Le 15 avril 2019, le Conseil de l’UE adoptait, en co-décision avec le Parlement européen, la directive sur le droit d’auteur dans le marché unique numérique (qui doit encore être publiée au Journal Officiel). Le communiqué de presse du Conseil de l’UE du 15 avril 2019 (à comparer avec celui de la Commission) indique sobrement que: “L’UE adapte les règles relatives au droit d’auteur à l’ère numérique”. Depuis le milieu des années 1990, le droit d’auteur doit s’adapter à l’environnement numérique. Rien de nouveau donc?

La question de savoir si le droit d’auteur est “fit” pour le monde numérique revient régulièrement dans l’actualité législative et aussi judiciaire. Le précédent bloc législatif, la directive 2001/29 sur le droit d’auteur dans la société de l’information, visait aussi à ajuster ce droit. L’équilibre à trouver entre les parties prenantes évolue avec les technologies, les applications et les marchés: au début de l’Internet, pas de YouTube ou Facebook, pas de connexion rapide, ni d’écrans multiples. Les récents modes de diffusion et de consommation requièrent de nouveaux équilibres entre les usagers, mieux servis en contenus et plus exigeants en termes d’accès, les créateurs, qui peinent à se faire rémunérer pour les usages démultipliés de leurs oeuvres, et les nouveaux intermédiaires que sont les plateformes de contenus.

Le législateur européen a-t-il trouvé le point d’équilibre entre la “freedom” souhaitée par les internautes et la “fairness” pour les créateurs et producteurs. “Free & fair” (voir la vidéo associée au communiqué de presse du Conseil) est décidément un slogan à la mode: cette formule magique est aussi invoquée pour la nouvelle génération d’accords commerciaux. Mais reste un point difficile à atteindre.

L’une des questions à vérifier est de savoir si l’article 17 relatif à “l’utilisation de contenus protégés par des fournisseurs de services de partage de contenus en ligne” est ce “texte équilibré” vanté par le Conseil. En tout cas, il n’est pas court:

Article 17. Utilisation de contenus protégés par des fournisseurs de services de partage de contenus en ligne

1. Les États membres prévoient qu’un fournisseur de services de partage de contenus en ligne effectue un acte de communication au public ou un acte de mise à la disposition du public aux fins de la présente directive lorsqu’il donne au public l’accès à des œuvres protégées par le droit d’auteur ou à d’autres objets protégés qui ont été téléversés par ses utilisateurs. Un fournisseur de services de partage de contenus en ligne doit dès lors obtenir une autorisation des titulaires de droits visés à l’article 3, paragraphes 1 et 2 de la directive 2001/29/CE, par exemple en concluant un accord de licence, afin de communiquer au public ou de mettre à la disposition du public des œuvres ou d’autres objets protégés.

2. Les États membres prévoient que, lorsqu’un fournisseur de services de partage de contenus en ligne obtient une autorisation, par exemple en concluant un accord de licence, cette autorisation couvre également les actes accomplis par les utilisateurs des services relevant du champ d’application de l’article 3 de la directive 2001/29/CE lorsqu’ils n’agissent pas à des fins commerciales ou lorsque leur activité ne génère pas de revenus significatifs.

3. Quand un fournisseur de services de partage de contenus en ligne procède à un acte de communication au public ou à un acte de mise à la disposition du public, dans les conditions fixées par la présente directive, la limitation de responsabilité établie à l’article 14, paragraphe 1, de la directive 2000/31/CE ne s’applique pas aux situations couvertes par le présent article.
Le premier alinéa du présent paragraphe n’affecte pas l’éventuelle application de l’article 14, paragraphe 1, de la directive 2000/31/CE à ces fournisseurs de services pour des finalités ne relevant pas du champ d’application de la présente directive.

4. Si aucune autorisation n’est accordée, les fournisseurs de services de partage de contenus en ligne sont responsables des actes non autorisés de communication au public, y compris la mise à la disposition du public, d’œuvres protégées par le droit d’auteur et d’autres objets protégés, à moins qu’ils ne démontrent que:

a) ils ont fourni leurs meilleurs efforts pour obtenir une autorisation; et
b) ils ont fourni leurs meilleurs efforts, conformément aux normes élevées du secteur en matière de diligence professionnelle, pour garantir l’indisponibilité d’œuvres et d’autres objets protégés spécifiques pour lesquels les titulaires de droits ont fourni aux fournisseurs de services les informations pertinentes et nécessaires; et en tout état de cause
c) ils ont agi promptement, dès réception d’une notification suffisamment motivée de la part des titulaires de droits, pour bloquer l’accès aux œuvres et autres objets protégés faisant l’objet de la notification ou pour les retirer de leurs sites internet, et ont fourni leurs meilleurs efforts pour empêcher qu’ils soient téléversés dans le futur, conformément au point b).

5. Pour déterminer si le fournisseur de services a respecté les obligations qui lui incombent en vertu du paragraphe 4, et à la lumière du principe de proportionnalité, les éléments suivants sont, entre autres, pris en considération:
a) le type, l’audience et la taille du service, ainsi que le type d’œuvres ou d’autres objets
protégés téléversés par les utilisateurs du service; et
b) la disponibilité de moyens adaptés et efficaces et leur coût pour les fournisseurs de services.

6. Les États membres prévoient que, à l’égard de nouveaux fournisseurs de services de partage de contenus en ligne dont les services ont été mis à la disposition du public dans l’Union depuis moins de trois ans et qui ont un chiffre d’affaires annuel inférieur à 10 millions d’euros calculés conformément à la recommandation 2003/361/CE de la Commission, les conditions au titre du régime de responsabilité énoncé au paragraphe 4 sont limitées au respect du paragraphe 4, point a), et au fait d’agir promptement, lorsqu’ils reçoivent une notification suffisamment motivée, pour bloquer l’accès aux œuvres ou autre objets protégés faisant l’objet de la notification ou pour les retirer de leurs site internet.

Lorsque le nombre moyen de visiteurs uniques par mois de tels fournisseurs de services dépasse les 5 millions, calculé sur la base de l’année civile précédente, ils sont également tenus de démontrer qu’ils ont fourni leurs meilleurs efforts pour éviter d’autres téléversements des œuvres et autres objets protégés faisant l’objet de la notification pour lesquels les titulaires de droits ont fourni les informations pertinentes et nécessaires.

7. La coopération entre les fournisseurs de services de partage de contenus en ligne et les titulaires de droits ne conduit pas à empêcher la mise à disposition d’œuvres ou d’autres objets protégés téléversés par des utilisateurs, qui ne portent pas atteinte au droit d’auteur et aux droits voisins, y compris lorsque ces œuvres ou autres objets protégés sont couverts par une exception ou une limitation.

Les États membres veillent à ce que les utilisateurs dans chaque État membre puissent se prévaloir de l’une quelconque des exceptions ou limitations existantes suivantes lors du téléversement et de la mise à dispostion de contenus générés par les utilisateurs sur les services de partage de contenus en ligne:
a) citation, critique, revue;
b) utilisation à des fins de caricature, de parodie ou de pastiche.

8. L’application du présent article ne donne lieu à aucune obligation générale de surveillance.

Les États membres prévoient que les fournisseurs de services de partage de contenus en ligne fournissent aux titulaires de droits, à leur demande, des informations adéquates sur le fonctionnement de leurs pratiques en ce qui concerne la coopération visée au paragraphe 4 et, en cas d’accords de licence conclus entre les fournisseurs de services et les titulaires de droits, des informations sur l’utilisation des contenus couverts par les accords.

9. Les États membres prévoient la mise en place par les fournisseurs de services de partage de contenus en ligne d’un dispositif de traitement des plaintes et de recours rapide et efficace, à la disposition des utilisateurs de leurs services en cas de litige portant sur le blocage de l’accès à des œuvres ou autres objets protégés qu’ils ont téléversés ou sur leur retrait.

Lorsque des titulaires de droits demandent à ce que l’accès à leurs œuvres ou autres objets protégés spécifiques soit bloqué ou à ce que ces œuvres ou autres objets protégés soient retirés, ils justifient dûment leurs demandes. Les plaintes déposées dans le cadre du dispositif prévu au premier alinéa sont traitées sans retard indu et les décisions de blocage d’accès aux contenus téléversés ou de retrait de ces contenus font l’objet d’un contrôle par une personne physique. Les États membres veillent à ce que des mécanismes de recours extrajudiciaires soient disponibles pour le règlement des litiges. Ces mécanismes permettent un règlement impartial des litiges et ne privent pas l’utilisateur de la protection juridique accordée par le droit national, sans préjudice du droit des utilisateurs de recourir à des voies de recours judiciaires efficaces. En particulier, les États membres veillent à ce que les utilisateurs puissent s’adresser à un tribunal ou à une autre autorité judiciaire compétente pour faire valoir le bénéfice d’une exception ou d’une limitation au droit d’auteur et aux droits voisins.

