Digital Platform Analysis PDF

3.C Digital platforms What do we study in this section? 1. Definitions and typology 2. Platform vs. Pipeline 3. Expectations matter! 4. Regulation issues 40 Digital platforms: A broad definition A platform is an entity that brings togeth...

3.C Digital platforms What do we study in this section? 1. Definitions and typology 2. Platform vs. Pipeline 3. Expectations matter! 4. Regulation issues 40 Digital platforms: A broad definition A platform is an entity that brings together economic agents, actively manages network effects among them, and, thereby, generates economic value. Key characteristic Platforms are “managers Presence of of network effects” NETWORK EFFECTS Source: Belleflamme and Peitz (2018). 41 Network effects Definition Network effects arise when the value of consuming a product or service depends on the consumption decisions of other users. Network effects are positive (negative) if the product or service becomes more (less) valuable as more people use it. 2 main categories … in her own group One agent’s decisions → Same-side network effects as to whether and how much to interact on the platform affect the well-being of other … in another group agents… → Cross-side network effects 42 Network effects (2) Illustration Same-side Cross-side network effects network effects More end-users More end-users → More communication possibilities → More value for app developers → More value for end-users More apps → More value for end-users The decisions of platform users are interdependent. 43 Mapping network effects on platform Cross-side network effect (A→B) Same-side network Positive? Negative? Negligible? effect (B→B) Positive? Negative? Negligible? Group A Group B Platform Same-side network effect (A→A) Positive? Negative? Cross-side network effect (B→A) Negligible? Positive? Negative? Negligible? 44  Network effects generate feedback loops The presence of positive direct network effects gives rise to a feedback loop. The higher the activity level of the group… the more attractive it becomes for each group member to increase her activity level… feeding back into the group’s overall activity level. 45 Online vs. Offline platforms Platforms are not specific to the digital economy Examples of digital platforms and their ‘brick-and-mortar’ equivalent Marketplaces Professional networks Social networks 46 Online vs. Offline platforms (2) Platforms have been around for a long time 47 Online vs. Offline platforms (3) Why are there way more platforms in the digital economy? Digital technologies considerably reduce transaction costs Mainly, costs linked to coordination of needs and trust issues. Example: Car sharing ‘Old days’ → Few people were giving Why? Lack of (accepting) rides to (from) strangers Coordination & Trust → No meeting between supply and demand → Underused asset 48 What is (and what is not) a platform? Recall the definition A platform is an entity that brings together economic agents, actively manages network effects among them, and, thereby, generates economic value. Applying the definition Note always easy: often a matter of degree and focus. Rule: If network effects are prominent and actively managed by some entities, then the platform aspect is central to analyse the economic environment. 49 What is (and what is not) a platform? Amazon 1994-1996 1996-2001 2001-today Sellers Sellers Sellers Sale of Sale of Affiliation goods goods Sale of Sale of Affiliation goods goods Buyers Buyers Buyers Pure reseller, no network effects Cross-side effects → Not a platform between the two sides (Pipeline) → ‘Two-sided’ Platform Same-side effects on the buyer side → ‘One-sided’ Platform 50 What is (and what is not) a platform? Examples 34 CHAPTER 1. PLATFORMS: DEFINITIONS AND TYPOLOGY Network effects are...... present?... managed? Platform? Original Facebook Yes (same-group) Yes Yes Wikipedia Yes (same-group) Yes Yes Uber Yes (cross-group) Yes Yes Esperanto Yes (same-group) No No Aldi No No No Tesco Yes (cross-group)? Yes? Yes? Individual blog No No No Group or ad-financed blog Yes (cross-group) Yes Yes Petrol stations No No No Notaries No No No Table 1.2: Platform or not? 51 Platforms vs. Pipelines PIPELINES PLATFORMS Value is created in a linear Value is created by facilitating way with centrally employed interactions between external staff and owned assets. providers and consumers. Transactions are controlled. Transactions are enabled. MOTIVATION & ADAPTATION Ability to drive professionals to exert costly efforts that lead to improve the customers’ experience. Capacity to adjust decisions to private information that professionals may have COORDINATION Capacity to internalize potential spillovers Remedy: Charge professionals a positive (negative) transaction fee if negative (positive) spillovers Source: https://medium.com/@wasan.makhlouf/pipeline-vs-platforms-e8540663c916 + Hagiu & Wright 52  Towards ‘plat-lines’ and ‘pipe-forms’? PIPELINES PLATFORMS Value is created in a linear Value is created by facilitating way with centrally employed interactions between external staff and owned assets. providers and consumers. Transactions are controlled. Transactions are enabled. Move to hybrid models Pipe-forms / Plat-lines 53 Platform and Pipeline: A question of dosage Example of Amazon Both control mode (reseller) and enable mode (marketplace) Which mode should be used for which products? Products that change Products that change relatively slowly relatively fast (Books) (Electronics) RESELLER MARKETPLACE (little need for adaptation) (high need for adaptation) Matches observations Data from January 2014 (Hagiu and Wright, 2015) Amazon was listed as a seller for 54% of the 20,5 million new books, but for only 1% of the 25,8 millions new ‘Electronics’ items it had in stock. 