Regulating Digital Piracy Consumption

Regulating Digital Piracy Consumption
∗ † ‡
Jieteng Chen Yuetao Gao T. Tony Ke
December 25, 2023
Abstract
Regulators across the globe have imposed penalties on consumers for digital piracy
consumption. Contrary to expectations, however, digital piracy consumption has con-
tinued to grow. We develop a simple model of competition between a copyright holder
and a pirate firm to offer a plausible account for this observation as well as actionable
guidelines for optimal regulation design. The core of our idea is to endogenize the pi-
rate firm’s strategic investment in anti-tracking technologies that help consumers evade
a regulator’s penalty. We find that as the penalty rises, piracy consumption can sur-
prisingly increase after decreasing first; relatedly, the copyright holder and the society
may suffer from tighter regulation. Depending on the cost of anti-tracking technologies
of the pirate firm, the regulator optimally sets the penalty to operate in two different
regimes. When the technology is available at a low cost, the regulator can achieve the
goals of maximizing social welfare and minimizing piracy consumption simultaneously
by setting a moderate penalty that maximizes consumers’ expected penalty and toler-
ates some level of piracy consumption. In contrast, when the technology is costly, the
regulator should set a relatively high penalty to completely impede piracy supply. Ad-
ditionally, we show that supply-side regulation does not substitute away demand-side
regulation, and educating consumers about copyright protection may unintentionally
lead to an increase in piracy consumption. Lastly, we identify complex non-monotonic
long-run effects of piracy consumption regulation on the copyright holder’s incentives
for content creation and copyright protection.
Keywords:
digital piracy, regulation, copyright protection
∗
Department of Marketing, Business School, The Chinese University of Hong Kong, Shatin, Hong Kong,
China. Email: jieteng.chen@link.cuhk.edu.hk
†
Department of Marketing, School of Management, Xiamen University, Xiamen, Fujian, China. Email:
ytgao@xmu.edu.cn
‡
Department of Marketing, Business School, The Chinese University of Hong Kong, Shatin, Hong Kong,
China. Email: tonyke@cuhk.edu.hk
Electronic copy available at: https://ssrn.com/abstract=4198295

INTRODUCTION
Digital piracy refers to the act of illegally downloading and distributing copyrighted
digital material over the Internet. Digital piracy is a serious problem that has witnessed an
unprecedented rise over the years (Brown and Holt 2021). According to Blackburn, Eisenach,
and Harrison Jr (2019), digital piracy annually eroded $29.2–71 billion revenue and drained
230,000–560,000 jobs of the U.S. economy. A report by an anti-piracy solution provider
(MUSO 2022) estimates that piracy websites attracted 182 billion visits globally in 2021—a
15.2% increase from 2020.
Pirated digital content is commonly downloaded by piracy users for no monetary cost
through peer-to-peer networks (such as torrent) using notorious piracy websites. Pirate
Bay is one such example, consistently topping the list of the most visited piracy websites
among torrenting users in recent years. In response to digital piracy consumption, regulators
across the globe have enacted or strengthened regulations by penalizing users who download
pirated digital content, with the aim of curbing these activities. For instance, in 2008, the
U.S. issued the PRO-IP Act, which boosted civil penalties for copyright infringement from
$500–100,000 to $1,000–200,000. Likewise, in 2021, France replaced the 13-year-old anti-
piracy agency HADOPI (High Authority for the Distribution of Works and the Protection of
Rights on the Internet) with a new regulatory body named ARCOM (Regulatory Authority
for Audiovisual and Digital Communication), aimed at producing tighter regulation and
stronger protection of access to digital content. These demand-side regulations are unique
to digital content, as digital consumption can be traced by regulators much more easily than
physical transactions.
Given the high penalty invoked by the above regulations, one might naturally intuit that
piracy users would be reluctant to visit Pirate Bay. Contrary to the intuition, however, user
visits to Pirate Bay, as measured by Google Trends, have actually risen since the inception of
these regulations, as shown in Figure 1.1 Similar trends have also been observed with respect
1
There is no intention to infer any causal relationship from these time series; rather, they show that these digital

to other leading piracy websites (specifically, Torrentz and IsoHunt around 2008 and 2009,
and YTS and 1337x around 2021). Indeed, Dejean, Pénard, and Suire (2010) evaluate the
impact of HADOPI law on piracy consumption and find that piracy consumption actually
rose by 3% since the inception of the regulation.2 In this paper, we aim to offer a plausible
account for this seeming inefficacy of demand-side piracy regulation and to provide actionable
guidelines for regulators to optimally design regulations against digital piracy.
100
The PRO-IP Act in the U.S.
80
60
40
Apr 13, 2008 Jul 13, 2008 Oct 13, 2008 Jan 13, 2009 Apr 13, 2009
(Effective Date)
100
The HADOPI Law in France
80
60
40
(Effective Date)
100
The ARCOM Law in France
80
60
40
(Effective Date)
Figure 1: Pirate Bay’s Google Trends Before and After Enforcement of New Anti-Piracy
Laws
Most piracy websites, including Pirate Bay, make money from advertising revenue, which
is positively associated with the traffic volume of piracy users. Pirate Bay thus suffers when
its users avoid surfing its website out of concerns about high regulatory penalties. To counter
this user reluctance, Pirate Bay constantly invests in new technologies or exerts efforts that
help piracy users evade regulators’ attempts to track their footage on its website. Table 1
piracy regulations are not effective (or not effective enough) to suppress the ongoing growth of piracy consumption.
Google Trends, which measures the popularity of search queries on Google, has been used as a proxy for piracy
consumption in the literature (e.g., Lu, Rajavi, and Dinner 2021).
2
This study was originally written in French, and its main findings were summarized in English in a news report
https://tinyurl.com/4nbzut8w (france24.com) [retrieved from waybackmachine; accessed April 7, 2023].

details a set of Pirate Bay’s anti-tracking actions in recent years.3 Pirate Bay’s anti-tracking
actions are also apparent from its most recent homepage in Figure 2, which facilitates its
visitors to hide their IP addresses to escape penalty by using the VPN service. These
countermeasures offer piracy users a sense of security with respect to their ability to avoid
penalties, which may explain why they are willing to visit Pirate Bay in greater numbers
since the inception of the regulations in Figure 1. From a different angle, the invention of
secure torrenting technologies on piracy websites appears effective in helping piracy users
evade penalties, because only €87,000 fines have been generated by the HADOPI law since
its enactment.4
Table 1: Pirate Bay’s Anti-Tracking Actions
Time Actions
2008 Supported SSL encryption in response to Sweden’s new wiretap-
ping law
2008 Inserted “random IP addresses” automatically to pollute the evi-
dence gathering of anti-piracy outfits
2010 Offered the IPREDator VPN service that keeps piracy users
anonymous and safe from being tracked by law enforcement
2012 Switched from torrents to untraceable magnet links, which makes
it more difficult to determine who has downloaded a file using the
website
2013 Released a free web browser, PirateBrowser, to circumvent inter-
net censorship
2020 Used the IP address operated by an anonymous VPN service
provider, OPVN, which does not store any logs
We formalize the above observations by developing a simple vertical competition model.

The model consists of a copyright holder offering a high-quality digital product strategically
sets a price to fight piracy, and a pirate firm offering a free, low-quality pirated version of
the product strategically invests in an anti-tracking technology to help consumers evade a
3
Source: https://linkmix.co/15876993 (theguardian.com, torrentfreak.com, readwrite.com, vpnoverview.com)
[retrieved from waybackmachine; accessed April 7, 2023]. Similar countermeasure actions have been taken by other
piracy websites. For instance, Kickass Torrents forces SSL encryption for all visitors to avoid surveillance.
4
See https://tinyurl.com/mpw4e9fm (torrentfreak.com) [retrieved from waybackmachine; accessed April 7,
2023].

Figure 2: Pirate Bay’s Homepage (accessed March 9, 2023)
regulator’s penalty for piracy consumption. Consumers differ in their preference regarding
product quality and choose between copyrighted and pirated products subject to a potential
penalty.
Our equilibrium analysis shows that when the pirate firm can develop anti-tracking tech-
nologies at a relatively low cost, piracy supply cannot be completely eliminated from the
market no matter how high the penalty imposed on piracy users is, because the pirate firm
will always participate in the market by investing in these technologies to help consumers
evade the penalty. More importantly, we find that piracy consumption exhibits a U-shape
with the penalty, where an increase in the penalty can surprisingly promote piracy consump-
tion when the penalty is relatively high. The rise in piracy consumption is again driven by the

pirate firm’s endogenous investment in anti-tracking technologies, since the expected penalty
for consumers’ piracy consumption depends not only on the penalty set by the regulator but
also on the likelihood of consumers being tracked—a factor controllable by the pirate firm via
anti-tracking technologies. Hence, the higher the penalty, the more the pirate firm invests
in anti-tracking technologies. Consumers’ expected penalty consequently exhibits an inverse
U-shape with the penalty. Put differently, we demonstrate that the pirate firm’s strategic
investment in anti-tracking technologies may dominate the direct effect of the penalty and
thus result in a lower expected penalty for piracy users and more piracy consumption.
As a result, tighter regulation can make both the copyright holder and the society worse
off due to the resulting more piracy consumption. We find that depending on the cost of
anti-tracking technologies of the pirate firm, the regulator optimally sets the penalty to
operate in two different regimes. On the one hand, when the technology is available at
a low cost, the regulator can reach the goals of maximizing social welfare and minimizing
piracy consumption simultaneously by setting a moderate penalty that maximizes consumers’
expected penalty and tolerates some level of piracy consumption. On the other hand, when
the technology is costly to acquire for the pirate firm, the regulator should set a relatively
high penalty to completely impede piracy supply.
Besides demand-side regulation, we also examine alternative policy tools that combat
piracy. We show that supply-side regulation does not substitute away demand-side regula-
tion, because shutting down the pirate websites could reduce market competition and result
in lower social welfare. Moreover, we find that educating consumers to adopt copyrighted
products may inadvertently prompt more consumers to use pirated products.
Lastly, we extend the main model in several directions. First, we study how piracy
consumption regulation impacts a copyright holder’s incentives to create content by endoge-
nizing its product quality. Second, we allow a copyright holder to take proactive protection
of its copyrighted content by endogenizing the quality of pirated products. In these two
extensions, we show that the penalty may produce complex non-monotonic effects on the

optimal quality of copyrighted and pirated products respectively. Third, we show that using
an Internet service provider (ISP) as an inspector to enforce the penalty may not alleviate
the piracy problem. Fourth, we demonstrate that our results can be extended to broader
regulation contexts.
Our paper contributes to the growing literature on digital piracy. Some research focuses
on the ways along which copyright holders’ initiated copyright protection may make them
earn lower profits, including intensifying price competition (Jain 2008), reducing downstream
competition (Vernik, Purohit, and Desai 2011), and discouraging consumer search (Guo
and Meng 2015). Some other research looks at the impact of regulation on digital piracy.
Copyright holders can be better off penalizing consumers for copyright infringement so as
to diminish their adoption of pirated products (Lahiri and Dey 2013; Dey, Kim, and Lahiri
2019) or punishing pirate firms to deter their entry to the market (Dey, Kim, and Lahiri
2019). Moreover, Tunca and Wu (2013) show that penalizing users who produce pirated
products for themselves may lower a copyright holder’s profit because doing so will lend
a commercial pirate firm a competitive advantage. Notably, all the above studies do not
model pirate firms’ strategic decisions as we do, and thus our research offers original and
novel insights to the digital piracy literature.
Broadly, our paper also enriches the recent studies on dishonest behaviors and unin-
tended consequences of regulations or the lack thereof across various contexts. Durbin and
Iyer (2009) find that bribes can make a good advisor more likely to offer a truthful recommen-
dation to its client. Wilbur and Zhu (2009) show that the revenue of a search engine may be
high in the presence of click frauds. Zhu and Dukes (2015) present that media firms’ competi-
tion makes consumers less informed about the truth of facts. Singh (2017) argues that lifting
surveillance of corruptible agents may encourage them to recommend undesirable choices to
their clients. Gao (2018) reveals that adopting complicated anti-counterfeiting technologies
to deter counterfeiter entry may result in more counterfeit purchases. Iyer and Singh (2018)
show that a firm may pursue product safety certification even when the firm and consumers

do not have contrasting views regarding product safety. Dai and Singh (2020) find that a
diagnostic expert with high ability does not seek testing for its client and only offers diagno-
sis. Wu and Geylani (2020) demonstrate that a high penalty imposed on firms for deceptive
advertising may harm consumers. Gao and Wu (2023) point out that penalizing retailers for
selling counterfeits may lead them to sell a large proportion of counterfeits. Wu, Gal-Or, and
Geylani (2022) uncover that tighter regulation may make native ads more opaque and thus
lower consumer surplus and social welfare. Zhou and Zou (2023) highlight that regulations
prohibiting platforms from using price information for product recommendations may hurt
consumers.
The remainder of the paper is organized as follows. Section 2 describes the model setup.
Section 3 presents the equilibrium analysis. Section 4 delivers the results. Section 5 offers the
policy implications. Section 6 considers several extensions. Section 7 provides conclusions.
All proofs are relegated to Web Appendix.
MODEL SETUP
Consider a consumer market of size one served by two firms. A legal firm holds the
copyright for a digital product with quality v > 0, and a pirate firm offers a pirated version
of the product with quality βv, where β ∈ (0, 1) can be seen as the copyright holder’s
copyright protection level, as copyrighted products are generally superior to their pirated
counterparts (Jain 2008; Guo and Meng 2015).5 Marginal production costs of the firms are
set as zero due to the nature of digital products. The copyright holder charges price pc for its
copyrighted products. Consistent with common practice and the literature (Dey, Kim, and
Lahiri 2019), we assume that the pirate firm offers its products for free to consumers and
earns revenue kDp from advertising, where k > 0 is the expected ad revenue per consumer
and Dp is consumer demand for pirated products.6
5
We will endogenize the copyright holder’s product quality as well as the quality differential between copyrighted
and pirated products in two extensions respectively.
6
This model specification captures the commonly observed practice that copyright holders and pirate firms
typically adopt different revenue models. Piracy websites, such as Pirate Bay, are essentially online platforms that

To protect copyright and guard against piracy, a regulator of the market tries to keep
track of piracy consumption and imposes penalty t on consumers who adopt pirated products
and get tracked.7 There are two types of consumers in the market (Chen and Png 2003;
Jain 2008). Specifically, γ ∈ [0, 1) fraction of them are ethical in that they have high
moral standards and only consider copyrighted products; the remaining 1 − γ fraction are
unethical in that they choose between copyrighted and pirated products based on expected
utility maximization. In response to regulation, the pirate firm invests in an anti-tracking
technology or exerts an effort that entails a cost of s(1 − x)2 /2 and results in probability
x ∈ [0, 1] that a consumer will be tracked, where s > 0 is a cost parameter.8 That is, without
this investment, unethical consumers engaging in digital piracy face penalty t imposed by the
regulator; by investing more, the pirate firm can reduce x, the likelihood of piracy users being
tracked. As a result, xt measures the expected penalty for piracy consumption of unethical
consumers.
Consumers’ utilities from consuming a copyrighted product and a pirated one are respec-
tively:
uc = θv − pc ,
up = θβv − xt,
where θ measures consumers’ preference for product quality and is assumed to follow a
uniform distribution in [0, 1] across consumers. Consumers also have an outside option of
monetize visitor traffic through sponsored ads; copyright holders, like software developers, in general profit from
selling their own products directly to consumers and do not accommodate ads from other firms. Nevertheless, we
allow the pirate firm to adopt pricing-based revenue model in an extension.
7
We focus on demand-side regulation in this paper. While the regulator can also rely on supply-side regulation
to fight piracy, it involves other complexities. For example, piracy websites, such as Pirate Bay, could locate their
servers anywhere globally and actively modify their domain names to evade detection. As such, it is difficult to shut
down piracy websites via supply-side regulation. Nevertheless, we will incorporate supply-side regulation formally in
Section of Policy Implications. Additionally, as in Becker (1968) and Wu and Geylani (2020), a higher t could result
from a higher punishment level and/or a higher enforcement level. In Section of Policy Implications, we consider
Internet service provider (ISP) as an intermediary between the regulator and consumers, who can influence the
enforcement level.
8
In an extension, we allow anti-tracking actions to be initiated by consumers instead of the pirate firm.

