This article was downloaded by: [128.230.137.

248] On: 07 March 2024, At: 10:29

Publisher: Institute for Operations Research and the Management Sciences (INFORMS)
INFORMS is located in Maryland, USA

Manufacturing & Service Operations Management

Publication details, including instructions for authors and subscription information:
[Link]

Asymmetric Information of Product Authenticity on C2C

E-Commerce Platforms: How Can Inspection Services
Help?
Linqiu Li, Xin Fang, Yun Fong Lim

To cite this article:

Linqiu Li, Xin Fang, Yun Fong Lim (2023) Asymmetric Information of Product Authenticity on C2C E-Commerce Platforms: How
Can Inspection Services Help?. Manufacturing & Service Operations Management 25(2):631-647. [Link]
msom.2023.1186

Full terms and conditions of use: [Link]

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@[Link].

The Publisher does not warrant or guarantee the article’s 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 © 2023, INFORMS

Please scroll down for article—it is on subsequent pages

With 12,500 members from nearly 90 countries, INFORMS is the largest international association of operations research (O.R.)
and analytics professionals and students. INFORMS provides unique networking and learning opportunities for individual
professionals, and organizations of all types and sizes, to better understand and use O.R. and analytics tools and methods to
transform strategic visions and achieve better outcomes.
For more information on INFORMS, its publications, membership, or meetings visit [Link]
 MANUFACTURING & SERVICE OPERATIONS MANAGEMENT
Vol. 25, No. 2, March–April 2023, pp. 631–647
[Link] ISSN 1523-4614 (print), ISSN 1526-5498 (online)

Asymmetric Information of Product Authenticity on C2C

E-Commerce Platforms: How Can Inspection Services Help?
Downloaded from [Link] by [[Link]] on 07 March 2024, at 10:29 . For personal use only, all rights reserved.

Linqiu Li,a Xin Fang,b,* Yun Fong Limb

a
International Institute of Finance, School of Management, University of Science and Technology of China, Hefei, Anhui 230026, China; b Lee
Kong Chian School of Business, Singapore Management University, Singapore 178899, Singapore
*Corresponding author
Contact: lilinqiu@[Link], [Link] (LL); xfang@[Link], [Link]
(XF); yflim@[Link], [Link] (YFL)

Received: November 22, 2021 Abstract. Problem definition: We consider a customer-to-customer (C2C) platform that
Revised: June 26, 2022; September 29, 2022 provides an inspection service. Uncertain about product authenticity, a seller sells a prod­
Accepted: October 1, 2022 uct through the platform. Before purchasing, a buyer obtains a signal of the product
Published Online in Articles in Advance: authenticity from the product’s price set by the seller. The platform’s inspection service can
January 27, 2023 detect a counterfeit with a probability. If the product passes the inspection, the platform
sends it to the buyer and charges the seller a commission fee. Otherwise, the platform
[Link] returns it to the seller and charges the seller a penalty fee. Methodology/results: We
develop a two-stage game-theoretical model. In the first stage, the platform designs a con­
Copyright: © 2023 INFORMS
tract specifying the commission and penalty fees. In the second stage, the seller signals the
product authenticity by setting a price and the buyer decides whether to purchase it. This
results in a contract design problem that governs a signaling game. We find that the effect
of inspection is beyond merely detecting counterfeits. The inspection, even an imperfect
one, changes the signaling game’s structure and incentivizes the seller whose product is
likely authentic to sell through the platform. This can only be achieved by carefully choos­
ing the commission and penalty fees. Moreover, a larger platform’s expected profit does
not imply a larger commission fee or price in equilibrium. Under some mild conditions,
the optimal commission increases but the optimal penalty decreases as the platform’s
inspection capability improves. Managerial implications: The inspection service is not
widely available among leading C2C platforms as it is considered imperfect and costly.
Our study suggests that its benefit may be underestimated in practice. Moreover, the
inspection can eliminate the seller’s information rent and generate more revenue for the
platform. This paper provides guidance on how to set commission and penalty fees when
the inspection service is provided.

Funding: L. Li is supported by the National Natural Science Foundation of China [Grant 72071198] and
the Hong Kong Polytechnic University Distinguished Postdoctoral Fellowship Scheme [Grant
1-YWC7]. X. Fang and Y. F. Lim are grateful for the support from the Lee Kong Chian School of
Business, Singapore Management University [Maritime and Port Authority Research Fellowship
and Retail Centre of Excellence Research Grant]. Y. F. Lim is supported by the Association of South-
East Asian Nations Business Research Initiative Grant [Grant G17C20421], the Research Grants
Council of Hong Kong [Grants 15501920 and 15501221], and the Key Program of National Natural
Science Foundation of China [Grant 71931009].
Supplemental Material: The online appendices are available at [Link]

Keywords: asymmetric information • counterfeit • inspection • platform

1. Introduction Most C2C platforms have little control over the au­
Customer-to-customer (C2C) platforms receive increas­ thenticity of the products displayed. As a result, com­
ing attention among e-commerce business models. Pro­ plaints about counterfeits are on the rise (Rivera 2018).
minent examples of C2C platforms include eBay and According to Suthivarakom (2020), with third-party sell­
Taobao, which act as an intermediary to match individ­ ers on Amazon, the keyword “counterfeit” or “fake”
ual sellers and buyers. Under this business model, the appears in 1.725% and 4.275% of all the Amazon’s
sellers display their products on a platform and send reviews in 2015 and 2019, respectively. Similarly, Taobao
purchased items to the buyers directly. The platform was found to have 240,000 vendors selling counterfeits
obtains revenue by charging a commission fee whenever in 2017 (Hawkins and Thorpe 2019). Recently, sneaker
a product is successfully sold. resale is becoming a fiercely contested market as some

631
 Li, Fang, and Lim: Asymmetric Information of Product Authenticity on Platforms
632 Manufacturing & Service Operations Management, 2023, vol. 25, no. 2, pp. 631–647, © 2023 INFORMS

shoppers collect rare merchandise and sell it at a markup than 7.5 million transactions, and its gross product value
price online (Patel 2019). With a lucrative profit, there reached $1.8 billion (Farrell 2021).
are more and more counterfeits in the sneaker market. The transactions on these platforms proceed as fol­
The fake Nike footwear in 2019 would be worth more lows. A seller first displays a product on a platform and
than $472 million if it were real (Rohrlich 2020). C2C sets a price. If a buyer decides to purchase the product,
platforms cannot guarantee the authenticity of listed the buyer pays the price to the platform. The seller then
products because they do not physically handle the pro­ sends the purchased item to the platform for inspec­
Downloaded from [Link] by [[Link]] on 07 March 2024, at 10:29 . For personal use only, all rights reserved.

ducts. In addition, it is difficult for buyers to judge the tion. If the product passes the inspection, it will be
authenticity because they cannot physically assess the shipped to the buyer. In this case, the platform charges
products before purchases. Thus, buyers are worried a fraction of the selling price as a commission fee and
about ending up with a counterfeit at a high price (Kim remits the remaining amount to the seller. For example,
2018, Suthivarakom 2020). StockX sets 8%–14.5% of the price as a commission fee
Being aware of the product’s origin and able to physi­ for different product categories, whereas Goat sets
cally assess the items, sellers generally know more about 9.5%–25% of the price as a commission fee. If the prod­
the authenticity of their products than buyers. Some pro­ uct fails the inspection, the platform returns the item to
fessional sellers on Amazon and Taobao may know the seller. In that case, the platform refunds the pay­
whether their products are authentic when they know ment to the buyer and may penalize the seller based on
the product source. However, many of the sellers on the selling price. For example, StockX asks for 15% of
C2C platforms are nonprofessional and may not be the price as a penalty fee, whereas Goat has zero pen­
completely sure about the authenticity of their products. alty to the seller. In practice, different platforms adopt
For example, high-end sneakers are usually offered in a different policies for the commission and penalty fees.
limited quantity and have a long trade tail as the shoes One of our goals is to identify the optimal policy for the
continue to be bought and sold on secondary markets platforms.
(Dennis 2019). Although the source is not completely We develop a game-theoretical model to capture the
reliable, most collectors purchase such sneakers through interactions among a platform, a seller, and a buyer.
the secondary markets instead of official channels. More­ Being uncertain about product authenticity, the seller
over, some fake items resemble the authentic ones so sells a product through the platform. Before purchasing
closely that it becomes very challenging to differentiate the product, the buyer has even less information about the
them (Koaysomboon 2018). Even skilled appraisers make product authenticity and only obtains a signal from the
mistakes sometimes (Clifford 2019). It requires profes­ product’s price set by the seller. The platform can detect
sional training and extensive experience to inspect the whether the product is a counterfeit with a probability,
products. Because of unreliable sources and the lack of which measures the platform’s inspection capability. If
professional inspection capability, many sellers on C2C the product passes the inspection, the platform sends
platforms are not completely certain about the authentic­ the product to the buyer and charges the seller a com­
ity of their products. mission fee. If the product fails the inspection, the plat­
To address the issue of product authenticity or qual­ form returns the product to the seller and charges the
ity, the literature (Baiman et al. 2000, Hwang et al. 2006, seller a penalty fee. The game has two stages. In the first
Babich and Tang 2012) has investigated ex ante and ex stage, the platform designs a contract specifying the
post inspections. The ex ante inspections certify sellers commission and penalty fees as fractions of the selling
prior to transactions, whereas the ex post inspections price anticipating the interaction between the seller and
refer to inspections by buyers after receiving the pro­ the buyer. In the second stage, given the contract, the
ducts. Different from the literature, we study interim interaction between the seller and the buyer forms a
inspections adopted by some C2C e-commerce plat­ signaling game. The seller signals the product’s authen­
forms, where products are inspected after buyers place ticity by setting the price. Given the price, the buyer
their orders but before receiving the products. Typical then decides whether to purchase the product.
C2C platforms implementing the interim inspections We analyze the equilibrium decisions of the three
include StockX, Goat, and Poizon (see [Link], goat. parties given that the platform provides the inspection
com, and [Link], respectively). These platforms im­ service, and the seller is uncertain about the product
prove their inspection capability by investing heavily in authenticity. Specifically, we would like to answer the
advanced technology and experienced inspectors. For following questions:
example, Goat enhances its platform’s inspection tech­ 1. How does the platform’s inspection affect the sell­
nology using machine learning (Kelly 2018). Although er’s pricing decision? What are the benefits for the plat­
the interim inspections may cause a delay in transac­ form to provide the inspection service?
tions, buyers value this service, making these platforms 2. How should the platform set the commission and
popular. For example, in 2020, StockX completed more the penalty fractions to maximize its profit? How does
 Li, Fang, and Lim: Asymmetric Information of Product Authenticity on Platforms
Manufacturing & Service Operations Management, 2023, vol. 25, no. 2, pp. 631–647, © 2023 INFORMS 633

the platform’s inspection capability affect its commis­ and concludes the paper. Online Appendix A provides
sion and penalty? additional results. Online Appendix B discusses two
3. What if the seller knows the product authenticity? extensions of our model. Online Appendix C presents all
How do the equilibrium results and managerial insights proofs.
differ from the case in which the seller is uncertain about
it? 2. Related Literature
We summarize our findings and contributions as fol­
Downloaded from [Link] by [[Link]] on 07 March 2024, at 10:29 . For personal use only, all rights reserved.

