You are on page 1of 33

Sourcing for Quality: Incentives, Cooperation, and Inspection

Hsiao-Hui Lee Cuihong Li


Faculty of Business and Economics, University of Hong Kong, Pokfulam Road, Hong Kong
School of Business, University of Connecticut, Storrs, Connecticut, 06269, USA
hhlee@hku.hk Cuihong.Li@business.uconn.edu

November 2012

Buyers can manage product quality sourced from suppliers in several ways: they can improve
the incoming quality directly by cooperating with suppliers to improve a process or product, they
can improve the incoming quality indirectly by incentivizing suppliers quality-improvement efforts,
or they can control the outgoing quality by inspecting incoming units. In this paper, we study the
buyers use of these three strategiescooperation, incentivization, and inspectionto manage sourced
quality. Specifically, we consider a general effect relationship between the buyer and supplier
efforts, allowing for substitutable or complementary quality-improvement effects of these efforts.
Our results establish a strategic relationship between these efforts, which may differ from their
effect relationship due to the inclusion of endogenous inspection. We also identify the ways in which
cooperation influences the incentive structure, which may drive cooperation positively or negatively
depending on the effect relationship of efforts. Finally, we compare the results with observable and
unobservable supplier efforts, characterizing the impact of supplier effort observability.
Keywords: Quality, Incentives, Inspection, Cooperation, Supply chain

1.

Introduction

Firms are increasingly challenged to manage the quality of their suppliers products, as manufacturing outsourcing has become a common practice for reducing costs and improving core competencies.
For example, most product recall actions in the automobile industry can be traced back to the outsourced parts deficiencies; of the top ten recall actions worldwide in 2002, nine were related to
1
Electronic copy available at: http://ssrn.com/abstract=2191754

deficiencies in the suppliers parts (Maurer et al. 2004). In recent years, concerns over quality have
been heightened by a list of high-profile quality scandals that were linked to the suppliers faults;
examples include scandals related to toys (Spencer and Casey 2007), tires (Welch 2007), pet food
(Myers 2007), and medicine (Layton 2010). In this paper, we study three important strategies that
a buyer may take to manage the quality of sourced products.
To ensure that a supplier delivers high quality, a buyer could provide the supplier with appropriate incentives. The supplier effort to improve quality (e.g., investing in quality control processes,
adopting superior production systems and materials, and hiring skilled workers) is usually hard to
verify by a third party (e.g., the court) and thus cannot be regulated by a contract. Therefore,
sufficient monetary incentives are necessary to motivate the supplier effort. In particular, squeezing
the suppliers margin may backfire, as the supplier may have to retreat to low-quality solutions in
order to meet the pressure to contain costs (Knowledge@Wharton 2009). For example, in the controversy surrounding Mattels Chinese-made toys, although the suppliers were blamed for resorting
to paints containing lead, many believed that the purchasers must have shouldered some of the
blames for aggressively pushing down the prices. The low price was a driving force behind the
suppliers use of low-quality materials, as in the end, they (purchasers) [would] get the quality of
work they [were] willing to pay for (Knowledge@Wharton 2007, 2009).
Instead of solely relying on the supplier, the buyer could directly cooperate with the supplier to
ensure high quality. For example, the buyer may send her own engineers to the suppliers site to
help solve technical problems, assign personnel to work with the supplier to improve his internal
operations, or provide education and training to the suppliers employees on quality management
(Sako 2004; Liker and Choi 2004; Maurer et al. 2004). Buyer cooperation is especially a worthy
effort when the supplier can deliver significant cost advantages, and yet still fall behind in quality
standards (e.g., many suppliers in developing countries). In hindsight of several Chinese-made
product quality scandals, practitioners realized that a key to addressing the quality challenge was
to invest in the suppliers to improve their quality (Knowledge@Wharton 2007, 2009).
Both the buyers cooperation effort and the suppliers own improvement effort lead to better
quality, and their effects may interact with each other. Specifically, the two efforts can be complements or substitutes, if the presence of one increases or decreases the effect of the other in the
quality improvement. For example, the buyers engineering support that improves the production
process and solves quality problems for the supplier reduces the need of the suppliers own effort
to address these issues; thus, the buyers support may be considered as a substitute to the supplier
effort. Improving the product design to reduce defect rates can also be a substitute for the sup2
Electronic copy available at: http://ssrn.com/abstract=2191754

plier effort, as a better design makes the product quality more robust to the production process
(Boothroyd et al. 2002). On the other hand, education and training programs transfer knowledge
on quality management for the suppliers employees instead of directly assisting the production;
such efforts may be considered as complements, since the information and skills acquired from such
programs allow the supplier to better deploy management and engineering resources to enhance
quality.
While cooperation and incentives could directly and indirectly improve the quality of incoming
products respectively, the buyer could also control the outgoing product quality by inspecting the
incoming units to identify defects before further processing and sales. Defects discovered after
sales could be extremely expensive; in addition to the direct material and labor costs, they may
entail items such as warranty services, liability charges, product recalls, and loss of customer loyalty.
Therefore, to prevent external failures, the buyer must identify problems internally with inspection.
For example, Foreign Tires Sales, the tire importer, did not have a formal schedule of tire testing
before it was requested to initiate a $90 million recall due to the faulty tires it sourced. After
the incident, it started to test one tire per month, despite the expensive inspection costs at about
$40,000 per tire (Welch 2007). Likewise, Mattel stepped up inspections after the massive Chinesemade toy recall in 2007, in spite of the operation delays and lowered sales caused by product testing
(Burke 2007).
In addition to preventing external failures, incoming inspection also helps to hold suppliers
responsible for quality defects they have caused. Without inspection, the buyer may fail to detect
the defects immediately upon delivery, which could allow the supplier to escape responsibility
for problems discovered later. Specifically, to produce the final product the buyer may need to
process and assemble parts from different suppliers that jointly influence the product function, and
hence it is often difficult to identify the root cause of any defect and attribute responsibility to
individual parties if any quality deficiency is found in the final product (Baiman et al. 2001). Also,
for quality problems reported by customers long after the suppliers delivery, tracing the quality
problem backward in the supply chain can be difficult because of the lack of visibility of material
flows.1 Further, even if the buyer can identify the supplier responsible for the defects, pursuing
1

In the Foreign Tires Sales (FTS) incident, FTS took six weeks to locate a tire imported from the supplier in FTSs

distributors facilities, before FTS could further examine and identify the quality problem (Welch 2007). As another
example, in 2007 some California grocery stores discovered that the gingers they sold contained a dangerous pesticide,
but the provider of the tainted ginger was never identified due to the complexity of the supply chain (Zamiska and
Kesmodel, 2007).

compensation from the supplier is not easy due to the complexity of the legal process and financial
constraints of the supplier, especially when the supplier is small and resides in a foreign country
(Babich and Tang 2012; Sherefkin and Armstrong 2003).
In this paper, we develop a model to understand the holistic use of incentives, cooperation, and
inspection for managing quality sourced from suppliers. In this model, we consider a supply chain
with a buyer and a supplier. The supplier produces a product for the buyer, but the production
process can yield nonconforming units. The conforming probability is determined jointly by the
buyers cooperation effort and the suppliers quality-improvement effort. The buyer inspects the
product upon delivery, but the inspection may not identify a defective product successfully; the
inspection accuracy (i.e., probability of successful inspection) is chosen by the buyer. In reality, a
buyer may observe a suppliers effort, such as the process adopted or material used, with careful
review, auditing, or site visits, even though the supplier effort is non-contractible. In order to
understand how such activities to observe supplier effort affect the buyers strategy and profits, we
consider two situations in which the buyer can and cannot observe the supplier effort before she
decides the inspection accuracy. We summarize our main findings below.
First, we reveal the strategic relationship between the buyer and supplier efforts. The two
efforts may shift in similar or opposite directions as the cost of either changes, making them
strategic complements or substitutes. Interestingly, the two efforts may serve as complementary
strategies even though they are substitutable on quality-improvement effects, for as higher quality
implies less need of inspection, the buyer and supplier efforts jointly reduce the inspection cost by
improving quality.
Second, we establish the way in which the buyers cooperation strategy influences the incentive
structure. Because the buyers cooperation effort affects the marginal quality improvement achieved
by the suppliers own effort, the buyer effort moderates the incentive price needed to motivate the
supplier effort. Depending on the effect relationship between the two efforts (complements or
substitutes), the buyer effort may decrease or increase the incentive price. Therefore, when the
supplier effort becomes more expensive, which in turn increases the incentive cost, the buyer may
respond by enhancing or reducing cooperation. In addition, increasing external failure costs can
also change the buyer cooperation effort in either direction, as it reduces the friction associated
with the incentive price.
Third, we identify the impact of supplier effort observability. When the buyer cannot observe
the supplier effort, she cannot make the inspection decision responsive to the supplier effort choice.
Thus, compared to cases with observable supplier effort, not observing supplier effort makes it more
4

difficult to motivate supplier effort, as it requires a greater incentive price. Moreover, it introduces
a negative effect of cooperation on the supplier incentive, as cooperation improves the product
quality, thereby reducing inspection that the supplier faces and thus further raising the incentive
price needed to motivate the supplier. As a result, observing supplier effort not only leads to higher
supplier effort choice (with the incentive provided by the buyer), but also increases the preference
for cooperation in many cases. Considering the impact on the buyer profit, we find that observing
the supplier effort is most beneficial when the supplier effort cost is neither too high nor too low;
moreover, there may exist multiple ranges of supplier effort costs, corresponding to different buyer
effort choices, that justify the investment in monitoring the supplier effort.