consuming neither product, which is normalized to zero.
The timing of the game is as follows. At the first stage, the copyright holder sets the
price, pc ; at the same time, the pirate firm determines the tracking probability, x. At the
second stage, consumers make their product choice, and then the payoffs of all parties are
realized.9 To study the effects of piracy consumption regulation, we will first treat penalty t
as exogenous in the analysis and then endogenize t by adding a pre-stage to the game where
the regulator sets t first.
Prior to the model analysis, we make the following assumption to focus on the most
meaningful parameter ranges.
Assumption 1.
1−β
γ< .
β
The assumption ensures that the fraction of ethical consumers is not too large; otherwise,
the copyright holder may only serve ethical consumers, unethical consumers may only adopt
pirated products, and consequently, there is no direct competition between the two firms.10
Notice that γ is allowed to take the value of zero under the assumption. Put differently, our
results do not rely on the existence of ethical consumers in the market.
EQUILIBRIUM ANALYSIS
We solve the game by backward induction. Given that the pirate firm’s products are of
lower quality and entail a potential penalty for piracy consumption, it may be unprofitable
for the pirate firm to participate in the market when facing competition with the copyright
holder. This implies that there are two subgames to study.
In subgame I, both firms operate in the market. Let us first analyze the demand of
unethical consumers, who choose between the two products and the outside option based
9
In an extension, we study two alternative timings of the game where the copyright holder moves either before
or after the pirate firm, and we find the equilibrium outcome does not change qualitatively.
10
The analysis for sufficient large γ is provided in Section of Policy Implications.

on expected utility maximization. Specifically, an unethical consumer buys a copyrighted
product if and only if the following constraints are satisfied:
pc − xt pc
uc ≥ up ⇔ θ ≥ , and uc ≥ 0 ⇔ θ ≥ ,
(1 − β)v v
where the first constraint reflects the copyright holder’s demand loss from unethical con-
sumers due to the existence of pirated products. That is, the demand loss can be measured
by the segment of unethical consumers with θ ∈ [pc /v, (pc − xt)/[(1 − β)v]) as they would
purchase copyrighted products in the absence of pirated products.
Analogously, an unethical consumer adopts a pirated product if and only if
pc − xt xt
up ≥ uc ⇔ θ ≤ , and up ≥ 0 ⇔ θ ≥ .
(1 − β)v βv
Since
pc − xt pc pc − xt xt xt
> ⇔ > ⇔ pc > ,
(1 − β)v v (1 − β)v βv β
there are two cases to consider depending on the comparison between pc and xt/β, as shown
in Figure 3. Note that when pc ≤ xt/β, pirated products generate no demand even when
offered. The left panel in Figure 3 demonstrates the consumer market segmentation where
patrons of piracy tend to have intermediate θ. To map this result to real-world observations,
notice that θ can be reinterpreted as a measure of consumer income.11 Demand for piracy
emerges due to copyright streaming services owning content that is typically unavailable in
other portals. Therefore, the left panel in Figure 3 implies that consumers with intermediate
income are most interested in accessing content across platforms via piracy, because their
income is not very low so that they are still willing to bear the risk of a potential piracy
penalty but not very high so that they are subject to budget constraints.
11
On page 97, Tirole (1988) defines u(I) as a consumer’s utility over her income I and shows that θ = u′ (I)−1 ,
which increases with I due to concavity of u(·).
10

1 1
uc = θv − pc Buy copyrighted product
pc −xt
uc = θv − pc Buy copyrighted product
(1−β)v
pc
up = θβv − xt Consume pirated product v
xt
βv
uo = 0 No Consumption
uo = 0 No Consumption
0 0
Case (a): pc > xt/β Case (b): pc ≤ xt/β
Figure 3: Demand Structure of Unethical Consumers in the Presence of Both Firms
Furthermore, notice that ethical consumers’ demand is always given by Case (b). Conse-
quently, we can write down consumer demand for copyrighted products and pirated products
respectively as:

h i
pc pc −xt xt

γ 1 −
 + (1 − γ) 1 − if pc > ,
v (1−β)v β
DcI (pc , x) =

1 −
 pc
v
otherwise;

h i
(1 − γ) pc −xt − xt xt


(1−β)v βv
if pc > β
,
DpI (pc , x) =


0 otherwise,
where we have used the superscript ”I” to denote subgame I.

In subgame II, only the copyright holder operates in the market. We have both types of
consumer demand given by Case (b) in Figure 3.
So far, we have determined consumer demand for subgame I and subgame II. The re-
maining question is to identify the condition that delimits the two subgames. That is, when
will the pirate firm participate in the market? In fact, when the pirate firm participates in
the market and gets positive demand, its profit is
s xt
ΠIp (pc , x) = kDpI (pc , x) − (1 − x)2 for pc > .
2 β
11

By maximizing ΠIp (pc , x) with respect to x, we have the optimal choice of x as

∗ (1 − γ)kt
x = max 1 − ,0 . (1)
β(1 − β)sv
Interestingly, the pirate firm’s best response function, x∗ , does not depend on pc .12 Cor-
respondingly, the pirate firm’s maximum profit under x∗ is,

h i
 (1−γ)k pc −
 t (1−γ)kt2

(1−β)v β
+ 2β 2 (1−β)sv
if 0 ≤ t < T,
ΠIp (pc , x∗ ) =

 (1−γ)kpc −
 s
otherwise;
(1−β)v 2
β(1−β)sv
where T ≡ (1−γ)k
. The pirate firm will participate in the market if and only if

t −
 (1−γ)kt2

β 2β 2 (1−β)sv
if 0 ≤ t < T,
ΠIp (pc , x∗ ) > 0 ⇔ pc > pec ≡

 (1−β)sv

otherwise;
2(1−γ)k
To summarize subgame I and subgame II along with the pirate firm’s participation con-
straint, we have consumer demand for two types of products as follows:

h i n o
pc pc −xt xt

γ 1 −
 + (1 − γ) 1 − if pc > max , pec ,
v (1−β)v β
Dc (pc , x) =

1 −
 pc
v
otherwise;

h i n o
(1 − γ) pc −xt − xt xt


(1−β)v βv
if pc > max β
, pec ,
Dp (pc , x) =


0 otherwise.
12
The key driving forces behind the independence of x∗ and pc are that (1) the pirate firm’s demand is linear in
x, which results from the uniform distribution of the consumer quality preference, and (2) the pirate firm’s profit
margin does not depend on pc nor x. Consequently, the marginal return from a lower x is a constant that does not
depend on pc . In other words, by investing in a better anti-tracking technology, the pirate firm gains some marginal
consumers who would have chosen the copyrighted product in the absence of the investment, and the number as well
as profitability of these marginal consumers do not depend on the current market share of the pirate firm and thus
do not depend on the competing copyright holder’s price, pc .
12

The two firms’ profits are respectively,
Πc (pc , x) = pc Dc (pc , x),

s
Πp (pc , x) = kDp (pc , x) − (1 − x)2 .
2
Proposition 1 solves the equilibrium price and tracking probability by a set of equations
as follows: 
 p∗ = arg max Πc (pc , x∗ ),

c pc
 x∗ = arg max Π (p∗ , x),


x p c
where given p∗c and x∗ , we can define the equilibrium demand Dj∗ ≡ Dj (p∗c , x∗ ) and the
equilibrium profit Π∗j ≡ Πj (p∗c , x∗ ) for j ∈ {c, p}. We will provide necessary derivations or
elaborations that lead to each proposition in the main text but relegate complete proofs to
Web Appendix.
Proposition 1 (Equilibrium Characterization). There exist thresholds s and t such that
(i) if 0 < s < s or if s ≥ s and 0 ≤ t < t, the pirate firm’s equilibrium anti-tracking
technology choice is

∗ (1 − γ)kt
x = max 1 − ,0 ,
β(1 − β)sv
and the copyright holder’s equilibrium price is
(1 − β)v + (1 − γ)x∗ t
p∗c = ;
2(1 − γβ)
(ii) otherwise, if s ≥ s and t ≥ t, the pirate firm does not participate in the market, and
the copyright holder’s equilibrium price is
n vo
p∗c = min pec , .
2
Proposition 1 implies that when the technology cost is relatively low with 0 < s < s or
13

the penalty is relatively low with 0 ≤ t < t, the copyright holder and the pirate firm can
co-exist in the market where ethical consumers buy copyrighted products, and for unethical
consumers, the equilibrium consumer segmentation is shown by Case (a) in Figure 3—
those with relatively high preference for quality buy copyrighted products and those with
intermediate preference go for pirated products. The equilibrium x∗ is exactly given by
Equation (1). On the other hand, when both s and t are relatively high, the pirate firm is
driven out of the market and, consequently, no piracy is supplied.
EFFECTS OF REGULATION
In this section, we examine the effects of piracy consumption regulation on the equilibrium
outcome in a series of propositions. To begin with, we are interested in how penalty t
influences the pirate firm’s anti-tracking technology choice and consumers’ expected penalty.
Proposition 2 states the results. By Proposition 1, when s ≥ s, we restrict ourselves to
0 ≤ t < t, because otherwise, x∗ is not well defined.
Proposition 2 (Effect of Penalty on Anti-Tracking Technology Choice and Expected Penalty).
(i) (a) If 0 < s < s, x∗ decreases with t for 0 ≤ t < T and stays at zero for t ≥ T ;
(b) otherwise, if s ≥ s, x∗ decreases with t for 0 ≤ t < t.
(ii) There exists a threshold s that does not depend on t such that
(a) if 0 < s < s, x∗ t increases with t for 0 ≤ t < t, decreases with t for t ≤ t < T ,
and stays at zero for t ≥ T ;
(b) if s ≤ s < s, x∗ t increases with t for 0 ≤ t < t and decreases with t for t ≤ t < t;
(c) otherwise, if s ≥ s, x∗ t increases with t for 0 ≤ t < t.
The relationship between x∗ and t has a clear intuition—as penalty t increases, the
pirate firm reduces x∗ to keep its products still attractive for unethical consumers, whose
consumption utility of pirated products, up , depends on the whole term of x∗ t, the expected
14

penalty. In particular, when 0 < s < s and t ≥ T , the pirate firm chooses x∗ = 0, which
leads to zero of the expected penalty as a result.
To see the relationship between x∗ t and t, given that x∗ t is the determinant of consumer
utility and choice, it is worthwhile to take a closer examination by decomposing the marginal
effect of t on x∗ t into two parts:
d(x∗ t) dx∗

∗ (1 − γ)kt (1 − γ)k
= x + ×t =1− + − ×t . (2)
dt |{z} dt{z } β(1 − β)sv β(1 − β)sv
direct effect | | {z } | {z }
strategic effect direct effect strategic effect
Here, we focus on the situation where x∗ > 0; otherwise, the expected penalty will never
change even if the regulator further increases the penalty. The first part in Equation (2) is
the positive direct effect. The second part is the negative strategic effect and comes from the
pirate firm’s strategic investment in improving its anti-tracking technology in response to
tighter regulation. It is somewhat surprising that the strategic effect can dominate the direct
effect when s is relatively low and t is relatively high. In fact, one may interpret penalty t
as a form of tax on piracy consumption, and further intuit that while the pirate firm has an
incentive to compensate consumers by reducing x∗ to mitigate the adverse effect of piracy
tax, the compensation is a second-order effect that will never override the first-order effect
of taxation per se. However, it turns out that this intuition is incorrect, as shown in Figure
4. This is because the pirate firm’s compensation for consumers is not a simple monetary
transfer; instead, the effective tax on consumers is the expected penalty, x∗ t, which is jointly
determined by the regulator and the pirate firm, who bears a technology development cost
of s(1 − x∗ )2 /2. Since the pirate firm’s marginal benefit of reducing x∗ , (1 − γ)kt/[β(1 − β)v],
increases with t, the pirate firm has strong incentives to lower x when the penalty is high.
As a result, the strategic effect can override the direct effect under a high penalty.
15

0.035
0.056
0.026
Expected Penalty
0.042
Expected Penalty
0.018 0.028
0.009 0.014
0.000 0.000
t T t t
Penalty Penalty
(a): 0 < s < s (b): s ≤ s < s

(v = 1, β = 0.55, γ = 0.12, k = 1, s = 0.5) (v = 1, β = 0.55, γ = 0.12, k = 1, s = 0.8)
Figure 4: Effect of Penalty on Expected Penalty
As illustrated in Figure 4, under a relatively low s and a relatively high t, as t further

increases, the pirate firm has an incentive to decrease x∗ t by investing in anti-tracking tech-
nologies that reduce x∗ , because a decreased x∗ t benefits not only piracy users but also the
pirate firm itself. Notice that the decreasing relationship between x∗ and t is quite general,
which does not depend on the copyright holder’s pricing strategy, as shown by Equation (1);
nor does it depend on the uniform distribution of consumer preference for quality.13 We
can also explain the decreasing relationship between x∗ t and t from Equation (2), where the
direct effect decreases with t and increases with s while the strategic effect increases with t
and decreases with s (in magnitude); as a result, the latter can dominate the former when t
is relatively high and s is relatively low. For simplicity, we do not plot the case with s ≥ s
in Figure 4 and other figures in the following analysis, as it can be seen as a degenerate case
of s ≤ s < s, where the interval of [t, t] shrinks to an empty set.
Next, we investigate the impact of regulation on piracy consumption. Proposition 3
and Figure 5 reveal rich non-monotonic relationships between piracy consumption and the
penalty. These U-shape relationships between Dp∗ and t follow directly from the inverse
U-shape between x∗ t and t. The abrupt drop to zero of Dp∗ at t = t when s ≥ s is due to
the pirate firm’s participation constraint—for t ≥ t, piracy is no longer supplied. In short,
13
One can further show that as long as the distribution function is convex, x∗ decreases with t.
16

Proposition 3 uncovers that when technology cost s is not high enough 0 < s < s, tighter
regulation may backfire by promoting more piracy consumption at a relatively high penalty.
Proposition 3 (Effect of Penalty on Piracy Consumption).
(i) If 0 < s < s, Dp∗ decreases with t for 0 ≤ t < t, increases with t for t ≤ t < T , and is
invariant in t for t ≥ T .
(ii) If s ≤ s < s, Dp∗ decreases with t for 0 ≤ t < t, increases with t for t ≤ t < t, and stays
at zero for t ≥ t.
(iii) Otherwise, if s ≥ s, Dp∗ decreases with t for 0 ≤ t < t and stays at zero for t ≥ t.
0.471 0.471
Piracy Consumption
Piracy Consumption
0.448 0.353
0.425 0.236
0.402 0.118
0.378 0.000
t T t t T
Penalty Penalty
(a): 0 < s < s (b): s ≤ s < s

(v = 1, β = 0.55, γ = 0.12, k = 1, s = 0.5) (v = 1, β = 0.55, γ = 0.12, k = 1, s = 0.8)
Figure 5: Effect of Penalty on Piracy Consumption
Furthermore, we study how regulation affects the copyright holder’s price and profit,
which are summarized in Proposition 4. When 0 < s < s, the copyright holder’s equilibrium
price and profit exhibit an inverse U-shape with t for 0 ≤ t < T and are invariant in t for
t ≥ T , which can also be explained by the inverse U-shape between x∗ t and t for 0 ≤ t < T
and invariance between x∗ t and t for t ≥ T as shown in Proposition 2. The similar inverse
U-shape exhibits for 0 ≤ t < t when s ≤ s < s. When s ≥ s, it is interesting to see that
the copyright holder drops its price discretely at t = t so as to edge the pirate firm out
of the market. As t further increases, the maximum price the copyright holder can charge
17

while still keeping the pirate firm out of the market will rise; as t gets sufficiently high, it
is possible for the copyright holder to freely exercise its monopoly power in the market by
pricing at v/2 and gain monopoly profit.
In summary, Proposition 4 shows that the copyright holder can use different pricing
strategies to combat piracy, and tighter regulation can help the copyright holder for a rela-
tively high or low penalty but hurt at an intermediate penalty.
Proposition 4 (Effect of Penalty on Copyright Holder’s Price and Profit).
(i) If 0 < s < s, p∗c increases with t for 0 ≤ t < t, decreases with t for t ≤ t < T , and is
(ii) If s ≤ s < s, p∗c increases with t for 0 ≤ t < t, decreases with t for t ≤ t < t, drops
discretely at t = t, increases with t for t ≤ t < T , and is invariant in t for t ≥ T .
(iii) Otherwise, if s ≥ s, p∗c increases with t for 0 ≤ t < t, drops discretely at t = t, increases
with t for t ≤ t < T , and is invariant in t for t ≥ T .
The relationship between Π∗c and t is the same as between p∗c and t except that Π∗c is continuous
in t with no drop at t = t if s ≥ s.
0.257 0.267
0.253 0.245
Price
Price
0.249 0.222
0.245 0.200
0.241 0.177
t T t t T
Penalty Penalty
(a): 0 < s < s (b): s ≤ s < s