This paper studies a C2C platform with an inspection

lows. We find that the effect of inspection is beyond service under asymmetric information about product
merely detecting counterfeit products. The inspection, authenticity. It is related to four streams of literature:
even an imperfect one, changes the structure of the sig­ quality management and inspections in a decentralized
naling game completely. It incentivizes only the high- supply chain, signaling asymmetric quality information,
product-authenticity seller to sell through the platform. counterfeit products, and platform business models.
However, this can only be achieved with carefully The quality management literature typically focuses
designed commission and penalty fees. Because the on inspection decisions and contract designs in a moral
inspection service is not yet widely available among hazard setting in which suppliers’ quality decisions can­
leading C2C e-commerce platforms, our results suggest not be observed. For example, Baiman et al. (2000) ana­
that the benefit of product authenticity inspection may lyze a variety of scenarios in which players’ actions or
be underestimated in practice. In addition, with the
product failures may or may not be contractible. Hwang
inspection service, the platform can eliminate the infor­
et al. (2006) compare buyers’ inspection and vendors’
mation rent earned by the seller. This generates more
certification. Babich and Tang (2012) consider three me­
revenue for the platform.
chanisms to deal with product adulteration problems:
To maximize its expected profit, we identify the optimal
deferred payment, product inspection, and combined
contract for the platform. We find that a larger platform’s
mechanisms. Seung and Taesu (2018) investigate several
expected profit does not imply a larger commission frac­
reward mechanisms for collaborative product quality
tion or a higher price in equilibrium. Under some mild
improvement in a buyer-driven supply chain. Mariya
conditions, the optimal commission fraction increases but
and Edieal (2018) consider buyers’ investment efforts to
the optimal penalty fraction decreases as the platform’s
improve supplier quality, monetary incentives to moti­
inspection capability improves. This implies that the plat­
vate suppliers’ quality improvement efforts, and inspec­
form should adjust the contract parameters accordingly if
its inspection capability changes because of the use of new tion on supplier delivery to control outgoing quality.
technology or new inspectors. If the inspection is perfect Lee and Li (2018) model the interaction between a buyer
such that the platform can identify all the counterfeits, it and a supplier using relational contracts in which penal­
is optimal for the platform to remove the penalty fee ties for quality failures are set so that the supplier volun­
completely. tarily pays them. Different from these papers, our paper
We find that the equilibrium results and managerial falls into the category of adverse selection in games with
insights do depend on whether the seller is certain incomplete information rather than moral hazard. Speci­
about the product authenticity. Specifically, if the seller fically, we consider that the seller who has better (but
knows whether the product is authentic or counterfeit, not perfect) information about product quality tries to
the platform can design a contract to screen out the signal that information to the buyer with or without
seller with a counterfeit. In this case, the product on the an inspection. This is not analyzed in this stream of
platform would be authentic, and the platform’s profit literature.
and optimal commission would not depend on the Our work is closely related to signaling quality infor­
inspection capability. In contrast, if the seller is uncer­ mation. Several different forms of quality signals have
tain about the product authenticity, products with dif­ been examined in the literature, including advertising
ferent authenticity levels may be sold through the (Kihlstrom and Riordan 1984, Milgrom and Roberts
platform, reflecting what happens on C2C platforms 1986), pricing (Bagwell and Riordan 1991, Stiving, 2000),
selling second-hand collectible goods (such as StockX, warranties (Lutz 1989, Boulding and Kirmani 1993),
Goat, and Poizon). money-back guarantee (Moorthy and Srinivasan 1995),
The rest of this paper is organized as follows. Section 2 umbrella branding (Wernerfelt 1988), scarcity (Stock and
reviews the related literature. Section 3 describes the Balachander 2005), queues (Debo et al. 2012), and brand
two-stage model. Section 4 solves the signaling game in extension (Moorthy 2012). Kirmani and Rao (2000) pro­
the second stage. Section 5 derives the platform’s optimal vide a comprehensive review of the literature on signal­
contract in the first stage. Section 6 conducts sensitivity ing quality. Yu et al. (2015) show that capacity rationing
analysis both analytically and numerically. Section 7 ana­ in advance selling can be an effective signal of quality
lyzes the case where the seller is certain about the prod­ and use Pareto dominance to select equilibrium. Differ­
uct authenticity. Section 8 discusses managerial insights ent from this stream of literature, we investigate how a
 Li, Fang, and Lim: Asymmetric Information of Product Authenticity on Platforms
634 Manufacturing & Service Operations Management, 2023, vol. 25, no. 2, pp. 631–647, © 2023 INFORMS

platform’s inspection affects quality signaling between a these papers, a manufacturer either produces authentic
seller and a buyer on the platform. To the best of our products or counterfeits. In contrast, the seller is not cer­
knowledge, our paper is the first to analyze the impact tain about the product authenticity in our model.
of inspection in a signaling setting. In addition, different Finally, our paper is also related to the literature on
from the literature, we consider a contract design prob­ platform business models. Some papers in this stream
lem governing a signaling game. The platform’s optimal focus on the tradeoff between platform business mod­
contract depends on the equilibrium of the signaling els and traditional alternatives: a marketplace versus a
Downloaded from [Link] by [[Link]] on 07 March 2024, at 10:29 . For personal use only, all rights reserved.

game between the seller and the buyer. This makes our reseller in Hagiu and Wright (2015) and Tian et al.
model technically challenging to solve. Finally, most of (2018), a platform versus a vertically integrated firm in
the quality signaling papers consider only two quality Hagiu and Wright (2018), and agency selling versus
types: a high type or a low type. In contrast, we assume reselling in Abhishek et al. (2016). Jiang et al. (2011) con­
the seller’s type is continuous, representing a probability sider that an e-retailer can function as a platform, and
of the product authenticity. This model element reflects at the same time, sell directly. They find that function­
sellers’ uncertainty about product authenticity on C2C ing as a platform enables the e-retailer to learn about
platforms especially when selling second-hand collect­ the demand for long-tail products sold by its enrolled
ible goods. sellers and then to cherry-pick the successful ones for
Although our model can be applied to general quality direct selling. Mantin et al. (2014) show that by intro­
problems on C2C platforms, product counterfeiting is a ducing the platform format, a traditional retailer will
major cause of these quality problems, which relates our create an outside option that improves its bargaining
paper to the literature on counterfeit products. Wilcox power in a negotiation with a manufacturer. Several
et al. (2009) consider how luxury brands express them­ papers consider peer-to-peer platforms where each
selves affects consumers’ desire for counterfeits. Zhang consumer can be a supplier or a buyer (Fraiberger and
et al. (2012) study how nondeceptive counterfeits affect Sundararajan 2015, Jiang and Tian 2016, Benjaafar et al.
the price, market share, and profitability of brand name 2018, Tian and Jiang 2018, Abhishek et al. 2021). To the best
products. Hu et al. (2013) consider a distribution channel of our knowledge, none of the papers in this stream ana­
consisting of a supplier who offers all-unit quantity dis­ lyze the contract design of a platform that offers an inspec­
counts and a reseller who may divert some goods to the tion service as in our paper. The seller in our model pays a
gray markets. Qian (2014) shows that fake products can
penalty if the product fails the inspection. Thus, the con­
benefit a manufacturer of high-quality products during
tract that we study has its particular structure including
the early stage of brand development. Cho et al. (2015)
both the commission and the penalty, which is different
examine the effectiveness of anticounterfeit strategies
from the settings in the literature.
against both deceptive and nondeceptive counterfeiters.
Gao et al. (2017) find that copycats with a high physical
resemblance but low product quality are more likely to 3. Problem Formulation
successfully enter the market. Pun and DeYong (2017) We consider a C2C e-commerce platform, a seller, and
examine the decisions of a manufacturer and a copycat a buyer. The platform (such as StockX) serves as an
firm who are competing for strategic customers. Yao online marketplace and provides a product inspection
and Zhu (2018) study the economic impact of anticoun­ service. The platform first designs a contract, and then
terfeit technology and explore measures to help increase the seller decides whether to accept it. If the seller
the authentic company’s profit. Pun et al. (2021) study accepts the contract, the seller sells a product through
how blockchain technology can be used to combat coun­ the platform by setting a price. After seeing the price,
terfeiting. Tang et al. (2020) study online platforms’ antic­ the buyer decides whether to purchase the product.
ounterfeit efforts to deter the entry of counterfeits under The seller’s product is either authentic (θ � 1) with pro­
different business models. Guan et al. (2020) examine the bability λ or counterfeit (θ � 0) with probability (1 � λ),
impact of supplier copycatting in the presence of infor­ where λ ∈ [0, 1]. As discussed in Section 1, the seller is
mation asymmetry on product fit. Yi et al. (2022) investi­ not certain about the product authenticity, but the seller
gate how counterfeits influence a global supply chain knows the probability λ as private information. We call
and how the supply chain should effectively take antic­ λ the type of the seller. Following the literature of sig­
ounterfeit actions. Chen and Papanastasiou (2021) exam­ naling games (Kirmani and Rao 2000), we assume that
ine the interaction between social-learning manipulation the platform and the buyer only know the probability
and equilibrium market outcomes and the impact of anti­ distribution of λ over [0, 1] with probability density
manipulation measures. Wang et al. (2020) provide a lit­ function (p.d.f.) f (·) and cumulative distribution func­
erature review on counterfeiting and gray market in tion (c.d.f.) F(·). We call this probability distribution the
operations management. None of these papers analyze common prior belief of the platform and the buyer about
how a platform’s inspection affects a seller’s signaling of λ. Except the seller’s type λ, all the information is pub­
product authenticity as in our paper. Furthermore, in lic to all the three parties.
 Li, Fang, and Lim: Asymmetric Information of Product Authenticity on Platforms
Manufacturing & Service Operations Management, 2023, vol. 25, no. 2, pp. 631–647, © 2023 INFORMS 635

Figure 1. Event Flow After the Buyer Purchases the Product

Downloaded from [Link] by [[Link]] on 07 March 2024, at 10:29 . For personal use only, all rights reserved.