2.

Literature Review

The literature on quality management in a decentralized supply chain has typically focused on
contract design and inspection decisions in a moral hazard setting concerning supplier qualityenhancement activities. Among these papers, Baiman et al. (2000) analyze a variety of scenarios
in which players actions or product failures may or may not be contractible, Hwang et al. (2006)
compare supplier appraisal and certification, Taylor and Wiggins (1997) investigate the relationship
between lot sizing and inspection in multi-period quality control, and Babich and Tang (2012)
compare the inspection mechanism with deferred payments, which is extended by Rui and Lai
(2012) to a multi-unit setting. While these papers all assume that the product quality is fully
determined by the supplier effort, we differ from this literature by considering buyer effort in
supplier quality improvement, which has a profound impact on contract design and inspection
decisions. In addition, we incorporate supplier effort observability to reveal its impact on the
optimal sourcing strategy.
Authors of other papers consider the buyer as a manufacturer performing part of the production
process; in these papers, suppliers and the buyers activities jointly determine the quality of the
final product. Baiman et al. (2001) compare two product architectures that make the buyers activity coupled with or separable from the suppliers activity. Balachandran and Radhakrishnan (2005)
focus on the penalty design for the supplier based on the internal or external failure information.
Chao et al. (2008) examine cost sharing of product recalls with a root-cause analysis. Dong et al.
(2011) compare the choice between supplier inspection and quality audit as mechanisms to control
quality in dyadic and multi-level supply chains. In these papers, the manufacturers activity affects
the external-failure probability but not the internal-failure one of the product delivered by the

supplier. Our paper differs in that the buyer effort improves the quality that the supplier delivers,
thereby affecting both internal and external failures.
Buyer investment in supplier quality improvement has received increasing attention in the recent
supply chain management literature. In particular, Zhu et al. (2007) incorporate both the buyer and
supplier investment in quality improvement, and show that the buyer cannot cede the responsibility
of quality improvement to the supplier in many cases. In their model, the buyer and supplier
efforts are substitutes, and no inspection is used to control quality. We allow a general relationship
between the efforts (i.e., substitutes or complements), highlighting the effort-relationship effect on
the incentive structure in the presence of inspection. Wang et al. (2012) considers a competitive
setting with two buyers, concerning spillover risks. They focus on the buyer improvement effort
alone and the mechanism to control spillover, without considering the supplier effort or inspection.
A stream of literature considers situations in which buyer investment results in improved properties other than quality, such as supplier cost (Iyer et al. 2005; Kim and Netessine 2012), supplier
reliability (Wang et al. 2010), supplier capacity (Li 2012), input consumption (Corbett et al.
2005), or service output (Roels et al. 2010; Roels 2012). We focus on product quality, which can
be uniquely controlled by inspection. Similar to our model, Iyer et al. (2005) and Roels (2012)
also consider a general effect relationship between the buyer investment and some variable on the
supplier side. Specifically, Iyer et al. study an adverse selection problem in which supplier capability (in cost reduction) is private information, and they focus on how the relationship between
buyer investment and supplier capability affects the information rent (i.e., the cost to acquire private information). In contrast, we analyze a moral hazard setting concerning the supplier effort
choice, and examine how the relationship between buyer investment and supplier effort affects the
incentive cost to motivate the desired supplier effort. In both papers, while the effect relationship
affects how the buyer investment may be used to mitigate the agency cost, the different source of
friction (tied to information or action) drives the result in different directions. Roels (2012) models
a general relationship between the buyer and supplier efforts on service outputs. With a focus on
the service rather than contract design, this study explores the optimal effort allocation between
the two parties without considering their strategic interactions.
There are papers that jointly consider product quality and distribution in a supply chain, such
as Xu (2009) and Kaya and Ozer (2009). These authors focus on the relationship between quality
and market revenue, without considering either inspection or buyer cooperation.
Finally, our paper is concerned with a supplier property (quality) that can be improved endogenously. This differs from the study of supplier qualification (e.g., Wan and Beil 2009) that focuses
6

on exogenous properties of suppliers. Chen and Deng (2012) consider supplier quality determined
by both an exogenous attribute and endogenous effort. They examine the impact of product certification along with screening contracts on supplier effort; neither inspection nor buyer cooperation
is a concern.

3.

Model Setup

Efforts: We consider a two-tier supply chain with a buyer and a supplier. The buyer sources a
product from the supplier, with the number of units normalized to one, as in Baiman et al. (2000)
and others. The defect rate of the product that the supplier delivers depends on both the supplier
quality-improvement effort and the buyer cooperation effort. We model buyer effort x as a bi-value
choice between high effort H and low effort L; that is, the buyer can choose to invest in supplier
training, provide engineering support, or enhance product design, among others (x = H), or not to
exert such effort (x = L). For the buyer, the cost of a high cooperation effort is , and the cost of
a low cooperation effort is normalized to zero. Similarly, supplier effort y is also modeled as a bivalue choice between high effort H and low effort L; that is, the supplier can choose to improve the
product quality by using better materials, investing in production processes, or enhancing quality
control, among others (y = H), or not to make such effort (y = L). For the supplier, the cost of a
high quality-improvement effort is , and the cost of a low effort is again normalized to zero.
Product quality: Given buyer effort x and supplier effort y, the probability that the supplier
will deliver a good-quality product is Pxy , x, y {H, L}, where PxL PxH and PLy PHy .
The efforts are complementary if the quality improvement achieved with the supplier effort is
greater when the buyer also exerts effort than when she does not (i.e., PHH PHL PLH PLL ).
Otherwise, the efforts are substitutable, if the buyer effort reduces the marginal quality improvement
achievable with the supplier effort (i.e., PHH PHL PLH PLL ). A neutral effect relationship
(PHH PHL = PLH PLL ) between the efforts is a special case of both types.
Inspection: The buyer inspects the product that the supplier delivers, and this inspection generates either a good or a bad signal. We follow Baiman et al. (2000) and assume that a good signal
is always observed when the product quality is good, but the buyer does not necessarily receive a
bad signal when the product quality is bad. Let Q [0, 1] be the probability of getting a bad signal
given a bad product, and call this probability inspection accuracy. The inspection accuracy depends on the number of resources (e.g., technology, labor hours) invested in this inspection. When
the accuracy is high, any further improvement of the inspection accuracy becomes more difficult

(i.e., requiring more incremental resources), and hence we assume that the total inspection cost c is
increasing and convex in inspection accuracy Q. We use function c (Q) log (1 Q) to model
the inspection cost for given Q, where > 0 measures the costliness of inspection. An internal
failure occurs when the product fails the inspection.
Revenue and costs: The buyer sells the final product at price r. If the product turns out to
be defective, then the buyer suffers loss s0 due to costs related to the product recall, litigation,
or goodwill loss, among others. An external failure occurs when a defective product is sold to a
customer. We follow Baiman et al. (2000) to assume that if the buyer rejects the intermediate
product from the supplier, she receives net revenue v that is between r and r s0 . Therefore, an
internal failure always leads to a product rejection.
Contract: We assume that the inspection result is based on objective measures and thus is
verifiable and contractible. This means that the buyer and supplier can write a contract with
transfers based on the inspection result. Specifically, let the contract be (w, u), where w 0 is
the price paid to the supplier if the product passes the inspection, and u 0 is the penalty that
the supplier pays if the product fails the inspection (i.e., an internal failure). We assume that the
external failure is not contractible; thus, the transfer is independent of whether an external failure
occurs. As explained in the introduction, this assumption is reasonable when it is hard to identify
the source of the defect or pursue compensation long after the delivery. We consider the extension
allowing external failures to also be contractible events at the end of 7.
Sequence of events: We model the interaction between the buyer and supplier as follows. The
buyer first announces her cooperation effort x and offers the supplier contract (w, u). The supplier
will accept the contract if it brings an expected profit higher than his reservation value (normalized
to zero). If he accepts the contract, he exerts his quality-improvement effort y and delivers the
product. Upon receiving the product, the buyer performs inspection at accuracy Q. She rejects
the product if it fails the inspection; otherwise, she accepts the product and sells it on the market.
We assume that the supplier can observe the buyers effort, as exemplified in the product design
or the number of engineering hours allocated to the supplier. Thus, the supplier makes his effort
decision contingent on the buyer effort choice. However, the buyer may not always observe the
supplier effort. We consider both cases with observable (4) and unobservable (5) supplier effort.
In the observable case, the buyer decides the inspection accuracy based on the supplier effort choice,
whereas in the unobservable case, the inspection accuracy is determined simultaneously with the
supplier effort decision.

4.

Observable Supplier Effort

In this section, we analyze the case with observable supplier effort, in which the buyers inspection
decision is responsive to the supplier effort but the contract and cooperation effort are committed
upfront. 4.1 analyzes the equilibrium structure, and 4.2 establishes the buyers optimal outcome.

