Content uploaded by Paul Belleflamme
Author content
All content in this area was uploaded by Paul Belleflamme on Jan 23, 2017
Content may be subject to copyright.
Review of Economic Research on Copyright Issues, 2016, vol. 13(2), pp. 1-24
THE ECONOMICS OF DIGITAL GOODS: A PROGRESS REPORT
PAUL BE LLEFL AMM E
Abstract. Ifirst review the theoretical apparatus that has been largely used so far to analyze in-
formation goods industries. I argued then that although this apparatus was fairly appropriate in
the analog era and in the early digital era, it now needs to be significantly updated. The advent of
streaming challenges indeed the main assumptions that underlie the existing models. This observa-
tion leads me to propose two main directions for future research eorts. First, one needs to better
understand, and model, how streaming modifies the way content is accessed and consumed. Second,
more attention should be given to the roles and strategies of streaming platforms, which become
inescapable intermediaries regarding the distribution and consumption of digital goods.
1. Introduction
Following Sh apiro and Var ian (1998) , information can be broa dly const rued as anything
that can be digitized, i.e., encoded as a series of ones and zeros. To be shared, information
needs to be formatted in some way, so as to be transformed into an ‘information good’. When
the chosen format is digital, the information good is called a ‘digital good’. To illustrate these
definitions, think of the novel that is built in a writer’s mind as information; for readers to
access this information, it must be written down in some form, which can be analog (a printed
book) or digital (an e-book); the same goes for music, images, films, television programs,
software, applications, games, etc.
Digital goods are inherently non-rival, as the consumption by one individual does not
prevent consumption by another individual. From the supply point of view, non-rivalry implies
that at any level of production of the good, the marginal cost of delivering it to an extra
consumer (i.e., the cost of reproducing) is close to zero. Absent any policy intervention,
digital goods are also largely non-excludable, as it is hard for anyone to exclude another
individual from consuming them; from the supply point of view, this is equivalent to saying
1
2PAULBELLEFLAMME
that the reproduction cost is negligible not only for the creator of the good, but also for
anyone else. If creators cannot exclude non-buyers, they will not be able to appropriate the
revenues of their creation, which will undermine their incentives to create (given that the
fixed costs of creation may be substantial and, mostly, sunk). This is the classical problem of
‘underproduction’ that plagues public goods: the number (and quality) of digital goods that
is produced through private forces may be inferior to what society would deem as optimal.
One solution to restore incentives so as to increase private production is to make information
goods excludable by legal means. We find here the economic rationale behind copyright,
which grants the creator of an original work exclusive rights for its use and distribution.
Like any Intellectual Property (IP) right, copyright can only play its economic role eciently
if it can be enforced at a reasonable cost. This is where the Internet and the digitization of
information goods have dramatically changed the game by allowing end users to reproduce
copyrighted works at very low cost and in almost perfect quality. ‘Digital piracy’ (i.e., the
illegal reproduction and distribution of copyrighted digital goods by end-users) became an
ubiquitous phenomenon, and deeply aected the interaction between copyright holders, tech-
nology companies and consumers. In particular, the existing business models, relying on the
distribution of information goods through controlled channels, were gradually replaced by new
innovative strategies, exploring alternative distribution modes and sources of revenues.
As usual, it takes some time for economists to describe and analyze properly the latest
technology and business developments. Although an abundant theoretical literature has ad-
dressed issues related to the pricing of information goods and to piracy (see Belleflamme and
Peitz, 2012 and 2014, for recent surveys), the modeling framework that this literature has
largely relied upon may now seem ill-suited to apprehend the current evolution in information
goods industries and, in particular, the development of streaming as an alternative channel
of distribution and consumption of digital goods. However, I will argue in this paper that the
existing modeling framework should not be discarded too quickly: it still provides researchers
THE ECONOMICS OF DIGITAL GOODS: A PROGRESS REPORT 3
with a parsimonious and tractable setting on which richer analyses can be built. Therefore,
Ifirst brieflyreviewtheexistingmodelinSection2.Then,inSection3,Idescribethemain
recent developments in digital goods industries, and I explain how they question the underly-
ing assumptions of the existing model. Finally, in Section 4, I propose two main directions for
future research: deepen the understanding of consumers choices and modes of consumption,
and focus on the roles and strategies of digital platforms. I conclude in Section 5.
2. Pricing digital goods: a simple model
The objective of this section is to brieflydescribeasimplemodelthathasbeenlargely
used in the industrial organization literature toanalyzethepricingofinformationgoodsin
the presence of end-user piracy (see, e.g., Yoon, 2002; Belleflamme, 2003; Bae and Choi, 2006).
Ifocushereonatechnicaldescriptionofthemodelanditsmainresults.Iwillamplydiscuss
the suitability of its assumptions and possible extensions in the next two sections.
2.1. The baseline setting in the analog era. Let us consider the market for a single
information good. The copyright holder faces a fixed cost of production 0and a constant
marginal cost of production 01.Amassoneofendusersarecharacterizedbytheir
valuation for the information good. For simplicity, we assume that is drawn from the
uniform distribution on the interval [01].Eachuserhasaunitdemandforthegood.Hence,
if the good is sold at price ,thetotaldemandforthegoodisequivalenttothemassofusers
for whom .Thatis,()=1.
Before digitization took place, illegal copies were not only expensive to make, but they
were also of drastically inferior quality than originals. It was therefore reasonable to assume
that copyright was (almost) perfectly enforced. In that case, the copyright holder behaves
as a monopolist and sets the price of the information good so as to maximize its profit
=()(1).Theoptimalpriceiseasilyfoundas=(1+)2;thecorresponding
(gross, per period) monopoly profitisthenequalto=(1)24.Giventhecopyright
4PAULBELLEFLAMME
length, the expected future demand for the good, and the copyright holder’s discount factor,
one can compute the present discounted value (PDV) of the flow of future profits, which
is denoted .Forthecopyrightholdertostarttheproductionoftheinformationgood,
it must be that .1As for end users, their (per period) surplus is computed as
=(1)28.
2.2. Eects of digitization. The eects of digitization on the pricing of information (now
digital) goods can be separated into two broad categories: the ‘static’ eects, which concern
the pricing of existing goods, and the ‘dynamic eects’, which concern the production of new
goods.