(v = 1, β = 0.55, γ = 0.12, k = 1, s = 0.5) (v = 1, β = 0.55, γ = 0.12, k = 1, s = 0.8)
Figure 6: Effect of Penalty on Copyright Holder’s Price
18

0.138 0.163
Pro-t 0.133 0.152
Pro-t
0.129 0.142
0.125 0.131
0.120 0.120
t T t t T
Penalty Penalty
(a): 0 < s < s (b): s ≤ s < s

(v = 1, β = 0.55, γ = 0.12, k = 1, s = 0.5) (v = 1, β = 0.55, γ = 0.12, k = 1, s = 0.8)
Figure 7: Effect of Penalty on Copyright Holder’s Profit
Lastly, we analyze the effect of regulation on consumer surplus and social welfare in
Propositions 5 and 6. We consider two measures in calculating consumer surplus and social
welfare. First, in calculating consumer surplus (CS) and social welfare (SW ), we exclude
consumer surplus from pirated products as well as penalty fees. This reflects the idea that
the regulator should not design policies aiming to maximize consumer surplus gains from
consuming illegal products. Second, we also calculate total consumer surplus (T CS) and
total social welfare (T SW ), which further include consumer surplus from consuming pirated
products and penalty fees.
Proposition 5 (Effect of Penalty on Consumer Surplus and Total Consumer Surplus).
(i) (a) If 0 < s < s, CS exhibits a non-monotonic relationship with t in general for
0 ≤ t < T and is invariant in t for t ≥ T .
(b) Otherwise, if s ≥ s, CS exhibits a non-monotonic relationship with t in general

for 0 ≤ t < t, jumps discretely at t = t, decreases with t for t ≤ t < T , and is
(ii) (a) If 0 < s < s, T CS decreases with t for 0 ≤ t < t, increases with t for t ≤ t < T ,
and is invariant in t for t ≥ T .
19

(b) If s ≤ s < s, T CS decreases with t for 0 ≤ t < t and increases with t for t ≤ t < t
(T CS = CS for t ≥ t).
(c) Otherwise, if s ≥ s, T CS decreases with t for 0 ≤ t < t (T CS = CS for t ≥ t).
0.251 0.338
0.251 0.316
Consumer Surplus
Consumer Surplus
0.251 0.294
0.250 0.272
0.250 0.250
t T t t T
Penalty Penalty
(a): 0 < s < s (b): s ≤ s < s

(v = 1, β = 0.55, γ = 0.12, k = 1, s = 0.5) (v = 1, β = 0.55, γ = 0.12, k = 1, s = 0.8)
0.319 0.338
Total Consumer Surplus
Total Consumer Surplus
0.313 0.325
0.308 0.311
0.302 0.297
0.296 0.283
t T t t T
Penalty Penalty
(c): 0 < s < s (d): s ≤ s < s

(v = 1, β = 0.55, γ = 0.12, k = 1, s = 0.5) (v = 1, β = 0.55, γ = 0.12, k = 1, s = 0.8)
Figure 8: Effect of Penalty on Consumer Surplus and Total Consumer Surplus
Consumer surplus (CS) is calculated by integrating over all consumers who buy copy-
righted products, where the surplus for each consumer equals her valuation of a copyrighted
product minus price p∗c . A higher expected penalty has two opposing effects on CS: (1) it
makes pirated products less attractive and thus increases consumer demand for copyrighted
products, and (2) it induces the copyright holder to charge a higher price. These two op-
posing effects result in a non-monotonic relationship between CS and x∗ t. Furthermore, as
20

shown by Proposition 2, the expected penalty, x∗ t, has a non-monotonic relationship with
penalty t. Consequently, the relationship between CS and t, as a combination of the rela-
tionship between CS and x∗ t and that between x∗ t and t, exhibits a complex non-monotonic
pattern for 0 < s < s and 0 ≤ t < T , or for s ≥ s and 0 ≤ t < t. This is illustrated by the
top two panels in Figure 8. We further zoom into CS in the range of 0 ≤ t < t as shown by
a box in the top right panel. On the other hand, when s ≥ s, the copyright holder reduces
its price discontinuously at t = t so as to drive the pirate firm out of the market, which
improves CS since more consumers obtain copyrighted products at a lower price. For t ≥ t,
all consumers buy copyrighted products. Raising penalty t leads to a higher p∗c (that is, the
first effect is muted), which results in lower CS, as shown in Figure 8(b).
Total consumer surplus (T CS) comprises CS and the surplus of unethical consumers
who adopt pirated products, the latter of which equals their valuation of pirated products
minus the expected penalty. When 0 < s < s or when s ≥ s and 0 ≤ t < t, an increase in the
expected penalty increases p∗c and decreases consumers’ total demand for copyrighted and
pirated products. As a result, T CS decreases as the expected penalty increases. Therefore,
the U-shape between T CS and the penalty directly follows the inverse U-shape between the
expected penalty and t. When s ≥ s and t ≥ t, pirated products are not supplied and T CS
is equal to CS.
Proposition 6 (Effect of Penalty on Social Welfare and Total Social Welfare).
(i) (a) If 0 < s < s, SW increases with t for 0 ≤ t < t, decreases with t for t ≤ t < T ,
(b) If s ≤ s < s, SW increases with t for 0 ≤ t < t, decreases with t for t ≤ t < t,
jumps discretely at t = t, decreases with t for t ≤ t < T , and is invariant in t for
t ≥ T.
(c) Otherwise, if s ≥ s, SW increases with t for 0 ≤ t < t, jumps discretely at t = t,

decreases with t for t ≤ t < T , and is invariant in t for t ≥ T .
21

(ii) (a) If 0 < s < s, T SW increases with t for 0 ≤ t < t, decreases with t for t ≤ t < T ,
(b) Otherwise, if s ≥ s, T SW exhibits a non-monotonic relationship with t in general

for 0 ≤ t < t (T SW = SW for t ≥ t).
0.389 0.484
0.384 0.456
Social Welfare
Social Welfare
0.380 0.427
0.375 0.399
0.370 0.370
t T t t T
Penalty Penalty
(a): 0 < s < s (b): s ≤ s < s

(v = 1, β = 0.55, γ = 0.12, k = 1, s = 0.5) (v = 1, β = 0.55, γ = 0.12, k = 1, s = 0.8)
0.447 0.484
Total Social Welfare
Total Social Welfare
0.445 0.473
0.443 0.462
0.442 0.451
0.440 0.440
t T t t T
Penalty Penalty
(c): 0 < s < s (d): s ≤ s < s

(v = 1, β = 0.55, γ = 0.12, k = 1, s = 0.5) (v = 1, β = 0.55, γ = 0.12, k = 1, s = 0.8)
Figure 9: Effect of Penalty on Social Welfare and Total Social Welfare
Social welfare (SW ) comprises CS and the copyright holder’s profit. When 0 < s < s,
as the expected penalty rises, more consumers buy copyrighted products and the copyright
holder’s profit grows, which in turn improves SW . This implies that the pattern of SW
follows that of the expected penalty, peaking at t = t. When s ≥ s, analogously, the
relationship between SW and t follows that between x∗ t and t for 0 ≤ t < t. However, the
22

copyright holder discretely drops its price at t = t, which leads to a discontinuous jump in
SW . At this point, SW is maximized since copyrighted products are provided at a lower
price. When t ≥ t, only copyrighted products exist in the market, and thus a higher price
due to an increase in the penalty will harm SW .
Total social welfare (T SW ) comprises T CS, the copyright holder’s profit, and penalty
fees. When 0 < s < s or when s ≥ s and 0 ≤ t < t, as the expected penalty increases,
although the sum of the consumer surplus from consuming copyrighted products and the
copyright holder’s profit increases, fewer unethical consumers adopt pirated products, causing
a reduction in the sum of the consumer surplus from consuming pirated products and penalty
fees. We find that the former positive effect can be dominated by the latter negative effect
only if s ≥ s. As a result, when 0 < s < s, T SW and the expected penalty exhibit the
same pattern with t for t ≥ 0. However, when s ≥ s, combined with the non-monotonic
relationship between x∗ t and t, T SW has a complex relationship with t for 0 ≤ t < t.
Furthermore, when s ≥ s and t ≥ t, T SW coincides with SW because there is no piracy
supply.
POLICY IMPLICATIONS
After solving the equilibrium and identifying the effects of regulation, we analyze a series
of policy tools that could help the regulator combat piracy. We start by solving the welfare-
maximizing regulation.
Optimal Regulation
Proposition 7 (Optimal Regulation).
(i) (a) If 0 < s < s, SW is maximized at t = t.
(b) Otherwise, if s ≥ s, SW is maximized at t = t.
(ii) (a) If 0 < s < s, T SW is maximized at t = t.
23

(b) If s ≤ s < s+ , T SW is maximized at t = t, where s+ > s.
(c) Otherwise, if s ≥ s+ , T SW is maximized at t = t+ , where 0 < t+ < t.
Proposition 7 follows directly from Proposition 6. Let’s start with the scenario where
the regulator wants to maximize social welfare (SW ). When the pirate firm can develop
anti-tracking technologies at a relatively low cost with 0 < s < s, SW is maximized at the
moderate penalty t = t, where the expected penalty hits the maximum (Proposition 2) and
piracy consumption touches the minimum but cannot be completely eliminated (Proposition
3). The reason is that a penalty higher than t will encourage more consumer adoption
of pirated products and thus hurt the copyright holder due to the pirate firm’s strategic
investment in anti-tracking technologies. Therefore, a moderate penalty that maximizes the
expected penalty can simultaneously maximize SW and minimize piracy consumption. This
suggests the following actionable guideline for policymaking in this scenario: To achieve the
social optimum, the regulator should simply resort to maximizing the expected penalty when
setting the penalty. On the other hand, we also find that when s ≥ s, SW is peaked at the
high penalty t = t that edges the pirate firm out of the market.
Next, we consider the scenario where the regulator tries to maximize total social welfare
(T SW ) that includes the consumer surplus from pirated products and penalty fees. The
optimal penalty that maximizes T SW is the same as that maximizing SW , except for the
case when s ≥ s+ . As shown in Figure 10, when s > s+ , T SW increases with t for 0 ≤ t < t+ ,
then decreases with t for t+ ≤ t < t, and reaches the maximum at t = t+ . In comparison,
SW increases monotonically with t for 0 ≤ t < t. The difference comes from the consumer
surplus from pirated products and penalty fees, which decrease with the expected penalty
and thus decrease with t given s > s+ . Consequently, the global maximizer of T SW could
be either t = t+ or t = t as shown in Figure 10. At t = t, the copyright holder will set a low
price to force the pirate firm out of the market, and this task is easier to achieve when the
anti-tracking technology is costly. As a result, copyrighted products’ price at t = t increases
with s, and hence, T SW at this point becomes lower under a higher s. Therefore, when
24

s ≥ s+ , T SW at t = t+ can exceed T SW at t = t.
Social Welfare Total Social Welfare

0.453
0.432
0.411
0.391
0.370
t+ t
Penalty
Figure 10: Effect of Penalty on Social Welfare and Total Social Welfare for s ≥ s+
(v = 1, β = 0.55, γ = 0.12, k = 1, s = 260)
Supply-Side Regulation
In practice, regulators can also explore supply-side regulations to combat piracy. Will
supply-side regulations completely replace the role of demand-side regulations? As we have
argued earlier, this is unlikely, as it may prove challenging to rely on supply-side regulations
solely to completely shut down piracy websites, who have the ability to host their servers
globally and constantly alter their domain names to avoid detection. We formally examine
the role of supply-side regulation and its relationship with demand-side regulation next.
Specifically, we extend the main model by allowing the regulator to exert an effort of
d·y 2 /2 to raid and take down the pirate firm with probability y ∈ [0, 1], where d > 0 is a cost
parameter. As a result, given the regulator’s supply-side investment, with probability y, we
end up with the copyright holder as the monopoly in the market, and with the complementary
probability of 1 − y, the market remains the same as in the main model.
If social welfare (SW ) is the regulator’s target, it will choose t and y to maximize ySW M +
(1−y)SW −d·y 2 /2, where SW M is social welfare when the copyright holder monopolizes the
market. On the other hand, if the regulator seeks to maximize total social welfare (T SW ),
its objective becomes yT SW M + (1 − y)T SW − d · y 2 /2, where by definition, T SW M =
SW M . Solving the regulator’s joint maximization problem, we obtain the optimal supply-
25

side regulation y ∗ and optimal demand-side regulation t∗ by the following proposition.
Proposition 8.
(i) When the regulator aims to maximize SW ,

n o
SW M −SW |t=t
(a) if 0 < s < min{ŝ, s}, y ∗ = min d
,1 and t∗ = t;
(b) if min{ŝ, s} ≤ s < s, y ∗ = 0 and t∗ = t;
(c) otherwise, if s ≥ s, y ∗ = 0 and t∗ = t.
(ii) When the regulator aims to maximize T SW , y ∗ = 0 and the optimal demand-side
regulation t∗ is the same as in Proposition 7.
Proposition 8 implies that even if the regulator has access to supply-side regulation, it
does not use it unless s is low. The intuition is that using supply-side regulation to shut down
the pirate firm softens market competition and thus hurts social welfare. When SW M < SW
(T SW M < T SW ), the regulator will optimally choose y ∗ = 0 to maximize (total) social
welfare. An active supply-side regulation is implemented only when s is sufficiently low such
that 0 < s < min{ŝ, s}, under which case, so much piracy is supplied in the market such
that too many consumers switch from the copyright holder to the pirate firm, which reduces
SW . In a nutshell, Proposition 8 suggests that supply-side regulation may be useful when
the anti-tracking technology is available at a sufficiently low cost, but it does not substitute
the role of demand-side regulation even when it is being actively used.
Educating Consumers about Copyright Protection
Will it help suppress privacy consumption by educating instead of penalizing consumers?

This is the question we try to answer in this section by investigating the effect of γ on
piracy consumption. The premise is that by educating the public about the rationale and
importance of copyright protection, the regulator could potentially raise the fraction of
ethical consumers in the population. To this end, we need to expand our main analysis by
26

relaxing Assumption 1 and extending the support of γ from [0, (1 − β)/β) to the full range
of [0, 1]. The following proposition characterizes how γ impacts piracy consumption.14
Proposition 9 (Effect of Fraction of Ethical Consumers on Piracy Consumption). Given

e such that Dp∗ jumps discretely at γ = γ
0 < t < t̂, there exists a threshold γ e and decreases
with γ for γ
e < γ < 1.
The condition of 0 < t < t̂ in Proposition 9 ensures that the penalty is not prohibitively
high so that the pirate firm remains active in the market. The proposition shows that a
higher γ could increase piracy consumption. This is also illustrated in Figure 11 below. The
intuition is that as the fraction of ethical consumers in the market increases, the copyright
holder may find it more profitable to give up all unethical consumers and serve only ethi-
cal consumers by setting the monopoly price. As a result, this would lead to an increase in
piracy consumption from unethical consumers, because all of them now turn to pirated prod-
ucts. Moreover, it can be verified that this result does not depend on whether the tracking
probability x is exogenous or endogenous. In summary, it seems educating consumers about
ethical and legal usage of copyrighted products could inadvertently increase piracy consump-
tion, and thus cannot replace the role of punitive regulation on piracy consumption.
0.535
Piracy Consumption
0.401
0.268
0.134
0.000
1!-
-
e
. 1
.
Figure 11: Effect of Fraction of Ethical Consumers on Piracy Consumption

(v = 1, β = 0.75, t = 0.05, k = 1, s = 0.5)
14
We assume that the pirate firm moves after the copyright holder to preclude the mixed-strategy equilibrium due
to the simultaneous move of the two firms for large γ. As shown by an extension below, sequential move produces
the same equilibrium outcome as simultaneous move for the main model under Assumption 1.
27