4.1

Analysis

From 3, the interaction between the buyer and supplier constructs a two-stage game. In the
first stage, the buyer decides contract (w, u) and her cooperation effort x. In the second stage,
the supplier decides his quality-improvement effort y and then the buyer decides her inspection
accuracy Q, sequentially. We solve this two-stage game backward in 4.1.1 and 4.1.2.
4.1.1

Second-stage Analysis

We analyze the equilibrium of inspection accuracy Q and supplier effort y in the second stage, given
contract (w, u) and buyer effort x from the first stage. Let (Q, y|w, u, x) and (Q, y|w, u, x) denote
the supplier and the buyer profit functions, respectively.
Given product quality Pxy and inspection accuracy Q, the probability of passing a bad quality
product (i.e., external failure) is E (Pxy , Q) (1 Pxy ) (1 Q). The probability of a product
passing inspection (including the events that the product is good, and that the product is bad but
not identified) is S (Pxy , Q) Pxy + E (Pxy , Q), and hence the probability of failing inspection is
1 S (Pxy , Q). Define t w + u as the sum of the price and penalty, and define IH 1 and IL 0
as indices of the effort level. The supplier profit can be expressed as
(Q, y|w, u, x) = wS (Pxy , Q) u (1 S (Pxy , Q)) Iy = u + tS (Pxy , Q) Iy ,

(1)

where the total expected transfer to the supplier is tS (Pxy , Q) u. The buyer profit is then
(Q, y|w, u, x) = u + v + (r v t) Pxy (t + s0 r + v) E (Pxy , Q) c (Q) Ix .

(2)

The buyer receives an equivalent profit margin of rvt for a good-quality product, but pays a cost
of t + s0 r + v for a bad-quality product that passed the inspection. Define s s0 r + v > 0, as
the total penalty for a bad-quality product sold to the market, namely the external failure cost. For
our analysis, we assume that the external failure cost is sufficiently large (i.e., s / (1 PHH )).
This assumption implies that the inspection accuracy will always be positive (Q > 0)2 .
2

If Q = 0, the product from the supplier will always be accepted, and consequently, the supplier will not exert

effort regardless of the contract.

Maximizing (Q, y|w, u, x) over Q, the buyers optimal inspection accuracy responding to y is
b (t, Pxy ) 1
Q

.
(t + s) (1 Pxy )

(3)

Equation (3) suggests that the inspection accuracy increases with t (for given product quality Pxy ).
This occurs because, with a higher t, passing a bad product becomes more costly (either for paying
a higher price or missing a larger compensation from the supplier) and hence the buyer needs
to inspect more carefully in order to avoid this cost. Expecting the buyer response, the supplier
b (t, Pxy ) , y|w, u, x). From Equation (1), we can see
chooses his effort y {L, H} to maximize (Q
that the supplier effort choice depends on the sum of the price and penalty, t, as the supplier
not only will lose revenue w, but also will pay penalty u, if the product fails to pass inspection.
Thus, t can be considered as the effective incentive price to the supplier. Lemma 1 summarizes the
second-stage equilibrium. All the proofs are in the Appendix.
Lemma 1 With observable supplier effort, given buyer effort x and incentive price t, the supplier
effort is y (t, x) = L if t /(PxH PxL ), or y (t, x) = H, otherwise, and the inspection accuracy
b (t, Pxy ) where y = y (t, x).
is Q (t, x) Q
The supplier chooses a high quality-improvement effort if and only if the incentive price is
greater than the threshold, / (PxH PxL ). This confirms the intuition that the buyer needs to
offer a sufficiently high incentive in order to induce a high supplier effort. Note that the threshold
decreases in PxH PxL . That is, the incentive needed to induce the supplier effort is lower when the
quality improvement achieved with the supplier effort is more significant; the prospect of quality
improvement provides some self-motivation for the supplier. Consequently, the effect relationship
between the buyer and supplier efforts affects the supplier effort choice. If the efforts are substitutes
(complements) (i.e., PHH PHL ()PLH PLL ), then the required incentive is lower, making it
easier to induce a high supplier effort, when the buyer effort is low (high).
4.1.2

First-stage analysis

In the first stage, the buyer decides her cooperation effort x and contract terms w and u (or
equivalently u and t) to optimize her profit. The decision is subject to the reservation constraint
that leads to nonnegative profits for the supplier. Note from (1) and (2) that, for given t and x,
increasing u only transfers profits from the supplier to the buyer without affecting the second-stage
equilibrium. Thus, penalty u must be chosen to make the reservation constraint binding, and it can
be shown that the resulting penalty is always positive. Also note from (2) and (3) that increasing
10

incentive price t without changing x or y results in more inspection without bringing any benefit for
the buyer. Therefore, the buyer will choose the lowest possible incentive price to induce a certain
level of supplier effort. That is, for given x, the consideration of t can be limited to txL 0 and
txH / (PxH PxL ), which lead to a low and a high supplier effort, respectively.
b (txy , Pxy ) be the inspection accuracy based on the optimal
Given buyer effort x, let Qxy Q
incentive price corresponding to supplier effort y. Then the buyer profit associated with x and y is
xy = v + (r v) Pxy s (1 Pxy ) (1 Qxy ) c (Qxy ) Ix Iy .

(4)

The buyers decision transforms to the choice of (x, y) to maximize xy ; the solution is (x , y ).
Although the buyer extracts all the profit in the supply chain, the internal transfer, achieved
with the incentive price, txy , indirectly affects the buyer profit through its influence on the inspection
b (0, Pxy ) for
decision. Indeed, the first-best inspection accuracy in a centralized system would be Q
incoming product quality Pxy , which is lower than Qxy (equal when y = L). Thus, decentralization
results in over-inspection, and this friction further distorts the effort choices from the first-best
solution as will be shown later in Proposition 3.

4.2

Choice of Efforts

We analyze the outcome of efforts based on supplier effort cost and buyer effort cost . The
(, ) space is divided into four areas, one for each effort combination. To identify the boundaries
of the areas, we define x as the critical value of at which xL = xH , x = L, H, and L the
value of at which LL = HL . Furthermore, let H (), S () and C () be the critical values
of for given at which LH = HH , LH = HL , and LL = HH , respectively. Using these
critical values, Proposition 1 establishes the effort outcomes depending on and .
Proposition 1 The conditions of and that lead to different effort choices are characterized in
Table 1, where H L if and only if the efforts are substitutable with PHL and PLH sufficiently
high or PLL and PHH sufficiently low, and H L arises when the effect relationship is neutral.
Figure 1 illustrates the effort choices in the (, )-space with numerical examples. By decreasing substitutability (or increasing complementarity) between the efforts, we have the left plot for
substitutable efforts with H L , the middle for substitutable efforts with H > L , and the
right plot for complementary efforts. The label shows the effort choice for each area, with the first
letter representing the buyer effort, and the second representing the supplier effort.

11

Table 1: Effort choices based on and


Effort choices

and conditions

x , y

H L

H > L

H,H

H , H ()

H , min (H () , C ())

H,L

H , min (L , S ())

H , L

L,H

L , max (H () , S ())

L , H ()

L,L

L , L

L , max (L , C ())

L,H

L,L

L,H

L,H

L,L

L,L

H,H
H,H

H,L

H,H

H,L
H,L

H,L

  0.6
(substitutable efforts)

  0.55
(substitutable efforts)

  0.47
(complementary efforts)

Figure 1: Effort outcome with observable supplier effort. r = 10, v = 0, s0 = 25 (s = 15), = 1,


PLL = 0.05, PHH = 0.95, PLH = m 0.1, PHL = m + 0.1. Efforts are complements if m 0.5, and
substitutes if m 0.5.

12

The outcome first leads to an intuitive result: the buyer cooperation effort, x , is L (H) when
cost is high (low), and the supplier quality-improvement effort, y , is L (H) when cost is high
(low). However, one partys effort cost not only affects its own effort choice, but also may have
a cross effect on the other partys. Indeed, varying one cost alone may cause both effort choices
to change simultaneously, resulting a boundary between LL and HH, or between LH and HL.
As shown in Table 1, when H L , LH and HL are adjacent effort combinations, sharing the
boundary = S () (see the left plot of Figure 1), whereas when H > L , LL and HH are
adjacent outcomes, sharing the boundary = C () (see the middle and right plots of Figure 1).
The former situation means that investing in cooperation and providing incentives to motivate the
supplier are substitutable strategies, as varying or drive both efforts to shift in the opposite
directions. In contrast, the latter situation means cooperation and incentive are complementary
strategies, as both efforts shift in the same direction. We call these the strategic relationships
between the buyer and supplier efforts. From Proposition 1, the condition for a substitutable
strategic relationship, H L , holds only if the efforts are sufficiently substitutable in their
quality-improvement effects. If the substitutability is weaker or the efforts are complementary in
their effects, then H < L , leading to a complementary strategic relationship.
The results suggest that the strategic relationship between the efforts is not necessarily consistent with their effect relationship based on quality improvement. Specifically, when the efforts
are weakly substitutable in their quality-improvement effects (i.e., substitutable with H > L ),
cooperation and incentives should be used as strategic complements. The reason is tied to the presence of inspection as an additional instrument for quality management. A high-quality product
requires a low inspection, and further improvement of the high-quality product releases even more
inspection burden than the improvement of a low-quality one. Thus, the fact that both efforts help
to improve the quality implies that the two efforts have a complementary effect on inspection cost
reduction. When the efforts are weakly substitutable, their effect relationship on the inspection
costs dominates the one on quality improvement, making them strategic complements. Note this
result only exists when the inspection accuracy is endogenously determined along with the cooperation and incentive strategies; if the inspection decision is exogenous, the strategic relationship
between the two efforts is consistent with their effect relationship on quality improvement.
With the boundaries for different effort choices established in Proposition 1, Proposition 2
characterizes important properties of the boundaries.
Proposition 2 (i) H () is decreasing (increasing) for substitutable (complementary) efforts.