2.2.1. Static eects. Digitization has been shown to aect the pricing of existing digital goods
in three contrasting ways. On the one hand, and most obviously, there is the displacement of
sales due to digital piracy. The digitization of information goods and the fast penetration of
the Internet have indeed led an increasing number of end users to copy and distribute digital
goods without the authorization of their legal owners. On the other hand, digitization also
aected the demand and the supply of digital goods in ways that contributed to mitigating
the negative impacts of piracy for copyright holders. I now show how these eects can be
integrated, one by one, into the previous model.
Piracy and sales displacement. The first eect of digitization is to enlarge the end-users’ set of
options. Besides buying a unit of the original information good, they now have the possibility
to access a pirated copy. Let us assume here that end users can access this copy free of
charge, but perceive it as an imperfect substitute for the original good. In particular, end
user attaches a value (1 )to the pirated copy, where (01) can be seen as the level
of ‘quality degradation’ of the copy with respect to the original. It is then easily seen that end
users are no longer willing to pay the full price to acquire a unit of the original product, but
1We note here tha t a longer copyright term incre ases and, thereby, raises the incentives for the creation of information
goods. This is the well-known dynamic eciency rationale of IP rights.
THE ECONOMICS OF DIGITAL GOODS: A PROGRESS REPORT 5
only the additional value that the original brings on top of the copy. That is, user prefers
the original to the copy insofar as (1 ),whichisequivalentto.Asaresult,
the demand facing the copyright owner decreases from ()=1to ( )=1.The
optimal price that maximizes =()( )is found as
1=(+)2.Aslongas,
the corresponding profitis
1=()24.Itisobviousthat1implies that
1
:
not only does the copyright holder sell fewer copies (as
1=()2
=(1)2)
but she also sells them at a lower price (
1=(+)2
=(1+)2). It follows that
digital piracy reduces the PDV of future profits and, hence, the incentives to create. In other
words, digital piracy has a negative impact on welfare from a dynamic eciency point of
view. However, by reducing the market power of the copyright holder and by allowing more
consumers to access the good, digital piracy enhances welfare from a static point of view.2
These contrasted findings illustrate the usual trade-oof IP rights: static eciency comes at
the expense of dynamic ineciency.
Lower marginal costs. As stressed by Wunsch-Vincent (2013) and Pénin (2015), digital tech-
nologies reduce the marginal cost of reproducing and distributing information goods. For
digital goods, this cost becomes negligible. If we set =0in the above analysis, we have
that the copyright owner facing digital piracy sets a price of
2=2for the original good,
and achieves a profitof
2=24.Puttingthetwoeects of digitization together (sales
displacement and lower marginal cost), we observe that digitization has a negative impact on
the copyright holder’s profits if and only if
2
,whichisequivalentto1.Unsur-
prisingly, the latter condition is more likely to be satisfied the higher the quality of pirated
copies (i.e., the lower the quality degradation )andthelowerthereductioninmarginalcost
(i.e., the lower the pre-digitization value of the cost, ). As for end users, the lower marginal
cost further increases their surplus.
2Welfare under piracy, noted
1,iscomputedastheadditionofthecopyrightholder’sprofit(
1), the surplus of the
consumers of the original, 1
1
1,andthesurplusoftheconsumersofthepiratedcopy,
1
0(1 ).It
is readily shown that
1
,where=+=3(1)28.
6PAULBELLEFLAMME
Mitigating factors. Even if digitization reduces the marginal cost of reproducing and distrib-
uting information goods, it also eases the access to high-quality pirated copies and, thereby,
tends to harm copyright holders. However, digitization entails other eects that may mitigate
this negative impact. The theoretical literature has identified three mechanisms by which
piracy may have a positive eect on the profits of copyright holders.3The first is a sampling
eect.Theargumentisthefollowing: insofarasinformationgoodsarecomplexexperience
goods, the copyright holder may want to use samples of its good so as to let consumers discover
their preferences before, hopefully, buying the good. Yet, producing samples and organizing
their distribution may be costly for the copyright holder, who may therefore prefer to let
pirated copies play the role of free samples. A larger availability of pirated copies may then
increase the demand of the original good: when consuming a pirated copy, users learn their
valuation of the good and if the latter is large, they decide to purchase the (higher-quality)
original.
The second mechanism stems from the network eects that numerous digital goods exhibit:
the more they are consumed, the larger the value they generate for their users. Sources of
network eects are, e.g., word-of-mouth in the case of cultural goods, or compatibility benefits
in the case of software. In the latter two examples, it is fair to assume that original and pirated
copies contribute equally to the network eects, which introduces another channel through
which digital piracy may improve the demand for the original good.
The third mechanism, called indirect appropriation,referstotheideathattheabilityto
make copies may increase a user’s willingness to pay for the original good and, thereby, allow
the copyright holder to capture (partially or even fully) the value of these copies. As argued
by Watt (2005) or Johnson and Waldman (2005), the digitization of information goods seems
to have rendered this mechanism inoperative, at least across users for a given digital good.
Asimilareect may, however, work across complementary goods for a given user: if pirated
3See the references in Belleflamme and Peitz, 2012 and 2014.
THE ECONOMICS OF DIGITAL GOODS: A PROGRESS REPORT 7
copies are consumed together with private goods that can hardly be copied, then they can
be used as a ‘loss leader’ so as to increase alternative sources of revenues. This may explain
why an artist may welcome the diusion of pirated copies of her work, as it may boost her
popularity and, thereby, the revenues from ticket sales and merchandising.
Aoversimplified way to introduce these eects into the previous model would be to denote
by 0the additional revenue that each copy allows the copyright owner to raise (either
because of increased demand of original goods through sampling or network eects, or because
of increased sales of complementary products). In the presence of digital piracy, we have seen
above that if the original good is priced at ,thenthequantityofpiratedcopiesisequalto
.Thecopyrightholderthenchoosesto maximize =(1 )+().Theoptimal
price is found as
3=(+)2and the corresponding profitis
3=(+)24.Comparing
to the initial situation absent piracy, we have that digital piracy harms the copyright holder
if and only if
3
,whichcanberewrittenas(1 ).Thelattercondition
is more likely to be met if (i) the revenues from the mitigating factors ()aresmall,and(ii)
the cost reduction induced by digitization ()issmall.Asfortheeects of the the quality
degradation of pirated copies (), it is ambiguous: higher quality copies displace more sales
of the original goods but generate larger complementary revenues.