La présente directive n’affecte en aucune façon les utilisations légitimes, telles que les utilisations relevant des exceptions ou limitations prévues par le droit de l’Union, et n’entraîne aucune identification d’utilisateurs individuels ni de traitement de données à caractère personnel, excepté conformément à la directive 2002/58/CE et au règlement (UE)2016/679.

Les fournisseurs de services de partage de contenus en ligne informent leurs utilisateurs, dans leurs conditions générales d’utilisation, qu’ils peuvent utiliser des œuvres et autres objets protégés dans le cadre des exceptions ou des limitations au droit d’auteur et aux droits voisins prévues par le droit de l’Union.

10. À compter du … [date d’entrée en vigueur de la présente directive], la Commission organise, en coopération avec les États membres, des dialogues entre parties intéressées afin d’examiner les meilleures pratiques pour la coopération entre les fournisseurs de services de partage de contenus en ligne et les titulaires de droits. Après consultation des fournisseurs de services de partage de contenus en ligne, des titulaires de droits, des organisations d’utilisateurs et des autres parties prenantes concernées, et compte tenu des
résultats des dialogues entre parties intéressées, la Commission émet des orientations sur l’application du présent article, en particulier en ce qui concerne la coopération visée au paragraphe 4. Lors de l’examen des meilleures pratiques, une attention particulière doit être accordée, entre autres, à la nécessité de maintenir un équilibre entre les droits fondamentaux et le recours aux exceptions et aux limitations. Aux fins des dialogues avec les parties intéressées, les organisations d’utilisateurs ont accès aux informations adéquates fournies par les fournisseurs de services de partage de contenus en ligne sur le fonctionnement de leurs pratiques en ce qui concerne le paragraphe 4.

Pour comprendre le dédale de ce texte et imaginer comment il pourra être mis en oeuvre, il faut se poser une série de questions:

• quels sont exactement les “fournisseurs de services de partage de contenus en ligne” visés? Car il y a plateforme et plateforme – la notion qui est passée dans le langage courant n’est du reste pas juridique. Ici il s’agit de plateformes de contenus partagés par les usagers.
• quand y a-t-il responsabilité directe de ces fournisseurs pour les contenus partagés?
• que faire pour éviter d’être tenu responsable?
• quels sont les actes des usagers qui sont spécifiquement et plus clairement autorisés par l’effet de l’article 17?
• quelles sont les conditions cumulatives pour l’exonération de responsabilité de ces fournisseurs?
• quelles sont les autres obligations pesant sur ces fournisseurs?

Au-delà de ces questions qui nécessitent une analyse du texte, il faudra plus généralement se demander:

• si l’article 17 répond à ce qu’on a appelé le “value gap” (voir aussi ici)?
• si l’article 17 écarte intégralement le régime exonératoire des intermédiaires en ligne défini aux articles 12 et suivants de la directive 2000/31 sur le commerce électronique?
• si l’article 17 peut être considéré comme dangereux pour la liberté d’expression, ce qui nécessite de comparer l’équilibre trouvé avec celui qui s’était établi à partir des anciens textes applicables, dont la directive 2001/29 et la directive 2000/31 ?
• si les campagnes de lobbying menées par les ayants droit, d’un côté, et les plateformes, de l’autre, ont gonflé certains aspects du débat au point de l’obscurcir (avec exemples à l’appui)?
• si un régime de responsabilité avec seuil, tel que celui prévu par l’article 17 pour les plateformes d’une certaine taille, se retrouve dans d’autres législations européennes relatives aux plateformes?
• comment s’assurer qu’un texte juridique puisse rééquilibrer le rapport de force sur le terrain?

Voilà quelques questions ouvertes qui nécessitent une recherche dans les textes et dans les têtes.

A la recherche d’un équilibre.

In a book soon to be published, Matthias C. Kettemann and Stephan Dreyer have invited 50 authors to

clear up misconceptions about the impact and reality of the Internet and explain fact-based, vividly and practically what science knows about Internet-based communication.

The book title summarises the objective quite well: Busted! The Truth About the 50 Most Common Internet Myths.

The task I was assigned was to bust the following myth: You can never overcome network effects. Here is my take on this.

Myth

Positive network effects arise when the value of a solution (product, service, platform) improves as more users adopt it. As early adoptions enhance future adoptions, a solution that manages to get ahead of the alternative ones will eventually—and irreversibly—dominate.

Busted

Network effects arise if users care about participation and usage decisions of other users when taking their own decision. If users belong to a unique group, network effects are ‘direct’, as more usage within the group directly affects each group member (as with communication devices). Network effects can also arise across users of distinct groups, as is the case on many digital platforms (Airbnb becomes more attractive for guests by having more hosts on board, and vice versa); here, network effects are indirect: an additional user affects the other users in her group not directly but via the increased participation in the other group.

In the presence of positive (direct or indirect) network effects, more usage enhances value, which triggers more usage, and so on. This self-reinforcing process is conducive to winner-takes-all situations (a unique solution eventually attracts most—if not all—users), followed by a form of lock-in (users are not willing to switch on their own because alternative solutions only become attractive in the unlikely event that all users switch together). Yet, a number of countervailing forces may curb the snowballing power of positive network effects. First, network effects are rarely positive all over: they may be restricted to small groups of users (such as preexisting friends on a social media), or negative within some group (like competing sellers on a trading platform), or become negative at some point (because the network infrastructure gets congested). Second, differentiation can play strongly against network effects: several solutions may coexist because they cater to the specific needs of different segments of users (like rival game consoles); also, a new and improved solution may displace a dominant one because it provides users with a sufficiently large value to overcome the lock-in (think of Facebook displacing Myspace).

So, even if winner-takes-all situations do exist (think of the current 95% market share of Google search engine in Europe), they cannot be explained solely by the presence of positive network effects: supply-side economies of scale or scope also contribute to their emergence, while anticompetitive conducts may play a part in their persistence.

Truth

Positive network effects generate self-reinforcing processes that may lead to winner-takes-all situations. Yet, there exist countervailing forces that makes it possible for solutions to coexist, and for once dominating solutions to be overcome by new, improved, ones.

Source

Paul Belleflamme and Martin Peitz (2018). Platforms and Network Effects. In Luis C. Corchon and Marco A. Marini (Eds). Handbook of Game Theory and Industrial Organization. Edward Elgar. Chelthenam, UK, Northampton, MA, USA. Working paper version available at SSRN.

Photo on Visual Hunt

This post, co-authored by Paul Belleflamme and Martin Peitz (whom IPdigIT is very happy to welcome in its team!) is an excerpt of a book, entitled ‘The Economics of Platforms‘, that the two authors are currently preparing for Cambridge University Press.

As already explained on this blog (se, e.g., here and here), network effects arise when the users of a solution (product, service, platform, …) care about participation and usage decisions of other users when taking their own decision. Network effects are positive if the value of the solution for each user increases the more users there are.

In the current economic and management literature, it is customary to distinguish between two categories of network effects, according to whether the users whose decisions influence one another belong to a unique group or to separate groups. If all users belong to a unique group, network effects are said to be ‘within-group‘: more usage within the group directly affects each group member; communication devices are the prototypical example. But there are also many environments in which users in one group mainly care about the participation of users belonging to another group. This is so on many digital platforms: BlaBlaCar’s value for drivers increases when more passengers use the platform, and vice versa; similarly,  authors on this blog enjoy a larger audience and we hope that you, readers, enjoy a larger set of authors. In these situations, network effects are said to be ‘cross-group‘.

The previous examples suggest that there is a clear dividing line between these two types of network effects. However, there exist many environments in which it may be tricky to determine whether the users of a platform belong to the same group or whether they should be split into separate groups. The objective of this post is to illustrate why it may be difficult to categorise network effects, and to discuss why it does—or does not—matter.

Who’s calling?

Telecommunication networks provide a nice illustration of the difficulty to draw a clear line between within- and cross-group network effects. Most of the economic literature on telecommunication networks assumes, for simplicity, uniform calling patterns, i.e., an equal likelihood for each subscriber to call and be called by any other subscriber.¹  This assumption of fully symmetric participants implies a single group exhibiting within-group network effects. Another simplifying assumption would be to consider that some people only make calls, while others only receive calls (e.g., restaurants and customers who want to order for delivery or make a reservation); in that case, there would be two distinct groups, with only cross-group network effects. The reality is naturally somewhere between these two extremes: subscribers are heterogeneous in their propensity both to make calls and to receive calls. Moreover, calling patterns are largely reported to be nonuniform: most subscribers have a ‘calling circle’, i.e., a subset of subscribers with whom they interact more frequently than with others.² Seen from an individual subscriber’s perspective, network effects are then mostly within-group (i.e., inside the calling circle), cross-group network effects (i.e., outside the calling circle) being relatively limited; yet, each subscriber makes a different distinction between the two types of network effects as calling circles differ.