Let µS and µB represent the reservation prices of the passes the buyer’s inspection, whereas a counterfeit
seller and the buyer, respectively, for an authentic prod­ product fails the buyer’s inspection with probability
uct. In contrast, for a counterfeit product, we assume PB ∈ [0, 1]. If the buyer finds that the received product
that the reservation prices of the seller and the buyer are is a counterfeit, an external failure occurs and the plat­
zero. Thus, the seller’s expected reservation price is λµS , form incurs an external failure loss ℓ, which includes
which depends on the type λ. If the seller chooses not to the loss due to customer dissatisfaction (Baiman et al.
sell through the platform, then the seller can earn a reve­ 2000). The product cannot be returned because it is dif­
nue that is equal to the expected reservation price through ficult for the buyer to provide evidences that the prod­
an outside option. Knowing the type λ, the seller makes a uct passing the platform’s inspection is a counterfeit.
pricing decision π(λ) : [0, 1] → [0, µB ] ∪ {∞}, where ∞ is For example, StockX does not allow product returns.
a formidable price representing that the seller does not Figure 1 illustrates the event flow after the buyer pur­
wish to sell the product on the platform. After observing chases the product.
the price π, the buyer makes a purchase decision I(π) : There are two stages in the decision-making process.
[0, µB ] → {0, 1}, where I(π) equals one if the buyer pur­ In the first stage, the platform chooses the contract
chases the product and zero otherwise. parameters α and β. After knowing the contract (α, β),
If the buyer decides to purchase the product, the the seller decides whether to sell the product through
buyer pays the price π to the platform. The seller then the platform by setting the price π based on the type λ
sends the product to the platform for an inspection. in the second stage. After observing the price π, the
Inspired by the quality inspection literature (Baiman buyer updates the belief about the seller’s type and
et al. 2000, Hwang et al. 2006, Lee and Li 2018), we decides whether to purchase the product. Figure 2
assume that an authentic product always passes the shows the sequence of the decisions.
inspection. In contrast, a counterfeit product fails the Given a contract (α, β), the interaction between the
inspection with probability PM, where PM ∈ [0, 1] repre­ seller and the buyer forms a signaling game. Following
sents the platform’s inspection capability. We assume a backward induction, we first analyze the equilibrium
that the inspection is conducted by an independent and of the signaling game in the second stage and then
professional team and thus cannot be manipulated by identify the platform’s optimal contract in the first stage
the platform. taking into account the impact of the contract on the
If the product passes the inspection, then the plat­ signaling game.
form sends the product to the buyer. The platform
keeps a fraction α of the payment π as a commission 4. Second Stage: Signaling Game
and remits the remaining fraction (1 � α) to the seller. If We first analyze the equilibrium of the second-stage
the product fails the inspection, the seller pays a frac­ signaling game, which is a dynamic game with incom­
tion β of π
to the platform as a penalty (see, for exam­ plete information. In this game, the seller uses the price
ple, [Link]). The product is then returned to the
seller and the platform refunds the payment π to the Figure 2. Decision Sequence
buyer. To make our model general, we assume α ∈
(�∞, 1] and β ∈ [0, ∞). For the inspection, the platform
incurs an inspection cost c to handle and inspect the
product. To avoid uninteresting equilibria, we assume
0 ≤ c < µB � µS .
After receiving the product, the buyer may inspect it.
Likewise, we assume that an authentic product always
 Li, Fang, and Lim: Asymmetric Information of Product Authenticity on Platforms
636 Manufacturing & Service Operations Management, 2023, vol. 25, no. 2, pp. 631–647, © 2023 INFORMS

to signal the product’s authenticity, and then the buyer and only if RB (1; π) ≥ RB (0; π) � 0, which is equivalent
decides whether to purchase the product. to
We use the concept of perfect Bayesian equilibrium (see (1 � PM )π
its definition in Online Appendix C) to analyze the E[λ | π] ≥ q(π)¢ , (1)
µB � PM π
game. Following the literature of signaling games (Bag­
R1
well and Riordan 1991), we focus on pure-strategy where E[λ |π]¢ 0 xm(x| π)dx is the seller’s expected
equilibria. Moreover, if multiple equilibria exist, we
Downloaded from [Link] by [[Link]] on 07 March 2024, at 10:29 . For personal use only, all rights reserved.

type from the buyer’s perspective after the buyer

apply Pareto dominance to select an equilibrium (Fuden­ observes the price π, and q(π) is the critical level of the
berg and Tirole 1991, Yu et al. 2015). Specifically, we seller’s expected type.
choose an equilibrium that Pareto dominates others
from the seller’s point of view (in this equilibrium, each Lemma 1 (Buyer’s Equilibrium Decision). After observing
seller type λ obtains a revenue no less than what it the price π, the buyer purchases the product (i.e., I∗ (π) � 1)
earns in any other equilibria). Because the seller makes if and only if the seller’s expected type E[λ |π] is no less
a decision first in the signaling game, it is reasonable to than its critical level q(π).
assume that the seller can choose the equilibrium most To analyze the seller’s decision, we need to determine
appealing to himself. Such an equilibrium is supported the seller’s expected revenue. Based on Figure 1, the
by behavioral experiments (Van Winden 1999). Last, to seller obtains the revenue (1 � α)π if the product passes
avoid trivialities, we focus on participating equilibria (Jans­ the inspection with probability λ + (1 � λ)(1 � PM ), but
sen et al. 2005, Yu et al. 2015). Specifically, we assume that pays the penalty βπ if the product fails the inspection
in an equilibrium the seller would not adopt price 0 or a with probability (1 � λ)PM . Thus, if the seller chooses
price that is rejected by the buyer. In these cases, it is more to sell the product with price π ∈ (0, µB ], the seller’s
profitable for the seller to choose not to sell through the expected revenue is RS (π; λ, I) � [(1 � PM + PM λ)(1 � α)
platform. Therefore, we focus on a set of equilibria in π � (1 � λ)PM βπ]I. If the seller chooses not to sell
which the seller either chooses not to sell through the plat­ through the platform (π � ∞), the seller’s expected revenue
form (π � ∞) or adopts a positive price (0 < π ≤ µB ) that is RS (π; λ, I) is equal to the expected reservation price λµS .
acceptable to the buyer.
Next, we derive the equilibrium of the second-stage
Different from the literature on price signaling, we
signaling game. We will prove in Section 5 that the plat­
investigate the effect of inspection on the signaling
form’s optimal contract does not choose a small penalty
game. We first present the equilibrium results with the
fraction (β < 1�P PM (1 � α)). Thus, we present the equilib­
M
platform providing the inspection service in Section
rium analysis under a small penalty fraction in Online
4.1. After that, we compare these results with a case in
Appendix A.1 and focus on the equilibrium under a
which the platform does not provide the inspection ser­
large penalty fraction (β > 1�P PM (1 � α)). Specifically, we
M
vice in Section 4.2.
follow the three steps illustrated in Figure 3 to derive
the equilibrium of the signaling game.
4.1. Equilibrium Results of the Signaling Game
We first analyze the purchase decision of the buyer.
4.1.1. Step 1: Reduce the Space of Possible Equilib­
According to Figure 1, the probability for the product to
ria. In contrast to signaling games with discrete types,
pass the platform’s inspection is λ + (1 � λ)(1 � PM ).
the seller’s type is continuous in our model. As a result,
Thus, the expected cost for the buyer to purchase the
the seller’s pricing decision is a function π(λ) : [0, 1]
product is (1 � PM + PM λ)π. Given π ∈ (0, µB ] and λ ∈
→ (0, µB ] ∪ {∞}. There is an infinite number of possible
[0, 1], the buyer’s utility is RB (I; π, λ) � [λµB � (1 � PM
functions mapping from [0, 1] to (0, µB ] ∪ {∞}. Thus,
+ PM λ)π]I, where λµB represents the buyer’s expected
we identify some necessary conditions of the seller’s
reservation price. After observing the price π, the buyer
equilibrium decision π∗ (λ) in Lemma 2, which will sub­
generates a posterior belief about the seller’s type λ, which
stantially simplify our subsequent analysis. Given (α,
is a probability distribution over [0, 1] with p.d.f. m(· |π). PM β�(1�PM )(1�α)
R1 β), define a constant λ̂¢ and a function
The buyer’s expected utility is RB (I; π) � 0 RB (I; π, x)m PM (1�α+β)
λ̂µS
(x | π)dx. The buyer purchases the product (I∗ (π) � 1) if A(λ̂)¢ P .
M (1�α+β)λ̂+(1�PM )(1�α)�PM β

Figure 3. Three Steps to Derive the Equilibrium of the Signaling Game

 Li, Fang, and Lim: Asymmetric Information of Product Authenticity on Platforms
Manufacturing & Service Operations Management, 2023, vol. 25, no. 2, pp. 631–647, © 2023 INFORMS 637

Lemma 2 (Structures of the Seller’s Equilibrium Deci­ F(x) � 1 � F(x). The buyer’s on-equilibrium belief m(· | π̂)
sion). If β > 1�P
PM (1 � α), the seller’s equilibrium decision
M
(the belief associated with the price adopted on the
has one of the following forms: equilibrium path) should satisfy the Bayesian rule:
(i) The price π∗ (λ) � ∞ if λ < λ ˆ and π∗ (λ) � π
ˆ if λ ≥ λ̂, m(x | π̂) equals f (x)=F(λ̂) if x ≥ λ̂, and zero otherwise.
where R1 R1
Thus, E[λ | π̂] � 0 xm(x| π̂)dx � λ̂ xf (x)dx=F(λ̂). Ac­
λ̂ < λ̂ ≤ 1, π̂ � A(λ̂): (2)
cording to Lemma 1, I∗ (π̂) � 1 if and only if E[λ | π̂]
Downloaded from [Link] by [[Link]] on 07 March 2024, at 10:29 . For personal use only, all rights reserved.

(ii) The price π∗ (λ) � ∞ for 0 ≤ λ ≤ 1. µ E[λ | π̂]