Amoreproactiveapproach: Freemium. Thepreviousanalysissuggeststhatdigitalpiracymay
not be as harmful for copyright holders as it appears as first glance. Theoretically, the net
eect of piracy could even be positive. Yet, this should not lead us to conclude that copyright
holders should welcome digital piracy. Such a conclusion would be based on an incorrect
counterfactual, as the mechanisms that were just described do not need piracy to work. The
copyright holder may indeed decide to market cheaper (possibly free) versions of her digital
good in order to oer samples, increase network benefits, and/or generate complementary
revenues. Clearly, this option is more costlythanlettingpiratedcopiesdothejob,butit
allows the copyright holder to regain the control over the distribution of the digital good.
8PAULBELLEFLAMME
This is the idea behind the ‘freemium’ strategy, which consists in oering a menu of ver-
sions of the digital goods; typically, a free version with limited features is proposed along a
premium version, which has the full features and is sold at a positive price. In the context of
digital piracy, this scheme is not a just a tacticof(second-degree)pricediscrimination: the
objective is also to propose a legitimate version of the digital good that directly competes with
pirated copies. As Halmenschlager and Waelbroeck (2014) nicely put it, the aim is “Fighting
piracy with free”. To depict this strategy within the above framework, let the original good
correspond to the premium version (with a net utility of ), and let the free version be
adegradedversionconferringanetutilityof(1 ).Itistheneasilyseenthatanyuser
prefers the legitimate free version to the pirated version if and only if (1 )(1 ),
or .Thatis,thecopyrightholdermustmakesurethatthequalityofherfreeversionis
slightly above the quality of pirated copies.4
2.2.2. Dynamic eects. Insofar as digital piracy negatively aects the copyright holders’ prof-
its, then incentives to produce new works or works of higher quality decrease. In other words,
although a static welfare analysis may see piracy in a favorable light, the dynamic eects are
more likely to be negative.5
However, digitization may also have some positive dynamic eects. As indicated above,
digitization contributes to reducing a number of costs related to the creation, production and
distribution of information goods. This implies that, for a given level of future profits, a
wider range of authors and creators may be incentivized to create new works (or works of
higher quality). In the model used so far, we could think of creators as being heterogeneous
in terms of their fixed cost of creation, ,andassumethatis drawn from some probability
distribution. Digitization would then make this distribution more skewed to the left, which
could somehow counterbalance the reduced incentives due to lower profits. In sum, although
4Naturally, to evaluate properly the overall profitability of that option, we would need to know the costs for the copyright
holder of organizing this form of menu pricing.
5See, e.g., Novos and Waldman (1984), Johnson (1985), or Bae and Choi (2006) for arguments along these lines.
THE ECONOMICS OF DIGITAL GOODS: A PROGRESS REPORT 9
the condition for creation to be profitable becomes more stringent (because profits decrease),
alargermassofcreatorswouldbeinapositiontosatisfyit(becausecostsdecreaseaswell).
The net dynamic eects of digitization are thus, a priori, ambiguous.6
3. Recent developments in digital goods industries
In this section, I brieflydescribethreeintertwinedinnovationsthathavedeeplyaltered
digital goods industries (in particular music and video) over the last decade. I then explain
how these innovations question the assumptions that underlie the models presented in the
previous section.
3.1. Three disruptive forces. Over the last decade, new intermediaries have proposed new
modes of distribution and consumption of digital goods, and developed new business models
to monetize them.
•The new intermediaries are digital companies such as Spotify, Deezer, Soundcloud,
Google Play Music, Appple Music, Tidal and YouTube for music, or Netflix, Hulu,
Amazon Instant Video, Showtime, HBO Now for video. As I will argue below, an
important feature of these companies is that they can be seen as platforms,asthey
bring together a large number of users who interact with each other, namely content
producers, consumers and, possibly, advertisers.
•The novelty in terms of distribution is that these platforms propose a streaming service;
that is, they supply content in real time over the internet. End users can thus access
the content of their choosing without the need to download files, which is a major
innovation as far as consumption is concerned. Compared to the purchase of physical
media or the download of (legal or illegal) files, streaming does not convey permanent
ownership of the content to the users but only provide them with a non-durable access
6Waldfo gel (2012) finds no evidence of a decline in recorded music quality since the advent of peer-to-peer sharing
(his results even suggest an increase in quality). Aguiar and Waldfogel (2016) revisit this question by using a more
comprehensive data set; they conclude that music quality “has increased in the eyes of consumers around the world.”
10 PAUL BELLEFLAMME
to this content. In exchange, streaming platforms oer a single point of access to huge
libraries of contents; they also provide users with a number of value-added services
(e.g., creation and sharing of playlists, synchronisation on several devices, recommen-
dation systems), which allow them to accesscontentanytimeandanywhere,aswell
as to discover new content more easily.
•The new business models used by streaming platforms are predominantly based on
subscription, embedded advertising, and freemium oers. In the subscription model,
users pay a flat rate for a given period of time, which allows them to consume the
content available on the platform on an unlimited basis during that period. In the
embedded advertising model, content can be accessed for free but is bundled with
advertising; the advertising nuisance can they be seen as the shadow cost of free con-
tent. Finally, as explained above, the Freemium model is a hybrid of the two previous
models: users can choose between the free version, with advertising embedded into it,
and the premium version, which oers ad-free content and other add-ons in exchange
for a flat fee.
As I now show, the model described in the previous section appears as ill-suited to analyze
the impacts of these three innovations.