Poker face … or poker faces?

Online games provide another environment for a platform with, at first sight, a single user group—in this case, the players—that is subject to network effects; yet, after a closer look, one observes a richer structure, featuring within-group and cross-group network effects, as illustrated by survey evidence of cross-group network effects in online poker.

Wimmer, Philander and Redona (2018) ran a survey among players of online poker in which they asked them to evaluate the presence of other players on the attractiveness of a poker site.³ Clearly, the more players are active on a site, the shorter the waiting time to a play. This suggests that there are positive within-group network effects. However, a closer look reveals that players are heterogeneous in their abilities. Poker, as all zero-sum games in which success depends on players’ abilities, has the property that winning probabilities increase in the ability of a player and decrease in the ability of other players. If we distinguish between low-skilled and high-skilled players, then every player finds a site more attractive if there are more low-skilled players. If the reduced waiting time from a larger pool is dominated by the reduced probability of winning, then every player finds a site less attractive if there are more high-skilled players. Thus, evaluated in a mixed pool, we would expect that there are positive within-group and cross-group network effects generated by low-skilled players and negative within-group and cross-group external effects generated by high-skilled players. In line with this hypothesis, the authors find that more participation by low-skilled players makes the poker site more attractive, while participation by high-skilled players does the opposite.

“Readers make writers and writers make readers”⁴

Since people can read and write on Wikipedia and no other users re involved, one may suspect that there is a single user group. However, some users are almost exclusively readers, whereas others may frequently update and add content. Thus, also in this case we can distinguish different groups, contributors and non-contributors. Natural experiments are an interesting way to gain insights into the presence and type of within-group and cross-group network effects. A natural experiment takes place if a system consisting of two or more groups that are linked through network effects is exposed to a shock that directly affects the user base of one or more groups.

Wikipedia is an example of successful private provision of a public good. While one may expect that Wikipedia suffers from the free-rider problem—that is, people are less willing to contribute the larger the number of users, Zhang and Zhu (2011) find the opposite using data on Chinese Wikipedia in 2005.⁵ The shutdown of Wikipedia in mainland China in October 2005 allows them to analyse the response by contributors from other places (e.g., Taiwan, Hong Kong, Singapore) to this shock in readership and number of contributors. Since the number of potential contributors has dropped, the free-rider problem suggests that non-blocked users are more likely to contribute. By contrast, Zhang and Zhu find that contribution levels of non-blocked users dropped by more than 40 percent with the shutdown. As they argue,

“contributors receive social benefits from their contributions, and the shrinking group size reduces these social benefits.” (p. 1601)

Distinguishing between contributors and non-contributors, this suggests that there are positive cross-group network effects from non-contributors to contributors. Clearly, we also expect positive cross-group network effects from contributors to non-contributors to be present, as the latter benefit from improved content on Wikipedia.

Union is strength

The previous examples show that many platforms that may be thought to cater to a single group of users are better described, upon closer examination, as serving multiple groups because of heterogeneous network effects. On the other hand, one should also note that if a platform facilitates the interaction between two (clearly distinct) groups of users, within-group and cross-group network effects are present and may be difficult to disentangle. Take the example of peer-to-peer marketplaces like Uber or Airbnb, which enable the interaction between providers and consumers of services; clearly, each group exerts positive cross-group network effects on the other group. Yet, the quality of the matching between peers from the two groups increases with the volume and reliability of data that the platforms collect from providers and consumers alike. Hence, also a form of within-group network effects appears: the larger the participation on one side, the more data is generated (about feedbacks, reputation, reviews, geo-localisation, etc.), which enhances the
quality of the platform’s service and, thereby, the utility of all users on the same side.

Does it matter?

In a nutshell, the distinction between within- and cross-group network effects may be hard to establish in some environments. This may not be a problem for a number of issues. For instance, if we are concerned with the evolution of competition among platforms, we understand that positive network effects–whether within or mutual across groups–generate a self-reinforcing process (big platforms tend to become bigger, while small platforms tend to become smaller). As previously explained on this blog, such self-reinforcing process may lead to winner-takes-all situations.

Yet, determining the exact nature of the network effects may prove much more crucial when it comes to design specific strategies. In the poker example, for instance, a platform that considers that network effects are of the within-group kind would deploy strategies to attract any player, irrespective of their skills. Yet, such strategy may lead to an excess of high-skilled players, which may eventually reduce the attractiveness of the platform for any player. From that point of view, it is key to identify more precisely the types of network effects that link users on the platform.

1 See, e.g., Armstrong, M. (1998). Network Interconnection in Telecommunications. Economic Journal 108, 545–564 / Laffont, J.-J., P. Rey, and J. Tirole (1998a). Network Competition: I. Overview and Nondiscriminatory Pricing. Rand Journal of Economics 29, 1–37 / Laffont, J.-J., P. Rey, and J. Tirole (1998b). Network Competition: II. Price Discrimination. Rand Journal of Economics 29, 38–56 / de Bijl, P. and M. Peitz (2002). Regulation and Entry into Telecommunications Markets. Cambridge: Cambridge University Press.
2 See, e.g., Hoernig, S., R. Inderst and T. Valletti (2014). Calling Circles: Network Competition with Nonuniform Calling Patterns, Rand Journal of Economics 45, 155–175.
3 Wimmer, B. S., Philander, K. S., and Redona, M. (2018). The Effects Network Externalities on Platform Value and Management: Evidence from Internet Poker Users. Available at SSRN.
4 Quote attributed to Carl McKever in Poetic Puberty: Developmental Stages of a Poet.
5 Zhang, X., and Zhu, F. (2011). Group Size and Incentives to Contribute: A Natural Experiment at Chinese Wikipedia, American Economic Review 101, 1601–1615.

Photo credits: 1. Visualhunt / 2. marcoverch on Visualhunt.com  CC BY / 3. @bastique on Visualhunt.com  CC BY-SA /

Le 19 février 2020, la Commission européenne dévoilait son Livre blanc sur l’intelligence artificielle (IA)[1], en même temps que sa « stratégie européenne pour les données »[2]. Les deux dossiers sont étroitement liés puisque l’apprentissage-machine sur lequel repose l’IA est alimenté par les données. Et les progrès de l’IA dépendent de l’essor d’une économie des données. Une consultation publique est ouverte jusqu’au 19 mai. Une proposition législative pourrait suivre fin 2020.

La stratégie européenne vise à faciliter l’accès à des données de qualité et à assurer le marché unique des données. Plusieurs axes sont envisagés: l’ouverture des données publiques selon l’axe G2B (du ‘gouvernement’ aux entreprises/business), le partage de données B2B (entre entreprises), mais aussi le B2G (des entreprises au profit des politiques publiques).

Sur le partage B2G, un groupe d’expert a déposé, le 19 février aussi, ses recommandations[3], par ex. créer des fonctions d’intendance des données ou tester des PPP (partenariats public-privé) sur l’échange de données.

Pour assurer l’excellence de l’IA, la Commission entend investir 20 milliards d’euros par an et mobiliser divers partenaires. Pour inspirer la confiance dans l’IA, elle propose d’établir un cadre juridique qui limite les risques pour les droits fondamentaux (vie privée et non-discrimination) et pour la sécurité.

Petite surprise : finalement la Commission n’envisage pas d’interdire l’usage des techniques de reconnaissance faciale. Pour la Commission, les applications d’IA à haut risque devraient être soumises à des obligations spécifiques (pour les données d’entraînement, leur conservation, la robustesse, etc.).

En matière de sécurité et de responsabilité[4], le cadre existant précède l’émergence des technologies comme l’IA, l’Internet des objets ou la robotique. Sa « neutralité technologique » assure son adaptabilité à de nouvelles avancées technologiques. Néanmoins les caractéristiques des nouveaux produits modifient la donne : leur connectivité peut compromettre la sécurité de manière indirecte (piratage d’un objet connecté pour enfants) ; l’autonomie des produits de l’IA peut générer des risques non prévus au départ, obligeant à revoir les conditions de sécurité ; les robots humanoïdes peuvent présenter des risques pour la santé mentale des personnes vulnérables. En outre, les règles sur la sécurité des produits ne traitent pas des risques liés à des données erronées ; or les outils intelligents sont dépendants des données.

S’agissant des logiciels, la directive 2006/42 relative aux machines y fait référence ; en revanche, il n’est pas clair que la directive 85/374 sur la responsabilité du fait des produits s’applique à tous les logiciels intégrés aux produits, ce qui justifie une clarification. Une autre suggestion est de faciliter la charge de la preuve de la défectuosité et de la faute du fabricant, en présumant celle-ci si des exigences de sécurité ne sont pas remplies. Face à l’opacité des systèmes d’IA (‘boîte noire’), d’autres mesures devraient faciliter la preuve et l’indemnisation des victimes. Les applications d’IA présentant un profil de risque élevé pourraient être soumises à une responsabilité stricte (couplée à une assurance obligatoire). C’est déjà le régime applicable aux véhicules, autonomes ou pas.