≥ q(π̂), which is equivalent to π̂ ≤ 1�PMB+PM E[λ | π̂]. There­
Lemma 2 implies that the seller, who chooses to sell, fore, we obtain
adopts the same equilibrium price π ˆ regardless of the R1
type. This is because if the seller chooses to sell, the sell­ µB λ̂ xf (x)dx
π̂ ≤ Π̂(λ̂)¢ R 1 , λ̂ ∈ [0, 1), (3)
er’s expected revenue RS (π; λ, I) increases with π as
λ̂ (1 � PM + PM x)f (x)dx
long as the price is acceptable to the buyer (i.e.,
I∗ (π) � 1). Thus, to maximize the expected revenue, the where Π̂(λ̂) represents the expected value of the prod­
seller, who chooses to sell, should adopt the highest uct from the buyer’s perspective given the posterior
price acceptable to the buyer regardless of the type. In belief about the seller type. The buyer is willing to pur­
other words, a strategy profile with different seller
chase the product with a price no higher than Π̂(λ̂)
types adopting more than one selling price is not an ˆ chooses to
equilibrium in our model, because the seller types when only the high-type seller with λ ≥ λ
adopting the lower prices always have incentives to sell. The definition of Π̂(λ̂) is not applicable to λ̂ � 1
deviate to the highest price. Thus, only semi-pooling because its denominator is zero at λ̂ � 1 (see (3)). For
equilibria (part 1 of Lemma 2) with one selling price or consistency, we define Π̂(1) � limλ̂→1� Π̂(λ̂) � µB . Intui­
pooling equilibria (part 2 of Lemma 2) may arise in this tively, when only the seller with type λ � 1 (authentic
signaling game. This is different from the literature on for sure) chooses to sell, the buyer is willing to purchase
separating equilibrium (Bagwell and Riordan 1991, Yu the product with a price no higher than the reservation
et al. 2015) because the seller’s revenue in our problem price µB for an authentic product. Therefore, we obtain
can be separated in terms of the seller type λ and the Lemma 3.
signal π (see the proof of Lemma 2 for details).
Lemma 3. In a semi-pooling equilibrium where π∗ (λ) � ∞
We further explain the structure of the equilibrium ˆ and π∗ (λ) � π
if λ < λ ˆ if λ ≥ λ̂, the equilibrium price satis­
pricing decision specified in Lemma 2 as follows. If the
fies π̂ ≤ Π̂(λ̂).
penalty fraction β is large, part 1 of Lemma 2 shows that
there may exist a semi-pooling equilibrium, in which the
4.1.2. Step 2: Find All the Equilibria Within the Reduced
seller with type larger than a threshold (λ ≥ λ̂) chooses Space. Lemmas 2 and 3 characterize necessary condi­
to sell with the same equilibrium price π̂, whereas the tions that an equilibrium needs to satisfy. This signifi­
seller with type smaller than the threshold refuses to cantly reduces the space of possible equilibria. Next, we
sell. In this semi-pooling equilibrium, the seller can par­ identify all the equilibria within this reduced space. Spe­
tially reveal the type by the pricing decision. Specifically, cifically, we prove that any pair of λ ˆ and π
ˆ satisfying
the equilibrium price π ˆ signals that the seller’s type is
no less than λ̂. Let R̂ S (λ) � RS (π̂; λ, I∗ (π̂)) � [(1 � PM +
PM λ)(1 � α) � (1 � λ)PM β]π ˆ represent the seller’s ex­ Figure 4. Seller’s Expected Revenue R̂ S (λ) by Selling with
pected revenue by adopting the equilibrium price π̂. In Price π
ˆ

Figure 4, the bold solid line represents R̂ S (λ), and the

thin solid line represents the seller’s expected reserva­
tion price λµS . The two lines intersect at λ̂, which is
implied by the condition π̂ � A(λ̂) in (2), where type-λ ˆ
seller is indifferent between selling with price π ˆ and not
selling. The seller with λ < λ ˆ refuses to sell through the
platform because the seller will earn less than the expected
reservation price by selling with price π̂. In contrast, the
seller with λ ≥ λ ˆ earns more than the expected reservation
price λµS by selling with price π̂, and the seller chooses to
sell through the platform.
Next, we identify the condition on the equilibrium
price π ˆ such that it is acceptable to the buyer. Denote
 Li, Fang, and Lim: Asymmetric Information of Product Authenticity on Platforms
638 Manufacturing & Service Operations Management, 2023, vol. 25, no. 2, pp. 631–647, © 2023 INFORMS

Conditions (2) and (3) can lead to a semi-pooling equilib­ There exists a unique semi-pooling equilibrium for
rium with appropriately specified off-equilibrium be­ the signaling game: The seller chooses to sell the product
∗
liefs, and a pooling equilibrium also exists. Because of with the same equilibrium price π̂ ∗ if λ ≥ λ̂ , but the
space limitations, we summarize all the equilibria for ∗
seller chooses not to sell through the platform if λ < λ̂ .
different contract parameter values in Lemma C.1 in ∗
The equilibrium price π̂ ∗ attains its upper bound Π̂(λ̂ )
Online Appendix C.
in Lemma 3. The equilibrium price π̂ ∗ reveals that the
Downloaded from [Link] by [[Link]] on 07 March 2024, at 10:29 . For personal use only, all rights reserved.

∗
seller’s type is no less than λ̂ and the on-equilibrium
4.1.3. Step 3: Refine the Equilibria by Pareto Domi­ belief m∗ (· | π) for π � π̂ ∗ satisfies the Bayesian rule. The
nance. Within the reduced space, still there are infinitely off-equilibrium belief is specified for the price π > π̂ ∗ so
many equilibria (see part 1 of Lemma C.1 in Online that the expected type of the seller conditioned on the
Appendix C). Among these equilibria, we choose an equi­ price π cannot be too high. Although the off-equilibrium
librium that Pareto dominates the others such that in this belief can be arbitrary for the price π < π̂ ∗ , the buyer’s
decision in (4) should be consistent with the buyer’s
equilibrium, each seller type λ obtains a revenue no less
belief. We provide additional details of the beliefs in
than what it earns in any other equilibria. We prove in
Online Appendix C.
Theorem 1 that after refining the equilibria by Pareto
Theorem 1 shows that the inspection, even an imper­
dominance, there exists a unique semi-pooling equilib­ ∗
∗ fect one, incentivizes the high-type seller with λ ≥ λ̂ to
rium determined by λ̂ , which is the unique root of
∗ sell through the platform. However, this can only be
A(λ̂) � Π̂(λ̂). Although the equilibrium threshold λ̂ can­ achieved by carefully choosing the commission fraction
not be expressed as a closed-form function of α and β, we α and penalty fraction β. Specifically, the equilibrium in
are able to derive a set of (α, β) that induces the equilib­ ∗
n which the seller with λ ≥ λ̂ chooses to sell is sustained
∗ ∗ µ ∗
rium with λ̂ . Define Ω(λ̂ ) � (α, β) |α < 1 � Π̂(λ̂S ∗ ) , β � by contract parameters (α, β) ∈ Ω(λ̂ ). Figure 5 illustrates
∗ ∗ ∗ o ∗ ∗ ∗
(1�PM +PM λ̂ )(1�α)Π̂(λ̂ )�λ̂ µS
∗ ∗
∗ ∗
for 0 < λ̂ < 1, and Ω(λ̂ ) � the set Ω(λ̂ ) on the α-β plane. Given λ̂ ∈ [0, 1], Ω(λ̂ ) is
(1�λ̂ )PM Π̂(λ̂ )
n o ∗
a ray (without the end point) in the dotted region of Fig­
µ
(α, β) |α � 1 � Π̂(λ̂S ∗ ) , β > 1�P
PM
M
(1 � α) for λ̂ � 1. Theo­ ure 5, where α is small and β is large. All (α, β) points on
∗
rem 1 summarizes the complete equilibrium result of the the ray Ω(λ̂ ) can induce the equilibrium in Theorem 1.
signaling game after the refinement. n o
µ
Define Ω(1+ ) � (α, β) | α > 1 � µS , β > 1�P PM (1 � α) ,
M
B
Theorem 1 (Refined Equilibrium of the Signaling Game). which corresponds to the shaded region in Figure 5.
If the platform provides the inspection service (PM > 0),
∗ ∗ Corollary 1 summarizes the equilibrium in Ω(1+ ).
under the contract (α, β) ∈ Ω(λ̂ ) and λ̂ ∈ (0, 1], there exists
a unique pure-strategy perfect Bayesian equilibrium that Corollary 1. If the platform provides the inspection service
Pareto dominates other equilibria from the seller’s point of (PM > 0), under the contract (α, β) ∈ Ω(1+ ), the seller will
view. We have the following results in this equilibrium. not sell through the platform with π∗ (λ) � ∞ for 0 ≤ λ ≤ 1.

The seller’s pricing decision is Figure 5. (Color online) Equilibria Under Different Contract
( ∗
∞, if λ < λ̂ ; Parameter Values
∗
π (λ) � ∗ ∗
π̂ ∗ ¢Π̂(λ̂ ), if λ ≥ λ̂ :
The buyer’s purchase decision is
8
< I(E[λ| π] ≥ q(π)),
> if 0 < π < π̂ ∗ ;
∗
I (π) � 1, if π � π̂ ∗ ; (4)
>
:
0, if π̂ ∗ < π ≤ µB :
The buyer’s beliefs m∗ (· |π) are determined as follows:
(i) If 0 < π < π̂ ∗ , then m∗ (x | π) is arbitrary.
(ii) If π � π̂ ∗ , then
8 ∗
>
< 0, if x < λ̂ ;
m∗ (x |π) � f (x) ∗
>
: ∗ , if x ≥ λ̂ :
F(λ̂ )
∗ ∗
R 1 (iii)∗ If π̂ < π ≤ µB , then m (x |π) satisfies E[λ |π] �
0 xm (x| π)dx < q(π).
 Li, Fang, and Lim: Asymmetric Information of Product Authenticity on Platforms
Manufacturing & Service Operations Management, 2023, vol. 25, no. 2, pp. 631–647, © 2023 INFORMS 639

Intuitively, a larger commission fraction α or penalty R λ̌

µB 0 xf (x)dx
fraction β leads to a smaller set of seller types that want Π̌(λ̌) � F(λ̌)
for 0 < λ̌ ≤ 1 and Π̌(0) � limλ̌→0+
to sell through the platform. For large α and β as given ∗
ˇ
Π̌(λ̌) � 0. Define λ̌ � max{λ̌ | (1 � α)Π̌(λ̌) ≥ λ̌µS , 0 ≤ λ
in the shaded region Ω(1+ ) in Figure 5, the seller earns ≤ 1}.
less than the expected reservation price by adopting
any price π ∈ (0, µB ]. Thus, the seller chooses not to sell Proposition 1. If the platform does not provide the inspec­
tion service (PM � 0), there exists a unique pure-strategy
Downloaded from [Link] by [[Link]] on 07 March 2024, at 10:29 . For personal use only, all rights reserved.

through the platform regardless of the type.

