin 1 Co ai 1 0, and promise not to
postqualify any bidder whose bid falls beyond the reserve
price for that bidder. When bidders have equal i s (symmetric prequalication), a common reserve price for all
bidders is optimal; see 4.4.
Adjusted Virtual Costs Account for Postqualication.
Proposition 1 indicates that the buyer should sequence
bidders for postqualication according to adjusted virtual
cost, which we dene to be xi + ai . The added
adjustment ai captures both the cost of postqualifying
bidder i and the risk that bidder i fails postqualication.
Intuitively, this adjustment should be lower if i already survived very strict prequalication, and indeed ai does
decrease in i . Thus, the optimal sequencing rule takes
into account not only virtual cost, but also favors bidders
who are more likely to be qualied by applying adjustments when sequencing the bids for postqualication. If
the buyer uses an identical prequalication hurdle for all
bidders (identical i s), the optimal sequencing rule simply
postqualies bidders according to cost xi (bidders identical ai s wash out, and is increasing); see 4.4, where
the case of symmetric prequalication is discussed.
Payment to Winner Depends on Competitors Virtual
Costs and Qualication Probabilities. Proposition 1
implies that to induce all bidders to bid their true costs, the
buyer nds it optimal to commit to pay zero to all bidders
save the contract
1winner, say bidder i, who is paid his cost xi
plus a markup xi i zi xi /i x dzi . This payment can
be rewritten (see the proof of Proposition 1) as
t

j=1
1 xij + aij ai ij

+ ri
t

1 ik if t 1 or
k=1
j1
k=1
1 ik
ri
if t = 0
(10)
where ij (j = 1 t) is the index of the jth bidder out

of t bidders sequenced after bidder i for postqualication
according to the optimal sequencing rule such that xij +
aij ri + ai . When the buyer reaches bidder i
in the postqualication stage, she has at least t + 1 options
available to her besides bidder i, namely, the t bidders
described above, plus the outside option. Bidder i could
have increased his bid and still been preferred by the buyer
to each of these t + 1 options. When computing the payment to bidder i, the buyer rewards bidder i for this. In particular, the value 1 xij + aij ai is exactly the
amount that bidder i could have bid and still
been preferred
to bidder ij . This value is weighted by ij j1
k=1 1 ik ,
exactly the probability that the buyer would end up contracting with bidder ij if the buyer were to, upon reaching
941
bidder i in the qualication list, forgo qualifying bidder i.

Similarly, the reserve price ri is the maximum amount bidder i could have bid and still been preferred to the out
side option, which is weighted by tk=1 1 ik (if t 1)
or 1 (if t = 0), exactly the probability that the buyer would
end up using the outside option if the buyer were to, upon
reaching bidder i in the qualication list, forgo qualifying
bidder i. As a check, when i = 1 for all i the payment
reduces to the smallest rejected bid xi1 (or the reserve price,
whichever is smaller), which is exactly the payment rule of
the traditional second-price auction with a reserve price.
4.2.2. Optimal Mechanism Implementations. Auction formats differ in the amount of visibility bidders have
on competitors bids. At one extreme is the sealed-bid format, where bidders submit a single bid known only to the
bidder and auctioneer. At the other extreme is an open-bid
format, where bidders can see competing bids and respond
by updating their own bid. In the following, we show that
the optimal auction can be implemented in either of these
two format types.
Sealed-Bid Format. As is straightforward to see given
Proposition 1, the buyer can conduct a sealed-bid auction
after directly announcing the optimal sequencing rule as
specied in Proposition 1 and the payment rule per (10)
above; this will lead to an equilibrium where all bidders
truthfully bid their costs.
Open-Bid Format. In our open-bid format, bidding
proceeds as follows. The buyer starts the auction by setting the current bid at Co . All those bidders who wish to
remain in the auction at the current bid signal their willingness to do so. The buyer then lowers the current bid
by some decrement, and all those bidders still willing to
remain in the auction at this new current bid again signal
their willingness to do so. We assume this decrement is
arbitrarily small, ensuring that bidders can bid down to any
value they wish to. For simplicity, we assume that reentry is disallowed: Once a bidder fails to signal willingness
to remain in the auction, he cannot resume bidding at a
lower price (the equilibrium remains if this assumption is
relaxed). We let bi denote bidder is dropout bid, that is,
bidder i remains in the auction up to bid bi , but drops out
when the buyer lowers the current bid below bi . The buyer
keeps lowering the current bid until at most one bidder
remains in the auction. After the auction, the buyer discards
bidders who did not participate in the auction (i.e., those
who never signalled a willingness to remain in the auction),
and postqualies the remaining bidders in order of ascending dropout bids (the bi s), breaking ties randomly.
This setup resembles a traditional clock auction
(e.g., Ausubel and Cramton 2006), with the crucial difference that bids can be interpreted as bids for positions in
the postqualication sequence. The dropout bids are used
to compute the contract payment in the following manner.
Suppose that bidder i is awarded the contract, that is, i is
the rst bidder that successfully passes postqualication.

Let ij and t be as dened immediately following (10). The
payment awarded to bidder i is precisely that given in (10),
with one small change: bij is used in lieu of xij + aij
within the bracketed term of the rst summation.
Because this is an open-bid format, during the auction,
in addition to the current bid, bidders are able to see competing bids remaining in the auction. In equilibrium, each
bidder i will simply remain in the auction unless the current bid dips below his true adjusted virtual cost xi +
ai , at which point he drops out. It is easy to check that
in equilibrium bidders are sequenced and paid exactly as
Proposition 1 species. In summary, we have:
Proposition 2. Under the open-auction format described
above, the strategy of dropping out of the auction at ones
true adjusted virtual cost constitutes a Bayesian Nash equilibrium under which the optimal mechanism from Proposition 1 is implemented.
4.3. Optimal Qualication Screening Strategy
The buyer seeks the overall optimal mechanism that minimizes the expected total cost PRE+PAY+POST+NT. For
a given set of prequalication thresholds, the previous subsection derived the buyers optimal auction and postqualication mechanism, minimizing PAY + POST + NT. Hence,
it remains only to proceed one layer up and investigate the
optimal selection of the prequalication thresholds (equivalently i s). The buyer faces four practical concerns: uncertain supplier costs, uncertain supplier qualication levels,
costs associated with verifying the qualication of a supplier, and an outside option cost. Comparisons of the latter two costs determine the buyers optimal prequalication
thresholds, for given uncertainties over supplier costs and
qualication levels. We rst identify the central trade-off
involved in the qualication strategy decision, followed by
Proposition 3, which characterizes the buyers optimal qualication screening strategy via three switching thresholds
for qualication cost.
Central Trade-Off of Qualication Screening Decision. Under the optimal auction and postqualication mechanisms described in 4.2, the more qualication due diligence the buyer performs before the auction, the more she
pays before the auction but the less she pays after the auction. That is,
PRE increases in i while PAY + POST + NT
decreases in i
for all i = 1 n
(11)
(This result is proved formally in the electronic companion.)