3.2. Inadequacy of previous modeling. In regards of the recent changes in digital goods
industries, a number of simplifying assumptions of the previous model no longer seem ap-
propriate. First, the model assumed a unique decision-maker, the copyright holder. This
was seen as a reasonable approximation of a longer value chain, where ob jectives were fairly
aligned; for instance, in the music industry, it was sensible to assume that record companies
and artists had similar views regarding, for instance, piracy. This may still be so nowadays
but, as just explained, a major dierence comes from the emergence of digital platforms,
which increasingly stand as inescapable intermediaries between content producers and end
users. There are at least two reasons for incorporating these platforms into the model. On
THE ECONOMICS OF DIGITAL GOODS: A PROGRESS REPORT 11
the one hand, platforms control what becomes the dominant distribution channel for content,
which gives them increasing bargaining power in terms of price fixing. On the other hand, the
objectives of platforms and copyright holders do not necessarily converge; for instance, in the
music industry, platforms relying on advertising revenues have a clear incentive to make the
access to content cheaper, which artists and music labels may resent.7
Second, the model considered a copyright holder in a monopoly position, selling a single
information good at a uniform price. The monopoly assumption was usually justified by the
large horizontal dierentiation that exists across digital goods. When the only (legal) way to
consume content was to acquire the ownership ofit,itwasalsoreasonabletoconsiderthata
consumer’s utility for several digital goods could be approximated by the sum of the utilities
of the dierent goods, which made it possible to consider separately the market for each good.
Finally, the uniform price assumption made sense as very little price discrimination was ob-
served. In the era of streaming platforms, theseassumptionslookobsolete. Streamingmakes
the demands for dierent digital goods interdependent, because of the subscription model
(the average cost of consuming one particular good decreases with the quantity consumed of
other goods) and of recommendation systems (the discovery of some goods depends on the
consumption of other goods). Also, platforms propose a menu of prices for their services (the
freemium model is nothing but a form of second-degree price discrimination).
Third, as indicated above, streaming enlarges the set of options for end users in a non-trivial
way. The simple binary choice between a legal and an illegal copy has to make way to a more
complex choice between permanent ownership of a limited set of goods and non-durable access
to a seemingly unlimited library of content.8To clari fy the latt er choi ce, users’ pr eferenc es
need to be elucidated, regarding not only the content itself, but also the way it is consumed
7For e xample, the Fi nanci al T ime s re por te d in Ma rch 20 15 that U niversal Mu sic Gro up “(was) u sing lice nce neg ot iatio ns
with Spotify to push for changes to the company’s free service, privately arguing that it is not suciently distinct from
the its paid-subscription tier.”
8What may further complicate users’ (and copyright holders’) choices is that legal and illegal content often coexist on
the same platform (as is, for instance, the case on YouTube).
12 PAUL BELLEFLAMME
(convenience, conditions of availability, search and discovery possibilities, interaction with
other users, etc).
4. Building new models
To summarize th e previou s discuss ion, there are two clear directions in which existing
models of digital goods need to be updated to account for the impacts of streaming: on
the demand side, users’ choices must be better understood; on the supply side, decision-
making must shift from copyright holders to platforms. In this section, I explore these two
issues by proposing my own thoughts and by reviewing some recent theoretical and empirical
contributions.
4.1. Understanding consumer choices. As already stressed, streaming extends users’ con-
sumption possibilities for digital content. It would be wrong, however, to consider streaming
as just one additional channel through which users can access content. That is, extending the
previous model by simply adding one option in the users’ utility maximization programme
would miss the major changes that streaming is bringing. First, the unit of consumption has
changed: instead of adding up demands for individual contents, users pay for accessing an
almost bottomless library of contents. Second, the conditions of access have changed as well:
instead of a guaranteed access through permanent ownership, streamingprovidesuserswith
atemporaryaccessthatceasesoncethesubscriptionhasexpired.Finally,andperhapsmore
importantly, streaming has the potential to reduce significantly consumers’ search costs for
content and, thereby, to enhance their ability to discover not only new content but also their
own tastes. This is due to the subscription model (for a monthly fee, the user has access
to any content available on the platform) and to the value-added services that the platform
oers (recommendation systems, custom radios, content sharing on the platform itself or on
social networks, etc).
THE ECONOMICS OF DIGITAL GOODS: A PROGRESS REPORT 13
Any modeling of digital industries in the presence of streaming should thus rely on a careful
analysis of how streaming aects individual consumption behavior. To address this issue, I
first review recent empirical contributions that try to estimate how streaming aects the
existing channels of content consumption. Next, I propose some preliminary thoughts as to
how streaming technology may transform music consumption.
4.1.1. Choice of consumption channel. Anumberofinsightfulobservationsemergefromrecent
empirical work regarding the impact of streaming on other consumption channels. As far as
the choice between renting (i.e., streaming) and purchasing is concerned, it is not entirely clear
whether the two channels are substitute or complement to one another. Aguiar and Waldfogel
(2015) suggest substitutability: according to their study based on song-level digital sales,
streaming displaces legal downloads. Wlömert and Papies (2016) reach a similar conclusion:
after observing a panel of more than 2500 music consumers repeatedly over more than one year,
they find that the adoption of free and paid streaming services cannibalizes consumers’ music
expenditures on other channels. The research by Hiller (2016) also indicates substitutability;
using, as a natural experiment, the removal from YouTube of Warner Music content (and
its restoration some time later), Hiller showsthatthedisplacementfromYouTubevideosis
significant, especially for albums that have a very successful debut. Noteworthy is the fact that
even though streaming cannibalizes other channels, the net eect on content producers’ profits
may still be positive. For instance, Wlömert & Papies (2016) estimate that the overall eect
of streaming on industry revenue is positive (the positive eect of paid streaming outweighs
the potentially negative eect of free streaming).
Other studies suggest, however, the existence of some form of complementarity between
streaming and other channels. Relying on individual-level clickstream data of a representa-
tive sample of 5,000 French Internet users, Aguiar (2015) finds that ad-supported streaming
services stimulate alternative channels of music consumption that oer mobility; an explana-
tion could be that some consumers first use the free streaming service to discover the existence
14 PAUL BELLEFLAMME
of and their match value for new content and next, turn to downloading channels to enjoy the
benefits of mobile consumption. Aguiar and Martens (2016) document a similar stimulating
eect of music streaming on digital music sales: using clickstream data on a panel of more
than 16,500 European consumers, they find a positive relationship between the use of licensed
streaming websites and licensed websites selling digital music. Finally, Kretschmer and Peuk-
ert (2014) are able to estimate that the free music videos available on YouTube trigger the
sales of music albums, but have no eect on the sales of individual songs; identification comes
from comparing Germany (where music videos are blocked because of an ongoing royalties
dispute between YouTube and the German collecting society) and other countries (where the
same content is easily accessible).