Theorem 1 and Corollary 1 analyze the equilibria of perfect Bayesian equilibrium that Pareto dominates other
the second-stage signaling game under a large penalty equilibria from the seller’s point of view. We have the fol­
lowing results in this equilibrium.
fraction (β > 1�P
PM (1 � α)), and next we analyze the equi­
M

The seller’s pricing decision is

libria of boundary cases (β � 1�P PM (1 � α)). The boundary
M
( ∗ ∗
cases consist of the following three parts: (i) Ω(0) � ∗ π̌ ∗ ¢Π̌(λ̌ ), if λ ≤ λ̌ ;
n o π (λ) � ∗
µS
(α, β) |α < 1 � Π̂(0) , β � 1�PM
(1 � α) , which is a ray ∞, if λ > λ̌ :
PM
∗ The buyer’s purchase decision is
(without its end point) shown in Figure 5; (ii) V(λ̂ ) �
n o 8 � �
µ
(α, β) |α � 1 � Π̂(λ̂S ∗ ) , β � 1�PM
(1 � α) , which is the end >
> π
PM
< I E[λ |π] ≥ µ ,
> if 0 < π < π̌ ∗ ;
∗ ∗ B
point of the ray Ω(λ̂ ), and the dashed line in Figure 5 I (π) �
∗ >
> 1, if π � π̌ ∗ ;
represents all such end points for 0 ≤ λ̂ ≤ 1; and (iii) >
:
n o 0, if π̌ ∗ < π ≤ µB :
µS
V(1+ ) � (α, β) |α > 1 � Π̂(1) , β � 1�P
PM
M
(1 � α) , which is
The buyer’s beliefs m∗ (· |π) are determined as follows:
the bottom boundary of the shaded region in Figure 5.
(i) If 0 < π < π̌ ∗ , then m∗ (x |π) is arbitrary.
Corollary 2 summarizes the equilibria of the boundary
(ii) If π � π̌ ∗ , then
cases. 8
Corollary 2. Suppose the platform provides the inspection < f (x)∗ , if x ≤ λ̌ ∗ ;
>
∗
service (PM > 0). m (x |π) � F(λ̌ )
>
: ∗
(i) If (α, β) ∈ Ω(0), the seller chooses to sell with price 0, if x > λ̌ :
π∗ (λ) � Π̂(0) for λ ∈ [0, 1]. (iii) If π̌ ∗ < π ≤ µB , then m∗ (x | π) satisfies E[λ | π] �
∗ R1
(ii) If (α, β) � V(λ̂ ), there exist multiple equilibria. Among ∗
(x |π)dx < µπ :
0 xm
these equilibria, the seller earns the same expected revenue for B

λ ∈ [0, 1]. If the platform does not inspect the product, then the
(iii) If (α, β) ∈ V(1+ ), the seller will not sell through the ∗
low-type seller with λ ∈ [0, λ̌ ], who likely holds a
platform with π∗ (λ) � ∞ for λ ∈ [0, 1]. counterfeit product, chooses to sell the product through
∗ the platform in the equilibrium. This is because for the
For (α, β) � V(λ̂ ), the unique equilibrium identified
∗ low-type seller with a small λ, it is easy to earn a reve­
by Theorem 1, in which π∗ (λ) � ∞ for λ < λ̂ and π∗ (λ)
∗ ∗ nue that matches the expected reservation price λµS .
� π̂ ∗ � Π̂(λ̂ ) for λ ≥ λ̂ , is in the set of equilibria in part This phenomenon is similar to the classical “adverse
2 of Corollary 2. We select this equilibrium as the refined selection.” That is, the quality of goods traded in a mar­
∗
equilibrium for (α, β) � V(λ̂ ) to facilitate the description ket can decay in the presence of information asymme­
of the optimal contract in Section 5. try (Akerlof 1970). In contrast to Akerlof (1970), who
only considers discrete quality levels, the seller’s prod­
4.2. Effect of the Inspection Service on the uct authenticity level λ is continuous in our model. It is
Seller’s Decision worth noting that only the low-type seller sells the
What is the impact of the platform’s inspection on the product in Proposition 1, implying that without the
seller’s decision? To answer this research question in Sec­ inspection service, there is a high chance for the buyer
tion 1, we analyze the case without the inspection service, to receive a counterfeit product. This is not equivalent to
and compare it with our base model with the inspection. only counterfeits are sold on the platform. Our result is
After the buyer purchases the product, without the consistent with the observations in Section 1.
inspection, the seller directly sends the product to the Recall from Theorem 1 that if the platform provides
buyer (for example, see [Link] and [Link]). This the inspection service and uses an appropriate contract
means PM � 0 and c � 0 in our model. The commission (α, β), only the high-type seller sells the product through
fraction α is the only contract parameter in this case. the platform, which is opposite to the result in Proposi­
Proposition 1 summarizes the equilibrium result tion 1. This is because without the inspection service, the
for the case without the platform’s inspection. Define platform screens the seller only through the commission
 Li, Fang, and Lim: Asymmetric Information of Product Authenticity on Platforms
640 Manufacturing & Service Operations Management, 2023, vol. 25, no. 2, pp. 631–647, © 2023 INFORMS

fraction α. In contrast, with the inspection service, the the seller. The following proposition analyzes how the
∗ ∗
platform can screen the seller with an additional contract platform’s contract parameters (α, β) ∈ Ω(λ̂ ) ∪ V(λ̂ )
parameter (penalty fraction β). If the platform uses a con­ affect the seller’s expected revenue R∗S (λ) �
tract (α, β) in the dotted region of Figure 5, Theorem 1 RS (π∗ (λ); λ, I∗ ) in the second-stage equilibrium.
shows that the seller with a low type cannot earn a reve­
Proposition 2. If the platform provides the inspection ser­
nue matching the expected reservation price and chooses
vice (PM > 0), then we have the following results.
not to sell. As the product authenticity level on the plat­
Downloaded from [Link] by [[Link]] on 07 March 2024, at 10:29 . For personal use only, all rights reserved.

∗
form improves, the buyer is willing to pay a higher price, 1. Given λ̂ ∈ [0, 1), as α decreases and β increases along
∗
which attracts a higher-type seller whose expected reser­ the ray Ω(λ̂ ), the seller’s expected revenue R∗S (λ) in the
∗
vation price is higher. This increases the authenticity level second-stage equilibrium increases for λ ∈ [λ̂ , 1].
∗ ∗
and selling price even further, thus attracting the seller 2. Given λ̂ ∈ [0, 1] and the contract (α, β) � V(λ̂ ), the
with an even higher type. Comparing Theorem 1 and seller’s expected revenue R∗S (λ) in the second-stage equilibrium
Proposition 1 shows that the effect of the inspection ser­ ∗
equals the expected reservation price λµS for λ ∈ [λ̂ , 1].
vice is beyond merely detecting counterfeit products. It
also improves product authenticity by altering the seller’s Proposition 2 shows that the platform with the in­
equilibrium decision. spection service can alter the seller’s expected revenue
The previous comparison implies that the effect of the R∗S (λ) in the second-stage equilibrium by adjusting the
inspection service may be larger than one would expect. contract parameters α and β. Specifically, as α
decreases
∗
Specifically, Babich and Tang (2012) show that even and β increases along the ray Ω(λ̂ ) in Figure 5, the
∗
with a sufficiently high inspection capability, the seller high-type seller with λ ∈ [λ̂ , 1] earns more revenue in
cannot be completely deterred from product adultera­ the equilibrium. Moreover, if the platform adopts the
∗
tion in a moral hazard setting. In contrast, in a signaling contract (α, β) � V(λ̂ ), which corresponds to the end
∗
setting, we find that the inspection can completely re­ point of the ray Ω(λ̂ ) in Figure 5, the seller’s expected
verse the seller’s decision even with a low inspection revenue equals the expected reservation price regard­
capability. Given that the inspection service is largely less of λ.
missing from leading C2C e-commerce platforms, our
Corollary 4. If the platform provides the inspection service
results suggest that the benefit of inspection in improv­
(PM > 0), the seller’s information rent is eliminated under
ing product authenticity may be underestimated in ∗
the contract (α, β) � V(λ̂ ).
practice.
Furthermore, if no inspection service is provided, the In contrast to Corollary 3, Corollary 4 shows that the
∗
low-type seller with λ ∈ [0, λ̌ ] earns more than the inspection service is a useful tool for the platform to
expected reservation price, which leads to the follow­ eliminate the seller’s information rent. Moreover, we
ing corollary. can easily prove that the inspection service can increase
the total revenue of the platform and the seller. There­
Corollary 3. If the platform does not provide the inspection
fore, by providing the inspection service, the platform
service (PM � 0), it has to pay information rent to the seller.
can generate more revenue for itself through increasing
The information rent determines how the revenue is the total “pie” and decreasing the share of the seller.
distributed between the platform and the seller. Section This may explain why some platforms begin to offer
5.1 investigates the effect of the inspection service on the inspection service in practice. For example, eBay
the platform in terms of the information rent and com­ recently starts to inspect sneakers, watches, and hand­
pares it with Corollary 3. bags from its sellers (Holt 2020).
With multiple seller types, the literature of contract
5. First Stage: Optimal Contract design considers a menu of contracts to extract the sys­
We first analyze how the platform’s contract para­ tem profit and reduce the information rent (Baron and
meters with the inspection service affect the seller’s Myerson 1982). However, we consider a single contract
expected revenue in Section 5.1 and then identify the (α, β) for all the seller types because it is impractical for
platform’s optimal contract in Section 5.2. the platform to choose continuous α and β for different
seller types in practice (see StockX and Goat). In con­
5.1. Effect of the Contract Parameters on the trast to the literature, by offering only a single contract
Seller’s Revenue for all the seller types, our platform can eliminate the
∗
Theorem 1 and Corollary 2 show that in the second- seller’s information rent if (α, β) � V(λ̂ ).
stage equilibrium generated by any contract (α, β) ∈
∗ ∗ ∗
Ω (λ̂ ) ∪ V(λ̂ ), the seller with λ ≥ λ̂ chooses to sell 5.2. Platform’s Optimal Contract
the product through the platform. However, differ­ Anticipating the equilibrium of the second-stage sig­
ent contracts lead to different expected revenues of naling game given each contract, we design the optimal
 Li, Fang, and Lim: Asymmetric Information of Product Authenticity on Platforms
Manufacturing & Service Operations Management, 2023, vol. 25, no. 2, pp. 631–647, © 2023 INFORMS 641
∗
contract of the platform. Recall from Theorem 1 and Lemma 4. Under the contract (α, β) � V(λ̂ ), the platform’s
∗ ∗
Corollary 2 that the contracts (α, β) ∈ Ω(λ̂ ) ∪ V(λ̂ ) expected profit is
lead to the equilibrium in which the seller with λ ∈ Z 1
∗ ∗ ∗
[λ̂ , 1] chooses to sell. For a given threshold λ̂ of the RM (λ̂ ) � ∗ [x(µB � µS ) � (1 � x)(1 � PM )PB ℓ � c] f (x)dx:
∗
seller type, we first find the optimal contract (α∗ (λ̂ ), λ̂
∗ ∗ ∗
β∗ (λ̂ )) in Ω(λ̂ ) ∪ V(λ̂ ) in Figure 5. Then, we identify (7)
∗∗
the optimal threshold λ̂ for the platform.
Downloaded from [Link] by [[Link]] on 07 March 2024, at 10:29 . For personal use only, all rights reserved.