13

(ii) H and L are increasing in s, L is independent of s, and for any given , H () increases
in s for substitutable efforts, and decreases in s for complementary efforts.
Proposition 2 (i) shows that H , the threshold value of that separates the low and the high
buyer efforts when the supplier effort is H, depends on the supplier effort cost, , exhibiting another
cross effect of effort costs on the effort choices. Specifically, H decreases (increases) in for substitutable (complementary) efforts, suggesting that increasing supplier effort cost (while remaining
low to keep y = H) leads the buyer effort to shift from H to L for substitutable efforts, and from L
to H for complementary efforts. In other words, in response to increasing supplier effort costs, the
buyer should invest less in cooperation if the efforts are substitutable, and more if the efforts are
complementary. This result occurs due to the influence of cooperation on the incentive price. A
higher increases the incentive price needed to motivate the supplier effort, and this consequence
is mitigated with lower (higher) cooperation under substitutable (complementary) efforts, as the
buyers move reduces the incentive price by improving the marginal quality-improvement effect of
the supplier effort. Note that when is sufficiently high such that the supplier effort is always low,
no incentive needs to be offered, and hence the buyer effort choice is independent of , resulting in
a constant L .
Proposition 2 (ii) offers insights about how external failure cost s affects the effort choices.
Increasing s always leads to a higher supplier effort (as both L and H increase in s); thus, the
buyer facing a higher external failure cost prefers the strategy of offering sufficient incentives to
motivate the supplier. However, the buyers own cooperation effort does not necessarily increase
in s. Specifically, for small that leads to a high supplier effort, as s increases, the buyer effort
can only shift from L to H (as H becomes higher) under substitutable efforts, but from H to L
(as H becomes lower) under complementary efforts. The latter is interesting as it suggests that
the buyer may refrain from investing in supplier improvement when facing a higher external failure
cost. Since the inspection accuracy increases with the sum of s and t, a higher external failure
cost s causes more inspection, thus reducing the friction of over-inspection tied to incentive price
t, as the influence of the incentive price on the inspection decision becomes less significant. As
a result, the value of cooperation, under complementary efforts, to mitigate the incentive price
becomes lower when s is higher. This causes the buyer to reduce cooperation and solely rely on
more stringent inspection to control quality, in response to an increasing external failure cost. Note
that L is independent of s; when the supplier effort cost is too high for the buyer to incentivize the
supplier, adjusting only the inspection policy is sufficient for the buyer in response to the change

14

of s, without altering the cooperation strategy.


Finally, we compare the outcome to the first-best solution. Recall that the first-best solution
can be obtained by maximizing (4) with txy replaced by 0. This results in a similar structure of
the buyer and supplier efforts based on the and values as shown in Table 1, except that the
threshold values are different. Proposition 3 compares the threshold values in the decentralized
system with their counterparts in the first-best solution.
Proposition 3 Compared to the values in the first-best solution, L , H , S and C are lower,
whereas H is lower for substitutable efforts and higher for complementary efforts, with L identical.
The first part suggests that motivating the supplier effort is less preferred in the decentralized
system than it is in the centralized system. This is intuitive as inducing supplier effort incurs
incentive costs. It is interesting, however, to note from the second part that the buyer cooperation
strategy (based on H ) can shift in either way: If the efforts are complements, then cooperation
will be more favorable than in the first-best solution, and the opposite is true if the efforts are
substitutes. This result occurs due to the effect of cooperation on the incentive price, which can
be positive or negative depending on the effort relationship. The way the cooperation strategy
deviates from the first best offers a contrast to Iyer et al. (2005); in that study, the authors
consider an adverse selection problem where the efficiency loss is caused by the lack of information
about the supplier type (cost-reduction capability), and the buyers resource investment moderates
the information rent for screening the supplier type. They show that if buyer resource and supplier
capability are complements (substitutes), then the buyers second-best investment will be smaller
(greater) than the first best. We focus on a moral hazard setting in which the friction is tied to the
lack of direct control of the supplier action (quality-improvement effort), and the buyer investment
affects the incentive cost for motivating supplier effort. Though both Iyer et al. and us illustrate
the strategic value of buyer investment in mitigating the friction of delegation, whether it is tied
to information or action, the difference in the results comparing the investment to the first best
suggests that the source of the friction plays a critical role in how this drives the buyer investment
strategy.

5.

Unobservable Supplier Effort

In the case with unobservable supplier effort, the buyer determines the inspection accuracy without
knowing the supplier effort. Since the supplier cannot observe the buyers inspection decision before
15

making his effort choice, the two players move simultaneously in the second-stage game. We use
the superscript u to differentiate the notations for the unobservable effort case.

5.1

Analysis

Similar to 4, we analyze the equilibrium structure backward. In the second stage, a pure strategy equilibrium may not exist for given first-stage decisions. Thus, we analyze a mixed-strategy
equilibrium. The supplier strategy is characterized by q [0, 1], the probability of choosing a high
effort. Given q and x {L, H}, the (expected) product quality (probability of being good quality)
becomes Px (q) qPxH + (1 q) PxL . The second-stage equilibrium is characterized in Lemma 2,
in which qb (t, x) (1 PxL ) / (PxH PxL ) ((PxH PxL /t) (t + s))1 .
Lemma 2 With unobservable supplier effort, given buyer effort x and incentive t, the probability
of the supplier choosing a high effort is

0,
if t < txH or qb (t, x) 0

u
q (t, x) =
qb (t, x) , if t txH and qb (t, x) (0, 1) ,

1,
if t txH and qb (t, x) 1

(5)

b (t, Px (q u (t, x))).


which is increasing in t, and the inspection accuracy is Qu (t, x) = Q
From Lemma 2, it remains true that the supplier effort increases with the incentive price.
However, now with unobservable supplier effort, t txH is only a necessary condition for a positive
probability of the supplier choosing a high effort, whereas with observable supplier effort, it ensures
that the supplier chooses a high effort with probability 1. This implies that motivating the supplier
effort is more difficult when it is not observable.
In the first stage, given buyer effort x and incentive t, the penalty, u, is still designed to deliver
zero profit to the supplier. Therefore, the buyers expected profit as a function of t and x is,
u (t, x) = v+(r v) Px (q u (t, x))sE (Px (q u (t, x)) , Qu (t, x))c (Qu (t, x))Ix q u (t, x) .
The buyer chooses t and x to maximize u (t, x).
Although a general t given x may lead to a mixed strategy of supplier effort in the second
stage, interestingly, we find that the optimal t will always generate a pure strategy, making the
supplier effort strategy, q, either 0 or 1. The buyer either provides zero incentive, resulting in a
low supplier effort, or makes the incentive large enough for the supplier to always choose a high
effort. Therefore, again, there are four possible equilibrium structures each corresponding to a
16


b tuxy , Pxy the
combination of efforts x, y {L, H}. Let tuxy be the incentive price and Quxy = Q
inspection accuracy corresponding to buyer effort x and supplier effort y. Lemma 3 characterizes
the equilibrium.
Lemma 3 With unobservable supplier effort, given buyer effort x {L, H}, the optimal incentive
price is set either to tuxL = 0, leading to q u = 0, or to tuxH , leading to q u = 1, where tuxH is the
b (tu , PxH ) = txH /tu , or equivalently tu = (s, txH , PxH ), in which
unique positive root of Q
xH
xH
xH
q
(6)
(s, t, P ) [ (s t /(1 P )) + [s /(1 P ) t]2 + 4st]/2.
tuxH increases with PxH , given PxH PxL . tuHH tuLH if and only if the efforts are complementary
with PHL sufficiently small or PLL sufficiently large.
From Lemma 3, tuxH is greater than txH ; in other words, given buyer effort x, the incentive price
required to induce the supplier effort is greater than the one in the observable case. This results
in QuxH QxH , so for the same incoming quality, the buyer needs to raise her inspection accuracy
due to effort unobservability.
Recall from Lemma 1 that the incentive price with observable effort, txH , depends on the
marginal quality improvement, PxH PxL , achieved from the supplier effort. Now with unobservable
effort, the incentive price, tuxH , depends not only on the marginal improvement but also on the target
quality, PxH , as shown in (6). While greater marginal improvement reduces the incentive price as in
the observable case, higher target quality increases the incentive price. The latter happens because
higher product quality leads to decreased inspection by the buyer, increasing the probability of
passing a defective product. This disincentivizes the supplier to improve quality, and thus a higher
incentive price is needed to motivate the supplier. When the supplier effort is observable, the threat
of more stringent inspection following the supplier low-effort choice serves as a stick to motivate a
high effort from the supplier (along with the incentive price). Now with unobservable effort, the
threat of inspection is credible only when the target quality is low.
Recall also from 4.1.1 that with observable supplier effort, cooperation increases the incentive
price (tHH tLH ) under substitutable efforts, and decreases the incentive price (tHH tLH )
under complementary efforts. Now with unobservable supplier effort, tuHH tuLH holds not only for
substitutable efforts but may also for complementary efforts: since higher target quality increases
the incentive price and cooperation improves the target quality (PHH PLH ), tuHH can be greater
than tuLH despite cooperation making the supplier effort more effective at improving quality. Only
if the effect complementarity between the efforts is strong enough (PHL sufficiently small or PLL
17

sufficiently large) does tuHH < tuLH arise. These results suggest that unobservability of supplier
effort introduces a negative effect of cooperation on the supplier incentive, regardless of the effort
relationship.
Based on the definition of Quxy and tuxy , the buyer profit given (x, y), x, y {L, H}, becomes
u
u
uxy = v + (r v) Pxy
s 1 Pxy



1 Quxy c Quxy Ix Iy .

The optimization problem again transforms to the search of x and y to maximize uxy .