As far as the impact of streaming on piracy is concerned, empirical evidence is limited.
Sinclair and Green (2016) conclude, from 35 in-depth qualitative interviews, that streaming
services are a more ecient way to tackle the problem of digital piracy than previous methods
of copyright enforcement. Yet, other studies suggest that music streaming acts more as a
complement than as a substitute of digital piracy. Borja, Dieringer and Daw (2015) use a
representative survey of 197 college students (with questions regarding their online shopping
habits); they find a positive correlation between frequent use of streaming services and illegal
downloading. Borja and Dieringer (2016) reinforce the previous result via their analysis
of 1052 surveys conducted on undergraduate students in two universities in South Florida.
The implication of the latter findings for theoretical works is that any analysis of streaming
services needs to be carried out in the shadow of piracy (even if piracy is no longer making
the headlines).
4.1.2. How does streaming aect consumption? In the previous discussion, the impact of
streaming on other consumption channels waslargelymadeundertheimplicitassumption
of stable consumption patterns. However, there are reasons to believe that streaming con-
tributes to modifying the way users consume content. To understand the changes brought
THE ECONOMICS OF DIGITAL GOODS: A PROGRESS REPORT 15
by streaming, Seabright (2016) compares streaming and downloading in terms of accessing
content once (“first play”) and accessing it repeatedly afterwards. As content is an experi-
ence good, the first play generates a discovery value for the user. In the download option
(involving permanent ownership), the cost of first play is high and creates an (uncertain)
discovery value, while guaranteeing an optionvalueofaccessingthecontentagainforever.
From t he point of view of the content pro ducer, revenue is invar iant t o the discovery value
of accessing the content. In contrast, in the streaming option, the subscriber faces a zero
marginal cost of first play, but gets value of first play, as well as a non durable option value
of accessing the content again (as it is limited to the expected lifetime of the subscription).
Here, the revenue generated is much lower for the first play but is potentially much larger for
content with higher discovery value. All in all, Seabright (2016) explains that there are three
ways in which streaming generates increased social value for the same average revenue: (i) it
considerably reduces the costs of the first play for users; (ii) it lowers the users’ incentive to
resort to alternative inecient discovery mechanisms (e.g., radio or television) so as to avoid
this high cost; (iii) it allows content producers to capture much larger revenues for content
with high discovery value.
In the same vein, Datta, Knox and Bronnenberg (2016) conjecture that streaming may aect
consumption behaviour in three potential ways.First,becausestreamingwidensthevariety
of accessible content, it may generate additional content consumption instead of displacing
it from other channels. Second, streaming may have an impact not only on the quantity of
content that is consumed but also on its variety. Like Seabright (2016), the authors argue that
in the streaming option, variety is free at the margin (whereas it is costly in the download or
purchase options). Finally, streaming may transform content discovery by allowing users to
find high-value content more eciently.
To test t hese hyp otheses, D atta, Knox and Bron nenberg (2016) con struct a un ique panel
data set of individual consumers’ listening behavior on digital music platforms (streaming
16 PAUL BELLEFLAMME
and ownership-based). Their results tend to validate the three hypotheses: after subscribing
to a streaming service, users increase their consumption of music both in quantity and in
variety; they also discover more new music, and tend to play repeatedly their best discoveries.
An important challenge for future theoretical research would be to build a model of con-
sumer behaviour that generates the same predictions, while shedding light on the underlying
mechanisms.
4.2. The rising power of streaming platforms. As streaming becomes a major mode
of content consumption, as well as a major source of revenues for content producers,9the
operators of streaming services secure an increasing market power in digital goods industries.
It is thus essential to understand correctly how they function, what are their roles, and
how they aect market outcomes. As indicated above, the streaming operators can be seen
as platforms,i.e.,asintermediaries that facilitate and manage the interaction between a large
number of users.Thisbroaddefinition encompasses multi-sided platforms, which facilitate the
interaction between users belonging to separate groups (or ‘sides’). More precisely, streaming
services are digital platforms,astheinteractiontakesplaceontheInternetandconcerns
digital goods.
The industrial organization literature studying the economics of platforms provides thus
important insights into the working of streaming services.10 Ifocushereonthreemainissues:
the identification of external eects, the strategies that platforms deploy to internalize these
eects, and the peculiarities of the competition among platforms.
4.2.1. Identifying cross- and within-group external eects. The main function of platforms
is to internalize the various external eects that the interaction between the users generate.
9Fried lan der (2016 ) rep orts th at stream ing b eca me in 2 01 5 “th e la rgest c omp one nt o f [the mu si c] indu stry revenues ,
comprising 34.3% of the market, just slightly higher than digital downloads.” Datta, Knox and Bronnenberg (2016) add
that “[a] similar shift from ownership-based to streaming-based business models is taking place in other copyright-related
industries (e.g., movies, games, books).
10For sem in al c ontribut io ns o n two-sided ma rket s, s ee R ochet and Tirole (2003) and A rm strong (200 6) . For a recent
review of the literature on these two topics, see Belleflamme and Peitz (2016).