Bref beaucoup de chantiers pour la nouvelle Commission qui lie les mesures de transition numérique aux mesures nécessitées par le « green deal » annoncé en décembre dernier (COM(2019) 640 final).

Graphe de données – Extrait du blog de Martin Grandjean

[1] COM(2020) 65 final. Son sous-titre révèle deux axes: « Une approche européenne axée sur l’excellence et la confiance ».

[2] COM(2020) 66 final.

[3] https://ec.europa.eu/digital-single-market/en/news/experts-say-privately-held-data-available-european-union-should-be-used-better-and-more .

[4] Voir le rapport spécifique aussi publié le 19/2/2020 (Rapport sur les conséquences de l’IA, de l’internet des objets et de la robotique sur la sécurité et la responsabilité, COM(2020) 64 final).

In this series of posts, my objective is to explain the basic economic mechanisms that shape the competition among platforms. In this episode, I argue that positive network effects tend to lead to situations of market dominance: one platform ‘wins’ the market, leaving almost nothing to its competitors. In the next episode, I will show that there also exist counteracting forces that may allow several competing platforms to stay on the market. Finally, in the third episode, I will discuss how interoperability and multihoming are game-changers, insofar as they determine the form of competition that exists among platforms.

(This series of posts combines elements taken from a book, entitled ‘The Economics of Platforms‘, that Martin Peitz and myself are preparing for Cambridge University Press).

Platforms can be defined as entities that enable interactions among users and generate value from these interactions (see Belleflamme and Peitz, 2019). Network effects are the common driver behind the two roles that platforms play. First, because network effects make the users’s decisions interdependent, they raise coordination issues that users can hardly address by themselves; an intermediary–the platform–is then needed to bring users together and make the interaction possible. Second, network effects make the value of the interactions depend on the number of participating users (typically, with positive network effects, users value more platforms that attract a large number of other users); through the actions they take to attract users and encourage their participation, platforms can leverage network effects and create value for their users. Furthermore, network effects crucially influence the competition among platforms.

A ‘winner-takes-all’ tendency

In many situations, users can choose among several platforms to achieve the benefits of interaction: languages, communication channels, video game consoles, search engines, etc. Positive network effects make this choice quite peculiar insofar as they often induce users to coordinate on joining a single platform, at the expense of all the other platforms. A so-called ‘winner-takes-all’ outcome may then emerge, with the following two interesting properties. First, even though the platforms to choose from appear to be ex ante symmetric, the market is likely to end up in an asymmetric situation. Second, the benefit of coordinating on a single platform may lead some users to adopt actions that do not give them the highest stand-alone utility; that is, some users may decide to go against their personal tastes.

To see this, let us develop a very simple model with two platforms, named A and B, and two types of users, named R-types and S-types (see Arthur, 1989). Users arrive randomly on the market and have to choose between the two platforms. In a world without network effects, users would only compare the stand-alone utilities that the two platforms offer. If we assume that R-types perceive the stand-alone utility of platform A as larger, while S-types do for platform B, and if the two types of users are equally represented in the population, then the two platforms should have an equal market share in the long run. At each period, there is an equal chance that any platform will be chosen. It is like tossing a fair coin: the law of large numbers predicts an equal number of tails and heads in the long run.

Things change dramatically in the presence of positive direct network effects that are platform-specific (i.e., they are limited to users who adopt the same platform). If users care not only about stand-alone utilities but also about the possibility to interact with other users, their choice is affected by the current market shares of the two platforms when their turn comes to make a choice. The natural tendency to choose the platform of one’s taste (i.e., platform A for an R-type or platform B for an S-type) may then be overturned. To give some structure to the model, suppose that users place a value of 10 on having their preferred stand-alone utility and a value of 1 on interacting with any other user. Then, when an R-type arrives on the market, she will decide to adopt platform B if she observes that platform B counts at least 10 more users than platform A: the advantage of platform B in terms of network effects outweighs the disadvantage it faces for R-types in terms of stand-alone utility. From then on (see the figure below), all users will adopt platform B irrespective of their type: platform B will increase its market share and platform A will never catch up. It is as if the coin we flip repeatedly became increasingly unfair: the more, say, heads are drawn, the more the weight distribution of the coin becomes asymmetric so that the probability of heads increases. This kind of process leads eventually to a state in which only one result obtains. In our example, this means that a single platform will eventually take the whole market.

This argument shows that positive network effects, and the attraction loops that they generate, have consequential impacts on the competition between incompatible platforms, as they lead to winner-takes-all situations. Markets with platforms have thus the flavour of natural monopolies: only one firm can reasonably survive while making a profit. In general, natural monopolies are due to supply-side economies of scale: because of large set-up cost (e.g., building the infrastructure to distribute the product), it is more efficient to have one large firms producing large quantities. Think, for instance, at the electricity or transport sectors. In contrast, in markets with platforms, increasing returns to scale are observed on the demand side: when a platform attracts more users (i.e., “produces more”), it is not the average cost per user that decreases but it is the average revenue per user that increases (because users are willing to pay more to be on a larger platform).

Arthur (1989) also uses his model to describe four common features of the competition between incompatible platforms:

1. Path-dependence: the final outcome depends on the way in which adoptions build up (i.e., on the path the process takes);
2. Inflexibility, or lock-in: the left-behind platform would need to bridge a widening gap if it is chosen by adopters at all;
3. Non predictability: the process locks in to the dominance of one of the two platforms, but which platform is not predictable in advance;
4. Potential inefficiency: the platform that “takes the market” needs not be the one with the longer-term higher payoff.

The following video illustrates how the current dominance of the QWERTY keyboard fits this description:

In the next post of this series, I describe several forces that may counteract the feedback loops induced by positive network effects.

Photo credits: Photo on Visualhunt.com

In this series, I explain the basic economic mechanisms that shape the competition among platforms. The first episode showed that positive network effects create feedback loops, which drive markets with platforms towards ‘winner-takes-all’ situations. This episode somehow back-pedals by identifying a number of forces that mitigate the impacts of network effects.

A few winners take most

Although positive network effects drive users to agglomerate on a single platform, there exist other forces that can lead users to prefer different platforms. There can also be situations that allow competing platforms to coexist on the same market. I review these forces and situations here.

Differentiation

In the model described in the first post of this series, all users (R-types and S-types alike) end up joining the same platform because the difference in network benefits eventually outweighs the difference in stand-alone benefits. This situation is more likely to arise when users have relatively similar preferences and when they visit a platform mostly to interact with other users. In contrast, if users differ in their preferences and when some of them care more about the stand-alone services that one particular platform offers, then chances are higher that two or more platforms will manage to stay profitably in the market. A smaller platform may compensate the lower network benefits that it generates by offering services that are sufficiently differentiated from those of its competitors. Differentiation can be horizontal (i.e., services appeal to a particular segment of users) or vertical (i.e., the quality of services is perceived as higher). It could also be that users are only interested in meeting or interacting with other users who are physically close to them; different platforms can then start their operations by focusing on different local markets, while trying to compete more globally at a later stage.

The sector of online dating nicely illustrates that platforms may coexist in a market. In 2015, the German competition authority accepted the merger between two major online dating platforms, arguing that this merger would not significantly impede competition (see here):

Multihoming

Multihoming refers to the possibility for users to visit more than one platform at a time, as opposed to singlehoming that describes situations in which users stick to a single platform. There are many situations in which users prefer to singlehome although reasons for this can be diverse.

• A first obvious reason is cost. When joining a certain type of platform is considered as expensive, users tend first to compare the options available in the market and then to select the one that best fits their needs. For instance, this is what occurs with phone, TV and Internet bundles. People will compare various packages from different providers and opt for only one solution. We can reasonably think that people at home have no more than one television set-top box and only one subscription to the Internet, probably purchased from the same provider. Similarly, most people do not own multiple smartphones.
• A second reason is convenience. Even if registering to several platforms would not represent an added expense, the fact of being an active member on several platforms simply does not make sense. In other words, the user here does not perceive any added value in joining another platform while remaining active on the original one.

Yet, users can also decide to multihome. Social networks represent the perfect example. People that are keen to share personal stories, professional news, pictures of family and friends, etc. on the Internet make use of social networks. And people today own several accounts, selecting what sounds to be the most appropriate channel given the nature of the content to be shared or the audience they target. This is confirmed by the GlobalWebIndex survey (see here): “social media users are comfortable in maintaining a presence across more than one platform; while digital consumers had an average of around 4 social accounts back in 2013, this figure has now [2017] doubled to around 8.”