∗ ∗ Lemma 4 shows that the platform’s expected profit

We first find the optimal contract (α∗ (λ̂ ), β∗ (λ̂ )) for a
∗ equals the gap of the expected reservation prices between
given threshold λ̂ such that the platform’s expected the buyer and the seller minus the expected external fail­
profit is maximized. To derive the platform’s expected ure loss and the inspection cost. Because a higher seller
profit, we first consider the total expected profit of both type results in a larger gap of the expected reservation
∗
the platform and the seller. In the equilibrium given λ̂ , prices and a smaller expected external failure loss, the
∗
the platform and the seller earn a total revenue of Π̂(λ̂ ) platform prefers the seller with a higher type. To maxi­
∗
(see Theorem 1) if the seller’s product passes the inspec­ mize its expected profit RM (λ̂ ), the platform should
∗
tion, which happens with probability 1 � PM + PM λ. On choose a proper λ̂ in (7). Theorem 2 determines the opti­
∗∗
the other hand, the platform incurs the external failure mal threshold λ̂ for the platform.
loss ℓ if the buyer finds that the product that passes the
Theorem 2 (Optimal Contract of the Platform). If the
platform’s inspection is a counterfeit, which happens platform provides the inspection service (PM > 0), the plat­
with probability (1 � λ)(1 � PM )PB (see Figure 1). More­ ∗∗ ∗∗ ∗∗
form’s optimal contract is (α∗ (λ̂ ), β∗ (λ̂ )), where λ̂ �
over, the platform incurs the inspection cost c for each (1�PM )PB ℓ+c
product inspection. Thus, in the equilibrium with thre­ (1�PM )PB ℓ+µ �µ . Furthermore, the platform’s optimal expected
B S
∗ ∗∗
shold λ̂ , the total expected profit of both the platform profit is R∗M � RM (λ̂ ).
R1 ∗
and the seller is λ̂ ∗ [(1 � PM + PM x) Π̂(λ̂ ) � (1 � x)(1 � ∗∗
Interestingly, the optimal threshold λ̂ is indepen­
PM )PB ℓ � c] f (x)dx: Note that (α, β) do not affect the total dent of the distribution F(·) of the seller type λ. This is
expected profit but determine how the total profit is dis­ appealing because the distribution F(·) may not be
tributed between the platform and the seller. accurately estimated in practice. Regardless of the dis­
According to Proposition 2, as α decreases and β tribution F(·), the platform should adopt a contract to
∗ ∗∗
increases along the ray Ω(λ̂ ) in Figure 5, the revenue attract the seller with λ
in the same set [λ̂ , 1].
∗
of the seller with λ ∈ [λ̂ , 1] increases. Because the total Because the contract in Theorem 2 is optimal for the
∗
expected profit is fixed given (α, β) ∈ Ω(λ̂ ), the plat­ platform, we have the following corollary.
form’s expected profit decreases. Therefore, the plat­ Corollary 5. Any contract with β < 1�P
PM (1 � α) (the bot­
M

form’s expected profit is the largest at the end point tom blank triangle of Figure 5) is strictly dominated by the
∗ ∗ ∗∗
(α, β) � V(λ̂ ) in Figure 5. That is, given λ̂ ∈ [0, 1], the optimal contract (α∗ , β∗ ) � V(λ̂ ) for the platform.
∗ ∗
optimal contract is (α∗ (λ̂ ), β∗ (λ̂ )), where
Some platforms, such as StockX and Goat, claim that
∗ µS their customers will receive 100% authentic products. To
α∗ (λ̂ ) � 1 � ∗ , (5) ensure that, the platform needs to have a perfect inspec­
Π̂(λ̂ ) tion service with PM � 1. In this case, the platform’s opti­
mal contract is given in the following corollary.
∗ (1 � PM )µS
β∗ (λ̂ ) � ∗ : (6) Corollary 6 (No Penalty for Perfect Inspection). If the
PM Π̂(λ̂ ) platform provides a perfect inspection service (PM � 1), then
∗∗ µ ∗∗
∗∗
c
λ̂ � µ �µ , α∗ � 1 � µS , β∗ � 0, and Π̂(λ̂ ) � µB .
Next, we identify the optimal threshold λ̂ for the plat­ B S B

form. According to Proposition 2, the seller’s expected Corollary 6 shows that the platform should remove
revenue is equal to the expected reservation price under the penalty if it has the perfect inspection service. This
∗
the contract (α, β) � V(λ̂ ). Thus, the platform’s expected result is interesting and can be explained as follows. If
∗
profit RM (λ̂ ) equals the two parties’ total expected the inspection service is perfect, an external failure will
profit minus the seller’s expected reservation price. That not happen. In this case, it is optimal for the platform to
∗ R1 ∗ remove the penalty completely by setting β∗ � 0 such
is, RM (λ̂ ) � λ̂ ∗ [(1 � PM + PM x)Π̂(λ̂ ) � (1 � x)(1 � PM )
∗ that it can attract as many seller types as possible. This
PB ℓ � c � xµS ] f (x)dx: Substituting Π̂(λ̂ ) in (3) into the benefits the platform because even the seller with low
∗
expression of RM (λ̂ ), we obtain Lemma 4. λ can potentially contribute to the platform’s profit if
 Li, Fang, and Lim: Asymmetric Information of Product Authenticity on Platforms
642 Manufacturing & Service Operations Management, 2023, vol. 25, no. 2, pp. 631–647, © 2023 INFORMS

the product turns out to be authentic. Nevertheless, the Therefore, a larger expected profit does not imply a larger
∗∗ c
seller with λ < λ̂ � µ �µ is still rejected by the plat­ commission fraction or a higher price in equilibrium.
B S
form. This is because the expected revenue generated
from such a low seller type cannot cover the platform’s 6.2. Sensitivity to Inspection Capability PM
inspection cost. According to Proposition 2, under the platform’s opti­
mal contract, the seller’s expected revenue equals the
Downloaded from [Link] by [[Link]] on 07 March 2024, at 10:29 . For personal use only, all rights reserved.

expected reservation price λµS , which does not depend

6. Sensitivity Analysis
on the inspection capability PM. Theorem 3 shows how
We now analyze how market conditions influence the
the other equilibrium results change with PM. Define
equilibrium results under the optimal contract and ans­ M )PB ℓ
∗∗ R1 R1
λ† � (1�P(1�P
M )PB ℓ+µ �µ
, ξ � F(λ̂ ) � λ̂ ∗∗ f (x)dx, and η � λ̂ ∗∗
wer our second research question in Section 1. Specifi­ B S

cally, we perform sensitivity analysis on the distribution xf (x)dx.

of λ and the platform’s inspection capability PM in Sec­ Theorem 3 (Sensitivity to PM).
tions 6.1 and 6.2, respectively. We conduct sensitivity ∗∗
1. The optimal threshold λ̂ decreases with PM. In con­
analysis on other parameters in Online Appendix A.2. ∗∗
trast, the equilibrium price Π̂(λ̂ ) increases with PM if and
only if
6.1. Sensitivity to the Seller-Type Distribution ∗∗ ∗∗ (ξ � η)η
Although Theorem 2 shows that the optimal threshold λ† (1 � λ̂ )f (λ̂ ) ≤ ∗∗ : (10)
∗∗ η � λ̂ ξ
λ̂ is independent of the seller-type distribution F(·),
the other equilibrium results under the platform’s opti­ 2. The platform’s optimal expected profit R∗M increases
mal contract depend on the distribution. Proposition 3 with PM. In contrast, the optimal commission fraction α∗
describes how the characteristics of F(·) affect these increases with PM if and only if (10) holds. The optimal pen­
equilibrium results. Consider markets 1 and 2 with alty fraction β∗ decreases with PM if and only if
seller-type distributions F1 (·) and F2 (·), respectively.
∗∗ ∗∗ P2M η2 + (1 � P2M )ξη
λ† (1 � λ̂ )f (λ̂ ) ≤ ∗∗ : (11)
Proposition 3 (Sensitivity to Seller-Type Distribution). (1 � PM )PM (η � λ̂ ξ)
1. The platform’s optimal expected profit can be written
R1
as R∗M � (µB � µS + (1 � PM )PB ℓ) λ̂ ∗∗ F(x)dx. Thus, R∗M is As the platform’s inspection capability PM increases,
larger in market 1 than in market 2 if and only if the seller faces a higher risk of paying the penalty. Intu­
Z 1 Z 1 itively, one may think that the seller is less likely to sell
F 1 (x)dx ≥ F 2 (x)dx: (8) the product through the platform. However, part 1 of
∗∗ ∗∗ ∗∗
λ̂ λ̂ Theorem 3 shows the opposite result: λ̂ is decreasing
2. The optimal commission fraction α∗ is larger, the opti­ in PM, implying that the seller is more likely to sell
mal penalty fraction β∗ is smaller, and the equilibrium price through the platform. Because it is more likely to screen
∗∗
Π̂(λ̂ ) is larger in market 1 than in market 2 if and only if out a counterfeit product as PM increases, the platform
∗∗
R1 R1 optimizes the contract to induce a lower threshold λ̂
λ̂ F 1 (x)dx
∗∗ F (x)dx
∗∗
2 of the seller type to increase the probability of a success­
∗∗ ≥ λ̂ ∗∗ : (9)
F 1 (λ̂ ) F 2 (λ̂ ) ful transaction.
Part 1 of Theorem 3 shows that the equilibrium price
∗∗
To satisfy (8), it is sufficient to have F 1 (x) ≥ F 2 (x) for all Π̂(λ̂ ) increases with PM if and only if (10) holds. To
x ∈ [0, 1] (i.e., F1 (·) is stochastically larger than F2 (·)). understand this, define input quality qI and output qual­
According to the discussion after Lemma 4, the platform ity qO as follows:
earns more profit from the seller with a larger λ. There­ R1
∗∗ xf (x)dx

fore, it is intuitive that the platform generates more profit q � λ̂

I
∗∗ , (12)
F(λ̂ )
in the market with a stochastically larger distribution. R1
According to part 1 of Proposition 3, the platform’s O λ̂ xf (x)dx
∗∗ qI
R1 q � R1 � : (13)
optimal expected profit is proportional to λ̂ ∗∗ F(x)dx. ∗∗ (1 � P
M + PM x)f (x)dx
1 � PM + PM qI
λ̂
Because the seller chooses to sell the product with prob­ The input quality qI represents the probability of receiv­
∗∗
ability F(λ̂ ), the platform generates an expected profit ing an authentic product from the seller by the platform
R1 ∗∗
proportional to λ̂ ∗∗ F(x)dx=F(λ̂ ) conditioned on the before its inspection. The output quality qO represents
seller chooses to sell. It is possible for two distributions the probability of sending an authentic product to the
F1 (·) and F2 (·) to satisfy (8) but not (9) (see the discussion buyer by the platform after its inspection. Note that
after the proof of Proposition 3 in Online Appendix C). qI ≤ qO . If PM � 0, then qO � qI ; if PM � 1, then qO � 1.
 Li, Fang, and Lim: Asymmetric Information of Product Authenticity on Platforms
Manufacturing & Service Operations Management, 2023, vol. 25, no. 2, pp. 631–647, © 2023 INFORMS 643

Figure 6. Impact of PM on the Platform’s Optimal Contract Parameters

(a) (b) (c)

Downloaded from [Link] by [[Link]] on 07 March 2024, at 10:29 . For personal use only, all rights reserved.