EXTENSIONS
Alternative Timings of the Game
This extension considers two alternative timings of the game. First, when the copyright
holder sets pc before the pirate firm sets x, we find that the equilibrium outcome is exactly
the same as in the main model. This is because the pirate firm’s optimal choice x∗ does
not depend on the copyright holder’s price pc , as shown by Equation (1). Second, when the
pirate firm sets x before the copyright holder sets pc , Proposition 10 states that our main
findings are robust.
Proposition 10. There exist thresholds sat and sat such that
(i) if 0 < s < sat , x∗ t increases with t for 0 ≤ t < tat , decreases with t for tat ≤ t < T at ,
and is invariant in t for t ≥ T at , while Dp∗ decreases with t for 0 ≤ t < tat , increases
with t for tat ≤ t < T at , and is invariant in t for t ≥ T at ;15
at
(ii) if sat ≤ s < sat , x∗ t increases with t for 0 ≤ t < tat and decreases with t for tat ≤ t < t ,
at
while Dp∗ decreases with t for 0 ≤ t < tat and increases with t for tat ≤ t < t ;
at
(iii) otherwise, if s ≥ sat , x∗ t increases with t for 0 ≤ t < t , while Dp∗ decreases with t for
at
0≤t<t .
Endogenous Quality of Copyrighted Products
In this extension, we consider the copyright holder to be the digital content creator
and explore the impact of regulation on its incentives to invest in quality improvement for
its products (content creativity). Specifically, we add a pre-stage to the game where the
copyright holder decides on quality v of its products, which entails a cost of mv 2 /2. The
copyright holder’s profit is Φc (v) = Π∗c (v)−mv 2 /2, where m > 0 is assumed to be sufficiently
15
The superscript ”at” stands for ”alternative timing”, and it signifies thresholds in this extension. For conve-
nience, we omit the superscript for all equilibrium variables in Extensions.
28

high so that Φc (v) is concave in v. The optimal quality v ∗ is determined by the first-order
optimality condition, Φ′c (v)|v=v∗ = 0.
Using the implicit function theorem, we have
dv ∗ ∂Φ′ (v)/∂t
= − c ′′ ,
dt Φc (v) v=v ∗
which is in the same sign as ∂Φ′c (v)/∂t, as Φ′′c (v) < 0 by concavity of Φc (v). It is more
convenient to explicitly signify the dependence of Φc (v) on t by introducing ϕc (v, t) =
Φc (v); furthermore, we can isolate the impact of t on ϕc (v, t) through x∗ by introducing
ϕec (v, t, x∗ (v, t)) = ϕc (v, t) = Φc (v). Then, we have
!
dv ∗ ∂Φ′c (v) ∂ 2 ϕc ∂ 2 ϕec ∂ 2 ϕec ∂x∗ ∂ 2 ϕec ∂ 2 ϕec ∂x∗ ∂x∗ ∂ ϕec ∂ 2 x∗
∝ = = + + + + ∗ . (3)
dt ∂t ∂v∂t ∂v∂t ∂v∂x∗ ∂t ∂x∗ ∂t ∂x∗2 ∂t ∂v ∂x ∂v∂t
The first term on the right-hand side in Equation (3) is a direct effect under exogenous
x, which can be shown to be negative (see details in Web Appendix). That is, raising the
penalty induces lower product quality because the copyright holder faces weaker competition
from the pirate firm and thus has fewer incentives to differentiate its products from pirated
products. The remaining terms consist of an indirect effect due to endogenous x, which can
be shown to be positive for 0 ≤ t < T when 0 < s < s, and for 0 ≤ t < t when s ≥ s.
Intuitively, the pirate firm’s investment in anti-tracking technologies intensifies competition
and thus prompts the copyright holder to invest more in quality improvement.
29

1.227 1.162
Optimal Product Quality
Optimal Product Quality

1.136 1.067
1.044 0.971
0.952 0.875
0.860 0.780
T t T
Penalty Penalty
(a): 0 < s < s (b): s ≤ s < s

(v = 1, β = 0.55, γ = 0.12, k = 1, s = 0.5, m = 0.14) (v = 1, β = 0.55, γ = 0.12, k = 1, s = 0.8, m = 0.14)
Figure 12: Effect of Penalty on Optimal Product Quality
In Figure 12, we show that the indirect effect can dominate the direct effect such that
v ∗ increases with penalty t. Moreover, the left panel in Figure 12 illustrates that when t is
sufficiently high, v ∗ is determined as if there is no regulation in the market because the pirate
firm always sets x∗ = 0, given that the copyright holder’s equilibrium profit is the function
of x∗ t. Lastly, the right panel in Figure 12 also reveals that v ∗ first drops discretely at t = t
and then increases with t. The drop is explained by the copyright holder’s strategic use of
a low price to force the pirate firm out of the market, whereas the subsequent increasing
relationship is due to stronger protection of monopoly profit with a higher penalty. Note
that Π∗c (v) is invariant in t for t ≥ T , so v ∗ stays the same for t ≥ T .
Endogenous Copyright Protection
In this extension, we add a pre-stage to the game where the copyright holder engages in
copyright protection by influencing the quality of pirated products. In other words, we allow
the copyright holder to control β. The lower the β, the stronger the copyright protection.
To this end, we specify the copyright holder’s profit as Φc (β) = Π∗c (β) − h(1 − β)2 /2, where
h > 0 is assumed to be sufficiently high so that Φc (β) is concave in β. The optimal copyright
protection β ∗ is determined by the first-order optimality condition, Φ′c (β)|β=β ∗ = 0. By
30

implicit function theorem, we have
dβ ∗ ∂Φ′ (β)/∂t
= − c ′′ ,
dt Φc (β) β=β ∗
which has the same sign with ∂Φ′c (β)/∂t because Φ′′c (β) < 0 by concavity of Φc (β). To
separate the impact of t on Φc (β) through x∗ , we introduce ϕec (β, t, x∗ (β, t)) = ϕc (β, t) =
Φc (β). Then, we have
dβ ∗ ∂Φ′c (β) ∂ϕc ∂ϕc ∂x∗

∂
∝ = + ∗
dt ∂t ∂t ∂β ∂x ∂β
!
∂ 2 ϕec ∂ 2 ϕec ∂x∗ ∂ 2 ϕec ∂ 2 ϕec ∂x∗ ∂x∗ ∂ ϕec ∂ 2 x∗
= + + + + ∗ , (4)
∂β∂t ∂β∂x∗ ∂t ∂x∗ ∂t ∂x∗2 ∂t ∂β ∂x ∂β∂t
where the first term on the right-hand side in Equation (4) is a direct positive effect under
exogenous x. That is, tighter regulation results in weaker copyright protection since the
copyright holder faces weak competition from the pirate firm. The remaining three terms
consist of an indirect effect under endogenous x, with the first term being negative. Intu-
itively, when penalty t is high, the pirate firm has an incentive to reduce x∗ to compete with
the copyright holder. To avoid intense competition, the copyright holder may reduce β to
enlarge vertical differentiation between the two firms to mitigate this incentive.
The direct effect or the indirect effect can dominate, so β ∗ can increase or decrease with
t, as shown in Figure 13 (a) and (b). The two panels also demonstrate that β ∗ is identical
when t = 0 and when t is sufficiently high since x∗ t = 0 for both cases. Moreover, Figure 13
(c) and (d) illustrate that β ∗ can jump (drop) discretely at t = t and then decrease (increase)
with t for the case of high s. That is, to drive the pirate firm out of the market, the copyright
holder can choose a combination of a higher pec and a lower β or a combination of a lower pec
and a higher β, depending on which one is more cost-effective. Furthermore, β ∗ is constant
for t ≥ T because Π∗c (β) is invariant in t for t ≥ T .
31

0.583 0.444
Optimal Copyright Protection

0.574 0.426
0.565 0.409
0.555 0.392
0.546 0.375
T T
Penalty Penalty
(a) (v = 1, γ = 0, k = 1, s = 0.5, h = 0.6) (b) (v = 1, γ = 0, k = 1, s = 0.5, h = 0.4)
0.641 0.444

0.619 0.421
0.597 0.398
0.575 0.375
0.554 0.352
t T t T
Penalty Penalty
(c) (v = 1, γ = 0, k = 1, s = 0.8, h = 0.6) (d) (v = 1, γ = 0, k = 1, s = 0.8, h = 0.4)
Figure 13: Effect of Penalty on Optimal Copyright Protection
Figure 14 below illustrates that piracy consumption can increase with the penalty when
the copyright holder endogenously determines β. On the one hand, as in the main model,
an increase in t directly induces the pirate firm to reduce x to help piracy users evade the
penalty. On the other hand, an increase in t also affects β ∗ , which can indirectly mitigate
the pirate firm’s incentives to reduce x. It turns out that the direct effect can dominate
the indirect effect. Consequently, our main results can qualitatively remain to hold in this
extension.
32

0.500 0.500
Piracy Consumption
Piracy Consumption
0.478 0.475
0.455 0.450
0.433 0.425
0.411 0.400
t T t T
Penalty Penalty
(a) (v = 1, γ = 0, k = 1, s = 0.5, h = 0.6) (b) (v = 1, γ = 0, k = 1, s = 0.5, h = 0.4)
0.500 0.500
Piracy Consumption
Piracy Consumption
0.375 0.375
0.250 0.250
0.125 0.125
0.000 0.000
t t T t t T
Penalty Penalty
(c) (v = 1, γ = 0, k = 1, s = 0.8, h = 0.6) (d) (v = 1, γ = 0, k = 1, s = 0.8, h = 0.4)
Figure 14: Effect of Penalty on Piracy Consumption under Optimal β ∗
Using ISP as Inspector to Enforce Penalty
In practice, regulators may not have direct access to piracy consumption records, and
they may have to rely on an Internet service provider (ISP) to garner the relevant information
and carry out punishment. In this case, the ISP can be viewed as an intermediary between
the regulator and consumers. How will the existence of such an intermediary impact piracy
demand regulation? We aim to look into this question here.
If a complete contract between the regulator and the ISP is available such that they
are able to share all the relevant information about piracy users’ activities and cooperate
to jointly devise penalties based on the information, the two parties can be treated as an
integrated agent. Our main model applies to this case directly without the need of any
33

modification. On the other hand, it is possible that the ISP may be unable to track or share
piracy users’ activities with the regulator due to users’ privacy control; moreover, the ISP
may have no right to penalize consumers for piracy consumption by throttling their Internet
connection speed, because this can be seen as violation of network neutrality. In this case,
we may end up with an incomplete contract between the regulator and the ISP. As such, we
study a setup where the regulator places joint liability on the ISP to enforce the penalty for
piracy consumption by punishing the ISP based on its enforcement level.
Specifically, we add a pre-stage to the game in which an ISP serves as an inspector who
maintains an enforcement level of z ∈ [0, 1] to monitor piracy users’ activities with a cost
rz 2 /2, where r > 0 is a cost parameter. The enforcement level z together with the pirate
firm’s anti-tracking technology x determine the probability that a piracy user is tracked,
which is parameterized as z · x. In line with the incomplete-contract literature (Iyer and
Villas-Boas 2003), we assume a linear contract, where the regulator monitors the ISP’s
enforcement level and penalizes the ISP l for lack of enforcement. Then, the ISP’s objective
is to choose enforcement level z so as to maximize its profit ΠI as follows:
r
ΠI = Π0I − l(1 − z) − z 2 ,
2
where Π0I is the ISP’s baseline profit. The maximization problem yields z ∗ = min {l/r, 1}.
The following proposition shows that our main result is robust for any l > 0.
Proposition 11.
(i) If 0 < l < r, the effects of t on the x∗ t and Dp∗ follow qualitatively the same relationship
as in the main model, except that the thresholds on t depend on l and r.
(ii) If l ≥ r, the equilibrium degenerates to that in the main model.
It is not surprising to find that for a sufficiently high punishment l, the ISP will choose
z ∗ = 1 so that we end up with the main model. Furthermore, Proposition 11 shows that
34

even with relatively low l, as long as being positive, our main result still holds qualitatively.
This is important, because in practice there could exist restrictions on l. For example, the
ISP’s participation constraint ΠI |z=z∗ ≥ 0 would imply an upper bound on l.
Consumer-Initiated Anti-Tracking Actions
Next, we consider an extension where anti-tracking actions are initiated by consumers

instead of the pirate firm. Then, consumers’ utility from consuming a pirated product is
up = θβv − xt − s(1 − x)2 /2, and the pirate firm’s profit is Πp = kDp . Proposition 12 states
that piracy consumption always decreases with t.
Proposition 12. There exists a threshold sca such that
(i) if 0 < s < sca , Dp∗ decreases with t for 0 ≤ t < s and is invariant in t for t ≥ s;16
ca ca
(ii) otherwise, if s ≥ sca , Dp∗ decreases with t for 0 ≤ t < t and stays at zero for t ≥ t .
To understand this, note that total piracy cost for piracy consumption is the sum of the
expected penalty x∗ t and the cost of taking anti-tracking actions s(1 − x∗ )2 /2. Although
consumers can reduce x∗ when penalty t increases, total piracy cost is nondecreasing in t.
Consequently, raising the penalty (weakly) diminishes piracy consumption.
Pricing-Based Revenue for Pirate Firm
This extension considers a pricing-based revenue model for the pirate firm. Formally,
the pirate firm sells pirated products at price pp . Then the pirate firm’s profit is Πp =
pp Dp − s(1 − x)2 /2. We characterize the equilibrium in Web Appendix and present the
impact of t on the equilibrium outcome by the following proposition.
Proposition 13. There exists a threshold spb such that

16
The superscript ”ca” stands for ”consumer-initiated anti-tracking”, and it signifies thresholds in this extension.
35

(i) if 0 < s < spb , x∗ t and x∗ t + p∗p increase with t for 0 ≤ t < tpb and decrease with t for
pb
tpb ≤ t < t , while p∗p and Dp∗ decrease with t for 0 ≤ t < tpb and increase with t for
pb
tpb ≤ t < t ;17
pb
(ii) otherwise, if s ≥ spb , x∗ t and x∗ t + p∗p increase with t for 0 ≤ t < t , while p∗p and Dp∗
pb
decrease with t for 0 ≤ t < t .
p$p + x$ t p$p x$ t
0.106
0.080
Piracy Costs
0.053
0.027
0.000
tpb t
pb
Penalty
Figure 15: Effect of Penalty on Cost of Piracy Consumption (for 0 < s < spb )
(v = 1, β = 0.55, γ = 0.12, s = 0.04).
Proposition 13 shows that the results of this extension are similar to those of the main
model. To understand this, note that total expected cost for piracy consumption is the sum
of price p∗p and the expected penalty x∗ t. The pirate firm’s equilibrium price, p∗p , exhibits
a U-shape with t, just the opposite of the expected penalty. Intuitively, when the expected
penalty is high (low), the pirate firm should set a low (high) price to compensate consumers.
However, because the dependence of p∗p on t is a strategic effect derived from the relationship
between x∗ t and t, the total effect is still governed by this relationship, as shown in Figure
15.
In particular, we would like to highlight that Proposition 13 can shed light on regulations
of other criminal activities, such as drug sales, prostitution, and tax evasion. This is because
illegal firms engaging in these activities make money from pricing and selling their products
17
The superscript ”pb” stands for ”pricing-based revenue”, and it signifies thresholds in this extension.
36

or services and also have strong incentives to help their consumers escape the penalty for
the consumption of illegal offerings. Therefore, the message of our paper can be applied
to broader contexts: tighter regulation may beget more consumption of illegal offerings,
regardless of whether illegal firms earn revenue from advertisements or sales.
CONCLUSIONS
Regulation of piracy consumption is a unique feature of digital piracy, because piracy

users’ activities can be relatively easy to track in comparison to those for physical goods.
Consequently, regulators around the world have been passing new laws to track and penalize
piracy consumption.
Through a simple model, we show that a high penalty from the regulator may not only
be ineffective in impeding piracy consumption but also, in a worse situation, inadvertently
promote piracy consumption by spurring the pirate firm’s investment in anti-tracking tech-
nologies. This adverse situation can happen when the pirate firm can access or develop anti-
tracking technologies less costly and the penalty is relatively high. Based on our analysis, we
devise the optimal piracy consumption regulation that depends on the cost of anti-tracking
technologies of the pirate firm. Specifically, when the cost is low, the regulator can mini-
mize piracy consumption and maximize social welfare simultaneously by setting a moderate
penalty that maximizes consumers’ expected penalty and tolerates some level of piracy con-
sumption. In contrast, when the cost is high, the regulator at optimum chooses a relatively
high penalty that completely thwarts piracy supply. Our model also produces additional
policy implications. First, supply-side regulation that raids and takes down piracy websites
does not replace demand-side regulation, as it may reduce market competition. Second, edu-
cating consumers about copyright protection may unintentionally encourage more consumers
to adopt pirated products.
Lastly, we make several remarks on the applicability of our study. First, our results can
be extended to the context where the copyright holder offers its products for free by adopting
37

the advertising-based revenue model. In practice, many digital content providers, such as
Spotify, adopt the freemium model, where consumers use the basic product for free but have
to bear with advertisements in between content. When consumers have limited attention,
the time spent on watching ads entails an opportunity cost, which can be thought of as the
de-facto price paid to enjoy free products. Second, in reality, sometimes the existence of
pirated products can increase consumer awareness of copyrighted products and thus boost
their demand. Mortimer, Nosko, and Sorensen (2012) document that in the context of
music, while piracy displaces CD sales, it increases concert revenue for less well-known artists.
Intuitively, this type of positive demand spillover does not diminish the pirate firm’s strategic
reactions to regulation, and thus we expect our main result that a high penalty can increase
piracy consumption to continue to hold. Moreover, with such positive demand spillover,
the copyright holder is more willing to accommodate the co-existence of the pirate firm in
the market. Therefore, we expect the regulator to have a higher tolerance for piracy when
devising the regulation to maximize social welfare. Third, our results cannot be bluntly
applied to understand the counterfeit context, because consuming counterfeits (think of
knockoff Louis Vuitton handbags) is generally not illegal. Counterfeit users may not know
that the product in use is counterfeit or can always claim that it is authentic even if they
know it is counterfeit. In contrast, the digital piracy context has less ambiguity about piracy
users knowingly committing the wrongdoing.
38