Therefore, in an optimal qualication screening strategy the
buyer trades off the prequalication cost against the total
postauction cost. Hereafter, we call this the central trade-off
of the qualication decision.
942
Optimal Qualication Screening Strategy Thresholds.
Although the central trade-off reveals the buyers key consideration when deciding the prequalication thresholds,
solving for the optimal that minimizes PRE + PAY +
POST + NT for xed K and Co is precluded by the
complexity of the total cost expression (see, for example,
Equation (10)). Instead, we wish to describe the optimal qualication screening strategy more qualitatively by
characterizing when various possible types of qualication strategies will be used. For any xed outside option
cost Co , we seek to show that there exist three qualication
cost thresholds dened as follows. Let K nt be the smallest qualication cost such that for any qualication cost
K K nt , forgoing the auction in favor of the outside option
is optimal. Let K post be the smallest qualication cost such
that, for any qualication cost K such that K post K < K nt ,
delaying all qualication screening until after the auction
(referred to as post-only for short) is optimal, that is, i =
for all i. Let K pre be the largest qualication cost such that
for any qualication cost K K pre , doing all qualication
screening before the auction (referred to as pre-only for
short) is optimal, that is, i = 1 for all i.
Proposition 3. Fix the outside option cost Co . Under
the buyers optimal (total-cost minimizing) qualication
screening strategy, K nt , K post , and K pre all exist, are positive, and are nite. In other words,
(i) prequalication only, that is i = 1 for all i, is optimal if K K pre ;
(ii) a mix of pre- and postqualication is optimal only if
K pre < K < K post ;
(iii) postqualication only, that is i = for all i, is
optimal if K post K < K nt ; and
(iv) it is optimal to forgo the auction in favor of the
outside option if and only if K nt K.
Moreover, the thresholds K nt , K post , and K all increase
in the outside option Co .
Part (i) of Proposition 3 shows that there exists a threshold K pre , separating the decision between complete versus partial prequalication. Keeping the outside option cost
constant, if the qualication cost is below this threshold,
the buyer prefers to completely prequalify suppliers before
the auction; furthermore, any increase in the outside option
cost while holding the qualication cost xed will result in
the buyer still preferring complete prequalication. Part (iii)
of Proposition 3 shows that a similar threshold K post exists
between the partial prequalication versus the postqualication only decision. Parts (i), (ii), and (iii) of the proposition together indicate that, keeping the outside option cost
xed, as qualication cost increases from zero the buyers
decision shifts from complete prequalication, to a mixture of pre- and postqualication, and then to the complete postqualication decision. Finally, the buyer prefers
to forgo the auction altogether if the cost of qualication is
too high, as shown in part (iv) of Proposition 3.

Proposition 3 indicates that despite the risk of nontransaction and larger contract payment, the buyer sometimes nds
it protable to postpone some or all qualication screening
until after the auction. The decision of how much qualication to postpone depends on the qualication cost because
the buyer has more incentive to risk nontransaction and
higher payments with qualication postponement if doing
so avoids high qualication expenses before the auction.2
4.4. Symmetric Prequalication Levels
Although it can be in the buyers interests to set unequal
i s for different bidders when K pre < K < K post , doing so
may raise fairness concerns about the procurement process.
Under the optimal mechanisms described in 4.2, bidders
who are more thoroughly prequalied are advantaged: First,
they are more likely to win the contract because they are
more likely to be ranked at the top of the postqualication
sequence given that their virtual costs require less upward
adjustment to capture postqualication costs (i.e., ai
decreases in i ); second, they will be paid more upon winning the contract, given that the payment to any winning
bidder i per formula (10) increases in i . In a symmetric
environment with no a priori differences among suppliers,
the buyer may wish to prequalify suppliers equally, even
though this might technically not be optimal, to avoid ill
will created by arbitrarily choosing one supplier to be more
prequalied than the others.
Symmetric prequalication also simplies the buyers
optimal auction and postqualication mechanism. When
bidders are equally prequalied, the optimal mechanisms of
4.2 use a common reserve price for all bidders. If is the
probability that any bidder in the auction is truly qualied,
this optimal reserve price is
r = maxmin 1 Co a 1 0
Moreover, because a common qualication cost adjustment
a is applied to all bidders, the optimal postqualication
sequencing rule reduces to simply ranking bidders according to ascending virtual costs, and in turn this is equivalent
to simply ranking bidders according to true costs (by the
virtual cost function increasing in true cost). In the following, we discuss a modied version of a standard sealed-bid
rst-price auction, where the contract winner is paid what
he bids.
Sealed-Bid First-Qualied-Price Auction. When bidders are equally prequalied, the buyer can conduct a
sealed-bid rst-qualied-price auction to implement the
optimal auction and postqualication mechanism. In such
an auction, the buyer sequences all bidders (who bid below
the common reserve price r) according to ascending bid
values, and awards the contract to the rst qualied bidder
(if any) with a payment equaling the bidders bid.
943
It is intuitive that bidders react to the competitors qualication probability in equilibrium. This reaction is captured
by the size of their price markup: For a xed
r reserve price
r (which is set by the buyer), the markup xi 1 F zi /
1 F xi n1 dzi decreases in .
Threshold Policy for Symmetric Prequalication Levels. With a restriction to symmetric prequalication levels, Proposition 3 still holds. We omit the proof of this;
however, the intuition is that the central trade-off (captured
by Equation (11)) remains with this restriction.
5. Numerical Illustrations
In this section, we illustrate the optimal qualication
screening strategy and show cost savings for a straightforward case in which the cost distribution is uniform and the
qualication level distribution has decreasing density. This
models a setting in which supplier cost types are evenly
dispersed and the buyer has an increasing marginal cost to
screen out each additional percent of unqualied suppliers.
We normalize the cost distribution F to U $500000
$1000000 to reect dollar values, and take H q = q