THE ECONOMICS OF DIGITAL GOODS: A PROGRESS REPORT 17
What are the main external eects at work on streaming platforms? Streaming platforms
connect up to three categories of users: content producers (artists, record companies, movie
studios, etc), consumers and, possibly, advertisers. External eects exist across and within
these three groups. As far as cross-group external eects are concerned, there are first obvious
positive eects across content producers and consumers: on the one hand, the larger the vari-
ety of content that a streaming platform distributes, the more attractive it is for consumers; on
the other hand, the more consumers subscribe to a streaming platform, the larger the revenues
that content producers can hope to gain if their content is available on this platform. Next,
eects across advertisers and consumers are likely to have opposite signs: advertisers clearly
appreciate the presence of more consumers on a given platform, but consumers generally see
advertising as a nuisance. That is, consumers exert a positive external eect on advertisers,
whereas advertisers exert a negative external eect on consumers. Note that the Freemium
business model exploits the fact that consumers dier in their valuation of this negative eect
of advertising: the free version is targeted at consumers with a relatively low disutility from
advertising, and the premium version at the other consumers. Finally, there does not seem to
be any direct cross-group external eect across content producers and advertisers. There are,
however, indirect eects: as advertisers care about consumers (i.e., eyeballs) and consumers
care about content variety, advertisers prefer platforms giving access to more content; as for
content producers, their attitude towards the presence of advertisers on the platform depends
on a number of factors: the eect of free access on consumers’ participation, the payments
(royalties) received from the platforms for, respectively, free and paying consumers, the share
of advertising revenues that content producers obtain, etc. As mentioned above, the fact that
some music labels put pressure on Spotify to reduce the amount of music that consumers can
access for free indicates that some content producers perceive the net eect of advertising as
negative.11
11This is probably due to Spotify’s price structure: the per-stream rate that Spotify pays to artists is lower for the free
service than for the premium service (see Singleton, 2016).
18 PAUL BELLEFLAMME
Turning now to within-group external eects,theyaremainlytobefoundwithinthegroup
of consumers. Positive network eects exist among streaming consumers for at least two
reasons. First, the benefits of interacting with other users (for instance, by sharing playlists)
increase with the size of the user base. Second, recommender systems (which allow users to
discover high-value content) gain in precision as more consumers use the platform. Within
the other two groups, external eects are likely to be negative as both content producers and
advertisers compete for the attention of consumers.
4.2.2. Platform strategies. How do streaming platforms manage user participation and vol-
umes of interaction? As Hagiu and Wright (2015) explain, they can choose between two basic
ways to function: either as a marketplace (allowing ‘suppliers’ to sell their product directly
to ‘buyers’) or as a reseller (purchasing products from ‘suppliers’ to resell them to ‘buyers’).
Actually, most streaming platforms (e.g., Spotify, Netflix) combine the two ways: they act
as resellers for content, but as marketplaces for advertising. As far as content is concerned,
streaming platforms buy it from content producers and resell it to consumers; that is, they
pay royalties to content producers for the right to make their content accessible to consumers.
As a result, there is no direct interaction between consumers and content producers (as would
be the case under the marketplace mode): consumers and content producers interact with the
platform only. The opposite prevails as far as advertising is concerned. Here, the platform
first sets participation fees for advertisers andforconsumersandthen,lettheminteractas
they please. Typically, advertisers pay a fee to have their ads seen along with content, and
consumers are given free access if they accept to consume content along with ads. In this
regard, streaming platforms resemble media platforms, which mainly rely on advertising as
asourceoffinancing but which increasingly turn to freemium-like models (e.g., paywalls).12
There is, however, a major dierence between the two types of platforms: media platforms
12For recent surve ys on the economics of media in the digital era, see A nd erson and Jullien (20 15 ), and Peitz and
Reisinger (2015).
THE ECONOMICS OF DIGITAL GOODS: A PROGRESS REPORT 19
produce the large part of their content themselves, whereas streaming platforms buy their
content from third-party producers.
In terms of price strategies, one major decision facing streaming platforms is the choice
of business model: subscription, embedded-advertising, or a combination of these two (i.e.,
freemium). Thomes (2013) addresses this issue by considering a monopoly streaming platform
that chooses to oer either a free service, which is ad-financed and of low quality, or a paid
service of higher quality. The platform will optimally choose the former service as long as
consumers are suciently tolerant to advertising; in that case, streaming is shown to be an
eective tool to fight piracy. Although this paper usefully introduces two-sided aspects, it
ignores the eects that streaming has on content consumption. In this regard, the analysis by
Carroni and Paolini (2016) is more innovative: they also examine a platform’s choice between
afreemiumandapuresubscriptionmodelandshowthattheconsumers’attitudewithrespect
to advertising plays a critical role; what they addisarichermodelingoftheconsumers’utility
function (utility increases with the number of contents that are available on the platform) and
of the content producers’ participation decisions (content producers, who are assumed to have
dierent outside options, deicide whether or not to let the platform distribute their content).
Streaming platforms also use non-price strategies to manage the external eects across and
within groups. I already mentioned the value-added services that streaming platforms pro-
pose: recommendation systems, sharing of playlists, portability across devices. These services
contribute to reinforce the positive network eects among consumers. If these services are
platform specific, they also create switching costs for consumers, which is a potential source of
competitive advantage for the platform (see the discussion below).13 Another common strat-
egy is to sign exclusivity agreements with some content producers so as to attract, and retain,
13For a r ev ie w of the im pacts of ne twork eects and switching costs on competition, see Farrell and Klemperer (2007).
20 PAUL BELLEFLAMME
consumers interested in this exclusive content.14 Finally, some platforms have started to pro-
duce their own content. A priori, this strategy may have contrasting eects for the platform:
it may attract additional consumers, while scaring away third-party content producers.15
4.2.3. Platform competition. As clearly explained by Armstrong (2006), the competitive game
among platforms may dier fundamentally according to whether users have the possibility or
not to ‘multihome’, i.e., to be active on dierent platforms at the same time (if not, they are
said to ‘singlehome’). As far as streaming platforms are concerned, it is generally the case that
content producers and advertisers multihome, while consumers singlehome. More precisely,
content producers usually sell the right to distribute their content to several platforms; as a
result, consumers just need to subscribe to a single platform to access all the content they may
want to consume; and because consumers singlehome, the only way for advertisers to reach
them all is to place their ads on several platforms. This logic is, however, challenged when
platforms sign exclusive deals with particular content producers. In that case, consumers may
be forced to multihome (i.e., to have multiple subscriptions to dierent streaming platforms)
if they do not want to forgo the access to some content they value.16
The literature on two-sided platforms explains that in a market with competing platforms
where one group of users multihomes while the other group singlehomes, platforms usually
tend to compete fiercely on the singlehoming side, while ‘milking’ the multihoming side.