∗∗
On the one hand, because the optimal threshold λ̂ zero (see Corollary 6). We provide the following num­
decreases with PM, we know from (12) that qI becomes erical examples to illustrate how Conditions (10) and
lower as PM increases. On the other hand, a counterfeit (11) can be satisfied. Example 1 gives a wide range of
product is more likely to be screened out as PM increases. distributions that satisfy the conditions.
If (10) holds, the positive effect of the improved inspec­
Example 1. Consider the seller type λ
follows a Beta
tion capability dominates the negative effect of the lower
distribution B(1, 2), B(2, 1), B(2, 2), or B(0:5, 0:5), with
input quality. In that case, the output quality qO im­
PB � 0:2, ℓ � 3, c � 0.2, µS � 5, and µB � 10. Figure 6(a)
proves as the inspection capability PM increases. Accord­
∗∗ shows the p.d.f. of each distribution. Figure 6, (b) and
ing to (3), the equilibrium price Π̂(λ̂ ) � µB qO . Thus, (c), shows that the platform’s optimal contract para­
∗∗
Π̂(λ̂ ) increases with PM under Condition (10). meters α∗ and β∗ , respectively, are monotonic in the
Part 2 of Theorem 3 shows that the platform generates inspection capability PM.
more profit if its inspection capability PM increases. This In contrast, if Conditions (10) and (11) are violated,
may sound intuitive, but in fact, there are several drivers the optimal contract parameters are no longer mono­
for this result. First, both the equilibrium price and the tonic in PM. We present such an example here.
commission fraction increase with PM under Condition
(10); thus, the platform earns more commission. Second, Example 2. Consider the p.d.f. of the seller type f (x) �
∗∗ φ(x, 0:25, 0:012 )+1
because the optimal threshold λ̂ decreases with PM, the 2 where φ(x, 0:25, 0:012 ) is the p.d.f. of a
,
seller is more likely to sell through the platform. Third, it truncated normal distribution on [0, 1] with mean 0.25
becomes less likely to have an external failure if PM and standard deviation 0.01. We set PB � 0:2, ℓ � 10, c �
increases, and thus the platform is less likely to suffer 0.2, µS � 6, and µB � 10. The density is 20.45 at x � 0.25,
from the external failure loss. In summary, the combina­ which violates Conditions (10) and (11). Figure 7, (a)
tion of these three drivers implies that the platform’s and (b), shows that the platform’s optimal contract pa­
expected profit increases with PM. rameters α∗ and β∗ , respectively, are not monotonic in
Part 2 of Theorem 3 also shows that the optimal com­ the inspection capability PM.
mission fraction α∗ and penalty fraction β∗ are not always In summary, α∗ increases but β∗ decreases as PM
monotonic in PM and the monotonicity is guaranteed increases under some mild conditions (Figure 6). How­
by Conditions (10) and (11), respectively. Under these ever, if the density of a certain type is high as in Exam­
conditions, the platform should adjust the contract ple 2, then α∗ and β∗ may not be monotonic in PM
parameters accordingly if its inspection capability in­ (Figure 7). Therefore, the platform should pay close
creases. Specifically, as PM increases to 100%, α∗ in­ attention to the seller-type distribution when choosing
µ
creases to its upper bound 1 � µS and β∗ reduces to the contract parameters.
B

Figure 7. Impact of PM on the Platform’s Optimal Contract Parameters When Conditions (10) and (11) Are Violated

(a) (b)
 Li, Fang, and Lim: Asymmetric Information of Product Authenticity on Platforms
644 Manufacturing & Service Operations Management, 2023, vol. 25, no. 2, pp. 631–647, © 2023 INFORMS

7. Model with Two Seller Types from our base model. For example, if the seller knows
Sellers may know whether their products are authentic the product authenticity, an external failure would not
when product sources are clear. To investigate such a happen because only the type 1 seller chooses to sell
setting, we consider a case in which λ is either one or under the platform’s optimal contracts. Interestingly,
zero such that the seller is certain about the product this result holds for any positive platform’s inspection
authenticity in this section. Comparing this case with capability. Therefore, the platform can consider adopt­
our base model, where λ is continuous, allows us to ing the inspection service because even a low inspec­
Downloaded from [Link] by [[Link]] on 07 March 2024, at 10:29 . For personal use only, all rights reserved.

investigate the impact of the seller’s uncertainty about tion capability would eliminate external failures with a
the product authenticity on equilibrium results. This carefully designed contract. In contrast, in our base
will answer our third research question in Section 1. model, the platform may receive a counterfeit from the
In this section, we assume that λ is either one or zero, seller who is not certain about the product authenticity,
which means the seller knows that the product is authen­ and thus an external failure may happen, which reflects
tic (type 1) or counterfeit (type 0), respectively. Again, λ the reality of the second-hand market of collectible
is the seller’s private information. The platform and the goods mentioned in Section 1.
buyer have a common prior belief about λ, which is a Theorem 4 also generates different insights from our
discrete probability distribution: P(λ � 1) � w and P(λ base model. First, in our base model (see Theorem 3),
� 0) � 1 � w. The rest of the setting is the same as the the platform has an incentive to improve its inspection
base model. In our base model, the buyer has two capability (Snow 2020) as it can increase the platform’s
“layers” of uncertainty about the product authenticity as expected profit. In contrast, if the seller knows the
the buyer has a belief about λ, which itself is a probability product authenticity (see Theorem 4), the platform
of the product authenticity. In contrast, in this section, would not have an incentive to enhance its inspection
the buyer has only one layer of uncertainty about the capability because it does not impact the platform’s
product authenticity because λ � 1 or 0 completely reveals expected profit. Second, in our base model (see Theo­
the product authenticity. Theorem 4 summarizes the equi­ rem 3), the optimal commission and penalty fractions
librium results if the seller is certain about the product au­ can be either monotonic or nonmonotonic in the inspec­
thenticity. The definitions of Ω(0), Ω(1), Ω(1+ ), Ω̃, and tion capability. The platform should adjust the commis­
V(1) are provided in Online Appendix C. sion and penalty fractions accordingly if its inspection
capability changes. In contrast, if the seller knows the
Theorem 4. Suppose the seller knows the product authen­ product authenticity (Figure 8), the optimal commis­
ticity (λ � 1 or 0) and the platform provides the inspection sion fraction is independent of its inspection capability
service (PM > 0). (α∗ � 1 � µS =µB ). Therefore, the seller-type distribution
1. Under a contract (α, β) ∈ Ω(1), only the type 1 seller (continuous or discrete) plays an important role in gen­
chooses to sell through the platform with price µB. Under a con­ erating different insights.
tract (α, β) ∈ Ω(0), the seller chooses to sell through the plat­
form with price µB w=(1 � PM + PM w) regardless of the type.
Under a contract (α, β) ∈ Ω(1+ ) ∪ Ω̃, the seller does not sell
through the platform. Figure 8. (Color online) Seller’s Decision in the Signaling
Game When the Seller Knows the Product Authenticity
2. The set of optimal contracts is V(1), in which only the
type 1 seller chooses to sell through the platform with price
µB. The platform’s expected profit is w(µB � µS � c).
Figure 8 illustrates the seller’s decision in the second-
stage signaling game for different contracts. In the
upper-left region Ω(1) where α is small and β is large,
only the type 1 seller chooses to sell. In the lower-left
region Ω(0) where both α and β are small, both seller
types choose to sell. In the upper-right region Ω(1+ ) and
lower-right region Ω̃, the seller does not sell. Part 2 of
Theorem 4 shows that it is optimal for the platform to
adopt a contract in V(1) where the penalty fraction β is
large. This screens out the type 0 seller with a counterfeit
and only the type 1 seller chooses to sell (Figure 8).
Theorem 4 shows that the inspection service still
incentivizes the high-type (type 1) seller to sell as long
as the platform chooses the commission and penalty
fractions carefully. However, other results are different
 Li, Fang, and Lim: Asymmetric Information of Product Authenticity on Platforms
Manufacturing & Service Operations Management, 2023, vol. 25, no. 2, pp. 631–647, © 2023 INFORMS 645

8. Conclusion in the inspection capability if the density of a certain

We study a C2C platform functioning as a marketplace seller type is high (see Figure 7). Therefore, the platform
that provides an inspection service. Selling a product should pay attention to the seller-type distribution when
through the platform, a seller is uncertain about the designing the contract. With a perfect inspection capabil­
product authenticity. The seller is characterized by type, ity, the platform can remove the penalty completely,
representing the probability of the seller’s product being attracting more sellers who are less confident about their
authentic. Compared with the seller, a buyer has even product authenticity (see Corollary 6). For example, Goat
Downloaded from [Link] by [[Link]] on 07 March 2024, at 10:29 . For personal use only, all rights reserved.

less information about the product authenticity. Before sets a zero penalty fee (see [Link]).
the buyer purchases the product, the buyer obtains a sig­ We also analyze a model with two seller types (λ � 1
nal of the product authenticity from its selling price. The or 0), which generates different insights from our base
platform’s inspection service helps to detect whether the model. Specifically, when the seller knows the product
product is counterfeit with a probability. We develop a authenticity, the platform can adopt a contract to screen
two-stage game-theoretical model capturing the inter­ out the seller with a counterfeit given asymmetric infor­
actions among the platform, the seller, and the buyer. mation of product authenticity (see Theorem 4). Thus,
In the first stage of the game, the platform designs a the platform would not receive a counterfeit and an
contract determining a commission fraction and a pen­ external failure would not happen. Moreover, the plat­
alty fraction. In the second stage, the seller first decides form’s expected profit and the optimal commission
whether to sell the product through the platform by set­ fraction are independent of the inspection capability. In
ting its price. Given the price, the buyer then decides contrast, in our base model, the seller who is uncertain
whether to purchase the product. Our model provides about the product authenticity may sell through the
guidance and insights into the platform’s contract de­ platform, reflecting what happens on platforms selling
sign and provision of the inspection service. second-hand collectible goods (such as StockX, Goat,
We have identified the following effects of providing and Poizon). Moreover, the platform’s expected profit
the inspection service. First, even with a very low inspec­ and the optimal contract parameters in our base model
tion capability, the inspection service completely alters depend on the inspection capability.
the seller’s decision and improves the product authentic­ Online Appendix B analyzes two extensions of our
ity (see Theorem 1 and Proposition 1). Our results suggest model. The first extension assumes that the buyer has a
that the benefit of inspection service may be underesti­ positive reservation price for a counterfeit. The second
mated in practice as the service is not widely available extension considers that the buyer has a private reser­
among leading C2C e-commerce platforms. Second, with vation price for an authentic product. Our main results
the inspection service, the platform can influence the sell­ continue to hold under the two extensions. Because our
er’s revenue and generate more revenue for itself by paper is the first to analytically investigate C2C plat­
adjusting the commission and penalty fractions and even forms providing the inspection service, more future
eliminate the seller’s information rent despite the asym­ research is possible. First, it may be interesting to con­
metric information of product authenticity (see Corollar­ sider a rating system on the seller’s product authentic­
ies 3 and 4). Finally, the platform’s profit increases with ity, where the belief about the seller type is updated
its inspection capability (see Theorem 3). This partially dynamically over time. Second, one may endogenize
explains why some C2C platforms invest heavily to the platform’s inspection capability to investigate its
improve their inspection capability (Snow 2020). This in­ incentive to manipulate the inspection result. Third, in
herent incentive for the platform to improve its inspection the second stage, considering the bargaining between
capability helps combat counterfeits. the seller and the buyer, or the buyer making a price
To maximize its expected profit, the platform should offer can be interesting.
carefully choose the commission and penalty fractions
(see Theorem 2). Interestingly, a larger platform’s ex­ Acknowledgments
pected profit does not imply a larger commission frac­ The authors thank the department editor, associate editor,
tion or a higher price in equilibrium (see Proposition 3). and two anonymous referees for valuable comments that
have substantially improved the paper. Portions of this paper
Under some mild conditions, the optimal commission
were presented in the 2021 INFORMS Annual Meeting and
fraction increases but the optimal penalty fraction de­
the 12th POMS-Hong Kong International Conference. The
creases as the platform’s inspection capability increases authors thank the audiences for many insightful comments
(see Theorem 3). This partially explains why platforms and stimulating questions.
without the inspection service (such as Taobao) usually
charge a lower commission fee than platforms with the
References
inspection service (such as StockX and Goat), where the Abhishek V, Guajardo JA, Zhang Z (2021) Business models in the shar­
commission fraction can be as high as 25%. However, ing economy: Manufacturing durable goods in the presence of
the optimal contract parameters may not be monotonic peer-to-peer rental markets. Inform. Systems Res. 32(4):1450–1469.
 Li, Fang, and Lim: Asymmetric Information of Product Authenticity on Platforms
646 Manufacturing & Service Operations Management, 2023, vol. 25, no. 2, pp. 631–647, © 2023 INFORMS