for tractability. We suppose that the buyer would like to
hold an auction with three bidders, and that the buyer has
a strict qualication requirement (q0 = 080) such that only
about 11% of all suppliers are truly qualied ( = 1
Hq0 = 11%, e.g., this could model the case in which
foreign suppliers are extremely unlikely to meet rigorous
qualication requirements set by the buyer). We say that an
asymmetric strategy allows bidders qualication probabilities to differ, whereas a symmetric strategy requires that
bidders have equal qualication probabilities.
Figure 1.
Illustration of regions where it is optimal to

use pre-only (A); a mix of pre- and postqualication (B); post-only (C); or forgo the auction (D). Panel (a) depicts regions in the
K Co plane. Panel (b) shows optimal bidder
qualication probabilities assuming the outside option cost is xed at Co = $1200000.
(a)
Qualfication cost K (in thousand dollars)
Proposition 4. Under symmetric prequalication, the

sealed-bid rst-qualied-price auction with reserve price
r = maxmin 1 Co a 1 0 implements the optimal mechanism from Proposition 1. In this auction, it is a
BayesianNash equilibrium for a bidder with cost xi r to
r
bid xi + xi 1 F zi /1 F xi n1 dzi .
200
K pre
K post
K nt
180
160
140
120
100
80
60
40
20
0
500
1,000
1,500
Outside option cost Co (in thousand dollars)

(b)
1.00
3*
0.14
0.10
56
1.00
2*
0.14
0.10
30.5
1.00
1*
0.14
0.10
17.5
K pre
170 183
K post K nt
Qualification cost K (in thousand dollars)
5.1. Optimal Qualication Probabilities

Asymmetric Strategy. Figure 1(a) illustrates Proposition 3: three thresholds K pre , K post , and K nt exist and are
increasing in the outside option cost Co . This gure depicts
the optimal qualication screening strategy: Pre-only is
optimal in region A, a mix of pre- and postqualication is
optimal in region B, post-only is optimal in region C, and
forgoing the auction is optimal in region D.
To further examine the optimal asymmetric qualication
screening strategy, we consider a specic outside option cost
of $1,200,000, which is 20% higher than the worst possible
supplier cost (e.g., this could model procurement with lowcost foreign suppliers). Figure 1(b) illustrates the switching
behavior of the optimal asymmetric qualication screening strategy for this outside option cost. When $17500 <
Notes. Plots assume n = 3 bidders, qualication threshold q0 = 080,

and supplier cost and qualication level distributions F U $500000
$1000000 and H q = q, respectively. In panel (b), K pre = $17500,

post
nt
K = $170000, K = $183000; furthermore, the buyer prefers to fully
prequalify three, two, or one bidder(s) if the qualication cost is between
zero and $17500, between $17500 and $30500, or between $30500 and
$56000, respectively.
K < $30500, the optimal qualication screening strategy

entails partially prequalifying one bidder but fully prequalifying the other two. Such a strategy balances the prequalication cost with the total postauction cost, given that it
saves prequalication cost (by postponing some qualication due diligence for one bidder) while guaranteeing that
the buyer nds a qualied bidder to contract with (by having two bidders fully prequalied). As the qualication cost
944
Figure 2.
Panel (a) plots the optimal symmetric bidder qualication probability for n = 3
bidders. Panel (b) plots the optimal number of bidders under symmetric qualication,
n n = 3, and the optimal symmetric qualication probability, , for n bidders.
(a) 1.0
K pre = $19,500
K post = $170,000
K nt = $183,000
0.9
0.8
0.7
* 0.6
0.5
0.4
0.3
0.2
0.1
0
K pre
K post K nt 200
100

(b) 3
n*
qualication strategy without restriction is not symmetric

for $17500 < K < $19500, whereas the optimal qualication strategy without restriction is symmetric for K near
K post ($170,000) and K nt ($183000). Figure 2(a) assumes
the buyer invites n = 3 bidders to the auction. With asymmetric prequalication thresholds, the buyer always nds
it optimal to invite up to n bidders; however, it turns out
that a buyer who is restricted to use symmetric prequalication might choose to invite n < n bidders to her auction. Figure 2(b) describes such a case: When $20000 <
K < $44000, the optimal number of bidders under symmetric prequalication, n = 2, is less than n = 3. In other
words, fully prequalifying two bidders is desirable, but fully
prequalifying a third bidder does not sufciently reduce
expected total costs to warrant the additional prequalication costs spent on the third bidder. Without the symmetric prequalication restriction, the buyer could simply use
postqualication on the third bidder; however, the buyer forgoes the third bidder if this bidder cannot be invited without
being prequalied to the same level as the other two bidders,
as is the case when restricting to symmetric prequalication. In summary, inviting fewer bidders is a realistic way in
which a buyer might cope with a symmetric prequalication
strategy restriction.
5.2. Cost Comparisons
Optimal Asymmetric Strategy vs. Pre-Only. Fully

prequalifying all bidders before the auction (pre-only) is
generally standard in practice. Let
1.0
*
0.5
pre-only strategy optimality gap
0.1
= 1
0
20
40
60
80
100
120
140
160
180
200