REFERENCES
Becker, Gary S (1968), “Crime and Punishment: An Economic Approach,” Journal of Political
Economy, 76 (2), 169–217.
Blackburn, David, Jeffrey A Eisenach, and David Harrison Jr (2019), “Impacts of Digital Video
Piracy on the US Economy,” theglobalipcenter.com https://tinyurl.com/2s4hjjyy, (re-
trieved from waybackmachine; accessed March 20, 2023).
Brown, Steven Caldwell and Thomas J Holt (2021), Digital Piracy: A Global, Multidisciplinary
Account Taylor & Francis Limited.
Chen, Yeh-ning and Ivan Png (2003), “Information Goods Pricing and Copyright Enforcement:
Welfare Analysis,” Information Systems Research, 14 (1), 107–123.
Dai, Tinglong and Shubhranshu Singh (2020), “Conspicuous by its Absence: Diagnostic Expert
Testing under Uncertainty,” Marketing Science, 39 (3), 540–563.
Dejean, Sylvain, Thierry Pénard, and Raphaël Suire (2010), “Une première évaluation des effets de
la loi Hadopi sur les pratiques des Internautes français,” marsouin.org https://tinyurl.com/
3ab2r7ea, (retrieved from waybackmachine; accessed April 7, 2023).
Dey, Debabrata, Antino Kim, and Atanu Lahiri (2019), “Online Piracy and the “Longer Arm” of
Enforcement,” Management Science, 65 (3), 1173–1190.
Durbin, Erik and Ganesh Iyer (2009), “Corruptible Advice,” American Economic Journal: Mi-
croeconomics, 1 (2), 220–42.
Gao, Yuetao (2018), “On the Use of Overt Anti-counterfeiting Technologies,” Marketing Science,
37 (3), 403–424.
Gao, Yuetao and Yue Wu (2023), “Regulating Probabilistic Selling of Counterfeits,” Management
Science, 69 (8), 4498–4517.
Guo, Liang and Xiangyi Meng (2015), “Digital Content Provision and Optimal Copyright Protec-
tion,” Management Science, 61 (5), 1183–1196.
Iyer, Ganesh and Shubhranshu Singh (2018), “Voluntary Product Safety Certification,” Manage-
ment Science, 64 (2), 695–714.
Iyer, Ganesh and J Miguel Villas-Boas (2003), “A bargaining theory of distribution channels,”
Journal of Marketing Research, 40 (1), 80–100.
Jain, Sanjay (2008), “Digital Piracy: A Competitive Analysis,” Marketing Science, 27 (4), 610–626.
Lahiri, Atanu and Debabrata Dey (2013), “Effects of Piracy on Quality of Information Goods,”
Management Science, 59 (1), 245–264.
Lu, Shijie, Koushyar Rajavi, and Isaac Dinner (2021), “The Effect of Over-the-top Media Services
on Piracy Search: Evidence from a Natural Experiment,” Marketing Science, 40 (3), 548–568.
Mortimer, Julie Holland, Chris Nosko, and Alan Sorensen (2012), “Supply Responses to Digital
Distribution: Recorded Music and Live Performances,” Information Economics and Policy,
24 (1), 3–14.
MUSO (2022), “2021 MUSO Discovering Piracy by Industry Data Review,” muso.com https:
//tinyurl.com/4w6272j9, (retrieved from waybackmachine; accessed April 7, 2023).
Singh, Shubhranshu (2017), “Competition in Corruptible Markets,” Marketing Science, 36 (3),
361–381.
Tirole, Jean (1988), The Theory of Industrial Organization MIT press.
39

Tunca, Tunay I and Qiong Wu (2013), “Fighting Fire with Fire: Commercial Piracy and the
Role of File Sharing on Copyright Protection Policy for Digital Goods,” Information Systems
Research, 24 (2), 436–453.
Vernik, Dinah A, Devavrat Purohit, and Preyas S Desai (2011), “Music Downloads and the Flip
Side of Digital Rights Management,” Marketing Science, 30 (6), 1011–1027.
Wilbur, Kenneth C and Yi Zhu (2009), “Click Fraud,” Marketing Science, 28 (2), 293–308.
Wu, Yue, Esther Gal-Or, and Tansev Geylani (2022), “Regulating Native Advertising,” Manage-
ment Science, 68 (11), 8045–8061.
Wu, Yue and Tansev Geylani (2020), “Regulating Deceptive Advertising: False Claims and Skep-
tical Consumers,” Marketing Science, 39 (4), 788–806.
Zhou, Bo and Tianxin Zou (2023), “Competing for Recommendations: The Strategic Impact of
Personalized Product Recommendations in Online Marketplaces,” Marketing Science, 42 (2),
360–376.
Zhu, Yi and Anthony Dukes (2015), “Selective Reporting of Factual Content by Commercial
Media,” Journal of Marketing Research, 52 (1), 56–76.
40

WEB APPENDIX
Proof of Proposition 1
Proof. In any equilibrium, the copyright holder chooses to either charge a high price such that
the pirate firm exists in the market (subgame I), or charge a low price such that the pirate firm
cannot survive in the market and consumers only buy copyrighted products (subgame II),
which is sustainable since the pirate firm cannot make profits taking the copyright holder’s
n o
x∗ t (1−γ)kt2 (1−β)sv
price as given (Tirole 1988). Since pec − β = min 2β 2 (1−β)sv , 2(1−γ)k ≥ 0, we have

p c 1 − p c


v
if 0 ≤ pc ≤ pec ,
Πc (pc ) =
n h io
pc −x∗ t

p γ 1 − pc + (1 − γ) 1 −

otherwise,
c v (1−β)v
n o
∗ (1−γ)kt
where x = max 1 − β(1−β)sv
,0 . Notice Πc (pc ) is a piecewise quadratic function. As pc
increases, Πc (pc ) drops discretely at pc = pec . For pc > pec , its unconstrained maximizer is
(1−β)v+(1−γ)x∗ t
denoted as pIc = 2(1−γβ)
.
(i) If pIc > pec , which is equivalent to 0 < s < (1−γ)k
1−γβ
and t ≥ 0, or s ≥ (1−γ)k
1−γβ
and 0 ≤
√ 2 2 2
(2−γβ−β)βsv− β sv [(2−γβ−β) s−4(1−γ)(1−β)k]
t < t′ ≡ 2(1−γ)k
, we have pec < v2 , which implies that pIc
and pec are two local maximizers of Πc . Denote ∆Πc (t) = Πc (pIc ) − Πc (pec ). If ∆Πc (t) > 0,
the equilibrium is subgame I in which p∗c = pIc and x∗ is given by Equation (1); otherwise,
∆Πc (t) ≤ 0, the equilibrium is subgame II and p∗c = pec .
d∆Πc (t) (1−γ)k
First, we show dt
≤ 0. For t ≥ T , ∆Πc is constant. When 0 < s < 1−γβ
and
(1−γ)k
0 ≤ t < T , or when s ≥ 1−γβ
and 0 ≤ t < t′ we have

d∆Πc (t) 2(1 − γ)kt 1 2 (1 − γ)kt
= M (t) 1 − − 1 − pec (t) 1 − ,
dt β(1 − β)sv β v β(1 − β)sv
n h o i
(1−γ) (1−γ) (1−γ)kt
where M (t) = 2(1−γβ) (1−β)v
1− t + 1 > 0. For 0 ≤ t < min{t, t′ }, where t =
β(1−β)sv
β(1−β)sv
, it suffices to show M (t)− β1 1 − v2 pec (t) < 0, which increases with t. Let us consider

2(1−γ)k
4(1−γ)k (1−γ)
two cases. When t ≤ t′ ⇔ s ≤ 3−2γβ−β , we have M (t) − β1 1 − v2 pec (t) < − 2(3−2γβ−β)

< 0;
41

4(1−γ)k (1−γ)k
when t′ < t ⇔ s > we have M (t′ ) − 1
1 − v2 pec (t′ ) < 0. For t ≥ t,

3−2γβ−β
> 1−γβ
, β
h i
d∆Πc 2(1−γ)kt
dt
≤ 0 due to 1− β(1−β)sv
≤ 0.
(1−β)v
Second, we have ∆Πc (0) = 4(1−γβ) > 0. When 0 < s < (1−γ)k
1−γβ
, if ∆Πc (T ) > 0 ⇔ s < s ≡
√
(1−γ)[(1−γβ)− β(1−γ)(1−γβ)]
(1−β)(1−γβ)
k, ∆Πc (t) > 0 for all t ≥ 0; otherwise, if s ≤ s < (1−γ)k
1−γβ
, there exists
a unique threshold t such that ∆Πc (t) > 0 for 0 ≤ t < t and ∆Πc (t) ≤ 0 for t ≥ t. When

β(1−γ)pec (t′ ) x∗ (t′ )t′
s ≥ (1−γ)k
1−γβ
, ∆Π c (t ′
) = − (1−β)v
p
e c (t′
) − β
< 0, which implies that there exists a
unique threshold t such that ∆Πc (t) > 0 for 0 ≤ t < t and ∆Πc (t) ≤ 0 for t ≤ t < t′ .
(1−β)v 1−β
Lastly, we have Πc (pIc ) ≥ 4(1−γβ)
> γ v4 due to γ < β
, where γ v4 is the maximum profit
by serving ethical consumers only. This means that the copyright holder never charges a
high price to serve only ethical consumers in equilibrium.
(1−γ)k
(ii) If pIc ≤ pec , which is equivalent to s ≥ 1−γβ
and t ≥ t′ , we consider two cases. When pec ≤
v
2
, Πc (pec ) > Πc (pc ) for pc > pec , and we also have Πc (pec ) ≥ Πc (pec (t′ )) > Πc (pIc (t′ )) > γ v4 . When
pec > v2 , the copyright holder’s optimal price is v2 . So Πc is maximized at p∗c = min pec , v2 .

Combining (i) and (ii), we conclude that when 0 < s < s, or when s ≥ s and 0 ≤ t < t,
the equilibrium is subgame I; when s ≥ s and t ≥ t, the equilibrium is subgame II.
dx∗ (1−γ)k
Proof. (i) We have dt
= − β(1−β)sv < 0 for 0 ≤ t < T and x∗ = 0 for t ≥ T .
d(x∗ t) 2(1−γ)kt β(1−β)sv
(ii) When 0 < s < s, we have dt
=1− β(1−β)sv
>0⇔0<t<t≡ 2(1−γ)k
. We also have
x∗ = 0 and x∗ t = 0 for t ≥ T . When s ≥ s, if t < t, x∗ t increases with t for 0 ≤ t < t and
then decreases with t for t ≤ t < t; otherwise, x∗ t increases with t for 0 ≤ t < t. As shown in
4(1−γ)k 4(1−γ)k
proof of Proposition 1, when s > 3−2γβ−β
, t < t′ < t always holds. When s ≤ s ≤ 3−2γβ−β
,
we can show that t < t if and only if s < s. Formally, we have
p
4(1 − γ)[(3 − 2γβ − β) − 3β(1 − γ)(1 − γβ)]
t < t ⇔ ∆Πc (t) > 0 ⇔ s < s ≡ k.
β 2 (1 − γ)2 + 9(1 − β)(1 − γβ)
42

dDp∗
h i ∗
(1−γ)
Proof. We have Dp∗ = 2(1−γβ)
1− 2−γβ−β
β(1−β)v
(x∗ t) , which implies dt
∝ − d(xdt t) .
Proof. First, we analyze the effect of t on p∗c . If 0 < s < s, or if s ≥ s and 0 ≤ t < t,
(1−β)v+(1−γ)x∗ t dp∗c d(x∗ t)
p∗c = 2(1−γβ)
, which implies dt
∝ dt
. If s ≥ s, p∗c drops discontinuously from
pIc to pec at t = t since pIc > pec for 0 ≤ t ≤ t, as shown in Proof of Proposition 1. For
t ≤ t < T , p∗c = min{e pc , v/2} is weakly increasing in t since pec increases with t. For t ≥ T ,
n o
(1−β)sv v
p∗c = min 2(1−γ)k , 2 is invariant in t.
Second, we analyze the effect of t on Π∗c . If 0 < s < s, or if s ≥ s and 0 ≤ t < t,
(1−γ)
Dc∗ = 1
2
+ 2(1−β)v
(x∗ t). Since both p∗c and Dc∗ are positively linear in x∗ t, Π∗c = p∗c Dc∗ has the
same relationship with t as that between x∗ t and t. If s ≥ s and t ≤ t < T , p∗c = min{pec , v/2}
is increasing in t but does not surpass v2 . Hence, Π∗c is increasing in t. If s ≥ s and t ≥ T ,
Π∗c is also invariant in t.
Proof. (i) First, we consider CS. If 0 < s < s, or if s ≥ s and 0 ≤ t < t, we have
dCS d(x∗ t)
R1 R1
CS = γ p∗c (θv − p∗c )dθ + (1 − γ) p∗c −x∗ t (θv − p∗c )dθ. Then we have dCS
dt
= d(x ∗ t) dt
and
v (1−β)v
(1 − γ)(βp∗c − x∗ t) d(x∗ t − p∗c ) 1 (1 − γ)x∗ t dp∗c

dCS
= − 1+ ,
d(x∗ t) (1 − β)2 v d(x∗ t) 2 (1 − β)v d(x∗ t)
| {z }| {z }
positive effect negative effect
where the first term is the positive effect of a higher x∗ t, which induces more consumers to
buy copyrighted products, and the second term is the negative effect of a higher pc on CS.
Notice that if 0 < s < s and t ≥ T , CSc is invariant in t since x∗ t = 0.
R1
If s ≥ s and t ≥ t, we have CS = p∗c (θv − p∗c )dθ. Notice that p∗c discretely drops from pIc
v
pIc −x∗ t
to pec at t = t. We observe that pec < pIc < (1−β)
, which implies that both the integrand and
integration range of CS discretely jump at t = t. Hence, CS discretely jumps at t = t. For
∗
p∗c dpc
t ≤ t < T , dCS
dt
= − 1 − v dt
≤ 0. For t ≥ T , CS is invariant in t as p∗c .
43

(ii) Second, we consider T CS. If 0 < s < s, or if s ≥ s and 0 ≤ t < t, we have T CS =
∗ −x∗ t
R p(1−β)v
c

R1 ∗
R1 ∗ ∗
γ p∗c (θv − pc )dθ + (1 − γ) p∗c −x∗ t (θv − pc )dθ + x∗ t (θβv − x t)dθ . We have
v (1−β)v βv
p∗c dp∗c p∗c − x∗ t dp∗c

∗
p c − x∗ t x∗ t d(x∗ t)

dT CS
= −γ 1 − − (1 − γ) 1 − + − ,
dt v dt (1 − β)v dt (1 − β)v βv dt
∗
which implies dT CS
dt
∝ − d(xdt t) .
If s ≥ s and t ≥ t, T CS = CS decreases with t for t ≤ t < T and is invariant in t for
t ≥ T . At t = t, because T SW , which is the sum of T CS, Π∗c and penalty fees, discretely
jumps and penalty fees drop to zero as shown in the proof of Proposition 6, T CS must
discretely jump too.
Proof. (i) First, we consider SW . If 0 < s < s, or if s ≥ s and 0 ≤ t < t, we have
R1 R1
SW = γ p∗c (θv)dθ + (1 − γ) p∗c −x∗ t (θv)dθ, and
v (1−β)v
dSW (1 − γ) d ∗ d
(1 − β 2 γ)p∗c − (1 + γ − 2γβ)x∗ t (x t) ∝ (x∗ t),