When users can multihome, it is easier for several platforms to coexist. That is, multihoming mitigates the tendency for network effects to lead to winner-takes-all situations. Jullien and Sand-Zantman (2019, pp. 11-12) explain why:

Interoperability

Interoperability (or compatibility) is another element that mitigates the agglomeration of users on a single platform. Two platforms are interoperable if the users of one are able to interact with the users of the other. Note that interoperability is often a matter of degree. Think, for example, of using your smartphone for its original function, that is to phone someone. To make a video call through Facetime, users must both have an iPhone because this application, developed by Apple, is not available to users of devices running on other OSs than iOS; in this case, there is no interoperability. Most `Voice over IP’ applications (e.g., Skype, Whatsapp or Viber) are available for all major OSs; yet, the performance of some apps may vary across OSs, which may cause problems (such as synchronization issues) when calls are made between different devices, meaning that interoperability is imperfect. Lastly, if you use the regular phone lines, you do not need to worry about which operating system (OS) your correspondant has on her smartphone: there is perfect interoperability.

Interoperability plays a similar role as multihoming. When platforms are interoperable, the implications of choosing a particular platform become less dramatic, as it remains possible to interact with users of the other platform. The same goes with multihoming agents. However, the main difference between interoperability and multihoming is that it is easier for platforms to prevent interoperability than multihoming. One needs thus to examine the platforms’ incentives to be interoperable with one another.

Interoperability has two major implications for the competition between platforms.

1. Interoperability increases the consumers’ willingness to pay, as consumers prefer to join a ‘common’ large network rather than having to choose between two potentially smaller networks. This effect tends to benefit all platforms, since it allows them, other things being equal, to set higher fees or to increase participation.

2. Interoperability decreases the quality difference between the platforms, as consumers’ adoption decisions no longer depend on the platforms’ relative network sizes. Whether this second effect is positive or negative for a platform depends on the size of its network under incompatibility: the level playing field that interoperability creates tends to make platforms with smaller networks better off and platforms with larger networks worse off.

Negative same-side network effects

In many economic environments, the attractiveness of a platform for the members of one group also depends on the participation of the members of the very same group. That is, there exist same-side network effects, which platforms have to take into account when choosing their strategies. Same-side network effects appear when the members of one group compete with one another to interact with the other group. For instance, given a set of buyers on the platform, the expected profits of sellers on eBay decrease in response to the entry of competing sellers. Similarly, if an additional competing shop opens in a shopping mall, the expected profits of existing shops decrease given a set of buyers in the mall. Another example is dating apps. These are characterized by positive cross-side network effects, as the app becomes more attractive the more it is used by people of the opposite gender. However, they are also characterized by negative same-side network effects, as more people of the same gender make it less likely that a match materializes for a particular person. Different from a buyer-seller context, there is typically no monetary transaction between users on a dating app.

Negative within-group effects may also arise because of congestion problems—for instance, sellers may compete for buyer attention, which is scarce. Consumption externalities are another instance of negative same-side network effects. For instance, referring to Airbnb visitors, Slee (2016) reports that “as their numbers grow, they erode the very atmosphere in which they bask and threaten the livability of the city for residents.” This arguably also applied to fellow visitors. Congestion effects may also be present on digital platforms with limited bandwidth such that, e.g., the delivery of research result is slowed down—this would imply that the stand-alone utility of the platform suffers from a lot of buyer participation. In the offline world, congestion problems appear when the platform’s physical venue is too crowded; for instance, shoppers may get stuck in a crowded shopping mall and, as a result, make fewer purchase attempts.

How do negative same-side network effects influence the competition among platforms? Belleflamme and Toulemonde (2009) examine the extent to which negative same-side network effects among sellers may help a new platform operator lure buyers and sellers away from an existing marketplace. In their model, the new platform faces a ‘chicken-and-egg’ problem: to attract sellers, it needs to attract buyers, but to attract buyers, it needs to attract sellers. One pricing strategy that addresses this problem is called the ‘divide-and-conquer’ strategy. With this strategy, the platform subsidizes the participation of one side (divide) and hopes to recoup the loss through the membership fee it sets on the other side (conquer). The question is whether the platform can make any profit with such strategy. The answer is ‘yes’ when the interaction among buyers and sellers only generates (positive) cross-side network effects. However, the presence of negative same-side network effects among sellers (e.g., because they offer substitutable products) blurs the picture. Competition among sellers turns out to be a mixed blessing for the new platform.

• The upside is that the sellers’ willingness to pay to join the new platform increases if only a few of them make the move; as a consequence, sellers are less sensitive to buyers’ participation to the new platform, which alleviates the ‘chicken-and-egg’ problem.
• Yet, the downside is that it will be more costly for the new platform to attract buyers if only a small subset of the sellers join.

The balance between the two effects depends on the relative strength of the same-side network effects (with respect to the cross-side network effects). There may be situations in which entry is not profitable.

Belleflamme and Peitz (2019) examine other ways by which seller competition influences the number of platforms carrying positive volume of trade.

Summary

The following table summarizes the forces that lead either to concentration or to fragmentation in markets with platforms.

In the next episode of this series, I will discuss why interoperability and multihoming are critical game-changers in markets in which network effects are at play.

In the first two episodes of this series, I examined what brings markets with platforms to be dominated by a single or by a few winners. I conclude here by discussing the critical influence of interoperability and multihoming on the competition that takes place in markets with platforms.

Competing ‘for’ or ‘in’ the market

As we learned in the previous post of this series, interoperability is a critical game-changer in network markets. Basically, when platforms agree to be interoperable with one another, it is as though they were agreeing to share the market, as network sizes stop conferring any sort of competitive advantage; they also accept somehow to level the playing field. In contrast, if platforms remain incompatible, their objective is to make network effects tip the scales in their favour, i.e., to grow their network at the expense of their competitors; eventually, one platform may succeed in capturing the whole market. Hence, the decision about interoperability is often a decision about how to compete: either compete ‘in’ the market if interoperability prevails, or compete ‘for’ the market otherwise.

Interoperability choices also determine if, when, and under which form standardization takes place on the market. If firms agree to make their platforms interoperable, one talks of ex ante (or ‘de jure’) standardization. Typically, this form of standardization follows from negotiations among firms, which often take place within Standard-Setting Organizations (SSOs). Otherwise, if platforms decide to remain incompatible, the combined adoption decisions of users will ultimately determine which platform gains widespread acceptance. Here, one talks of ex post (or ‘de facto’) standardization if one platform eventually dominates. Note that if incompatibility prevails, several platforms may well coexist in the long run if, e.g., platforms are sufficiently differentiated and/or network effects are not too strong.

Given the far-reaching consequences of interoperability decisions, it is important to understand how competing platforms strategically make them. Compatibility can indeed hardly be achieved in a unilateral way, as platforms can use technical measures or invoke their intellectual property rights to keep their platform incompatible.

The following case illustrates some of the strategic considerations that shape decisions about interoperability.

The battle of voice assistants. The most popular voice assistants are currently Apple’s Siri, Google’s Assistant, Amazon’s Alexa and Microsoft’s Cortana. While Apple and Google have made their voice assistant incompatible with any other system, Amazon and Microsoft have been working together to let their systems communicate with each other, so that users can access the services related to Alexa by summoning Cortana, and vice versa. What is the motivation behind this unusual collaboration between two rival companies?Wingfield (2017) explains that the CEOs of the two companies “are concerned that keeping assistants from working together could hold them back. The way they see it, each assistant has unique strengths that could benefit the other assistants.” In other words, as users should get more of the two voice assistants if they interoperate, users should also have a higher willingness to pay for these assistants. Pushing this logic one step further, the two CEOs also declared that they would welcome Apple and Google if they decided to join in the effort. However, both Apple and Google have probably more to lose than to win by opening their ecosystem. Both companies are indeed keen to use their proprietary voice assistant as a selling point of their smartphone operating system (iOS for Apple and Android for Google). By contrast, explains the same article, “Alexa is mostly used on Echo speakers that sit around a home, and Cortana is largely used on PCs. (…) The two companies have struggled in the smartphone business, which makes it hard to get people using Alexa and Cortana outside homes and offices.”

The impacts of multihoming on competition among platforms

(This section borrows from Belleflamme and Peitz, 2020.)

In many markets with platforms, users differ in their ability to be active on several platforms at the same time. As discussed in the previous post, this ability to multihome depends on which side a user belongs to. Often, one side is able to multihome, while the other is restricted to singlehome. To fix ideas, let us identify the two sides as “buyers” and seller” and let us assume that sellers are able to multihome while buyers are not. This situation is identified in the economic literature as “competitive bottlenecks.” The analogy with bottlenecks comes from the fact that a seller who wants to interact with a particular buyer has no other way than using the platform on which this singlehoming buyer is active. In the market environment with competitive bottlenecks, platforms compete for buyers (who are restricted to singlehome) but do not compete for sellers (who have the freedom to multihome). If sellers are restricted to singlehome as well, then platforms compete on both sides of the market. Then, comparing the latter environment (called “two-sided singlehoming”) with the former gives insights about the impacts that multihoming can have on the competition among platforms.