Note. All graphs are plotted against qualication cost, and assume outside option cost Co = $1200000, qualication requirement q0 = 08,
and supplier cost and qualication level distributions F U $500000
$1000000 and H q = q, respectively.
increases, it is optimal to postpone more due diligence to

save prequalication cost upfront. This explains why the
prequalication level for the partially prequalied bidder,
1 , decreases as K increases from $17500 to $30500.
However, when K exceeds $30500, having two bidders
fully prequalied is too expensive to be optimal. Instead, it
is optimal to fully prequalify only one bidder and partially
prequalify the other two when $30500 < K < $56000, and
it is optimal to partially prequalify all three bidders when
$56000 < K < $170000.
Symmetric Strategy. Figure 2(a) plots the optimal bidder qualication probability under symmetric qualication
( ). Threshold K pre in Figure 2(a) ($19500) is larger than
that in Figure 1(b) ($17500), and the other two thresholds
K post and K nt equal those in Figure 1(b) because the optimal
total cost with optimal asymmetric qualication screening

total cost with pre-only
(12)
In Figure 3(a), in addition to the previous scenario where
the buyer has a strict qualication requirement (q0 = 080),
we also examine the optimality gap for a scenario where the
buyer has a more lenient qualication requirement (q0 =
020). The maximal optimality gap is around 16% when
q0 = 080, and 30% when q0 = 020. For both qualication
requirement scenarios, the optimality gap peaks when qualication screening becomes too costly to make the auction
with pre-only worthwhile. To the left of this peak, the rate
of cost savings is small when K is small because the preonly strategy is (or is close to) optimal. To the right of this
peak, the rate diminishes as K increases to K nt because the
optimal qualication strategys total cost increases to Co ,
whereas the pre-only strategys cost is xed at Co for K
exceeding $40500 (when q0 = 08), or $120000 (when
q0 = 02).
When the qualication cost is not too expensive
($17500 < K < $40500), the optimality gap is greater
under the stricter qualication requirement. For example,
if supplier qualication costs $40000 (perhaps $20000 is
945
Figure 3.
Pre-only strategy optimality gap (%)
of the optimal qualication strategy to be sustained for

larger qualication costs K, eventually reaching 30%. The
relative cost savings are more quickly capped in the stricter
requirement case, because prequalication becomes prohibitively expensive for the buyer, who abandons the auction and resorts to the outside option at a smaller K.
n = 3, q0 = 0.2, Co = $1,200,000
Symmetric vs. Asymmetric Strategies. Given that a

symmetric prequalication strategy is fairer and simpler, a buyer might prefer it if it can closely approximate
the optimal asymmetric strategy. Let
n = 3, q0 = 0.8, Co = $1,200,000
30
20
10
17.5 25.0 39.5
120.0
200
100

(b)
Symmetric strategy optimality gap (%)
(a)
Panel (a) compares the optimal asymmetric

and pre-only qualication strategies costs (per
Equation (12)). Panel (b) compares the optimal asymmetric and optimal symmetric qualication strategies costs (per Equation (13)).
n = 3, q0 = 0.2, Co = $1,200,000
n = 3, q0 = 0.8, Co = $1,200,000
30
n = 8, q0 = 0.8, Co = $1,200,000
n = 3, q0 = 0.8, Co = $5,000,000
20
10
0
0
20
40
60
80
100
120
140
160
180
200

Note. All graphs are plotted against qualication cost, and assume
supplier cost and qualication level distributions F U $500000
$1000000 and H q = q, respectively.
spent to purchase and test supplier products, $15000 to

send buyer employees to inspect supplier facilities abroad,
and $5000 on time-intensive meetings with stakeholders
throughout the buyers company), about 154% of total
procurement costs are saved by postponing qualication
checks when q0 = 080, compared to about 34% when
q0 = 020. When q0 = 08, only about 11% of all suppliers
are truly qualied, making the full prequalication policy
very costly. Consequently, optimally postponing a portion
of qualication can be very benecial.
A more lenient qualication requirement of q0 = 020
makes prequalication more affordable and hence mitigates
the relative benets of aggressive postqualication. However, under the more lenient qualication requirement, the
affordability of pre-only allows the relative cost savings
symmetric strategy optimality gap

total cost with optimal asymmetric strategy
(13)
=1
total cost with optimal symmetric strategy
From Figure 3(b), we see that the optimal symmetric strategy approximates the optimal asymmetric strategy well
when the outside option cost Co , the number of bidders n,
and the qualication requirement q0 are small.3 In the range
of qualication cost K considered, this rate is no more
than 02% when Co = $1200000, n = 3, and q0 = 02.
The maximal rate observed is about 5% 7% when q0
increases to 08. However, the symmetric strategy approximates the optimal asymmetric strategy poorly when the outside option cost is large and qualication is expensive, but
not prohibitively so (making an auction still worthwhile).
Intuitively, a buyer will always nd it prudent to heavily use
prequalication to avoid the downside risk of an onerous
outside option. However, a buyer who can employ asymmetric prequalication can also enjoy the upside benet
of potentially reduced contracting costs from speculatively
inviting long shots to the auction, bidders admitted with
much or all of their qualication screening postponed. For
example, the optimality gap can reach about 32% when
Co increases to $5000000 in Figure 3(b). Therefore, in a
scenario where the buyer cannot produce in-house and the
contract is extremely importantfor example, u vaccine
procurement by a governmentthe buyer could be much
better off with the exibility to apply asymmetric prequalication thresholds to bidders, to simultaneously avoid the
downside risk of nontransaction while still enjoying speculative benets from casting a wider net for bids.
In summary, regardless of whether the buyer uses asymmetric or symmetric prequalication, these results suggest
that the buyer should seriously consider postponing some
or all of the supplier qualication process, especially when
the outside option cost is high, and either the qualication
requirement is strict and the qualication cost is moderate,
or the qualication requirement is lenient but qualication
is expensive.
6. Practical Considerations and