= 2
dt 2(1 − γβ)(1 − β) v dt dt
because [(1 − β 2 γ)p∗c − (1 + γ − 2γβ)x∗ t] > (1 − γβ)(βp∗c − x∗ t) > 0, which is positive since
x∗ t
p∗c > pec > , as shown in the proof of Proposition 1.
β
R1
If s ≥ s and t ≥ t, we have SW = p∗c (θv)dθ. At t = t, p∗c discretely drops from pIc to
v
pIc −x∗ t
pec and pec < pIc < (1−β)
, which implies that the integration range of SW ∗ discretely jumps .
∗ ∗
Hence, SW ∗ discretely jumps at t = t. For t ≤ t < T , dSW
dt
= − pvc dpdtc ≤ 0. For t ≥ T , SW
is invariant in t.
(ii) Second, we consider T SW . If 0 < s < s, or if s ≥ s and 0 ≤ t < t, we have T SW =
∗ −x∗ t
R p(1−β)v
c

R1 R1
γ p∗c (θv)dθ + (1 − γ) p∗c −x∗ t (θv)dθ + x∗ t (θβv)dθ . We have
v (1−β)v βv
dT SW (1 − γ)[β(1 − β)v − (4 − 3β − γβ)x∗ t] d(x∗ t)

= .
dt 4β(1 − β)(1 − γβ)v dt
44

β(1−β)sv β(1−β)sv
If 0 < s < s, we have max(x∗ t) = 4(1−γ)k
< 4(1−γ)k
, which implies β(1 − β)v − (4 −
d(x∗ t)
3β − γβ)x∗ t > β(1 − β)v − (4 − 3β − γβ) β(1−β)sv
4(1−γ)k
> 0, and dT SW
dt
∝ dt
.
If s ≥ s, β(1 − β)v − (4 − 3β − γβ)x∗ t can be positive or negative for 0 ≤ t < t. For
t ≥ t, we have T SW = SW , which decreases with t for t ≤ t < T and is invariant in t for
t ≥ T . At t = t, we have
∗ 2 ( ∗ 2 ∗ 2 )
p c − x∗ t

v pc v xt
lim T SW = lim γ 1 − + (1 − γ) 1 − (1 − β) −β
t↗t t↗t 2 v 2 (1 − β)v βv
I 2 2
v pc |t=t v pec |t=t
< 1− < 1− = T SW t=t ,
2 v 2 v
where the second inequality is due to pIc > pec at t = t. So T SW discretely jumps at t = t.
Proof. (i) First, we consider social welfare. If 0 ≤ s < s, or if s ≥ s, Proposition 6 applies
immediately. If s ≤ s < s, SW has two local maximizers t and t. Since pec t=t
< pIc t=t
<
pIc −x∗ t
pIc t=t
< (1−β) t=t
, we have
Z 1 Z 1 Z 1
SW |t=t = γ (θv)dθ + (1 − γ) (θv)dθ < (θv)dθ = SW |t=t .
pIc pIc −x∗ t p
ec
v t=t (1−β)v t=t v t=t
(ii) Second, we analyze total social welfare. If 0 < s < s, T SW is maximized at t = t as

shown in Proposition 6. If s ≤ s < s, we have p∗c t=0
> pec t=t
, which implies that T SW is
maximized at t = t because for 0 ≤ t < t,
∗ 2 ∗ 2 2

v pc v pc |t=0 v pec |t=t
T SW < 1− < 1− < 1− = T SW t=t
2 v 2 v 2 v
(1−β)v (1−β)v
We prove pec t=t
< p∗c t=0
by contradiction. Suppose pec t=t ≥ 2(1−γβ)
= 2(1−γβ)
, we
√
(1−γ)k (1−γβ)s− (1−γβ)s[(1−γβ)s−(1−γ)k]
must have 1−γβ
< s < s and t ≥ (1−γ)(1−γβ)k
β(1 − β)v > t. We have
(1−β)v(4k+βs)2 (1−β)v(4k+βs)2 (1−β)(1+β−2γβ)v
Πc (pIc ) t=t
< Πc (pIc ) t=t
= 64(1−γβ)k2
≤ 64(1−γβ)k2
< 4(1−γβ)2
≤ Πc (pec ) t=t
, which
contradicts with ∆Πc (t) = Πc (pIc ) t=t
− Πc (pec ) t=t
= 0.
45

β(1−β)v
If s > s, x∗ t increases with t for 0 ≤ t < t. When x∗ t t=t
≤ 4−3β−γβ
, T SW increases
β(1−β)v
with t for 0 ≤ t < t and thus is maximized at t = t; when x∗ t t=t > 4−3β−γβ , T SW has two
√
(4−3β−γβ)s− (4−3β−γβ)s[(4−3β−γβ)s−4(1−γ)k]
local maximizers. The small one is t+ = 2(1−γ)(4−3β−γβ)k
β(1 − β)v
β(1−β)v
such that x∗ t t=t+
= 4−3β−γβ
, and the large one is t = t. The rest shows that there exists a
β(1−β)v
threshold s+ such that if s > s+ > s, then x∗ t t=t
> 4−3β−γβ
and T SW
> T SW t=t . t=t+
h i
(1−γ)kt
First, we prove that x∗ t at t = t is increasing in s. Denote e(t) = x∗ t = 1 − β(1−β)sv t.
For 0 ≤ e < e(t) < e(t), we rewrite t and pec as functions of e:
p
β(1 − β)sv −β(1 − β)sv[β(1 − β)sv − 4(1 − γ)ke]
t(e) =
2(1 − γ)k
p
β(1 − β)sv − β(1 − β)sv[β(1 − β)sv − 4(1 − γ)ke] + 2(1 − γ)ke
pec (e) = .
4β(1 − γ)k
Then, e(t) can be determined by
[(1 − β)v + (1 − γ)e]2

pec (e)
∆Πc (e, s) = − pec (e) 1 − = 0.
4(1 − β)(1 − γβ)v v
∂∆Πc ∂∆Πc ∂e ∂∆Πc

We have ∂t
= ∂e ∂t
. < 0 as shown in proof of Proposition 1 and ∂e
Since ∂t ∂t
> 0,

we must have ∂∆Πc
∂e
< 0. We also have ∂∆Π
∂s
c
= − 1 − 2pecv(e) ∂ pe∂s
c (e)
> 0. Then, e(t) is
de(t) ∂∆Πc /∂s
increasing in s since ds
= − ∂∆Π c /∂e
> 0.
β(1−β)v
We show that there exists a threshold s+ +
1 such that if s > s1 , then e(t) > 4−3β−γβ
and thus
(1−γβ)2 +(2+β−3γβ)(1−β)
T SW has a local maximum for 0 ≤ t < t at t = t+ with T SW t=t+
= 2(1−γβ)(4−3β−γβ)
v.
β(1−β)v β(1−β)v
Consider two possibilities. If e(t)|s=s ≥ 4−3β−γβ
, then e(t) > 4−3β−γβ
for all s > s, which

β(1−β)v β(1−β)v
implies that s+
1 = s. Otherwise, if e(t)|s=s < 4−3β−γβ ⇔ ∆Πc 4−3β−γβ
, s < 0, we have

β(1−β)v (1−β)v β(1−β)v
lim pec 4−3β−γβ = (4−3β−γβ) and lim ∆Πc 4−3β−γβ , s > 0, which implies that there
s→+∞ s→+∞
+ β(1−β)v +
must exist a threshold s1 such that ∆Πc 4−3β−γβ , s1 = 0 since ∆Πc increases with s.
Second, we prove that pec (t) increases with s. We can rewrite t and e = x∗ t as functions
46

of pec :
p
β 2 (1 − β)sv[(1 − β)sv − 2(1 − γ)k pec )]
β(1 − β)sv −
t(pec ) =
(1 − γ)k
p
β(1 − β)sv − β 2 (1 − β)sv[(1 − β)sv − 2(1 − γ)k pec ]
e(pec ) = − + 2β pec .
(1 − γ)k
Then, pec (t) is determined by
[(1 − β)v + (1 − γ)e(pec )]2

pec
∆Πc (pec , s) = − pec 1 − = 0.
4(1 − β)(1 − γβ)v v
∂Πc [(1−β)v+(1−γ)e(pec )] ∂e(pec ) c ∂ pec

We have ∂s
= 2(1−β)(1−γβ)v ∂s
> 0 and ∂∆Π
∂ pec
c
= ∂∆Π∂t
/ ∂t < 0, which implies that
2
dpec (t) ∂∆Πc /∂s
ds
= − ∂∆Π c /∂ pec
> 0. Hence, T SW t=t
= v2 1 − pecv(t) decreases with s.
(1−β)v
We show that when s > s+ ec (t) is sufficiently high such that pec (t) > √
2 ,p ⇒
(1−γβ)(4−3β−γβ)
(1−β)v (1−β)v
T SW > T SW t=t . As shown in (ii), we have pec (t)|s=s < 2(1−γβ)
t=t+
< √ ,
(1−γβ)(4−3β−γβ)

which means ∆Πc √ (1−β)v , s < 0. We have lim ∆Πc √ (1−β)v , s > 0,
(1−γβ)(4−3β−γβ) s→∞ (1−γβ)(4−3β−γβ)
(1−β)v
which implies that there exists a threshold s+
> s such that pec (t) > √
2 for
(1−γβ)(4−3β−γβ)
s > s+ +
2 , and s2 is uniquely determined by ∆Πc √ (1−β)v , s+
2 = 0 since ∆Πc
(1−γβ)(4−3β−γβ)
increases with s.
Lastly, we conclude that T SW t=t+
> T SW |t=t if s > s+ ≡ max{s+ +
1 , s2 }.
Proof. (i). Let us consider social welfare. If 0 < s < s, when the maximum social wel-
fare is lower than SW M , which is equivalent to SW |t=t < SW M = 38 v ⇔ s < ŝ =
√
4(1−γ)k[1−β 2 γ−(1−γβ) 1+2βγ−3β 2 γ]
n M o
∗ SW −SW |t=t
(3γ+1)γβ 3 −2γ(3+γ)β 2 +(1+3γ)β
, the regulator should choose y = min d
, 1 and
t∗ = t; otherwise, when ŝ ≤ s < s, the optimal demand-side regulation is still t and
SW |t=t ≥ SW M , which yields y ∗ = 0. If s ≥ s, the maximum social welfare is always higher
R1
than SW M since SW |t=t = pec |t=t θvdθ > SW M , which also implies that y ∗ = 0 and the
v
optimal demand-side regulation is t.

(ii). Let us consider total social welfare. If 0 < s < s, we have T SW ≥ T SW |t=0 =
47

(3+β−4γβ)v R1
8(1−γβ)
> 38 v = T SW M . If s ≥ s, we have max T SW ≥ T SW |t=t = ec |
p
t=t
θvdθ > T SW M
v
since pec |t=t < v2 . Then, maximizing W (y) yields y ∗ = 0 and the optimal t∗ is the same as in
Proposition 7.
n o
1
β(1−β)sv
Proof. The analysis is performed under 0 < t < t̂ ≡ min β− 2
v, β(1 − β)v, k
.
We use backward induction to solve this game. Given pc , the pirate firm’s profit is

 h i
xt pc −(1−β)v
− − − 2s (1 − x)2 if 0 < x ≤



 k(1 γ) 1 βv t
,


 h i
Πp (pc , x) = k(1 − γ) pc −xt − xt − s (1 − x)2 if pc −(1−β)v
<x≤ βpc
,

 (1−β)v βv 2 t t



− s (1 − x)2 βpc


2
if x > t
.
Then, we have the pirate firm’s optimal choice as follows: when 0 < pc ≤ pec , the pirate
firm does not participate in the market; when pc > pec , the pirate firm joins the market and
the optimal choice is

 h i
k(1−γ)t k(1−γ)t
1− if pc > (1 − β)v + 1 − βsv t



 βsv



x∗ =
h i h i
pc −(1−β)v k(1−γ)t k(1−γ)t
 t
if (1 − β)v + 1 − β(1−β)sv t < pc < (1 − β)v + 1 − βsv
t



 h i

1 − k(1−γ)t k(1−γ)t

β(1−β)sv
if pec < pc < (1 − β)v + 1 − β(1−β)sv t.
Given the pirate firm’s participation decision and optimal choice, we have

 pc

p 1 − if 0 ≤ pc ≤ pec ,

c



 v


k(1−γ)t
pc −[1− β(1−β)sv ]t
h i
Πc (pc ) = γpc 1 − c + (1 − γ)pc 1 −
p
k(1−γ)t
 v (1−β)v
if pec < pc ≤ (1 − β)v + 1 − β(1−β)sv
t,



 h i
k(1−γ)t

γpc 1 − pc

if (1 − β)v + 1 − t < pc ≤ v.

v β(1−β)sv
h i
k(1−γ)t
Notice that when pc > (1 − β)v + 1 − β(1−β)sv
t, only ethical consumers buy copyrighted
48

products.
Next, we solve the copyright holder’s optimal price p∗c for 1−ββ
≤ γ < 1. We have
h i
k(1−γ)t
pec < βt < (1 − β)v + 1 − β(1−β)sv t < v2 , which implies that Πc increases with pc for
0 ≤ pc ≤ pec . We also have pec < pIc . Let us discuss two cases.
k(1−γ)t
(1−β)v+(1−γ)[1− β(1−β)sv ]t
h i
k(1−γ)t
(i). If pIc = 2(1−γβ)
< (1 − β)v + 1 − β(1−β)sv
t, which is equivalent to 1−β β
<
n √ o
2βkt2 +(1−β) −β(2β−1)svt−2β 2 (1−β)sv 2 + 4k2 t2 +4β[(2β−1)(v+t)−2β 2 v]ksvt2 +β 2 [t−2(t+v)β+2β 2 v]2 s2 v 2
′
γ <γ ≡ 2(2β−1)kt2
,
then, as pc increases, Πc increases for pec < pc ≤ pIc , decreases for pIc ≤ pc < (1 − β)v +
h i h i
k(1−γ)t k(1−γ)t
1 − β(1−β)sv t, increases for (1 − β)v + 1 − β(1−β)sv t ≤ pc < v2 , and decreases for pc ≥
v
2
.This means that the copyright holder serves two types of consumers if and only if
v
ec )−Πc ( v2 ).

max ∆ΠIc (γ), ∆ΠII I I II
c (γ) > 0, where ∆Πc (γ) = Πc (pc )−Πc ( 2 ) and ∆Πc (γ) = Πc (p
First, we have
H(γ)
∆ΠIc (γ) = where H(γ) = b1 γ 4 + b2 γ 3 + b3 γ 2 + b4 γ + b5 ,
4β 3 (1 − γβ)(1 − β)3 s2 v 3
b1 = k 2 t4 ,
b2 = −2kt3 [2kt − β(1 − β)sv],
b3 = 6k 2 t4 − 2β(1 − β)kst2 v[3t + (1 − β)v] + s2 v 2 β 2 (1 − β)2 [t2 + β(1 − β)v 2 ],
b4 = −4k 2 t4 + 2β(1 − β)kst2 v[2(1 − β)v + 3t] − β 2 (1 − β)2 s2 v 2 [(1 − β)v 2 + 2(1 − β)tv + 2t2 ],
b5 = {kt2 − β(1 − β)sv[t + (1 − β)v]}2 .
We have H ′′ (γ) = 12b1 γ 2 + 6b2 γ + 2b3 > 0 since its discriminant 12(3b22 − 8b1 b3 ) < −48β 2 (1 −

4 2 2 2 4 4 ′ 1−β
β) (2 − β )k s t v < 0. Since H β
< 0, H(γ) should either decrease with γ or first

1−β ′ 1−β
decrease and then increase with γ for β < γ < γ . Given H β > 0, and H(γ ′ ) < 0,
1−β
there exists an unique threshold γ
e1 such that for β
e1 , ∆ΠIc (γ) > 0 and for
< γ < γ
e1 ≤ γ < γ ′ , ∆ΠIc (γ) ≤ 0.
γ

d∆ΠII
c (γ) 2pec kt2
Second, we have ∆ΠII
c (γ) decreases with γ because dγ
= 1− − v4 <
v 2β 2 (1−β)sv

kt2
2β 2 (1−β)sv
− v
4
< − (2β−1)v
4
II ′ II 1−β
< 0. We also have ∆Πc (γ ) < 0. Therefore, if ∆Πc β
> 0,
49