5.2

Choice of Efforts

Similar to 4, we also define a set of thresholds that identify the boundaries of different effort
outcomes based on and : let xu be the critical value of at which uxL = uxH , and xu (),
u (), u () be the values of at which u = u , u = u , and u = u , respectively,
C
S
Lx
Hx
LH
HL
LL
HH

for given and x = L, H. The structure of the effort outcome is characterized in Proposition 4.
Proposition 4 With unobservable supplier effort, the conditions of and that lead to different
u , and
effort choices are characterized in Table 1 with x , x , C , and S replaced by xu , xu , C
u u if and only if the efforts are substitutable with
Su , respectively, for x = L, H, where H
L
u u arises when the effect relationship is
PHL sufficiently large or PLL sufficiently small, and H
L

neutral.
Figure 2 illustrates the structure using the same numerical setting as for Figure 1.
L,H

L,L

L,H

L,H

L,L

L,L

H,H
H,L

H,H

H,H

H,L

H,L

  0.6
(substitutable efforts)

  0.55
(substitutable efforts)

  0.47
(complementary efforts)

Figure 2: Effort outcome with unobservable supplier effort. Parameters are the same as those for
Figure 1.

18

u and u still determines the strategic relationFrom Proposition 4, the comparison between H
L

ship between the buyer and supplier efforts: investing in cooperation to improve quality directly
and providing incentives to achieve quality improvement indirectly (through the supplier effort)
u u ; otherwise they
are substitutable strategies (i.e., LH and HL are adjacent choices) if H
L

are complementary strategies (i.e., LL and HH are adjacent choices). Similar to the observable
case, a complementary strategic relationship exists not only for complementary efforts but also for
u > u (see the right and middle plots of Figure 2).
weakly substitutable efforts that renders H
L
u u ) in their quality-improvement
Only when the efforts are sufficiently substitutable (so that H
L

effects should they be used as substitutable strategies (see the left plot of Figure 2). Proposition 5
characterizes some properties of the boundaries.
u () is decreasing (increasing) if P
Proposition 5 (i) H
HL is sufficiently large (small) or PLL is

sufficiently small (large). Under a neutral effect relationship of efforts, there exists b such that
u () decreases with
b and increases with > .
b
H
u and u are increasing in s, u is independent of s, and u increases in s if and only
(ii) H
L
L
H

if tuHH tuLH .
It is known with observable supplier effort that increasing the supplier effort cost drives up
(down) cooperation, with H increasing (decreasing) in , for complementary (substitutable) efu () is more complex. As shown in part (i), only
forts. With unobservable effort, the shape of H
u always increasing or decreasing
if the efforts are sufficiently complementary or substitutable is H
u () is quasi-convex, first decreasing and then increasing as increases. The
in . Otherwise, H
u () is driven by the effect of the target quality on the incentive price.
quasi-convex shape of H

Recall that cooperation improves the target quality, driving up the incentive cost. This effect is
aggravated with a larger , causing the cooperation strategy to shift from H to L (by decreasing
u ). As further increases, the incentive price under cooperation would become so large that its
H

incremental change no longer has a significant impact on the buyer profit. This makes the buyer
profit insensitive to the change of under cooperation, while its sensitivity to can still be significant without cooperation, when the incentive price is lower. Therefore, the cooperation strategy
u becoming higher) as further increases.
shifts back from L to H (with H

With unobservable supplier effort, it remains true that a higher external failure cost makes a high
u and u increase in s), and its effect on the cooperation strategy
supplier effort more favorable (H
L
u increases in
can be either way, for increasing the external failure cost leads to more cooperation (H

s) if and only if cooperation increases the incentive price. As the condition for tuHH tuLH includes
19

not only the substitutable efforts but also the weakly complementary efforts, a high external failure
cost discourages cooperation only when the efforts are sufficiently complementary. This differs
from the observable case, in which the external failure cost has a positive (negative) effect on
the cooperation strategy as long as the efforts are substitutable (complementary). Despite this
difference, the reason for the negative effect is the same: increasing s reduces the friction (overinspection) caused by the incentive price, making cooperation less valuable as a tool for mitigating
the incentive price.

6.

Impact of Effort Observability

Having analyzed both the observable and unobservable cases, we now compare the two scenarios to
reveal the impact of effort observability. We first investigate how it changes the sourcing strategy.
Proposition 6 compares the thresholds for different effort choices under the two scenarios, and
Figure 3 illustrates the comparison by assembling the thresholds identified for observable effort in
Figure 1 and those for unobservable effort in Figure 2.
u , u , u () () and u () ().
Proposition 6 (i) H
H
L
S
C
L
S
C
u () () except when is sufficiently small with P
(ii) H
H
HL low or PLL high enough, and

is sufficiently large with PHL high or PLL low enough. When the efforts have a neutral effect
u () () for any .
relationship, H
H

L,H

L,L

L,H

L,H

L,L

L,L

H,H
H,L

H,H

H,H

H,L

  0.6
(substitutable efforts)

H,L

  0.55
(substitutable efforts)

  0.47
(complementary efforts)

Figure 3: Comparison of effort outcome between the observable (solid lines) and unobservable
(dashed lines) supplier effort scenarios. Parameters are the same as those for Figure 1.

20

First, the buyer will (weakly) prefer incentivizing a higher supplier effort when the supplier effort
u () ()
becomes observable. The result xu x , x = L, H, along with Su () S () and C
C

in part (i) implies that improving effort observability is associated with an expansion of areas for
high supplier effort choices (including LH and HH). This suggests that the supplier effort can
only shift from L to H, when it becomes observable. Therefore, motivating the supplier effort is
a more favorable strategy if it is observable. This is intuitive as the buyer would have to offer a
larger incentive priceand thereby incurring more agency coststo motivate the effort when it is
unobservable.
Next we show that improving supplier effort observability generally leads to a higher cooperu and
ation effort, except for some special cases. The comparison between H
H in Proposition 6

(ii) indicates that, when the two efforts are not highly dependent of each other on their qualityimprovement effects (as exemplified in a neutral relationship), or is neither too high nor too
u () ()), as shown in all
low, making the supplier effort observable enhances cooperation (H
H

three plots in Figure 3. This result is driven by the fact that with unobservable supplier effort,
cooperation increases the incentive cost (as it improves the target quality); thus, cooperation is
more valuable for observable supplier effort. However, if the efforts get sufficiently complementary
(PHL low or PLL high) with a small , or sufficiently substitutable (PHL high or PLL low) with a
u () > ()) can happen. In the former case, because of strongly
large , then the opposite (H
H

complementary efforts and small , the incentive price is very small under cooperation even if the
effort is unobservable, and in the latter case, because of strongly substitutable efforts and large ,
the incentive price under cooperation is very large even if the supplier effort is observable. In both
cases, changing effort observability has little impact on the profit when the buyer cooperates; monitoring the supplier effort delivers significant benefit only when the buyer does not cooperate. Thus,
in such cases it is more beneficial to employ cooperation with unobservable supplier effort, and
non-cooperation with observable effort. Though such cases exist theoretically, we observe them
rarely in empirical studies. Therefore, we conclude that observing the supplier effort enhances
cooperation generally.
Clearly, observing the supplier effort, if possible, always benefits the buyer. Our results show
how the buyer should revise her sourcing strategy accordingly when such observation becomes
possible. In reality, the information about the supplier effort may be acquired through activities
such as on-site visits and auditing, which could be costly. Therefore, It is important to understand
when the benefit of observing supplier effort justifies the cost for such activities. To provide insights
into this question, next we study the profit difference between the two scenarios with observable
21

and unobservable supplier effort, considering the impact of and .


The effect of buyer effort cost is straightforward. Since the buyer profit decreases linearly in
under the cooperation strategy, the profit difference is independent of if observability does not
affect the buyer effort choice; otherwise, the profit difference decreases in if it leads the buyer
effort choice to shift from L to H, and increases in with the opposite change. Thus, either a low
or high may justify the investment in monitoring the supplier effort, depending on how the effort
observability affects the cooperation strategy, which has been previously discussed.
The effect of supplier effort cost on observability benefits is summarized in Proposition 7. It
characterizes how the profit difference changes with for different effort outcomes.
Proposition 7 (i) xL uxL is independent of for x = L, H.
(ii) xH umL decreases in for x, m = L, H.
(iii) xH uxH first increases and then decreases in , for x = L, H.
(iv) If the efforts have a neutral effect relationship, then HH uLH first increases and then
decreases in . If PHH PHL is sufficiently large (for given PLL and PLL ), then HH uLH
always increases in .
(v) If the efforts have a neutral effect relationship, then LH uHH first increases and then
decreases in . If PLH PLL is sufficiently large (for given PHL and PHH ), then LH uHH
always increases in .
Part (i) indicates that if the supplier effort remains low (regardless of its observability), does
not affect the buyer profitability, which is intuitive. Part (ii) suggests that if effort observability
induces the supplier effort to switch from L to H, regardless of the buyer effort, the profit difference
always decreases in , as a higher implies larger incentive costs for motivating the supplier effort.
The cases when the supplier effort remains high are discussed in parts (iii), (iv) and (v), for different
change paths of the buyer effort. Among these cases, if the buyer effort keeps constant regardless
of the supplier effort observability (part iii), then the profit difference will first increase and then
decrease as increases, for the incentive disadvantage for the unobservable scenario is aggravated
with increasing up to a point, and then weakens as the profit becomes less sensitive to the
incentive price when the incentive price gets large. If the buyer effort changes (in either direction)
with the supplier effort observability, such effects of remain, but under the influence of the effect
relationship between the efforts. Specifically, when improving supplier effort observability induces
the buyer effort to shift from L to H (part iv), the profit difference will monotonically increase with
if the efforts are sufficiently complementary (i.e., PHH PHL being sufficiently high); otherwise,
22