Extensions
6.1. A More General Supplier Cost Model
Our analyses up to this point have assumed that suppliers
can privately and perfectly observe their own costs prior
946
to the auction and that these costs are statistically independent. However, one can easily imagine a situation in which
suppliers costs are related due to similar or shared cost
drivers. Moreover, suppliers might form a rough estimate
of their cost on their own, and utilize information from
competing suppliers bids during the auction to update this
estimate. In this subsection, we discuss how 4s results
can be adapted to a more general cost model that allows
for these possibilities, culminating in Proposition 5.
Suppose that each supplier i possesses private information about his idiosyncratic status, for example, production
technology, inventory level, order backlog, etc., which is
denoted by $i 0 1. Supplier is cost to fulll the contract
is then modelled by a function xi % X$i c %,
where $1 $n represents the collection of all bidders private information, c characterizes common cost
factors that depend on , and % characterizes exogenous
and publicly observable factors that could impact costs
directly (via X) and/or indirectly through affecting the
distribution of bidders private information (the $i s). We
assume that xi is strictly increasing in $i .
As an illustration, consider the following simple model
in which costs are correlated and suppliers adjust their
cost estimates after
seeing competitors bids. Let X$i
c % = $i + j e$j + %, where e 0 1 is an
increasing function. This model is essentially that which
appears
in Myerson (1981) without the term %. The summation j e$j could represent cost estimate revisions
based on anticipated demand in the supply base for specialized components, with suppliers private information (the
$i s) incorporating their own anticipated component demand
based on their on-hand component inventory and current
order backlog. In a similar vein, % could reect prevailing market prices for commodity inputs whose prices are
unaffected by the suppliers aggregate usage, being only
a negligible fraction of the commodities overall demand.
Although this illustrative example includes the exogenous
factor % in a simple additive fashion, the general cost model
can handle richer structures; for example, suppliers costs
can depend on % in a nonadditive way in order to model
heterogeneous efciencies in using commodity inputs. We
now return to describing the general model.
So that suppliers are ex ante symmetric, we assume
that c is exchangeable in all elements of , and $i has
a commonly known conditional distribution F % with
a positive and continuous density f %. Furthermore,
we assume that $i s are conditionally independent given
any xed publicly observable %. Hereafter, for notational
simplicity, we will suppress the variable % when writing
F , f , X, and xi s. By treating % as implicit, the general
supplier cost formulation introduced two paragraphs above
reduces to the formulation of correlated bidder costs in
Branco (1997).
With this cost model, our analysis of the optimal auction
and postqualication mechanism needs one key modication, which is to redene the virtual cost as a function of

the collection of all bidders private information and the

exogenous factor, that is,
F $i )xi

i xi +
f $i )$i
In the electronic companion, we derive the optimal mechanism assuming that (i) ) i /)$i > 0 for all i at all ;
and (ii) ) i /)$i > ) j /)$i for all i and j = i at all .
Whereas (i) ensures that bidder is virtual cost i strictly
increases in his type, (ii) ensures that, although increases
in is type can also increase the virtual costs of j = i,
such increases are smaller than the increase caused to is
own virtual cost. Condition (i) is satised, for example, if
xi is increasing and convex in $i , and F $i /f $i is
increasing in $i . Condition (ii) is then satised, for example, if )xi /)$i > )xj /)$i and ) 2 xi /)$i )$j 0.
The insight is that under conditions (i) and (ii) the redened
virtual cost i under the general cost model inherits a
key property of the virtual cost xi under the independent
private cost model, that is, any increase (decrease) of bidder is report to the buyer encourages the buyer to demote
(promote) bidder i in the postqualication sequence.
Under conditions (i) and (ii) on the general cost model,
it remains optimal to sequence bidders for postqualication according to ascending value of their adjusted virtual
costs i + ai , as we show in Proposition 5 below.
As before, the contract winner is the rst bidder (if any)
deemed qualied by postqualication. Furthermore, echoing Equation (10), incentive compatibility can again be
ensured by paying the contract winner, say bidder i, according to the amounts i could have increased his bid by and
still been preferred to the t bidders sequenced after him for
postqualication, and the reserve price. However, there is
one slight difference when actually computing the payment
under the general cost model: Because an increase in is
bid can affect other bidders virtual costs, when computing
the payment for bidder i, the buyer must take into account
how the postqualication sequence of the t bidders would
have been reshufed if i had increased his bid.
We now dene the notation needed to generalize Equation (10). Let i be the vector excluding the ith element,
and dene ri = maxminz 1 0, where z is such that
i z i + ai = Co . Let zij be such that i zij i +
ai = ij zij i + aij and zi1 zi2 zit ,
where ij (j = 1 t) are the indices of the t bidders such
that i + ai ij + aij and zij $i ri . Note
that zij and ri , representing the maximal amount that bidder i could have bid and still been preferred to bidder ij
and the outside option, respectively, are well dened due
to conditions (i) and (ii).
Proposition 5. Under the general cost model described
above, the optimal auction and postqualication mechanism proceeds as follows: only postqualify a bidder i if
i +ai Co ; sequence all such bidders for postqualication according to ascending value of i + ai ,
947
breaking ties randomly; award the contract to the rst

bidder i (if any) deemed qualied by postqualication with
a payment of
t1
j=1
xi zij i ij
t

1 ik
k=1
j1
1 ik + xi ri i
k=1
if t 1
or xi ri i
if t = 0
and pay nothing to all other bidders. Additionally, Proposition 3 characterizing the optimal qualication strategy
continues to hold.
The key reason that Proposition 3 continues to hold
under the general cost model is that the central trade-off
described in Equation (11) does not rely on the format of
the bidders cost function xi . In summary, Proposition 5
implies that the main results of 4the optimal mechanism (Proposition 1) and qualication thresholds (Proposition 3)continue to hold under the general cost model
described above, albeit with a bit more care needed when
calculating the contract winners payment to account for
correlation among bidders costs.
6.2. Value of Credible Reserve Price
Our optimal mechanism derivation in 4.2 assumed that the
buyer could credibly commit to not awarding the contract
to any bidder bidding above the reserve price set for him.
As Milgrom (1987) points out, an auctioneer who cannot
credibly commit to throwing away bids between the reserve
price and the auctioneers own valuation is disadvantaged:
The auctioneer cannot achieve the optimal ex ante expected
prots because bidders will ignore the announced reserve
price.
Postqualication of bidders places additional importance on reserve price credibility. Proposition 1 characterizes the optimal reserve price for bidder i as ri = max
min 1 bimax 1 0, where bimax Co ai is the
maximum bid from bidder i that the buyer would nd profitable to postqualify. Because x > x, we can have ri <
bimax ; i.e., in the optimal mechanism the buyer promises to
ignore bidder is bid if it is in ri bimax , even though she
actually would nd it protable to postqualify such a bid
after the auction.
If this promise is uncredible, the buyer is forced to set
bimax as the reserve price for bidder i and cannot apply an
optimal mechanism (optimal reserve prices). This is particularly problematic when bimax is greater than one, the
upper bound of supplier cost. If all other bidders fail postqualication and bidder i is the last remaining bidder, bidder i can make a take-it-or-leave-it offer of bimax to the
buyer. For example, suppose that all bidders have a symmetric qualication probability equal to . Per Proposition 4, a
fully credible buyer would nd it optimal to run a sealed-bid
rst-qualied-price auction with an optimal reserve price r
(recall 1 r). If the buyers lack of credibility forces her to