As platforms compete for both buyers and sellers under two-sided singlehoming, but only for buyers under competitive bottlenecks, one may be tempted to conclude that, when moving from two-sided singlehoming to competitive bottlenecks, sellers face higher prices and obtain a lower surplus, while buyers face lower prices and obtain a higher surplus. Also, one may expect platforms to achieve higher profits due to the reduced competition on the seller side. This is indeed the view expressed in a number of reports. For instance, the German Cartel Office wrote the following:

“In [the competitive bottleneck] scenario, the platforms were competing for users on the single-homing side. Accordingly, on the multi-homing side, platforms provided monopolistic access to single-homing users who were members of the platform. Regarding the framework of the model reviewed, this led to a monopolistic price on the multi-homing side, while the price on the single-homing side would be fairly low as a result of platforms competing for users on this side. In this respect, this may result in an inefficient price structure despite potentially intensive platform competition (on the single-homing side).”

Yet, Belleflamme and Peitz (2019) show that the effects of forcing one side to singlehome instead of letting it multihome (or the other way around) are less straightforward than what may be perceived in general. While it is true that platforms exert monopoly power over the multihoming side, users on this side may actually benefit from multihoming. In addition, platforms may find the two-sided singlehoming environment more profitable than the competitive bottlenecks.

The key intuition behind these results is that sellers (the users who could multihome) may pay a low price to start with in the competitive bottleneck case. This seems counterintuitive as we expect a monopolist (as each platform is vis-à-vis sellers in the competitive bottleneck case) to set higher prices than do competing duopolists (as platforms are vis-à-vis sellers under two-sided singlehoming). However, the economic literature has explained why the reverse may happen. Two economic effects are at work: the market share effect drives duopolists to set a price below the monopoly price because, at this price, they sell to fewer consumers than the monopolist; the price sensitivity effect works in the opposite direction, as it incentivizes duopolists to raise price above the monopoly level because they face a steeper demand curve. In the present setting, the latter effect outweighs the former in the absence of cross-side networks effects.

In sum, the move from two-sided singlehoming to competitive bottlenecks induces two contrasting effects on the seller fee. First, as just explained, letting sellers multihome drives platforms to reduce the price they charge to sellers in the absence of network effects. But, on the other hand, when buyers derive benefits from interacting with sellers, platforms have an incentive to raise the sellers’ fee when sellers can multihome. The reason is that an additional multihoming seller is less valuable regarding competition on the buyer side than an additional singlehoming seller, as the latter—but not the former—is attracted at the expense of the competing platform.

Three important lessons can be drawn from this analysis:

1. It is well possible that buyers, sellers and platforms are all better off when sellers are allowed to multihome.
2. Whenever platforms benefit from imposing exclusivity, doing so may benefit or hurt sellers, but definitely hurts buyers; that is, the situation may arise that the side which is subject to a contractual restriction benefits from this restriction, but that the other side suffers. As a result, in an environment with potential seller multihoming, an agency should prohibit the use of exclusivity on the seller side if its aim is to maximize buyer surplus.
3. Whenever buyers suffer from seller multihoming, platforms and sellers benefit from it.

To conclude

In this series of post, I have explained the basic economic mechanisms that shape the competition in markets with digital platforms. The presence of positive network effects naturally leads to concentration in these markets: there is only room for a limited number of firms, because users benefit from coordinating their choices. So, the stronger the positive network effects, the smaller the number of platforms that can profitably stay on the market. It may even be that only a single platform can make a sustainable profit– a ‘winner-takes-all’ situation.

Yet, a few firms may share the market when positive network effects are weaker and/or in the presence of negative network effects, either across groups of users (as in the case of ad-financed media) or within a particular group (for instance, when sellers compete with one another on a platform). Also, it is more likely to have several winners rather than just one if users can connect to more than one platform at a time. This could be because platforms are interoperable or because users can multihome. Interoperability and multihoming appear thus as game-changers, as they condition the type of competition that exists on markets with platforms. Yet, platforms can control these two dimensions (especially interoperability) and it appears that their preferences diverge as to whether multihoming and interoperability should be implemented or not. In general, large platforms prefer incompatibility, as it gives them the chance to become the winner that takes it all. In contrast, smaller platforms fear that they would loose if they compete ‘for the market’; they prefer thus situations of interoperability and multihoming that make platform compete ‘in the market’.

Imagine you moved from Denmark to Malaysia and you want to subscribe to Spotify Premium from your new home. Being new to the country, you do the math of converting the currency from Malaysian Ringgits back to Danish Crowns to understand what you will have to pay. Getting the math right, the $18.42 you were used to as a monthly fee from home differ from the $4.61 charged in Malaysia (Singers, 2014). How can that happen?

Matias Singers who moved from Denmark to Malaysia had a similar experience. Noticing a large price difference inspired him to create a so called “Spotify International Pricing Index” in honour of the famous “Big Mac Index”. Indeed, the price for a subscription to Spotify Premium varies substantially across the world. In 2014, it ranged from $2.93 on the Philippines to $18.42 in Denmark (Singers, 2014). Considering this large price spread, it should not come as a surprise that many consumers noticed it and that some consumers found it “unfair” (as some comments on Spotify’s online forum suggest).

In this post, we want to understand the economic rationale behind Spotify’s pricing policy. After a brief description of Spotify’s cost side, we explain how Spotify generates revenues by using various differential pricing strategies. Finally, we examine how these pricing strategies affect Spotify users.

Spotify as the Robin Hood of the music industry?

Spotify was launched in Sweden in 2008 as a music streaming service amid a difficult time for the music industry. Since the turn of the millennial, digitization and, allegedly, music piracy lead to declining sales in the music industry for over a decade (The Economist, 2016). Initially, also Spotify was accused of hurting creatives in the music industry, as part of its business strategy is to propose “free” subscriptions (see below). However, in contrast with sites offering pirated material, Spotify paid—and still pays—royalties to the music labels, which in turn pass some of the proceeds on to the musicians. According to David Israelite of the National Music Publishers Association each stream earns a fraction of a cent for the artist (The Economist, 2019). But adding earnings up over many streams by 271 million active users each month gives a beefy number. Since its launch back in 2008, Spotify has paid out over 15 billion Euros of royalties (Spotify, 2020); many industry experts claim that Spotify actually helped the music industry survive (Shaw, 2018).

How does Spotify earn the money that it, at least partly, passes on? To optimize its revenues, Spotify charges different prices to different users for its services. This practice—known as differential pricing—requires, on the part of the seller, some knowledge of what different groups of consumers are willing to pay for the service. The idea, of course, is to charge a larger price to those users who are willing to pay more.

The question, then, is how well does Spotify know its potential consumers? The term ‘potential’ is important here. For sure, Spotify knows quite well its actual consumers. The company can indeed monitor the activity of its users in a very detailed way. By collecting and processing these vast amounts of data, Spotify is able to drive its users to listen to more music. As Gershgorn (2019) puts it, “[t]he company has created algorithms to govern everything from your personal best home screen to curated playlists like Discover Weekly, and continues to experiment with new ways to understand music, and why people listen to one song or genre over another.” (In a previous post on this blog, we explain how reviews, ratings, and recommendations are central for digital platforms.)

Yet, Spotify knows much less about the consumers and their willingness to pay before they join the platform. To remedy this problem, the company employs two strategies. First, it uses a freemium model—a specific form of ‘versioning’—to induce consumers to reveal if they are willing to pay any money at all to use the platform. Second, it segments the consumers who are willing to pay a positive price into distinct groups and charges different prices to each of these groups; the country-specific prices described above correspond to this practice of ‘group pricing’. We now describe these two strategies.

A Freemium offer as a self-selection device

In most countries, Spotify offers two basic plans for consumers to choose from. Penny pinchers can go for an advertising-based free subscription, called “Spotify free”. This version, however, comes not only with the annoyance of getting ads in between the songs, but also with limited app functionality and limited audio quality. Users can obtain higher quality by signing up to a fee-based “Spotify premium” account.

This freemium strategy (freemium being a portmanteau combining ‘free’ and ‘premium’) is a form of versioning, as consumers essentially choose from different versions of a streaming experience offered at different prices. For a given set of content, the royalty rate does not differ according to whether streaming is part of Spotify Free or Spotify Premium. While premium users are likely to engage in more streaming, differences in per-user costs are unlikely to fully explain the differences in revenues per user. Instead, the different versions are offered as self-selection devices such that users interested in high quality streaming opt for the premium version, whereas occasional and less quality-sensitive users take the basic version at zero monetary cost.