it remains a quasi-concave shape. With strongly complementary efforts, a higher increases the
value of cooperation, due to the negative effect of cooperation on the incentive price. Likewise,
if the buyer effort shifts in the other direction from H to L (part v), the benefit of observability
will always increase in if the effort substitutability is strong enough (i.e., PLH PLL sufficiently
high); in such cases, a higher reduces the value of cooperation, as cooperation leads to a higher
incentive price under substitutable efforts.
Combining the results in Proposition 7 sheds light on how the benefit of observability is affected
by over the entire range. Recall from Proposition 6 that the supplier effort can only shift from
L to H as it becomes observable. Thus, for being sufficiently high or low, keeping the buyer
effort choice at L or H independent of , Proposition 7 suggests that the benefit of observability
increases in up to a certain point, after which it (weakly) decreases in . In such cases, investing
in monitoring the supplier effort should be considered when the supplier effort cost is neither too
high nor too low. For in an intermediate range, in which the buyer effort choice may change with
, we observe that the profit difference can have more than one peak as a function of ; in this case,
there can be several (disconnected) value ranges, corresponding to different buyer effort choices,
that justify the investment in monitoring the supplier effort.

7.

Conclusions

A buyer sourcing critical production activities from a supplier can use different strategies to manage
the quality of the sourced product. On one hand, she can improve the incoming quality directly
by cooperating with the supplier to improve the process and/or product, or she can improve the
incoming quality indirectly by providing incentives to the supplier to exert his quality-improvement
effort. On the other hand, she can control the outgoing quality by inspecting the incoming units.
This paper investigates how to use these three strategiescooperation, incentives, and inspection
in a holistic manner. We consider a general effect relationship between the buyer and supplier
efforts, allowing them to be substitutable or complementary on their quality-improvement effects.
For both observable and unobservable supplier efforts, we identify the structure of effort outcomes
along with the inspection decisions, and compare the results in the two scenarios.
Existing research in quality and supply chain management has typically focused on some, but
not all, of the three strategies. By considering all three strategies together, we establish the strategic
relationship between the buyer and supplier efforts, which may differ from their effect relationship
due to the consideration of inspection costs. Our analysis reveals the value of the cooperation strat-

23

egy at influencing the incentive structure, highlighting the difference to adverse selection problems
(Iyer et al. 2005) of the buyer using (non)cooperation to mitigate agency costs. By considering
both scenarios with and without supplier effort observability, we identify a unique negative effect
of cooperation on the supplier incentive for unobservable effort, thereby establishing the impact of
effort observability on both the strategy and profit.
Our results offer several important managerial insights. First, when facing both options of
directly and indirectly (by motivating the supplier effort) improving the supplier quality, the buyer
should not only consider their impact on the quality but also the effect on the inspection. Also,
the buyer often needs to embrace both options, even if they may be substitutable on their quality
effects. This is consistent with the observation in practice that buyers are paying more attention to
supplier development while recognizing the importance of offering appropriate incentives without
overly squeezing the suppliers. Second, when choosing the cooperation strategy, the buyer should
consider not only its direct effect but also its indirect effect (through the incentive) on the quality
outcome; while the former is always positive, the latter can be positive or negative depending
on the effort relationship. Thus, the buyer may adjust her cooperation strategy in either way
in response to increasing supplier effort costs or external failure costs; instead of stepping up on
cooperation, strengthening the inspection alone may suffice. Third, the sourcing strategy should
be customized based on supplier effort observability. If the buyer is able to observe the supplier
effort, she should prefer the strategy of providing incentives to motivate the supplier effort, and
often consider cooperation more favorably.
Finally, we consider the extension that allows the buyer to penalize the supplier for external
failures after sales, relaxing the assumption that external failures are non-contractible events. The
contract then includes a new termexternal failure penaltyin addition to transfer price and internal
failure penalty as in the main model. When both internal and external failures are contractible, it
is not surprising, as shown in Baiman et al. (2000), that the first-best outcome can be obtained if
there is no constraint on the contract terms. For both the observable and unobservable cases, the
optimal external penalty would be set equal to the optimal incentive price. This cancels out the
impact of the incentive price on the inspection decision, removing the friction of delegation and
achieving the first-best result.
In reality, there is usually a limit to how much the buyer can recover from the supplier for a
defective product due to reasons such as fairness concerns (Balachandran and Radhakrishnan 2005),
the difficulty of pinpointing the default responsibility (Zhu et al. 2007), and the suppliers financial
constraints. With this limit, the first-best result may not be achieved if the optimal incentive price
24

is greater than the limit, which can be the case when the quality improvement by the supplier
is relatively small or supplier effort cost is high. In this case, the external penalty cannot fully
cancel out the friction caused by the incentive price; over-inspection continues to exist, though it is
mitigated by the external penalty. For both observable and unobservable efforts, all the structural
results from the main model remain, though they are brought closer to the first-best solution.

References
Babich V., C. Tang. 2011. Managing opportunistic supplier product adulteration: deferred payments, inspection, and combined mechanisms. Manufacturing & Service Operations Management, forthcoming.
Baiman, S., P.E. Fischer, M.V. Rajan. 2001. Performance measurement and design in supply
chains. Management Science 47(1) 173-188.
Balachandran, K.R., S. Radhakrishnan. 2005. Quality implications of warranties in a supply chain.
Management Science 51(8) 1266-1277.
Boothroyd, G., P. Dewhurst, W. Knight. 2002. Product Design for Manufacture and Assembly.
CRC Press, New York, NY.
Burke, H. 2007. Mattel net falls on Chinese-made toy recall costs. Bloomberg News. October 15.
Chao, G., S. Iravani, C. Savaskan. 2009. Quality improvement incentives and product recall cost
sharing contracts. Management Science 55(7) 1122-1138.
Chen, Y.-J., M. Deng. 2012. Product certification and quality investment in supply chains. Naval
Research Logistics, forthcoming.
Corbett, C., G. DeCroix, A. Ha. 2005. Optimal shared-savings contracts in supply chains: Linear
contracts and double moral hazard. European Journal of Operational Research 163 653-667.
Dong, Y., X. Wan, K. Xu, Y. Xu. 2011. Quality audit and supplier inspection in multi-level supply
chains. Working paper.
Hwang, I., S. Radhakrishnan, L. Su. 2006. Vendor certification and appraisal: Implications for
supplier quality. Management Science 52(10) 1472-1482.
Iyer, A.V., L.B. Schwarz, S.A. Zenois. 2005. A principle-agent model for product specification and
production. Management Science 51(1) 106-119.
Kaya, M., O. Ozer. 2009. Quality risk in outsourcing: Noncontractible product quality and private
25

quality cost information. Naval Research Logistics 56 669-685.


Kim, S. and S. Netessine. 2012. Collaborative Cost Reduction and Component Procurement Under
Information Asymmetry. Management Science, forthcoming.
Knowledge @ Wharton. 2007. Sourcing in China: Behind the media frenzy, there is much to
explore. November 7.
Knowledge @ Wharton. 2009. Raising the bar: can China meet the quality challenge? June 3.
Layton, L. 2010. Johnson & Johnson division recalls 43 OTC medicines for infants and children.
Washington Post May 2.
Li, C. 2012. Sourcing for supplier effort and competition: Design of the supply base and pricing
mechanism. Management Science, forthcoming.
Liker, J., T. Choi. 2004. Building deep supplier relationships. Harvard Business Review 82(12)
104-113.
Maurer, A., F. Dietz, N. Lang. 2004. Beyond cost reduction: Reinventing the automotive OEMsupplier interface. The Boston Consulting Group.
Myers, R. 2007. Food fights. CFO Magazine June 1.
Roels, G. 2012. The Economics of Joint Production in Services. Working paper.
Roels, G., U. S. Karmarkar, S. Carr. 2010. Contracting for Collaborative Services. Management
Science. 56(5) 849-863.
Rui, H., G. Lai. 2012. Managing Product Adulteration with Deferred Payment and Inspection:
The Effects of Procurement Quantity and Lead Time. Working paper.
Sako, M. 2004. Supplier development at Honda, Nissan and Toyota: comparative case studies of
organizational capability enhancement. Industrial and Corporate Change 13(2) 281-308.
Sherefkin, R., J. Armstrong. 2003. Suppliers wary of increasing warranty charges. Automotive
News 77(6035) 17-18.
Spencer, J., N. Casey. 2007. Toy recall shows challenge China poses to partners. Wall Street
Journal August 3.
Taylor, C.R., S.N. Wiggins. 1997. Competition or Compensation: Supplier Incentives Under the
American and Japanese Subcontracting Systems. American Economic Review 87(4) 598-618.
Wan, Z., D. Beil. 2009. RFQ Auctions with supplier qualification screening. Operations Research
57(4) 934-949.
26

Wang, Y., W. Gillandy, B. Tomlin. 2010. Mitigating supply risk: Dual sourcing or process improvement? Manufacturing & Service Operations Management 12(3) 489-510.
Wang, Y., Y. Xiao, N. Yang. 2012. Sourcing with Original Design: Pricing and Spillover Risk.
Working paper.
Welch, D. 2007. Made in China: Faulty tires. Business Week July 12.
Xu, X. 2009. Optimal price and product quality decisions in a distribution channel. Management
Science 55(8) 1347-1352.
Zamiska, N., D. Kesmodel. 2007. Tainted gingers long trip from China to U.S. stores. Wall Street
Journal November 19.
Zhu, K., R.Q. Zhang, F. Tsung. 2007. Pushing quality improvement along supply chains. Management Science 53(3) 421-436.