instead use b max = Co a > 1 as a reserve price, a cost
type 1 bidder i expects to earn b max 11 n1 from
the auction. In contrast, note that when = 1 for all bidders, as is assumed in classical auction theory, cost type 1
bidders expect to earn zero prots in the auction, provided
the auction has at least two bidders.
Thus, two factors inate the costs of an uncredible buyer
in our setting: forgone price discrimination opportunities,
as seen in classical auction theory; and the being held
hostage by the last remaining bidder effect, which allows
even a worst-type bidder to expect positive prots from
the auction and is a consequence of postqualication that
to our knowledge is new to the literature. Continuing the
symmetric s example above, when Co a > 1 = r,
only the being held hostage effect exists; when 1 Co
a > r, only the price discrimination effect exists; and
when Co a > 1 > r, both effects exist.
These observations suggest that postqualication should
be used carefully when the buyer has little or no negotiating
clout with suppliers and must rely solely on competition
among suppliers for price concessions. In such situations
a supplier can command a very high price in a one-onone negotiation with the buyer, a damaging scenario for the
buyer that is risked by postqualication. Intuitively, this can
push the buyer to employ more stringent prequalication
screening. On the other hand, the buyer is between a rock
and a hard place if qualication screening is prohibitively
expensive, in which case she might nd it too costly to
prequalify extensively enough to avoid getting squeezed by
the suppliers. For a simple illustration of both of these possibilities, consider the setup of Figure 2. For qualication
cost $30000, the optimal symmetric qualication probability for a credible buyer is 0.66, which is lower than 0.93,
the symmetric qualication probability that can be (using
numerical analysis) shown to be optimal for an uncredible
buyer (who uses a sealed-bid rst-qualied-price auction
with reserve price Co a). However, at a higher qualication cost of $45000 the opposite happens: The credible
buyers optimal symmetric qualication probability of 0.27
exceeds 0.11, the optimal probability for the uncredible
buyer. Although we will leave a more complete exploration
of credibility and qualication screening interactions to
future work, the above suggests that despite the importance
of credibility in avoiding cost ination under postqualication, even an uncredible buyer might prefer heavy use of
postqualication if qualication is very expensive.
7. Conclusions
When issuing an RFQ for competitive bid, nding a supplier truly qualied to fulll the contract is often as
important as price concerns. Costly supplier qualication
processes are virtually ubiquitous in industry to help buyers
proactively avoid problems and expenses associated with
supplier nonperformance, e.g., recalls and product liability
948
issues. This paper explicitly models and suggests optimal
policies for both the supplier qualication and competitive
price negotiation processes together, and to our knowledge
is the rst study of optimal supplier qualication processes
in the operations management and auction-theoretic literatures. To save on total supplier qualication and contracting costs, we allow the buyer to delay all or part of the
qualication process until after the competitive price negotiation (an auction) and then home in on the lowest (virtual adjusted cost) bidders. Although delaying qualication
is not to our knowledge common practice in industry, our
study provides a mathematical framework that suggests that
such a postqualication stage can indeed be benecial.
In particular, we nd that the prequalication-only strategy is optimal solely when supplier qualication is relatively cheap. Because postponing qualication means that
some attractive bids in the auction may be disqualied, it
makes sense to completely prequalify suppliers if doing so
is cheap. However, for moderate-sized qualication costs
the buyer can do much better if some costly qualication
is delayed until after the auction because reduced qualication costs with judicious postqualication can more than
offset expected increases in the contracting costs (as determined by our auction-theoretic analysis). Figure 3(a) shows
total (qualication plus procurement) cost savings of about
3% 15% for a contract worth $12 M to the buyer when
qualifying a supplier costs $40,000, supplier cost types
are evenly dispersed, and the qualication level distribution has a decreasing density. More generally, Proposition 3
partitions the two-dimensional qualication cost (K) and
outside option cost (Co ) plane into regions where either
traditional prequalication only, our novel postqualication
only, or our novel mix of the two are optimal (illustrated
in Figure 1).
Although operations management analyses such as ours
may merely galvanize a reconsideration of current procurement policies, as supply chains lengthen and supply sources
become globalized and more varied, the increase in potential
new suppliers and the growing number of RFQ events could
make the standard prequalication-only strategy prohibitive
for resource-constrained procurement departments that cannot possibly fully prequalify all suppliers invited to all bidding events. Postqualication might eventually be used out
of sheer necessity to accommodate constrained qualication
resources, but fortunately our study shows that postqualication can be part of an optimally balanced supplier qualication strategy even without such resource constraints.
Our study is built on classical auction theory, and extends
the auction theory literature by developing an optimal auction and postqualication mechanism. We characterize the
optimal mechanism in Proposition 1 and propose both
sealed-bid and open-bid formats to implement it, including Proposition 4 showing that a simple variant of a standard sealed-bid auctiona sealed-bid, rst-qualied-price
auctionimplements the optimal mechanism when all bidders are prequalied up to the same level.