1−β
there exists an unique threshold γ
e2 such that for β
e2 , ∆ΠII
< γ < γ c (γ) > 0, and for
1−β
e2 ≤ γ < γ ′ , ∆ΠII
γ c (γ) ≤ 0. This further implies that for β < γ < γe ≡ max{e e2 },
γ1 , γ

max ∆ΠIc (γ), ∆ΠII c (γ) > 0 and copyright holder serves two types of consumers; for γ e≤γ<
γ ′ , max ∆ΠIc (γ), ∆ΠII

c (γ) ≤ 0 and thus copyright holder serves only ethical consumers. If

1−β 1−β 1−β
∆ΠII c β
≤ 0, then, ∆ΠIIc (γ) < 0 for all β < γ < 1, which implies that for β < γ < γ e1 ,
e1 ≤ γ < γ ′ , max ∆ΠIc (γ), ∆ΠII

max ∆ΠIc (γ), ∆ΠII c (γ) > 0 ; for γ c (γ) ≤ 0.
h i
k(1−γ)t
(ii) If pIc ≥ (1 − β)v + (1 − β)v + 1 − β(1−β)sv t ⇔ γ ′ ≤ γ < 1, Πc increases with pc
h i
k(1−γ)t
for pec < pc < (1 − β)v + 1 − β(1−β)sv t, implying ∆ΠIc (γ) < 0 for γ ′ ≤ γ < 1. We also
II ′

have ∆ΠII I II
c (γ) ≤ ∆Πc (γ ) < 0, which implies that max ∆Πc (γ), ∆Πc (γ) ≤ 0 and Πc is
maximized at pc = v2 .
1−β
Combing (i) and (ii), we find that for β
<γ<γ
e, copyright holder charges either pec or
pIc to serve both ethical consumers and unethical consumers; for γ
e ≤ γ < 1, copyright holder
charges v
2
e, p∗c jumps from pec or pIc
to serve only ethical consumers. Consequently, at γ = γ
(1−γ)[(1−γ)kt2 −βsvt+β 2 sv 2 ]
to v2 , so Dp∗ jumps either. For γ
e ≤ γ < 1, Dp∗ = β 2 sv 2
decreases with γ.
Proof. First, we state the equilibrium in Lemma A1.
at
Lemma A1 (Equilibrium Characterization). There exist thresholds sat and t such that
at
(i) if 0 < s < sat , or if s ≥ sat and 0 ≤ t < t , the pirate firm’s equilibrium anti-tracking
n o
technology choice is x∗ = max 1 − (2−γβ−β)(1−γ)kt
2β(1−β)(1−γβ)sv
, 0 , and the copyright holder’s equi-
(1−β)v+(1−γ)xat∗
librium price is p∗c = 2(1−γβ
;
at
(ii) otherwise, if s ≥ sat and t ≥ t , the pirate firm does not participate in the market, and
the copyright holder’s equilibrium price is p∗c = v2 .
Proof. We use backward induction to solve the game. If the pirate firm participates in the
market, at the second stage, the copyright holder’s problem is maxpc Πc = pc DI (pc , x), where
50

DI (pc , x) is indeed the demand structure in the main model. So the optimal pc given x is

 (1−β)v+(1−γ)xt β(1−β)v


2(1−γβ)
if 0 ≤ xt < 2−γβ−β
,
p∗c (xt) =
 n o
min xt , v

otherwise.
β 2
At the first stage, taking p∗c (xt) into consideration, the pirate firm’s profit is

h i
 (1−γ)k 1 − 2−γβ−β
− 2s (1 − x)2 β(1−β)v


2(1−γβ) β(1−β)v
xt if 0 ≤ xt < 2−γβ−β
;
Πp (x) =

− s (1 − x)2

otherwise.
2
n o
∗ (2−γβ−β)(1−γ)kt
Maximizing Πp (x) yields x = max 1 − 2β(1−β)(1−γβ)sv
,0 . The pirate firm participates
(1−γ)k
in the market if and only if Πp (x∗ ) > 0, which is equivalent to 0 < s < sat ≡ (1−γβ) , or
n √ o
at 2β(1−β)v (1−γβ)s− (1−γβ)s[(1−γβ)s−(1−γ)k]
s ≥ sat and 0 ≤ t < t ≡ (1−γ)(2−γβ−β)k
; otherwise, if s ≥ sat and
at
t ≥ t , the copyright holder monopolizes the market by charging v/2.
Next, we analyze the effect of regulation in Proposition 10. If 0 < s < sat , we have
d(x∗ t) β(1−β)(1−γβ)sv 2β(1−β)(1−γβ)sv
dt
> 0 ⇔ 0 ≤ t < tat ≡ (1−γ)(2−γβ−β)k
, and x∗ = 0 for t ≥ T at ≡ (1−γ)(2−γβ−β)k
. If s ≥ sat ,
similar to the proof of Proposition 2, we have
at 4(1 − γ)k − 3(1 − γβ)s 4(1 − γ)k

tat < t ⇔ Πp (x∗ ) t=tat
= > 0 ⇔ s < sat ≡ .
8(1 − γβ) 3(1 − γβ)
h i
(1−γ)k at (1−γ) 2−γβ−β ∗
Lastly, if 0 < s < sat ≡ (1−γβ)
, or s ≥ sat and 0 ≤ t < t , we have Dp∗ = 2(1−γβ)
1− β(1−β)
xt
dDp∗ ∗
and dt
∝ − d(xdt t) .
Proof. The equilibrium derivation directly follows the proof Proposition 1, except that we
need to replace t with z ∗ t. We only state the equilibrium in Lemma A2 and omit the proof.
Lemma A2.
51

(i) If 0 < s < s or if s ≥ s and 0 ≤ t < t/z ∗ , the pirate firm’s equilibrium anti-
n ∗ t)
o
tracking technology choice is x∗ = max 1 − (1−γ)k(z
β(1−β)sv
, 0 , and the copyright holder’s
(1−β)v+(1−γ)x∗ t
equilibrium price is p∗c = 2(1−γβ)
;
(ii) Otherwise, if s ≥ s and t ≥ t/z ∗ , the pirate firm does not participate in the market,
and the copyright holder’s equilibrium price is p∗c = min pec , v2 , where


 z∗ t − (1−γ)k(z ∗ t)2
if 0 ≤ t < T /z ∗ ,


β 2β 2 (1−β)sv
pec ≡

 (1−β)sv

otherwise;
2(1−γ)k
Then, we have following comparative statics.
Lemma A3.
(i) If 0 < s < s, x∗ t increases with t for 0 ≤ t < t/z ∗ , decreases with t for t/z ∗ ≤ t < T /z ∗ ,
and stays at zero for t ≥ T /z ∗ .
(ii) If s ≤ s < s, x∗ t increases with t for 0 ≤ t < t/z ∗ and decreases with t for t/z ∗ ≤ t <
t/z ∗ .
(iii) Otherwise, if s ≥ s, x∗ t increases with t for 0 ≤ t < t/z ∗ .
The proof of comparative statics largely follows that of Proposition 2 and thus is omitted.
Notice that when 0 < l < r, 0 < z ∗ < 1 and the effect of t on x∗ t is the same as the main
model, except that thresholds on t are different. When l ≥ r, we have z ∗ = 1, and thus the
equilibrium and comparative statics degenerate to the main model.
Proof. Maximizing up with respect to x yields unethical consumers’ optimal technology
choice x∗ = max{1 − st , 0}, which implies total piracy cost tca = x∗ t + 2s (1 − x∗ )2 =
52

n o
t2 s
min t − ,
2s 2
. Consumer demand for copyrighted products is

h i
pc pc −tca tca

γ 1 −
 + (1 − γ) 1 − if pc > ,
ca v (1−β)v β
Dc (pc , t ) =

1 −
 pc
v
otherwise.
Then, maximizing pc Dc (pc , tca ) yields the copyright holder’s equilibrium price as:

ca
 (1−β)v+(1−γ)t if 0 ≤ tca < β(1−β)v


2(1−γβ) 2−γβ−β
,
p∗c =
 n o
min tca , v

otherwise.
β 2
Then we have

h i
(1−γ) 2−γβ−β ca β(1−β)v
if 0 ≤ tca <



2(1−γβ)
1− β(1−β)v
t 2−γβ−β
,
Dpca∗ =


0 otherwise.
Notice that tca increases with t for 0 ≤ t < s and stays at s/2 for t ≥ s. If 0 < s <
2β(1−β)v s β(1−β)v
sca ≡ 2−γβ−β
, we have tca ≤ 2
< 2−γβ−β
, and thus Dpca∗ decreases with t for 0 ≤ t < s
ca
and is invariant in t for t ≥ s. Otherwise, if s ≥ sca , there exists a threshold t such that
q
ca ca ca
t < 2−γβ−β for 0 ≤ t < t and t ≥ 2−γβ−β for t ≥ t , where t = s − s2 − 2β(1−β)sv
ca β(1−β)v ca β(1−β)v
2−γβ−β
.
ca ca
So Dpca∗ decreases with t for 0 ≤ t < t and stays at zero for t ≥ t .
Proof. We solve the game and present the equilibrium outcome in the following lemma.
pb
Lemma A4 (Equilibrium Characterization). There exists a threshold t such that
pb
(i) if 0 ≤ t ≤ t , the pirate firm’s equilibrium price and anti-tracking technology choice
53

are
β(1 − β)[β(1 − β)v − (2 − γβ − β)t]sv

p∗p = ,
(4 − 3γβ − β)β(1 − β)sv − (2 − γβ − β)(1 − γ)t2
(1 − γ)[β(1 − β)v + (1 − γβ − β)t]t
x∗ = 1 − ,
(4 − 3γβ − β)β(1 − β)sv − (2 − γβ − β)(1 − γ)t2
and the copyright holder’s equilibrium price is
(1 − β)[2β(1 − β)sv + (1 − γ)(βs − t)t]v

p∗c = ;
(4 − 3γβ − β)β(1 − β)sv − (2 − γβ − β)(1 − γ)t2
pb
(ii) otherwise, if t > t , the pirate firm does not participate in the market, and the copyright
holder’s equilibrium price is

t v
p∗c = min , .
β 2
Proof. Similar to main model, consumer demands for copyrighted and pirated products are

h i

γ 1 − pc
pc −pp −xt pp +xt

v
+ (1 − γ) 1 − (1−β)v
if pc > β
,
DcI (pc , pp , x) =

1 −
 pc
v
otherwise;

h i
(1 − γ) pc −pp −xt − pp +xt pp +xt


(1−β)v βv
if pc > β
,
DpI (pc , pp , x) =


0 otherwise.
Similar to the main model, we consider two subgames. The copyright holder chooses
to either charge a high price and the pirate firm participates in the market (subgame I) or
charge a low price such that the pirate firm cannot earn a positive profit and consumers only
buy copyrighted products (subgame II). When the pirate firm participates in the market
with positive demand, its profit is

pc − pp − xt pp + xt s pp + xt
Πp (pc , pp , x) = (1 − γ)pp − − (1 − x)2 for pc > .
(1 − β)v βv 2 β
54

The pirate firm’s best response functions can be solved by the following first-order conditions:
 
 dΠp =
 (1−γ) p∗p (pc ) =
 β(1−β)(βpc −t)sv

dpp β(1−β)v
(βpc − 2pp − xt) = 0, 
2β(1−β)sv−(1−γ)t2
,
⇒ (5)
 
 dΠp = − (1−γ)tpp + s(1 − x) = 0;
 x∗ (p ) = 1 −
 (1−γ)(βpc −t)t
dx β(1−β)v c 2β(1−β)sv−(1−γ)t2
,
which depend on pc . Correspondingly, consumer demand for pirated products is Dp∗ (pc ) =
(1−γ)s(βpc −t) (1−γ)s(βpc −t)2
2β(1−β)sv−(1−γ)t2
and the pirate firm’s maximum profit is Π∗p (pc , p∗p (pc ), x∗ (pc )) = 2[2β(1−β)sv−(1−γ)t2 ]
.
Therefore, the pirate firm will participate in the market if and only if Π∗p (pc ) > 0 and
Dp∗ (pc ) > 0, which is equivalent to pc > βt . Consequently, in subgame II, the optimal price is
n o
t v pII
pII
c = min ,
β 2
, and the copyright holder’s profit is Π II
c = p II
c 1 − c
v
.
In subgame I, the copyright holder’s profit is Πc = pc DcI , and we have
dΠIc

2pc 2pc − pp − xtc
=γ 1− + (1 − γ) 1 − . (6)
dpc v (1 − β)v
Solving Equations (5) and (6), we have
(1 − β)[2β(1 − β)sv + (1 − γ)(βs − t)t]v

pIc = ,
(4 − 3γβ − β)β(1 − β)sv − (2 − γβ − β)(1 − γ)t2
β(1 − β)[β(1 − β)v − (2 − γβ − β)t]sv
pIp = ,
(4 − 3γβ − β)β(1 − β)sv − (2 − γβ − β)(1 − γ)t2
(1 − γ)[β(1 − β)v − (2 − γβ − β)t]t
x∗ = 1 − ,
(4 − 3γβ − β)β(1 − β)sv − (2 − γβ − β)(1 − γ)t2
which yields the copyright holder’s profit as
(1 − β)(1 − γβ)[2β(1 − β)sv + (1 − γ)(βs − t)t]2 v

ΠIc = .
[(4 − 3γβ − β)β(1 − β)sv − (2 − γβ − β)(1 − γ)t2 ]2
q
t (1−γ)β(1−β)v 2β(1−β)sv
(i) If pIc > β
, which is equivalent to (a) 0 < s ≤ 2(2−γβ−β)2
and 0 ≤ t < 1−γ
or (b)
(1−γ)β(1−β)v β(1−β)v t
s> 2(2−γβ−β)2
and 0 ≤ t < 2−γβ−β
, we have β
< v2 . We denote ∆Πc (t) = ΠIc − ΠII
c , the
55

equilibrium is subgame I if and only if ∆Πc (t) > 0, where
[β(1 − β)v − (2 − γβ − β)t]G(t)

∆Πc (t) = ,
β 2 v[(4 − 3γβ − β)β(1 − β)sv − (2 − γβ − β)(1 − γ)t2 ]2
where G(t) = a0 + a1 t + a2 t2 + a3 t3 + a4 t4 + a5 t5 and a0 = 4β 3 (1 − β)2 (1 − γβ)s2 v 3 ,

a1 = −β 2 (1 − β)[8 − 8β(1 + γ) + β 2 (1 + 6γ + γ 2 )]s2 v 2 , a2 = −4β 2 (1 − β)(1 − γ)(1 − γβ)sv 2 ,
a3 = 2β(1 − β)(1 − γ)(4 − 3γβ − β)sv, a4 = β(1 − γ)2 (1 − γβ)v, a5 = −(1 − γ)2 (2 − γβ − β).
q
(a) When 0 < s ≤ (1−γ)β(1−β)v
2(2−γβ−β)2
and 0 ≤ t < 2β(1−β)sv
1−γ
, since G′′′ (t) is an open down
quadratic function of t and G′′′ (0) > 0, G′′ (t) can either increase or first increase and then
q q
decrease with t for 0 ≤ t < 2β(1−β)sv
1−γ
. Combined with G ′′
(0) < 0 and G ′′ 2β(1−β)sv
1−γ
> 0,
we have that G′ (t) first decreases and then increases with t. Given G′ (0) < 0, G(t) can either
q
decreases or first decreases and then increases with t for 0 ≤ t < 2β(1−β)sv
1−γ
.
q
2β(1−β)sv
We also have G(0) > 0 and G 1−γ
< 0, then the continuity of G(t) implies that
q
pb pb
there exists a unique threshold t ∈ (0, 2β(1−β)sv 1−γ
) such that for 0 ≤ t < t , G(t) > 0 and
q
pb
for t ≤ t < 2β(1−β)sv
1−γ
, G(t) ≤ 0.
(b) When s > (1−γ)β(1−β)v
2(2−γβ−β)2
β(1−β)v
and 0 ≤ t < 2−γβ−β , we still have G′′′ (0) > 0 and G′′ (0) < 0.

β(1−β)v
If G′′′ 2−γβ−β > 0, then G′′ (t) increases with t and thus has at most one root for 0 ≤ t <

β(1−β)v ′′′ β(1−β)v ′′ β(1−β)v
2−γβ−β
. If G 2−γβ−β
< 0 ⇒ G 2−γβ−β
> 0, G′′ (t) first increases and then decreases
β(1−β)v
with t and thus has only one root for 0 ≤ t < 2−γβ−β
. Therefore, G′ (t) either first decreases
and then increases with t or always decreases with t. Let us consider two subcases.
(1−γ)β(1−β)v (1−γ)β(1−β)v
If 2(2−γβ−β)2
< s < (2−γβ−β)2
, since G′ (0) < 0, G(t) can either decreases or first
β(1−β)v
decreases and then increases with t for 0 ≤ t < 2−γβ−β . Together with G(0) > 0 and

β(1−β)v pb pb
G 2−γβ−β < 0, we can still determine the unique threshold t by G(t ) = 0.