Appendix: Proofs
b Pxy ) = Q.
b Under the buyers best response (i.e.,
Proof of Lemma 1. For simplicity, let Q(t,
b the suppliers profit is (Q,
b y) = (u + t/ (t + s)) + tPxy Iy . Although the buyers
Q = Q),
b does not. The supplier will choose
best response depends on the supplier effort, the term E(Pxy , Q)
the effort level that yields the highest tPxy Iy , since the first two terms are independent of y.
Proof of Proposition 1. Define x = max (xL , xH ), which is a function of . As increases,
H drops while L remains constant. Thus H intersects with L when is sufficiently large.
When H L , the boundary definition implies that (H, H) is the optimal choice if H ,
H (), and C (). But in this region, C () (implies that HH LL ) is
redundant, as H () implies that HH LH and H L implies that LH LL for
H . Similarly, we can remove the redundant constraints and complete Table 1.
Define g(z) = s/ (z + s) + log (z + s) for the rest of the appendix. (1) g(z) increases with z; (2)
g(z) increases with s for s 0; and (3) g(z)/s decreases with z, as 2 g(z)/sz < 0, for s 0.
Next, we compare L and H , by defining x = xH xL , which + (g(0) g(tx )) plus
some constant terms. x / = 1 t2xH < 0 implies that as we increase from 0 to x , the
difference between the profits shrinks from a positive value to zero. Moreover, when the efforts are
complements (substitutes), i.e., tHH ()tLH , L (H ) decreases faster than H (L ).
Therefore, if L |=0 < H |=0 for the complementary case, H > L , whereas if L |=0

27

H |=0 for the substitutable case, H L . Because H |=0 L |=0 is


(r v) ((PHH PHL ) (PLH PLL )) + log (((1 PHL ) / (1 PHH )) ((1 PLL ) / (1 PLH ))) ,
with txH = 0 being replaced in the equation as = 0, H > L for complementary efforts, as
PHH PHL > PLH PLL implies the term in logarithm is positive. Similarly, for neutral efforts,
H L again holds. When the efforts are substitutes, although the first term is positive, we
require (1 PHH ) /(PHH PHL ) (1 PLH ) /(PLH PLL ) (i.e., sufficiently high PHL and PLH
or sufficiently low PLL and PHH ) to guarantee that H |=0 L |=0 and thus H L .
Proof of Proposition 2. We abbreviate txH to tx . (i) H () satisfies that LH = HH , i.e.,
H () = (r v) (PHH PLH ) + (g(tL ) g(tH )) + log((1 PLH ) / (1 PHH )), and hence,


t2L
t2H
H ()

s (tL tH ) (2tL tH + s (tL + tH ))


=

=
.

(tL + s)2 (tH + s)2

(tL + s)2 (tH + s)2


It is non-positive (nonnegative) if and only if the efforts are substitutable (complementary), thus
leading to H () decreasing (increasing) under substitutable (complementary) efforts.
(ii) When y = L, t = 0, and when y = H, t = tx . By using Equations (3) and (4) and simple
algebra, we see that the first derivative of xH mL with respect to s as (g(0) g(tx )) /s,
which is positive by using Property 3 of the function g, as tx > 0. As a result, let x = m, we have
x increases with s, as x satisfies xH xL = 0. Second, we know HL LL is independent
of s, and hence L is independent of s. Finally, let x = H and x = L, and we have the first
derivative of HH LH respect to s as (g(tL )/s g(tH )/s). If the efforts are substitutes
(i.e., tH tL ), then the first derivative is non-negative, as g(z)/s decreases with z, and hence
HH LH increases with s, and thus so as H (). For complementary efforts, the opposite is
true.
Proof of Proposition 3. That the first-best incentive is 0 but txL = 0 and txH > 0 implies that
L is identical whereas L , H , S and C in the first best are larger. Using the implicit function
of H in Proposition 2, H is larger (smaller) for substitutable (complementary) efforts. 
b Px (q)).
Proof of Lemma 2. Similar to Equation (3), the best response of Q given q is Q(t,
Given Q, a mixed strategy of the supplier (0 < q < 1) is admitted if the buyers profit is the same
for y = H and y = L, i.e., tQ = txH . Otherwise, q = 1 if tQ > txH , and q = 0 if tQ < txH .
b Px (q)) = txH , gives qb(t, x). Note that t < tx guarantees
Solving the two simultaneously, i.e., tQ(t,
that tQ < txH and hence q u = 0. Thus, the equilibrium is obtained. At the optimum q u = qb,
Qu = txH /t, and hence Qu / t = Qu /t < 0; the inspection accuracy decreases with t.

28

Next, for qb (0, 1), by replacing Qu = txH /t, we can have


1
qb
[Qu /t (t + s) + (1 Qu )]
[Qu s/t + 1]
1
=
0,
=
2
2
t
PxH PxL
PxH PxL (1 Qu )2 (t + s)2
(1 Qu ) (t + s)
implying that qb increases with t. When q u = 0 or 1 are non-decreasing with t, and thus we prove
the (weakly) increasing property.
Proof of Lemma 3. We abbreviate txH to tx and tuxH to tux . First, if q u = 0, tu = 0, and if
q u = 1, tu is the smallest t that satisfies qb 1, as the buyers profit decreases with t. Then we will
show that when q u = qb, the optimal solution can either be q u = qb(tx , x), tu = tx or q u = qb = 1,
b ux , PxH ), or
tu = tux , where tux is the solution of the simultaneous equations, Qu = tx /tux = Q(t
equivalently hx (t) = (t)2 + t [s /(1 PxH ) tx ] stx = 0, leading to Equation (6). Note that
qb = 1 are included, due to the continuity of the profit function, and although there are two roots
of hx (t), only the positive one is feasible, because the root with a negative sign will be negative.
Again, by the continuity of the buyers profit function, we know that q u = 0, tu = 0 weakly
dominates q u = qb(tx , x) and q u = qb = 1 also weakly dominates the case when q u = 1 and qb 1.
As a result, the equilibrium happens at either q u = 0, tu = 0 or q u = 1, tu = tux .
Next, given PxH PxL (or namely tx ), and by using the implicit differentiation, we have

tux
h/PxH
tux / (1 PxH )2
=
=
0,
PxH
h/t t=tux
(2tux tx ) + (s /(1 PxH ))
because txu tx and s /(1 PxH ).
The third part starts with a neutral system (tL = tH ), in which tuH tuL can be replaced by
(s, tH , PHH ) (s, tH , PLH ). After simple algebra, we have (s, t, P )/P 0, which implies
that tuH tuL , as PHH PLH . Because tuH /PHL 0 (which can be obtained by using implicit
differentiation), we know tuH tuL is valid for all systems with substitutable efforts.
Finally, we prove the optimal solution where q u = qb [b
q (tx , x), 1]. Given x, the buyer profit
becomes a function of t. The first derivative of u (t, x) with t is


 u
u (t, x)

q
t
= (PxH PxL ) r v tx

t
1 Px (q u )
t
(t + s)2



(r v tx ) (1 + Qu ) + s + Qu s + t + (s2 /t) (1 Qu )
=
(1 Qu )2 (t + s)2

(A1)
(A2)

in which we use /(1 Px (q u )) = (1 Qu ) (t + s) and tx = /(PxH PxL ) in the derivation. We


next separate the space of rv into (1) rv < tx /(1Px (q u )), (2) tx /(1Px (q u )) rv <
tx , and (3) r v tx . The condition in region (1) and q u /t 0 implies that u (t, x) /t 0
(Equation A1), showing that the optimal solution happens at q u = qb(tx , x). In region (3), Equation
(A2) implies u (t, x) /t > 0, showing that the optimal solution happens at q u = 1.
29

In region (2), the sign of u (t, x) /t depends on the denominator in Equation (A2), in which
the first term increases with t, as its first derivative is 2 (r v tx ) (sQu /t2 ) > 0, and the


2 u
2
u
second term has the following first derivative, ((s/t) + 1) Q 2(s/t) (1 Q ) Qu . If the
second terms first derivative also is positive, then the big bracket in Equation (A2) increases
with t (or the profit function is convex), indicating that the optimal solution locates on the either
boundary. If the second terms first derivative is negative, i.e.,
(s/t + 1)2 Qu 2s2 (1 Qu ) /t2 0,

(A3)

we show that the first derivative of the profit function is positive, implying q u = 1.