The spirit of our results are broad in the sense that they
characterize the trade-off between costly prequalication
and increased likelihood that an attractive bid will be tenable (the bidder is truly qualied for the contract). For general auction formats not studied in this paper, we expect
that our main insight that postqualication can be an effective cost-reduction strategy for the buyer will continue to
hold. However, for auction mechanisms with extremely rich
strategy spaces, postqualication can add additional complexities to the bidding equilibrium analysis. For example,
were postqualication used in conjunction with a standard
reverse English auction, bidders may nd it optimal to drop
out of the auction before reaching their true costs in a way
that intricately depends on all previous bids in the auction
up to that point. Introducing postqualication in various
auction formats presents many opportunities for research
on competitive bidding.
Section 6 discusses several possible extensions to our
model, which although fairly general does make some
important assumptions to keep the analyses focused and
tractable. Other extensions are also possible. This paper
shows that the buyer can protably postpone qualication and employ asymmetric prequalication policies, even
under our assumption that suppliers are ex ante symmetric.
The ex ante asymmetric suppliers case, in which H , K,
and F are allowed to be supplier specic, is an interesting
possible direction for future work; although the analysis
would be more complex, we suspect that the main insights
of this paper would be preserved because ex ante supplier
asymmetry would likely make our asymmetric prequalication screening policies even more tting.
In this paper, we also assume that suppliers are privately
informed about their costs, and the buyer and each supplier
are equally unsure of the suppliers likelihood of meeting
the buyers qualication requirements. One interesting and
challenging direction for future work is the multidimensional asymmetric information case in which superior prior
knowledge of the qualication level is held on the supplier
side. The ensuing optimal procurement mechanism analysis
would involve information asymmetry that is multidimensional, a setting that is generally seen (e.g., Rochet and
Stole 2003) as a current challenge in mechanism design
research.
Our analysis assumes an innite supplier pool, which
ensures that the buyer is able to ll the auction. However,
in practice the number of suppliers who show interest in the
RFQ is nite and could be small for specialized purchases.
With a nite supplier pool, the number of suppliers passing
prequalication and entering the auction could be nondeterministic. Presumably the buyer could dynamically adjust the
prequalication threshold per the remaining supplier pool
size. We leave such extensions to our future work.
8. Electronic Companion
An electronic companion to this paper is available as part
of the online version that can be found at http://or.journal.
informs.org/.

Endnotes
1. Because the entire notion of being qualied is based
on the buyers specic requirements for the contract up
for bid, being deemed qualied does not necessarily imply
absolutely zero risk of nonperformance. For example, when
purchasing a simple, cheap, and noncritical indirect good
(such as ofce supplies), a buyer might be satised with
relatively light screening of the suppliers status and abilities prior to contracting.
2. Although our analyses assume that the outside option
cost is nite, if Co were instead allowed to be innite (to
capture a case in which the buyer absolutely must transact),
it is straightforward to see that the buyer would nd it
optimal to fully qualify at least one bidder in the auction to
eliminate the possibility of nontransaction.
3. Note that in calculating the optimality gap in Figure 3(b),
we have included the symmetric strategys exibility to optimally invite fewer than n bidders.
Acknowledgments
The authors are grateful to the associate editor and three
referees for valuable comments that greatly improved this
paper. They also thank Abigail Hohner of Hohner and Associates (formerly in Mars, Inc. procurement research group)
for helpful discussions, and seminar participants at the Ross
School of Business, Rotman School of Management, and
Amazon.com for their insightful comments and suggestions.
Helpful input on an early draft of this paper was provided
by students in the Fall 2005 OMS 899 course at Ross.
References
Ausubel, L. M., P. Cramton. 2006. Dynamic auctions in procurement.
N. Dimitri, G. Piga, G. Spagnolo, eds. Handbook of Procurement.
Cambridge University Press, Cambridge, UK.
Bagnoli, M., T. Bergstrom. 2005. Log-concave probability and its applications. Econom. Theory 26 445469.
Beil, D. R., L. M. Wein. 2003. An inverse-optimization-based auction
mechanism to support a multiattribute RFQ process. Management
Sci. 49 15291545.
Branco, F. 1997. The design of multidimensional auctions. RAND J.
Econom. 28 6381.
Braynov, S., T. Sandholm. 2003. Auctions with untrustworthy bidders.
Proc. IEEE Conf. Electronic Commerce, Newport Beach, CA, 2427.
Calveras, A., J.-J. Ganuza, E. Hauk. 2004. Wild bids. Gambling for resurrection in procurement contracts. J. Regulatory Econom. 26 4168.
Carr, S. M. 2003. Note on online auctions with costly bid evaluation.
Management Sci. 49 15211528.
Che, Y.-K. 1993. Design competition through multidimensional auctions.
RAND J. Econom. 24 668680.
Chen, F. 2007. Auctioning supply contracts. Management Sci. 53
15621576.
949
Dasgupta, P., P. Hammond, E. Maskin. 1979. The implementation of social
choice rules: Some general results on incentive compatibility. Rev.
Econom. Stud. 46 185216.
Elmaghraby, W. J. 2000. Supply contract competition and sourcing policies. Manufacturing Service Oper. Management 2 350371.
Feinberg, F. M., J. Huber. 1996. A theory of cutoff formation under imperfect information. Management Sci. 42 6584.
Gibbard, A. 1973. Manipulation of voting schemes: A general result.
Econometrica 41 587601.
Green, J., J. J. Laffont. 1977. Characterization of satisfactory mechanisms
for the revelation of preferences for public goods. Econometrica 45
427438.
Harstad, R. M., J. H. Kagel, D. Levin. 1990. Equilibrium bid functions
for auctions with an uncertain number of bidders. Econom. Lett. 33
3540.
Jap, S. D. 2003. An exploratory study of the introduction of online reverse
auctions. J. Marketing 67 96107.
Krishna, V. 2002. Auction Theory. Academic Press, San Diego.
Leenders, M. R., P. F. Johnson, A. E. Flynn, H. E. Fearon. 2006. Purchasing
and Supply Management, 13th ed. McGraw-Hill/Irwin, New York.
McAfee, R. P., J. McMillan. 1987. Auctions with a stochastic number of
bidders. J. Econom. Theory 25 119.
McAfee, R. P., J. McMillan. 1988. Search mechanisms. J. Econom. Theory
44 99123.
Milgrom, P. 1987. Auction theory. T. Bewley, ed. Advances in Econometrics: Fifth World Congress. Cambridge University Press, Cambridge,
UK, 132.
Milgrom, P. 2004. Putting Auction Theory to Work. Cambridge University
Press, Cambridge, UK.
Myers, R. 2007. Food ghts. CFO Magazine 23 7074.
Myerson, R. B. 1981. Optimal auction design. Math. Oper. Res. 6 5873.
Rochet, J.-C., L. A. Stole. 2003. The economics of multidimensional
screening. M. Dewatripont, L. P. Hansen, S. J. Turnovsky, eds.
Advances in Economics and Econometrics: Theory and Applications,
Eighth World Congress, Vol. 1, Chapter 5. Cambridge University
Press, Cambridge, UK.
Rothkopf, M. H. 1991. On auctions with withdrawable winning bids. Marketing Sci. 10 4057.
Snir, E. M., L. M. Hitt. 2003. Costly bidding in online markets for IT
services. Management Sci. 49 15041520.
Spencer, J., N. Casey. 2007. Toy recall shows challenge China poses to
partners. Wall Street Journal (August 3) A1.
U.S. Department of Commerce, Bureau of the Census. 2005. Statistics for
Industry Groups and Industries: 2004, Annual Survey of Manufactures. M04(AS)-1, pp.1155, and Appendix at http://www.census.gov/
prod/2005pubs/am0431gs1.pdf.
Waehrer, K. 1995. A model of auction contracts with liquidated damages.
J. Econom. Theory 67 531555.
Welch, D. 2007. Made in China: Faulty tiresHow a Chinese suppliers bad
decision turned into one importers worst nightmare, and may mean the
end of his business. Business Week (July 12) http://www.businessweek.
com/bwdaily/dnash/content/jul2007/db20070711_487422.htm.
Ye, L. 2007. Indicative bidding and a theory of two-stage auctions. Games
Econom. Behav. 58 181207.
Zheng, C. Z. 2001. High bids and broke winners. J. Econom. Theory 100
129171.