If s ≥ (1−γ)β(1−β)v
(2−γβ−β)2 , we have G ′
(t) < 0 because G ′
(0) < 0 and G ′ β(1−β)v
2−γβ−β
< 0, which

β(1−β)v pb β(1−β)v
further implies that G(t) ≥ G 2−γβ−β ≥ 0. This means that t = 2−γβ−β in this subcases.
q
(ii). If pIc ≤ βt , which is equivalent to 0 < s ≤ (1−γ)β(1−β)v
2(2−γβ−β)2
and t ≥ 2β(1−β)sv
1−γ
, or s >
(1−γ)β(1−β)v β(1−β)v
2(2−γβ−β)2
and t ≥ 2−γβ−β
, the equilibrium is subgame II.
56

pb
Combining (i) and (ii), we conclude that there always exists a threshold t such that
pb pb
for 0 ≤ t < t , the equilibrium is subgame I and for t ≥ t , the equilibrium outcome is
subgame II.
Analysis of Comparative Statics

pb
Next, we analyze the impact of t on the equilibrium outcome for t ≤ t . We have
d(x∗ t) β(1 − β)(4 − 3γβ − β)sv

= ϕ(t),
dt [(4 − 3γβ − β)β(1 − β)sv − (2 − γβ − β)(1 − γ)t2 ]2
d(pp + x∗ t) 2β(1 − β)(1 − γβ)sv
= ϕ(t),
dt [(4 − 3γβ − β)β(1 − β)sv − (2 − γβ − β)(1 − γ)t2 ]2
dp∗p β(1 − β)(2 − γβ − β)sv
=− ϕ(t),
dt [(4 − 3γβ − β)β(1 − β)sv − (2 − γβ − β)(1 − γ)t2 ]2
dDp∗ (1 − γ)(2 − γβ − β)s
=− ϕ(t),
dt [(4 − 3γβ − β)β(1 − β)sv − (2 − γβ − β)(1 − γ)t2 ]2
where ϕ(t) = (1 − γ)(2 − γβ − β)t2 − 2β(1 − β)(1 − γ)vt + β(1 − β)(4 − 3γβ − β)sv.
β(1−β)(1−γ)v
(i). When s ≥ (4−3γβ−β)(2−γβ−β)
, the discriminant of ϕ(t) is always smaller than or equal to
pb
zero. This implies that ϕ(t) ≥ 0 for all 0 ≤ t < t .
β(1−β)(1−γ)v
(ii). When s < (4−3γβ−β)(2−γβ−β)
, ϕ(t) < 0 if t > tpb ,where
p
pb β(1 − β)(1 − γ)v − β(1 − β)(1 − γ)v[β(1 − β)(1 − γ)v − (2 − γβ − β)(4 − 3γβ − β)s]
t (s) = .
(1 − γ)(2 − β − γβ)
pb
Similar to the proof of Proposition 2, if tpb ≥ t , then ϕ(t) is always positive for all
pb pb pb
t ∈ (0, t ); if tpb < t , ϕ(t) ≥ 0 for 0 ≤ t ≤ tpb and ϕ(t) < 0 for tpb < t ≤ t . Moreover,
q
pb
if 8β(1−β)(1−γ)v
(8−5γβ−3β)2
≤ s ≤ β(1−β)(1−γ)v
(4−3γβ−β)(2−γβ−β)
, we have tpb
≥ 2β(1−β)sv
(1−γ)
> t , which means that
8β(1−β)(1−γ)v
we only need to focus on the case where 0 < s < (8−5γβ−3β)2
and prove that there exists a
pb pb
threshold spb ∈ (0, 8β(1−β)(1−γ)v
(8−5γβ−3β)2
) such that for 0 < s < spb , tpb < t and for s ≥ spb , tpb ≥ t .
Equivalently, we prove that ∆Πc (s, tpb (s)) is positive if and only if s < spb .
57

To begin with, we show that ∆Πc (s, tpb (s)) is monotonically decreasing in s. We have
d∆Πc (s, tpb (s)) ∂∆Πc (s, tpb (s)) ∂∆Πc (s, tpb (s)) dtpb (s)
= +
ds ∂s ∂t ds
pb pb
∂∆Πc (s, t (s)) (1 − γ) dt
< − ,
∂s 4 − 3γβ − β ds
in which the ” < ” is due to
d∆Πc 2β(1 − γβ)(1 − γ)spIc ϕ(t) βv − 2t

= 2 2
−
dt [(4 − 3γβ − β)β(1 − β)sv − (2 − γβ − β)(1 − γ)t ] β 2v
2(1 − γ)(1 − γβ) 1
< −
(2 − γβ − β)(4 − 3γβ − β) β
1−γ
<− ,
4 − 3γβ − β
(1−β) ϕ(t) 1
where the first ” < ” is due to pIc < 2−γβ−β
and [(4−3γβ−β)β(1−β)sv−(2−γβ−β)(1−γ)t2 ]2
< β(1−β)(4−3γβ−β)sv
.
dtpb ∂∆Πc (s,tpb (s))
Then we have ds
and ∂s
as follow:
dtpb β(1 − β)(4 − 3γβ − β)v

= ,
ds 2(1 − γ)[β(1 − β)v − (2 − γβ − β)tpb ]
∂∆Πc (s, tpb (s)) 2β(1 − γβ)(1 − γ)2 pIc (tpb )[β(1 − β)v − (2 − γβ − β)tpb ](tpb )2
= .
∂s [(4 − 3γβ − β)β(1 − β)sv − (2 − γβ − β)(1 − γ)(tpb )2 ]2
dtpb ∂∆Πc (s,tpb (s))

Substituting ds
and ∂s
into the following equation, we have
∂∆Πc (s, tpb (s)) (1 − γ) dtpb

− <0
∂s 4 − 3γβ − β ds
2β(1 − γ)2 (1 − γβ)pIc (tpb ){[β(1 − β)v − (2 − γβ − β)tpb ]tpb }2 β(1 − β)v
∝ −
[(4 − 3γβ − β)β(1 − β)sv − (2 − γβ − β)(1 − γ)tpb2 ]2 2
2 I pb
β(1 − γ) (1 − γβ)pc (t ) β(1 − β)v
= −
2(1 − γ)2 2
2
β(1 − β) v
<− < 0,
2(2 − γβ − β)
β(1−β)(1−γ)v
which implies that ∆Πc (s, tpb (s)) decreases with s for 0 < s < (4−3γβ−β)(2−γβ−β)
.
4(1−β)(1−γβ)v 8β(1−β)(1−γ)v
When s = 0, ∆Πc (s, tpb (s)) s=0
= ∆Πc (0, 0) = (4−3γβ−β)2
> 0. When s = (8−5γβ−3β)2
,
58

q q
pb 2β(1−β)sv 8β(1−β)(1−γ)v 2β(1−β)sv
t (s) = (1−γ)
. We can have ∆Πc (s, tpb (s)) = ∆Πc (8−5γβ−3β)2
, (1−γ)
< 0 as
shown in the proof of Lemma A4.
Therefore, there exists a threshold spb such that when 0 < s < spb , ∆Πc (s, tpb (s)) > 0
pb
and the equilibrium variables change non-monotonically with t for 0 ≤ t ≤ t ; otherwise,
pb
when s ≥ spb , the equilibrium variables change monotonically with t for 0 ≤ t ≤ t .
Analysis of Endogenous Quality of Copyrighted Products

If 0 < s < s and 0 ≤ t < T , or if s ≥ s and 0 ≤ t < t, we have x∗ > 0 and thus
∂ 2 ϕec [(1 − γ)x∗ ]2 t

=− < 0.
∂v∂t 2(1 − γβ)(1 − β)(v ∗ )2
The indirect effect is

(1 − γ)2 ktL(t)
,
2β 2 (1 − β)3 (1 − γβ)s2 (v ∗ )4
where L(t) = −5(1 − γ)2 kt2 + 4β(1 − β)(1 − γ)sv ∗ t + 2β(1 − β)2 s(v ∗ )2 , which is an open-down
4β(1−β)sv ∗
quadratic function of t and thus L(t) > 0 for 0 ≤ t ≤ 5(1−γ)k
.
∗
If 0 < s < s and 0 ≤ t < T , we should have t < β(1−β)sv
(1−γ)k
; otherwise, x∗ will be zero in
∗
2 ∗ 2
the following stage. We have L(t) > L β(1−β)sv
(1−γ)k
= β(1−β) s(vk ) (2k−βs) > 0 since s < s < 2k
β
.
β(1−β)sv ∗
If s ≥ s and 0 ≤ t < t, we should have t < t(v ∗ ) ≤ (1−γ)k
, where t(·) is defined as
the main model given v ∗ ; otherwise, x∗ will not be well defined in the following stage. If

β(1−β)sv ∗ 4(1−γ)k
2k
s ≤ s < β , then L(t) > L (1−γ)k > 0. If s ≥ 2k
β
> 3−2γβ−β , then t(v ∗ ) < t′ (v ∗ ) <
4β(1−β)sv ∗
t(v ∗ ) < 5(1−γ)k
and L(t) > 0, where t′ (·) and t(·) are defined as the main model given v ∗ .
Analysis of Endogenous Copyright Protection

The first term is positive as
∂ 2 ϕec (1 − γ)x∗ [(1 − β ∗ )2 γv + (1 − γ)(1 + γ − 2γβ ∗ )x∗ t]

= > 0.
∂β∂t 2(1 − β ∗ )2 (1 − γβ ∗ )2 v
59

The rest terms are:
∂ 2 ϕec (1 − γ)x∗ [(1 − β ∗ )2 γv + (1 − γ)(1 + γ − 2γβ ∗ )x∗ t]

= > 0,
∂β∂t 2(1 − β ∗ )2 (1 − γβ ∗ )2 v
∂ 2 ϕec (1 − γ)t[(1 − β ∗ )2 γv + (1 − γ)(1 + γ − 2γβ ∗ )x∗ t]
= > 0,
∂β∂x∗ 2(1 − β ∗ )2 (1 − γβ ∗ )2 v
∂ 2 ϕec ∂ 2 ϕec ∂x∗ (1 − γ)W (t)
∗
+ ∗
= ∗ > 0,
∂x ∂t ∂(x ) ∂t 2 2β (1 − β ∗ )2 (1 − γβ ∗ )sv 2
∂ ϕec (1 − γ)[(1 − β ∗ )v + (1 − γ)x∗ t]t
= > 0,
∂x∗ 2(1 − β ∗ )(1 − γβ ∗ )v
∂x∗ (1 − γ)kt
= ∗ (1 − 2β ∗ ),
∂β [β (1 − β ∗ )]2 sv
∂ 2 x∗ (1 − γ)k
= ∗ (1 − 2β ∗ ),
∂β∂t [β (1 − β ∗ )]2 sv
where W (t) = −3(1−γ)2 kt2 +2β ∗ (1−β ∗ )(1−γ)svt+β ∗ (1−β ∗ )2 sv 2 > 0. Let us focus on the
∗ ∗
cases when the pirate firm joins the market and x∗ > 0. If s ≤ βk∗ , we have 0 < t < β (1−γ)k (1−β )sv
∗
(1−β ∗ )sv ∗ ∗ 2 2 ∗
and thus W (t) > W β (1−γ)k = β (1−β ) ksv (k−β s) ≥ 0. Otherwise, if s > βk∗ , we have
√
∗ ′ ∗ (1−β ∗ )v[β ∗ s+ β ∗ s(β ∗ s+3k)]
0 < t < t(β ) < t (β ) < , which implies W (t) > 0 since W (0) > 0
√ 3(1−γ)k
(1−β ∗ )v[β ∗ s+ β ∗ s(β ∗ s+3k)]
and W 3(1−γ)k
= 0, where t′ (·) and t(·) are defined in the main model
∂x∗ ∂ 2 x∗
given β ∗ . Notice that if β ∗ > 12 , ∂β
and ∂β∂t
are negative and thus the three terms of indirect
effect are all negative.
60

Regulating Digital Piracy Consumption

Uploaded by

Copyright:

Available Formats

Regulating Digital Piracy Consumption

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Regulating Digital Piracy Consumption

Uploaded by

Copyright:

Available Formats

Regulating Digital Piracy Consumption

Electronic copy available at: https://ssrn.com/abstract=4198295

Electronic copy available at: https://ssrn.com/abstract=4198295

Electronic copy available at: https://ssrn.com/abstract=4198295

Table 1: Pirate Bay’s Anti-Tracking Actions

We formalize the above observations by developing a simple vertical competition model.

Electronic copy available at: https://ssrn.com/abstract=4198295

Electronic copy available at: https://ssrn.com/abstract=4198295

Electronic copy available at: https://ssrn.com/abstract=4198295

Electronic copy available at: https://ssrn.com/abstract=4198295

Electronic copy available at: https://ssrn.com/abstract=4198295

Electronic copy available at: https://ssrn.com/abstract=4198295

Electronic copy available at: https://ssrn.com/abstract=4198295

Electronic copy available at: https://ssrn.com/abstract=4198295

Case (a): pc > xt/β Case (b): pc ≤ xt/β

Figure 3: Demand Structure of Unethical Consumers in the Presence of Both Firms

where we have used the superscript ”I” to denote subgame I.

Electronic copy available at: https://ssrn.com/abstract=4198295

Electronic copy available at: https://ssrn.com/abstract=4198295

Πc (pc , x) = pc Dc (pc , x),

 x∗ = arg max Π (p∗ , x),

Proposition 1 (Equilibrium Characterization). There exist thresholds s and t such that

and the copyright holder’s equilibrium price is

Electronic copy available at: https://ssrn.com/abstract=4198295

Proposition 2 (Effect of Penalty on Anti-Tracking Technology Choice and Expected Penalty).

(b) otherwise, if s ≥ s, x∗ decreases with t for 0 ≤ t < t.

(c) otherwise, if s ≥ s, x∗ t increases with t for 0 ≤ t < t.

Electronic copy available at: https://ssrn.com/abstract=4198295

Electronic copy available at: https://ssrn.com/abstract=4198295

(a): 0 < s < s (b): s ≤ s < s

Figure 4: Effect of Penalty on Expected Penalty

As illustrated in Figure 4, under a relatively low s and a relatively high t, as t further

Electronic copy available at: https://ssrn.com/abstract=4198295

Proposition 3 (Effect of Penalty on Piracy Consumption).

(a): 0 < s < s (b): s ≤ s < s

Figure 5: Effect of Penalty on Piracy Consumption

Electronic copy available at: https://ssrn.com/abstract=4198295

Proposition 4 (Effect of Penalty on Copyright Holder’s Price and Profit).

(a): 0 < s < s (b): s ≤ s < s

Figure 6: Effect of Penalty on Copyright Holder’s Price

Electronic copy available at: https://ssrn.com/abstract=4198295

Pro-t 0.133 0.152

(a): 0 < s < s (b): s ≤ s < s

Figure 7: Effect of Penalty on Copyright Holder’s Profit

Proposition 5 (Effect of Penalty on Consumer Surplus and Total Consumer Surplus).

(b) Otherwise, if s ≥ s, CS exhibits a non-monotonic relationship with t in general

Electronic copy available at: https://ssrn.com/abstract=4198295

(c) Otherwise, if s ≥ s, T CS decreases with t for 0 ≤ t < t (T CS = CS for t ≥ t).

(a): 0 < s < s (b): s ≤ s < s

Total Consumer Surplus

(c): 0 < s < s (d): s ≤ s < s

Figure 8: Effect of Penalty on Consumer Surplus and Total Consumer Surplus

Electronic copy available at: https://ssrn.com/abstract=4198295

Proposition 6 (Effect of Penalty on Social Welfare and Total Social Welfare).

(c) Otherwise, if s ≥ s, SW increases with t for 0 ≤ t < t, jumps discretely at t = t,

Electronic copy available at: https://ssrn.com/abstract=4198295

(b) Otherwise, if s ≥ s, T SW exhibits a non-monotonic relationship with t in general

(a): 0 < s < s (b): s ≤ s < s

Total Social Welfare

(c): 0 < s < s (d): s ≤ s < s

∗ 2 ∗ 2 2