This is because the platforms are in a monopoly position on the multihoming side (as they
control the access to the singlehoming users that have registered with them); yet, to acquire
and reinforce this monopoly position, platforms have to court users on the singlehoming
14For insta nc e, in 2016, T idal and Ap ple Music secured t he e xclusive rel ea se of, resp ectively, Beyoncé’s an d Frank
Ocean’s latest album (see Behr, 2016). For economic analyses of this strategy and its impact on competition, see, e.g.,
Armstrong and Wright (2007), Lee (2013), and Stennek (2014).
15The typical example of a streaming platform investing in in-house content is Netflix with series like House of Cards
or Farg o.Thomes(2015)analysesthisstrategyinthevideogameindustry.
16Currently, in the music streaming industry, exclusive deals are only temporary: an album is released on a single
platform for some period of time, before being made available on other platforms as well (a tactic called ‘windowed
release’). Permanent exclusivities are very rare, probably because most consumers are reluctant to subscribe to more
than one platform.
THE ECONOMICS OF DIGITAL GOODS: A PROGRESS REPORT 21
side. Applying this intuition to streaming platforms, we would thus expect consumers to
pay a low price and content producers, a high price to access the platform. However, as I
indicated above, streaming platforms do not operate as marketplaces but as resellers: instead
of selling access to the platform to content producers, they buy their content and resell it
to consumers. The analogy with two-sided platforms is thus imperfect, but not entirely
irrelevant: consumers do pay relatively low prices (especially in the freemium model), and
content producers often complain that they receive too low compensations (with some of them
even deciding to withdraw their content from streaming services).17
The various positive external eects that are at work on streaming platforms are likely
to create positive feedback loops: a bigger platform can buy more content and, thereby,
attract more users; this, in turn, raises the platform’s revenues (either through subscriptions
or through advertising revenues as more users also attract more advertisers), which allows the
platform to grow. Moreover, platforms propose value-added services, whose benefits cannot
easily be ported to other platforms; the resulting switching costs also tend to benefitmoreto
large platforms than small ones. One expects thus the competition between platforms to lead
to strong dominant positions (‘winner-takes-most’ types of situations). It is too early to tell
whether the main streaming industries (music and videos) are indeed following this trend, but
earlier entrants (such as Spotify and Netflix) seem to benefitfromafirst-mover advantage
and currently lead their respective market. It also appears that later entrants are forced to
adopt dierentiation strategies to try and gain market shares.18
5. Conclusion
Istartedthispaperbybrieflyreviewingthetheoreticalapparatusthathasbeenlargelyused
so far to analyze information goods industries. I argued then that although this apparatus was
fairly appropriate in the analog era and in the early digital era, it needs now to be significantly
17For instance, in 2014, the singer Taylor Swift decided to remove all of her albums from Spotify.
18For insta nc e, T idal entered the music stre am ing mar ket in 201 4, p rop os in g a b et te r so und quality and cla im in g to pay
higher royalties to music artists than the rival platforms.
22 PAUL BELLEFLAMME
updated. The advent of streaming challenges indeed the main assumptions that underlie the
existing models. This observation lead me to propose two main directions for future research
eorts. First, one needs to better understand, and model, how streaming modifies the way
content is accessed and consumed, as evidenced by recent empirical work. Second, more
attention should be given to the roles and strategies of streaming platforms, which become
unescapable intermediaries regarding the distribution and consumption of digital goods; to
this end, useful insights can be gained from the recent literature on intermediation, two-sided
platforms and advertising-financed media.
Hopefully, the thoughts that I assembled in this paper will inspire other scholars to revive
the theoretical research on digital goods industries. This is, I believe, necessary to grasp
the significance of the fascinating developments that these industries currently undergo and,
thereby, to inform business practices and public policy alike.
References
Aguiar, L. (2015), “Let the Music Play? Free Streaming, Product Discovery, and Digital Music Consumption”, JRC
Workin g Pape r.
Aguiar, L. and Martens, B. (2016),“DigitalMusicConsumptionontheInternet: EvidencefromClickstream
Data”, Information Economics and Policy,34;27—43.
Aguiar, L. and J. Waldfogel (2015),“StreamingReachesFloodStage: DoesSpotifyStimulateorDepressMusic
Sales?”, National Bureau of Economic Research.
Aguiar, L. and J. Waldfogel (2016),“EventheLosersGetLuckySometimes:NewProductsandtheEvolutionof
Music Quality since Napster”, Information Economics and Policy,34;1—15.
Anderson, S.P. and B. Jullien (2015),“TheAdvertising-FinancedBusinessModelinTwo-SidedMediaMarkets”,
in S. P. Anderson, J. Waldfogel and D. Strömberg (Eds.), Handbook of Media Economics (Volume 1), Amsterdam:
North-Holland.
Armstrong, M. (2006),“CompetitioninTwo-sidedMarkets”,Rand Journal of Economics,37;668—91.
Armstrong, M. and J. Wright (2007),“Two-sidedMarkets,CompetitiveBottlenecksandExclusiveContracts”,
Economic Theory,32;353—80.
Bae, S.-H. and J.P. Choi (2006),“AModelofPiracy”,Information Economics and Policy,18;303—20.
Behr, A. (2016),“FromFrankOceantoBeyonce,theProblemwithExclusiveStreamingDeals”,The Independent
(01/09/2016). Available at http://ind.pn/2c70pNB [accessed: 24 November 2016].
Belleflamme, P. (2003),“PricingInformationGoodsinthePresenceofCopying”,inGordon,W.J.andR.Watt
(Eds.), The Economics of Copyright. Developments in Research and Analysis,CheltenhamUK:EdwardElgar.
THE ECONOMICS OF DIGITAL GOODS: A PROGRESS REPORT 23
Belleflamme, P. and M. Peitz (2012),“DigitalPiracy: Theory”,inPeitz,M.andJ.Waldfogel(Eds.),The Oxford
Handbook of the Digital Economy,Oxford:OxfordUniversityPress.
Belleflamme, P. and M. Peitz (2014),“DigitalPiracy”,inBackhaus,J.G.(Ed.),Encyclopedia of Law and Eco-
nomics,NewYork: SpringerScience+BusinessMedia.
Belleflamme, P. and M. Peitz (2016),“PlatformsandNetworkEects”, forthcoming in L.C. Corchon and M.A.