RFQ Auctions With Supplier Qualification Screening

Hochgeladen von

Dokumentinformationen

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

RFQ Auctions With Supplier Qualification Screening

Hochgeladen von

Copyright:

Verfügbare Formate

This article was downloaded by: [14.139.128.

19] On: 28 January 2016, At: 20:50

RFQ Auctions with Supplier Qualification Screening

To cite this article:

Vol. 57, No. 4, JulyAugust 2009, pp. 934949

RFQ Auctions with Supplier Qualication Screening

for lengthy global supply chains, while price pressures and

Wan and Beil: RFQ Auctions with Supplier Qualication Screening

Operations Research 57(4), pp. 934949, 2009 INFORMS

paper analyzes how much costly supplier qualication

Wan and Beil: RFQ Auctions with Supplier Qualication Screening

Wan and Beil: RFQ Auctions with Supplier Qualication Screening

Operations Research 57(4), pp. 934949, 2009 INFORMS

that the buyer must purchase to run a test sample. A more

the buyer to consider them qualied for contract award.

Wan and Beil: RFQ Auctions with Supplier Qualication Screening

and postqualication process when suppliers are possibly

Wan and Beil: RFQ Auctions with Supplier Qualication Screening

Operations Research 57(4), pp. 934949, 2009 INFORMS

or to the buyers outside option. The buyer designs this

(incentive compatibility) (3)

Ui xi  is the expected prot of bidder i in equilibrium.

Solving for mi , the buyers expected payment to bidder i

Wan and Beil: RFQ Auctions with Supplier Qualication Screening

Operations Research 57(4), pp. 934949, 2009 INFORMS

The value xi , which equals bidder is true cost, xi , plus

subject to mi xi  = i xi xi +

and verify that at this solution i zi  is decreasing in zi

 1 xij  + aij  ai ij

where ij (j = 1     t) is the index of the jth bidder out

Wan and Beil: RFQ Auctions with Supplier Qualication Screening

Operations Research 57(4), pp. 934949, 2009 INFORMS

bidder i in the qualication list, forgo qualifying bidder i.

the rst bidder that successfully passes postqualication.

(This result is proved formally in the electronic companion.)

Wan and Beil: RFQ Auctions with Supplier Qualication Screening

Wan and Beil: RFQ Auctions with Supplier Qualication Screening

Operations Research 57(4), pp. 934949, 2009 INFORMS

$1000000 to reect dollar values, and take H q = q

Illustration of regions where it is optimal to

Proposition 4. Under symmetric prequalication, the

Outside option cost Co (in thousand dollars)

Qualification cost K (in thousand dollars)

5.1. Optimal Qualication Probabilities

Notes. Plots assume n = 3 bidders, qualication threshold q0 = 080,

$1000000 and H q = q, respectively. In panel (b), K pre = $17500,

K < $30500, the optimal qualication screening strategy

Wan and Beil: RFQ Auctions with Supplier Qualication Screening

Operations Research 57(4), pp. 934949, 2009 INFORMS

Qualification cost K (in thousand dollars)

qualication strategy without restriction is not symmetric

Optimal Asymmetric Strategy vs. Pre-Only. Fully

pre-only strategy optimality gap

Qualification cost K (in thousand dollars)

$1000000 and H q = q, respectively.

increases, it is optimal to postpone more due diligence to

total cost with optimal asymmetric qualication screening

Wan and Beil: RFQ Auctions with Supplier Qualication Screening

Operations Research 57(4), pp. 934949, 2009 INFORMS

Pre-only strategy optimality gap (%)

of the optimal qualication strategy to be sustained for

Symmetric vs. Asymmetric Strategies. Given that a

17.5 25.0 39.5

Qualification cost K (in thousand dollars)

(incentive compatibility) (3)

Ui xi is the expected prot of bidder i in equilibrium.

The value xi , which equals bidder is true cost, xi , plus

subject to mi xi = i xi xi +

and verify that at this solution i zi is decreasing in zi

1 xij + aij ai ij

where ij (j = 1 t) is the index of the jth bidder out

$1000000 to reect dollar values, and take H q = q

Notes. Plots assume n = 3 bidders, qualication threshold q0 = 080,

$1000000 and H q = q, respectively. In panel (b), K pre = $17500,

K < $30500, the optimal qualication screening strategy

$1000000 and H q = q, respectively.

$1000000 and H q = q, respectively.

spent to purchase and test supplier products, $15000 to