Abhishek V, Jerath K, Zhang ZJ (2016) Agency selling or reselling? Jiang B, Tian L (2016) Collaborative consumption: Strategic and economic
Channel structures in electronic retailing. Management Sci. implications of product sharing. Management Sci. 64(3):1171–1188.
62(8):2259–2280. Jiang B, Jerath K, Srinivasan K (2011) Firm strategies in the “mid
Akerlof G (1970) The market for “lemons”: Quality uncertainty and tail” of platform-based retailing. Marketing Sci. 30(5):757–775.
the market mechanism. Quart. J. Econom. 84(3):488–500. Kelly H (2018) This sneaker-selling startup is trying to stomp out fakes.
Babich V, Tang C (2012) Managing opportunistic supplier product adul­ Retrieved January 5, [Link]
teration: Deferred payments, inspection, and combined mechan­ technology/goat-sneaker-app/[Link].
isms. Manufacturing Service Oper. Management 14(2):301–314. Kihlstrom RE, Riordan MH (1984) Advertising as a signal. J. Political
Downloaded from [Link] by [[Link]] on 07 March 2024, at 10:29 . For personal use only, all rights reserved.

Baiman S, Fischer PE, Rajan MV (2000) Information, contracting, Econom. 92(3):427–450.

and quality costs. Management Sci. 46(6):776–789. Kim D (2018) How to avoid counterfeits when looking for deals?
Bagwell K, Riordan M (1991) High and declining prices signal prod­ Retrieved August 16, [Link]
uct quality. Amer. Econom. Rev. 81(1):224–239. blog/avoid-counterfeits-deals/
Baron DP, Myerson RB (1982) Regulating a monopolist with unknown Kirmani A, Rao AR (2000) No pain, no gain: A critical review of the
costs. Econometrica 50(4):911–930. literature on signaling unobservable product quality. J. Market­
Benjaafar S, Kong G, Li X, Courcoubetis C (2018) Peer-to-peer prod­ ing 64(2):66–79.
uct sharing: Implications for ownership, usage, and social wel­ Koaysomboon T (2018) Things you only know if you are … a lux­
fare in the sharing economy. Management Sci. 65(2):477–493. ury product inspector. Retrieved June 1, [Link]
Boulding W, Kirmani A (1993) A consumer-side experimental exam­ com/bangkok/news/things-you-only-know-if-you-are-a-luxury-
ination of signaling theory: Do consumers perceive warranties product-inspector-060118.
as signals of quality? J. Consumer Res. 20(1):111–123. Lee H, Li C (2018) Supplier quality management: Investment, inspec­
Chen L, Papanastasiou Y (2021) Seeding the herd: Pricing and wel­ tion, and incentives. Production Oper. Management 27:304–322.
fare effects of social learning manipulation. Management Sci. Lutz NA (1989) Warranties as signals under consumer moral haz­
67(11):6734–6750. ard. RAND J. Econom. 20(2):239–255.
Cho SH, Fang X, Tayur S (2015) Combating strategic counterfeiters Mantin B, Krishnan H, Dhar T (2014) The strategic role of third
in licit and illicit supply chains. Manufacturing Service Oper. party marketplaces in retailing. Production Oper. Management
Management 17(3):273–289. 23(11):1937–1949.
Clifford T (2019) The RealReal CEO says: ‘We keep changing’ to fend off Mariya B, Edieal P (2018) Dynamic relational contracts for quality
fakes. Retrieved November 13, [Link] enforcement in supply chains. Management Sci. 65(3):1305–1321.
[Link]? Milgrom P, Roberts J (1986) Price and advertising signals of product
qsearchterm=luxury. quality. J. Political Econom. 94(4):796–821.
Debo L, Parlour C, Rajan U (2012) Signaling quality via queues. Moorthy S (2012) Can brand extension signal product quality? Mar­
Management Sci. 58(5):876–891. keting Sci. 31(5):756–770.
Dennis G (2019) These are the most expensive sneakers to be resold Moorthy S, Srinivasan K (1995) Signaling quality with a moneyback
in the last year as the sneaker-resale market explodes. Retrieved guarantee: The role of transaction costs. Marketing Sci. 14(4):442–466.
April 17, [Link] Patel S (2019) Sneaker marketplace StockX seeks new customers
sneakers-sold-last-year-resale-2019-4. with Bleacher Report deal. Retrieved September 17, https://
Farrell M (2021) StockX valued at $3:8 billion and lets employees sell [Link]/articles/sneaker-marketplace-stockx-seeks-new-
shares. Retrieved April 7, 2021, [Link] customers-with-bleacher-report-deal-11568216633
stockx-valued-at-3-8-billion-and-lets-employees-sell-shares-1161 Pun H, DeYong GD (2017) Competing with copycats when customers
7830099. are strategic. Manufacturing Service Oper. Management 19(3):403–418.
Fraiberger S, Sundararajan A (2015) Peer-to-peer rental markets in Pun H, Swaminathan JM, Hou P (2021) Blockchain adoption for
the sharing economy. Working paper, New York University. combating deceptive counterfeits. Production Oper. Management
Fudenberg D, Tirole J (1991) Game Theory (MIT Press, Cambridge, MA). 30:864–882.
Gao SY, Lim WS, Tang C (2017) Entry of copycats of luxury brands. Qian Y (2014) Counterfeiters: Foes or friends? How counterfeits affect
Marketing Sci. 36(2):272–289. sales by product quality tier. Management Sci. 60(10):2381–2400.
Guan X, Pun H, Mao Y, Chen YJ (2020) Strategic communication in Rivera A (2018) What is C2C? Retrieved July 21, 2018, [Link]
a supply chain with copycat threat. Working paper. [Link]/[Link].
Hagiu A, Wright J (2015) Marketplace or reseller? Management Sci. Rohrlich J (2020) How $470 million worth of fake Nikes get into the
61(1):184–203. US. Retrieved January 2, [Link]
Hagiu A, Wright J (2018) Controlling vs. enabling. Management Sci. [Link]
65(2):577–595. Seung HY, Taesu C (2018) Quality improvement incentive strategies
Hawkins A, Thorpe J (2019) Buyer beware: Chinese shoppers’ weapon to in a supply chain. Transportation Res., Part E Logist. Transporta­
dodge cheap knock-offs. Retrieved March 21, [Link] tion Rev. 114:331–342.
content/aaf2b2ae-3108-11e9-80d2-7b637a9e1ba1. Snow J (2020) AI is a new weapon in the battle against counterfeits.
Holt K (2020) EBay will authenticate sneakers from resellers. Retrieved Retrieved August 10, [Link]
October 13, [Link] new-weapon-in-the-battle-against-counterfeits-11596805200
[Link]. Stiving M (2000) Price-endings when prices signal quality. Manage­
Hu M, Pavlin JM, Shi M (2013) When gray markets have silver lin­ ment Sci. 46(12):1617–1629.
ings: All-unit discounts, gray markets, and channel manage­ Stock A, Balachander S (2005) The making of a “hot product”: A
ment. Manufacturing Service Oper. Management 15(2):250–262. signaling explanation of marketers’ scarcity strategy. Manage­
Hwang I, Radhakrishnan S, Su L (2006) Vendor certification and ment Sci. 51(8):1181–1192.
appraisal: Implications for supplier quality. Management Sci. Suthivarakom G (2020) Welcome to the era of fake products. Retrieved
52(10):1472–1482. February 11, [Link]
Janssen M, Moraga-Gonzalez J, Wildenbeest M (2005) Truly costly counterfeit-fake-products/
sequential seach and oligopolistic pricing. Internat. J. Industry Tang Q, Chu J, Lim WS, Zhang X (2020) Online platform’s optimal
Organ. 23:451–466. efforts to combat counterfeits of luxury brands. Working paper.
 Li, Fang, and Lim: Asymmetric Information of Product Authenticity on Platforms
Manufacturing & Service Operations Management, 2023, vol. 25, no. 2, pp. 631–647, © 2023 INFORMS 647

Tian L, Jiang B (2018) Effects of consumer-to-consumer product Wilcox K, Kim HM, Sen S (2009) Why do consumers buy counterfeit
sharing on distribution channel. Production Oper. Management luxury brands? J. Marketing Res. 46(2):247–259.
27(2):350–367. Van Winden F (1999) Experimental studies of signaling games.
Tian L, Vakharia AJ, Tan Y, Xu Y (2018) Marketplace, reseller, or Luini L, Ed. Uncertain Decisions, Bridging Theory, and Experiments
hybrid: Strategic analysis of an emerging e-commerce model. (Kluwer Academic, Boston), 147–173.
Production Oper. Management 27(8):1595–1610. Yao S, Zhu K (2018) The role of anticounterfeit technology in com­
Wang Y, Lin J, Choi TM (2020) Gray market and counterfeiting in bating counterfeit products. Working paper.
supply chains: A review of the operations literature and impli­ Yi Z, Yu M, Cheung KL (2022) Impacts of counterfeiting on a global sup­
Downloaded from [Link] by [[Link]] on 07 March 2024, at 10:29 . For personal use only, all rights reserved.

cations to luxury industries. Transportation Res., Part E Logist. ply chain. Manufacturing Service Oper. Management. 24(1):159–178.
Transportation Rev. 133:101823 Yu M, Ahn H-S, Kapuscinski R (2015) Rationing capacity in
Wernerfelt B (1988) Umbrella branding as a signal of new product advance selling to signal quality. Management Sci. 61(3):560–577.
quality: An example of signaling by posting a bond. RAND J. Zhang J, Hong LJ, Zhang RQ (2012) Fighting strategies in a market
Econom. 19(3):458–466. with counterfeits. Ann. Oper. Res. 192:49–66.