Marini (Eds.), Handbook of Game Theory and Industrial Organization,Chelthenam,UKandNorthampton,MA:
Edward Elgar.
Borja, K., S. Dieringer and J. Daw (2015),“TheEect of Music Streaming Services on Music Piracy among
College Students”, Computers in Human Behavior,45;69—76.
Borja, K. and S. Dieringer (2016), “Streaming or Stealing? The Complementary Features between Music Streaming
and Music Piracy”, Journal of Retailing and Consumer Services,32;86—95.
Carroni, E. and D. Paolini (2016),“ContentAcquisitionbyStreamingPlatforms: PremiumvsFreemium”,Mimeo.
Datta, H., G. Knox, G. and B.J. Bronnenberg (2016), “Changing Their Tune: How Consumers’ Adoption of
Online Streaming Aects Music Consumption and Discovery”, Mimeo.
Farr ell , J. an d P. K l emp e r er (2 0 07 ) ,“CoordinationandLock-In:CompetitionwithSwitchingCostsandNetwork
Eects”, in Armstrong, M. and R. Porter (Eds.), Handbook of Industrial Organization (Volume 3), Amsterdam: North-
Holland.
Frie dla nde r , J.P. ( 2 01 6 ),“NewsandNoteson2015RIAAShipmentandRevenueStatistics”.Availableathttp:
//bit.ly/1ofh5WA [accessed: 23 November 2016].
Hagiu, A. and J. Wright (2015),“MarketplaceorReseller?”,Management Science,61;184—203.
Halmenschlager, C. and P. Waelbroeck (2014),“FightingFreewithFree:Freemiumvs.Piracy”,Mimeo.
Hiller, R.S. (2016),“SalesDisplacementandStreamingMusic: EvidencefromYouTube”,Information Economics
and Policy,34;16—26.
Johnson, J.P. and M. Waldman (2005),“TheLimitsofIndirectAppropriabilityinMarketsforCopiableGoods”,
Review of E conomic Research on Copyright Issues,2;19—37.
Johnson, W.R. (1985),“TheEconomicsofCopying”,Journal of Political Economy,93;158—74.
Kretschmer, T. and C. Peukert (2014),“VideoKilledtheRadioStar? OnlineMusicVideosandDigitalMusic
Sales”, CEP Discussion Papers dp1265,CentreforEconomicPerformance,LSE.
Lee, R.S. (2013),“VerticalIntegrationandExclusivityinPlatformandTwo-SidedMarkets”,American Economic
Review,103;2960—3000.
Novos, I.E. and M. Waldman (1984),“TheEects of Increased Copyright Protection: An Analytic Approach”,
Journal of Political Economy,92;236—46.
Peitz, M . and M . Reisinger (2015),“TheEconomicsofInternetMedia”,inAnderson,S.P.,J.WaldfogelandD.
Strömberg (Eds.), Handbook of Media Economics (Volume 1), Amsterdam: North-Holland.
Pénin, J. (2015),“L’ÉconomieduDroitD’AuteurFaceAuxDéfisdelaNumérisation”,inInstitutNationaldela
Propriété Industrielle (ed.), La Propriété Intel lectuel le et la Transformation Numérique de l’Économ ie,Paris: Institut
National de la Propriété Industrielle.
Rochet, J.C. and J. Tirole (2003),“PlatformCompetitioninTwo-sidedMarkets”,Journal of the European
Economic Association,1;990—1024.
24 PAUL BELLEFLAMME
Seabright, P. (2016),“Discussionof‘StreamingReachesFloodStage’,byLuisAguilarandJoelWaldfogel”,Confer-
ence on the Economics of Intellectual Property, Software and the Internet, Toulouse, 08/01/2016.
Shapiro, C. and H. Varian (1998),Information Rules: A Strategic Guide to the Network Economy,Boston:
Harvard Business School Press.
Sinclair, G. and T. Green (2016), “Download or Stream? Steal or Buy? Developing a Typology of Today’s Music
Consumer”, Journal of Consumer Behaviour,15;3—14.
Singleton, M. (2016),“DoesSpotifyNeedtoGoafterExclusiveContent?”,The Verge (18/02/2016). Available at
http://bit.ly/1ovbkFk [accessed: 24 November 2016].
Stennek, J. (2014), “Exclusive Quality — Why Exclusive Distribution May BenefittheTV-Viewers”,Information
Economics and Policy,26;42—57.
Thomes, T.P. (2013),“AnEconomicAnalysisofOnlineStreamingMusicServices”,Information Economics and
Policy,25;81—91.
Thomes, T.P. (2015),“In-HousePublishingandCompetitionintheVideoGameIndustry”,Information Economics
and Policy,32;46—57.
Wat t, R . (2 00 5),“IndirectAppropriability20YearsOn”,Review of Economic Research on C opyright Issues,2;1—4.
Wun sch-Vinc ent, S. (2 01 3) , “The Economics of Copyright and the Internet: Moving to an Empirical Assessment
Relevant in the Digital Age”, WIPO Economics & Statistics Series,EconomicResearchWorkingPaperNo.9.
Wal df og el , J. ( 2012),“CopyrightProtection,TechnologicalChange,andtheQualityofNewProducts: Evidence
from Recorded Music since Napster”, Journal of Law and Economics,55(4);715—40.
Wlömert, N. and D. Papies (2016),“On-DemandStreamingServicesandMusicIndustryRevenues—Insights
Fro m Spotif y’ s Mar ke t Entry”, International Journal of Research in Marketing,33;314—27.
Yoon , K . (2 002 ),“TheOptimalLevelofCopyrightProtection”,Information Economics and Policy,14;327—48.
Pau l B el l efl amm e : CO R E an d L ou vain S c ho o l of M a na gem e nt , U n iv e rs i té cat h ol i qu e d e Lo u vai n , 3 4
Voie du Rom an Pays, B- 1348 Louvain la Neuve, Belgium, Paul .Be lle fl a mm e@u clo uva in. be ; a l so a ffi liat ed
with CESifo. Acknowledgements: I am extremely grateful to Elias Carroni for useful comments on a
previous draft.