Importantly, Spotify only needs little information about its potential consumers to implement this strategy. All it needs to figure out is the dimensions of the service that some consumers are willing to pay more for than others. The next steps consist in designing different versions along these dimensions and letting consumers pick the version that best suits their taste, thereby revealing their preferences. Music lovers are unlikely to accept their favourite album being interrupted by a blunt ad for groceries, and thus choose a premium subscription. Spotify additionally “damages” the free version by withholding some features to push consumers into the Premium version.

To summarize, Spotify induces subscribers to self-select by how much they are willing to pay for the streaming service. In determining its optimal pricing policy, Spotify may also take into account network effects: Users of the “free version” are not only valuable because of the advertising bucks they generate, but possibly also because of additional data they generate. To the extent that these data allow Spotify to make better recommendations to all users (and, in particular, the subscribers of the Premium version), this gives Spotify an additional incentive to sign up many users on the free version.

Demographics-based group pricing

For the previous tactic to work fine (that is, to make sure that high-end consumers do not content themselves with the free version), the price of the premium version must be carefully chosen. This is where group pricing comes in. By relying on easily observable (and verifiable) characteristics that are correlated to the willingness-to-pay, Spotify can segment consumers into distinct groups and charge a higher price to those groups who are willing to pay more for music. Among these characteristics, one finds the user’s country of residence, the age, and the family status.

As noted above, the fees for premium accounts vary significantly from country to country. These differences are evidence of group pricing if the prices are adapted to different demand conditions and not just to different costs conditions. Is this so? Let us look at potential sources of differences in marginal costs across countries. First, marginal cost differences can arise owing to different royalty rates across space and artists. As Ovide (2018) reports, royalties per stream depend on the country of the recipient and the artist. Second, the content that is available with a subscription varies somewhat from country to country and, thus, can lead to different average royalty rates. Third, the revenue contribution of a subscriber of the free version may differ across countries (the ad revenue per user is likely to be higher in high-income than in low-income countries); the lost revenue from converting a consumer using the “free” version to the premium version differs thus, which affects the optimal premium price. However, following Waldfogel (2020), it is reasonable to assume that these differences in marginal costs remain relatively small and hence, cannot explain fully the differences in Spotify’s prices.

Thus, group pricing it is and the rationale behind this practice is intuitive. If it were selling contracts to all consumers around the globe at the same price, Spotify would either lose many consumers in markets with many price-sensitive consumers when charging too high a price or offer its library too cheaply in higher-income countries.

By the same logic, Spotify charges a lower price to students and to families (by letting related users access the service through a unique account). The student discount is justified by the larger price-sensitivity of students. As for the family plan, it follows the logic described above: Spotify is willing to use lower tariffs to increase traffic on the platform, as traffic rhymes with data and data are precious to improve the service.

Beware of arbitrage

Arbitrage is the archenemy of differential pricing. It can be of two forms. First, personal arbitrage occurs when consumers pretend to be somebody else to benefit from a lower price than the one they would have to pay otherwise. As far as versioning is concerned, personal arbitrage is taken care of by designing the price structure correctly. In particular, if the premium version is too expensive, the majority of consumers will stay with the free version, implying that the self-selection mechanism does not wok properly. To encourage consumers to switch to the Premium version, Spotify started, in May 2020, to offer a free trial period, a marketing practice common to sell subscription-based services.

In the case of group pricing, personal arbitrage can occur if users pretend to belong to groups benefiting of lower prices. For instance, consumers in more expensive countries could in theory use a VPN to sneak around the geographic price discrimination. Yet, as Snow (2019) reports, most people find it too cumbersome to do so or are not even aware of that possibility. Consumers could also pretend to be students, but this form of misreporting is limited by the fact that only people subscribing from an educational email address are eligible to student discounts.

Finally, users may also be tempted to abuse the family subscription option; to prevent this, Spotify updated its terms and conditions in August 2019 by announcing that it could ask members of family plans to prove that they are all living at the same address.

Who wins and who loses?

At the end of the day, some consumers are likely to benefit from Spotify’s group pricing and versioning, while others are likely to be better off without such differential pricing strategies. In a recent study, Joel Waldfogel finds that overall geographic price discrimination benefits Spotify’s subscribers in low-income countries while harming those in higher-income countries (Waldfogel, 2020).

The overall benefit to Spotify is currently hard to assess. While Spotify managed to grow an impressive user base and brand name over the last years, it is still losing money. Taking a long-term perspective, attracting more subscribers through low prices may well be profit-maximizing; the market valuation of Spotify suggests that many investors believe this to be true. Spotify may also further monetize on the vast amounts of user data it is collecting, for instance, by targeting advertising with its free offer. It also charges artists for promoting their music through playlists. Thus, playing with the consumers is just part of Spotify’s larger game.

This, in the end, might not just allow Spotify to be profitable, but also change the very music it plays: already now, the rules of streaming made the intros of songs become shorter than in the old days of “Joshua Tree” (The Economist, 2019). Thus, it is not just about whether the price for music streaming is right, but that with the advance of music streaming, the rules of the game in the music industry are changing.

Dans son Numéro 133 (Eté 2020), la revue En Question (éditée par le Centre Avec; voir aussi ici) publie un dossier intitulé “Peut-on encore partir en vacances?“, dossier brûlant s’il en est dans la période de pandémie que nous traversons.

Vous trouverez, via ce lien, un tiré-à-part de l’article que nous avons écrit pour ce numéro. Dans cet article, nous examinons les impacts qu’ont les plateformes de l’économie collaborative sur l’environnement économique et social dans lequel elles évoluent. Nous nous concentrons sur le secteur du tourisme et de l’hébergement. Nous commençons par expliquer comment ce secteur a été profondément affecté par l’émergence récente de nombreuses plateformes, la plus célèbre et prospère d’entre elles étant Airbnb. Nous résumons ensuite les résultats d’études qui ont estimé les impacts que ces plateformes ont, plus largement, sur les destinations touristiques et leurs habitants. Enfin, nous évaluons comment la crise sanitaire actuelle est susceptible de remettre cette analyse en cause.

N’hésitez pas à découvrir les autres articles de ce numéro très intéressant!

The post Airbnb: A qui profite l’économie collaborative? appeared first on IPdigIT.

The so-called sharing (or collaborative) economy is generally presented as comprising activities that involve sharing under-used resources, in that the owners of these resources (the providers) make them available to other individuals (the consumers).¹ Although this definition remains vague (there are many nuances in the ways in which resources are shared and the degree to which they are used), observers agree that activities in the collaborative economy share three important characteristics:

• Digital platforms are essential to the large-scale development of these activities. By leveraging digital technologies and data analysis techniques, these platforms reduce transaction costs and make the interaction between suppliers and consumers viable. Unlike traditional companies (also known as ‘pipelines’) that control transactions by producing goods and services with their own means of production, platforms facilitate transactions by linking consumers to independent service providers.
• Because sharing economy activities are decentralised (and sometimes informal), their organisation requires innovative governance mechanisms that use digital tools, often data intensive, to replace traditional economic interactions (e.g., rating systems to build trust between individuals who cannot interact face-to-face and/or repeatedly).
• As the sharing economy expands in various sectors of society, it is increasingly perceived as disruptive, as it provides an alternative offer in many industries, raises conflicts and tensions, and exposes many stakeholders to new types of risks.

It is this third characteristic that we develop in this article. In particular, we examine the impacts that sharing economy platforms have on the economic and social environment in which they operate. We focus on the tourism and accommodation sector. We begin by explaining how this sector has been profoundly affected by the recent emergence of many platforms, the most famous and successful of which being Airbnb. We then summarise the results of studies that have estimated the impacts that these platforms have, more broadly, on tourist destinations and their inhabitants. Finally, we assess how the current health crisis is likely to call this analysis into question.

Impacts of platforms on the tourism sector

Until recently, the travel and accommodation industry was almost entirely organised in a linear fashion, in the sense that services were provided by companies with their own means of production (means of transport, accommodation facilities) and hiring their own staff (drivers, hoteliers). An intermediate sector complemented the industry, consisting of travel agencies that booked and distributed seats and rooms, and helped travellers find their way through multiple alternatives.

Today, while most of these intermediaries have gone online, a whole new breed of intermediaries has emerged, offering travellers services similar to those of linear businesses, but in a very different way, which could be described as ‘circular’. The most prominent examples of these new intermediaries are Airbnb in the short-term accommodation sector and Uber in the point-to-point transport sector. Both companies share the characteristic of not having the means to provide their service (no accommodation infrastructure for Airbnb, no car fleet for Uber). Instead, they operate as platforms: they do not control the production of a service, but they make it possible by facilitating the interaction between those who want to consume a service (travellers) and those who can perform it (and who act mainly as independent contractors).