If s t, the condition that s + Qu s + t + s2 /t (1 Qu ) tQu (1 + Qu s/t), implies a
positive u (t, x) /t, i.e.,
u (t, x)
[(r v tx + tQu ) (1 + Qu s/t)]
(r v) (1 + Qu s/t)

=
0,
t
(1 Qu )2 (t + s)2
(1 Qu )2 (t + s)2
in which in the equality we use Qu = tx /t. If s t, the condition that s + Qu s + t + s2 /t



(1 Qu )

s (1 + Qu s/t), implies a positive u (t, x) /t, i.e.,




(1 + Qu s/t)
(1 + Qu s/t)

u (t, x)
(r v tx + s)

t
+
0,
x
t
1 Px (q u ) (1 Qu )2 (t + s)2
(1 Qu )2 (t + s)2
in which in the second inequality we use s /(1PHH ) /(1Px (q u )). By using Equation(A3),
the two conditions can be proved with some simple algebra, and hence is left to the readers.
Proof of Proposition 4. See the proof in Proposition 1 for effort choices. For the remainder of
the proof, we use uxL = xL , tuxL = 0, and we abbreviate txH to tx and tuxH to tux .
Next, given x, xu satisfies that xL = uxH , i.e.,
xu = (r v) (PxH PxL ) (g(tux ) g(0)) + log ((1 PxL ) / (1 PxH )) ,

(A4)

and the fixed-point theorem has a unique solution, as xL is independent of , whereas uxH
increases with , i.e., uxH / = (g(tux )/tux ) (tux /tx ) (tx /) 1 0, as g(z) increases
with z, tx = tux (1 /((1 PxH ) (s + tux )), and
tx
s
=1
.
u
tx
(1 PxH ) (s + tux )2

(A5)

When efforts are neutral, we will show that the right-hand side of Lu (Equation A4) is always
higher than the right-hand side of Lu for any , and the right-hand side function decreases with
u intersects with the increasing left. Therefore, the decreasing right-hand side function for H

hand side function at a higher than the one for Lu . We will also show that xu /PxH 0 and
30

u u , leading to u u for
xu /PxL 0. As a result, for systems with neutral efforts, H
L
H
L
u increases) or P
u
systems with complementary efforts, i.e., either PHH (H
LL increase (L decreases)
u increases) or P
u
u
u
or PHL (H
LH decrease (L decreases), H L always hold. Conversely, when
u u requires a sufficiently small P
the efforts are substitutable, H
HL or a sufficiently large PLL .
L

To complete the proof, we fill two missing components. First we know that (s + tuL ) (1 PLH )
(s + tuH ) (1 PHH ) 0, as the difference can be represented as

tuL

tuH tuL
tu
tuL

 > 0,
u H
= tH u
tH tH
tL
tL tL tuH tH

in which we used tL = tH and tx = tux (1 /((1 PxH ) (s + tux )) again. Next, the comparison of
the right-hand side for the neutral case can be simplified to




s(tuH tuL )
s + tuL 1 PHL 1 PLH
1 PHL 1 PLH
u
u

 +log
g(tL )g(tH )+log
=
> 0.
1 PHH 1 PLL
s + tuH 1 PHH 1 PLL
s + tuL s + tuH
Finally, xu /PxH 0, as
xu
PxH

!

tux
tux

tux
tux
> 0,
1+
=
(r

v)
+

1 PxH
PxH (tux + s)2
(tux + s)2 =xu

in which the last two terms are positive after simple algebra. Similarly, xu /PxL 0.
Proof of Proposition 5. We abbreviate txH to tx and tuxH to tux . (i) For a neutral effort system
u () / = 0 can be easily proved. Next, we decrease
with PLH = PHH , tuL = tuH and hence H

PLH while keeping PLH PLL (or tL ), PHH PHL and PHH (or tuH ) the same, and hence,
u () /)
(H
tH tuL (tuL )2 + s2 s/ (1 PLH )
=
.

PLH
PLH ( tu + s 2 s/ (1 PLH ))2
L
u () /)/P
Thus, there exists a b 0 so that (tuL )2 + s2 s/ (1 PLH ) 0 (or (H
LH 0),

b When we decrease PLH , when > ,


b then u () / 0, and when ,
b
if and only if .
H
u () / 0. Similar result can be derived when we move P
then H
HH . Thus, for every system
u () is quasiconvex.
with neutral efforts, H

Based on the quasiconvex property of the neutral efforts, we now change the effort relation. We
start with a system with neutral efforts, and then increase PLL while keeping the remaining three
u () /) /P
the same, we have (H
LL as (/) (tL /PLL ) multiplying the following term,
!
tuL
tuL (1 /((1 PLH ) (s + tuL )) (tuL )2 + s2 s/ (1 PLH )
+
2
2

2 .
2
tuL + s s/ (1 PLH )
1 s/((1 PLH ) s + tuL )
tuL + s s/ (1 PLH )

After some algebra, we can see that its sign depends on


((tuL + s)2 s/ (1 PLH ))2 + ((s + tuL )2 (s + tuL ) /((1 PLH ))( (tuL )2 + s2 s/ (1 PLH )).
31

Thus, if (tuL )2 + s2 s/ (1 PLH ) 0, then it is positive; otherwise, its sign is larger than
((tuL + s)2 s/ (1 PLH ))2 + ((s + tuL )2 s/((1 PLH ))( (tuL )2 + s2 s/ (1 PLH ))


= ((tuL + s)2 s/ (1 PLH )) 2tuL s + 2s2 2s/ (1 PLH ) 0.
u () /) /P
Similarly, we can prove (H
HL 0. The two conditions combined imply that as the
u () / is decreasing,
efforts become more substitutable (i.e., decrease PLL or increase PHL ), H
u () / is decreasing.
whereas when PHL is sufficiently small or PLL is sufficiently large, H

(ii) The proof is similar to the one of Proposition 1, except tux /s is not zero now. Therefore,
we only focus on proving the sign of the first derivatives, instead of repeating all the arguments.
(uxH mL )/s is (dg(0)/ds dg(tux )/ds), in which dg(tux )/ds is increasing, as
(2tux + s) (tux + s)2 (tux )2 / (1 PxH ) + (2tux + s) s/ (1 PxH )
dg(tux )
=
0,
ds
(tux + s)2 ((tux + s)2 + s/ (1 PxH ))
but (dg(z)/ds) /z is decreasing, as its sign depends on


(z + s)3 (z + 2s) (z + s)2 + z (z 3s) / (1 PxH ) 2s2 z (z + s) (/ (1 PxH ))2


(z + s)3 (z + 2s) (z + s)2 + z (z 3s) (z + s) 2s2 z (z + s) (/ (1 PxH ))2
= s (z + s)4 (5z + 2s) 2s2 z (z + s) (/ (1 PxH ))2 0
u is independent of s, and u increases with
As a result, we prove that xu decreases with s, and L
H

s if and only tuH tuL , as (uHH uLH )/s is (dg(tuL )/ds dg(tuH )/ds). 
Proof of Proposition 6. Again, abbreviate txH to tx and tuxH to tux . (i) When x = L, uLy Ly ,
and when x = H, uLH LH and uHH HH , as tux tx . Therefore, this part is proved.
(ii) Define uH = uLH uHH and H = LH HH . For given PLL , PLH , and PHH ,
(uH H )/PHL is (tuH /PHL ) multiplies the following term,
h

i
u tu + 2s + (/(1 P
u ))2 s/(s + tu )
t
))
(1

s/(s
+
t
HH
H
H
H
H
tH
g 0 (tuH ) g 0 (tH ) u =
.
2
2
u
tH
(1 PHH ) (tH + s) s + t
H

Because tuH /PHL 0, the sign of (uH H )/PHL depends on the terms in the big
bracket. The sum of the first two terms decreases linearly with and the sum of the latter two
terms increases concavely. Thus, we can find one and only one fixed point solution, (PHL ), for
uH /PHL = H /PHL , as when = 0 the big bracket is 2s / (1 PHH ) > 0 and when
= the big bracket is negative infinity. As a result, uH /PHL ()H /PHL , if and
only if ()(PHL ). Similarly, for given PHL , PLH , and PHH , we increase PLL . Because
tLu /PLL 0, we can find that uH /PLL ()H /PLL , if and only if ()(PLL ).
32

With neutral efforts, we have tH tL tuL tuH . Because g(z) increases with z, g(tuH )g(tH )
u () (). As a result, for (P
g(tuL ) g(tL ) and hence uH H , i.e., H
H
HL )

(for (PLL )), if uH H , decreasing PHL (increasing PLL ) improves the probability
u () (). For (P
that uH H , or equivalently H
H
HL ) (for (PLL )), if
u () ().
uH H , increasing PHL (decreasing PLL ) improves the probability that H
H

Proof of Proposition 7. Abbreviate txH to tx and tuxH to tux . The proofs of (i) and (ii) are
obvious; (v) can be proved by following the procedures in (iv). Thus, we only show (iii) and (iv).
(iii) (xH uxH ) / equals to (tux /) multiply g 0 (tux ) g 0 (tx ) (tx /tux ), which determines its sign, as txu / 0. Similar to the proof of Proposition 6 (ii), we know that (xH uxH ) /
0 if and only if 1 . Note that the effort choice remains the same, as g (tux ) g (tx ).
(iv) (HH uLH ) / equals to (tuL /) multiply g 0 (tuL ) g 0 (tH ) (tH /tuL ), which is
"
!#
u
2
t
t
1
s
t
H
L
H
1
,
g 0 (tuL ) g 0 (tH ) u =
2
2
tL
(tH + s)2 tL
tu + s
(1 PLH ) tu + s
L

which is decreasing in tH , thus increasing in PHH PHL . Define p1 as the value of PHH PHL at
which g 0 (tuL ) g 0 (tH ) (tH /tuL ) = 0; if it is greater than 1 PLL , set p1 equal to 1 PLL . Then
(HH LHu ) / 0 if and only if PHH PHL p1 . If the efforts are neutral, we can replace
tH by tL , and using the results in (iii) gives us the second part of (iv).

33

You might also like