In general, platforms such as Airbnb and Uber can create value by connecting users when they are not able to organise on their own an interaction that would be beneficial to them. Behind these two conditions lies a common driver called ‘network effects‘. Network effects mean that the more participants there are in the interaction, the more beneficial the interaction is for each participant. Network effects therefore increase the value of the interaction but also make it more difficult to organise. Indeed, when potential participants decide whether or not to interact, they generally do not consider the effects of their decisions on other agents. It is then likely that even if all agents find the interaction useful if it were to take place, none of them is sufficiently motivated to initiate the process. This is precisely where platforms can make a difference. By reducing a variety of transaction costs, platforms coordinate the decisions of participants; they make them perceive the value they generate for each other, thereby increasing their motivation to participate. In summary, platforms make a difference by managing network effects.

The huge advantage of network effects for platforms that manage them is that they generate virtuous circles. In the case of Airbnb, travellers value the platform all the more as the number of hosts is large, and vice versa. Therefore, if the platform succeeds in attracting more guests, it will attract more hosts who, in turn, will attract more guests, and so on.² This explains why platforms that succeed in becoming established have a natural tendency to grow rapidly. Founded in 2008, Airbnb has quickly become a key player in the tourism sector. In some markets, the platform is even considered the standard for accommodation demand, especially among young people, technophiles and price-sensitive travellers. Even if the start was rather slow, the development quickly accelerated from 2011 to reach a base of 150 million users and more than 7 million listed accommodations in 65,000 cities worldwide. In terms of capacity, Airbnb has surpassed the five largest hotel chains in the world. In view of its upcoming IPO (originally planned for 2019 but postponed due to the health crisis), the value of the firm has been estimated at $42 billion (compared to $31 billion in 2017).

Even if Airbnb has contributed, through its innovative offer, to develop the short-term accommodation market, its current market shares have been acquired, to a large extent, to the detriment of incumbent companies. Several studies have tried to estimate the effects of the increase in Airbnb’s supply of rental accommodation on the turnover per room and/or sales volume in the hotel industry. These studies show that Airbnb’s impacts differ according to the segments of the industry: Airbnb’s offer competes very seriously with budget hotels and motels, not at all with high-end hotels, and seems to compete on a level playing field with mid-range hotels. Thus, depending on the composition of the sample used by the researchers, the conclusion is that the increase in Airbnb’s offer reduces the turnover of the hotels (in the case of low-end hotels), does not affect it at all (high-end hotels) or has ambiguous effects (mid-range hotels). The industry professionals who were interviewed confirm these results from quantitative analyses; according to them, the major hotel chains do not consider Airbnb to be an immediate threat.

Impacts of the platforms on tourist destinations

The growing success of Airbnb (and other similar platforms) has generated a series of socio-economic impacts that go far beyond the tourist accommodation industry alone. Not only are sectors such as catering, culture or entertainment (which are complementary to the tourism sector) affected, but also, more fundamentally, the living conditions of the inhabitants of the destinations popular with Airbnb users.

Over the last five or six years, many researchers have taken up these issues. The majority of studies have relied on quantitative data (there is relatively little theoretical work or work using qualitative data). The basic data come from the Airbnb website; they are either collected by the researchers themselves (through specially written programs; see here) or acquired from specialised firms. These data are then compared with other indicators concerning, for example, the performance of hotels or the demographic, economic and social characteristics of tourist destinations. The aim of these studies is to establish causal relationships–or at least correlations–between Airbnb’s growth (represented by the increase in the number of ads on the site, for example) and various socio-economic aspects.

Sometimes the answers given to the same research question diverge. This is not surprising given that the phenomenon being studied is recent and that the approaches and databases used are varied. Nevertheless, recent work seems to converge on a few stylised facts that we summarise here.

• Concerning local economies, in general, there are positive impacts of Airbnb on the catering, tourism and entertainment sectors. However, the effects are not homogeneous. One has to distinguish between the heart of the cities (or tourist attractions) and the periphery. Since a large proportion of the properties listed on Airbnb are located in areas where the hotel offer is less developed (e.g., in the suburbs), the development of Airbnb helps to redistribute the wealth created by tourist activity in a more egalitarian way within the cities.
• Airbnb’s impacts on cities are not always that positive, however. The most problematic consequence of Airbnb’s growth (and one that is attracting the attention of municipalities and governments) is the scarcity of long-term rental properties. This has a crowding-out effect: for reasons of profitability, landlords turn more to renting for short-term stays, made possible by Airbnb, than to renting for long-term occupancy. Studies carried out in France estimate that it is sufficient to rent a property for ten days a month on Airbnb to reach the monthly income of a traditional long-term rental. The development of Airbnb is therefore associated with a more difficult access to long-term rentals, because the number of properties rented is decreasing and also, in some cities, because rents are being pushed up. Studies show, not surprisingly, that the victims of this development are mainly middle-class workers, students and young employees. In response to this problem, some local authorities (for instance, in New York) have imposed various regulations, including limiting the number of days a host can rent their property for short stays. Further research is needed to assess the extent to which these regulations are working.
• Finally, the development of Airbnb also affects life in the tourist areas, as these tend to be emptied of their ‘true’ inhabitants. There are two reasons for leaving these areas. On the one hand, as mentioned above, less housing is available and the rents for the remaining accommodations are rising. On the other hand, the remaining inhabitants suffer from the nuisance of living together with Airbnb guests (because the latter are just passing through and are therefore less likely to be attentive to the constraints of living together or to local conventions). Some studies show, however, that these cohabitation problems are not only felt by the locals but also by some Airbnb guests themselves. This effect of desertification of tourist areas therefore partly turns against Airbnb because it may lead travellers seeking personal contact with the local population to turn away from the platform.

Comparative impacts of the health crisis

At the time of writing, none of the impacts described above are really being felt. Indeed, the covid-19 pandemic has significantly affected Airbnb’s operations by drastically reducing the movement of people around the world. Of course, the entire tourism sector is being hit hard: just as Airbnb decided to lay off a quarter of its employees, many hotels are in a precarious situation, some even being forced to close down (such as the Hotel Métropole in Brussels). The question is, however, which of Airbnb (and collaborative platforms in general) or the hotel chains will be better able to recover once the crisis is over.

On the one hand, it can be argued that platforms are less at risk since they do not have their own properties and therefore, unlike hotel chains, do not have to pay the fixed costs associated with these properties. Airbnb should therefore get away with less damage.³ However, it must be stressed that Airbnb’s hosts are being hit hard by the crisis and, in particular, those who have invested in the development of properties for the sole purpose of renting them out for short stays on the platform. To try to recoup their investment, these hosts have no choice but to put their property on the long-term rental or sale market. And it is easy to understand that the sudden increase in supply that these movements cause prices to fall on both markets.

Hence, it is a safe bet that these hosts (whose properties are often highly valued by guests) will not try the Airbnb experience again anytime soon. It is also likely that candidate hosts will now be more reluctant to embark on entrepreneurial projects related to Airbnb. But if hosts are becoming scarce, the platform becomes less attractive for guests; and if fewer guests use it, it becomes less attractive to hosts. In short, these same network effects that facilitate the rapid growth of a platform can also accelerate its decline.

Faced with this critical situation, Airbnb did not sit idly by. On its new home page, the site now highlights monthly rentals. These ‘medium-term’ rentals are intended, for example, for people who have to spend a few months in another city for professional reasons. Airbnb’s interest in this kind of service is not new, but we suspect that the health crisis has prompted the platform to accelerate its entry into this segment. Given the size of Airbnb, this strategy is likely to have an impact on local rental markets. Of course, it is too early to estimate this impact and, in any case, it will be difficult to separate it from the effect of other factors (including the pandemic itself) which are currently hitting the markets.

In the end, the pandemic could be good news for Airbnb’s traditional competitors and for cities or neighbourhoods suffering from the inconvenience. For travellers, it may be bad news as their range of choices may shrink.

1 This post is an extended translation of an article that we published in French in the journal 'En Question' (see here). It contributes to the research project PROSEco (Platform Regulation and Operations in the Sharing Economy).

2 It is precisely by using this strategy that Airbnb has established its success. Unlike its competitor HomeAway (its predecessor in the market), Airbnb adopted a pricing structure that was more favourable to hosts than to guests (charging guests a booking fee of 9 to 12% of the value of the transaction, while hosts paid only a 3% service charge). Airbnb thus initiated the virtuous circle on the host side. However, Airbnb has reversed the trend in 2019 by increasing the bill for hosts (the fee is now 14% of the transaction value).

3 Airbnb CEO remains optimistic regarding the platform's recovery after the crisis; an IPO in 2020 does not seem to be ruled out. 

The post Sharing economy and tourism: who wins and who loses? appeared first on IPdigIT.