54 Takeaway The enabling mode (platform) fares better in terms of motivation (inducing professionals to exert efforts) and adaptation (letting professionals adjust their actions according to their private information). The controlling mode (‘pipeline’, i.e., vertical integration, resell) fares better in terms of coordination (internalizing spillovers across professionals, goods and services). 55 Demand and network effects A simple model 2 users: Xena and Yuri Unit demand for a good Intrinsic valuation for the good: >' and >( Assume: >' > >) Selling (access) price: + (given for now) Demand function absent network effects (benchmark) + > >' ⟹ @ = 0 >( < + ≤ >' ⟹ @ = 1 The quantity demanded is an inverse + ≤ >( ⟹ @ = 2 function of price (Law of Demand) At each price + corresponds a single quantity @. 56 A simple model (2) Network effects Each user gets some additional benefit C > 0 when the other user also purchases the good. Examples: Sharing of folder on Dropbox, of playlists on Spotify. Additional network benefit is conditional on the decision of the other user. → Users’ decisions become interdependent! → To derive the demand, we must characterize the Nash equilibrium of the game played by the two users. Assumption: ‘Strong’ network effects → C > >' − >( 57 A simple model (3) Adoption game N.E. if rY + b ≤ A ≤ rX (impossible if b > rX – rY ) N.E. if A ≤ rY + b Never N.E. N.E. if A ≥ rX 58 A simple model (4) Adoption game if strong network effects (C > >' − >( ⟺ >( + C > >' ) rY rX rY + b rX + b A X always buys X buys if Y buys X never buys Y always buys Y buys if X buys Y never buys Both buy Both buy or none None buys Demand function # > N+ + O ⟹ Q = 0 N3 < # ≤ N+ + O ⟹ R = ; or R = 6 # ≤ N3 ⟹ Q = 2 The two equilibria conform to a self-fulfilling prophecy. 59 Self-fulfilling prophecy 60 A simple model: Summary ‘Strong’ network effects 2 major transformations in the demand function No price such that only one user buys the good. Network effects make The same price may give rise to two different quantities. users’ purchasing decisions Consequences interdependent. As a result, there can be several levels Unpredictability: For intermediate fees, no way to know of demand for the same a priori whether the good will be adopted or not fee, making it hard to Potential inefficiency: predict a priori what users # ≤ N+ + O → Both users are better off if they buy. will decide and giving rise to potential inefficiencies, # > N3 → Users may not buy at equilibrium. as users may fail to purchase the product although this would make them all better off. 61 Should dominant platforms be regulated? Image: https://www.flickr.com/photos/dullhunk/53516146526 / Quote: https://pluralistic.net/2023/01/21/potemkin-ai/ 62 “How platforms die” (according to Corry Doctorow) 1. They are good to their users; Common way 2. They abuse their users to to address the chicken-and- make things better for their egg problem business customers; 3. They abuse those business customers to claw back all More questionable the value for themselves. Condition (a) Condition (b) Market power Quality degradation Image: https://m.wikidata.org/wiki/Q110436 / Quote: https://pluralistic.net/2023/01/21/potemkin-ai/ 63 1 & 2. The “chicken-and-the-egg” problem The conundrum of network effects o Group A is willing to participate only if Group B participates … o But Group B is willing to participate only if Group A participates! Offer chickens a deal they can’t refuse → Subsidize the participation of a whole group Common launching strategies Lay the first eggs yourself → Start as a pipeline and turn into a platform later 64 How do digital platforms gain market power? § Tipping § “Killer acquisitions” o Network effects o Elimination of possible generate positive sources of future feedback loops. competition o Feedback loops lead o Acquisition of patent to market § Switching costs portfolios and highly concentration. o Collective switching talented workers costs o Most interactions take o Ecosystem expansion place on a dominant o Use of personal data to platform. customise services o Imperfect data or reputation portability o Addictive design o Ecosystems 65 What affects platforms’ market power? Forces favoring Forces favoring platform dominance platform coexistence Positive network effects Negative network effects (cross-side and same-side) (congestion, competition among users, Switching costs mismatch, nuisance) Scale economies Platform differentiation Data self-reinforcing effects Multihoming Interoperability Data portability 66 3(b). Quality degradation by dominant platforms? 1 Distortion of search engines/recommender systems 2 Behavioural biases and obfuscation 3 Insufficient control of negative network effects The goals of the advertising business model do not always correspond to providing quality search to users. [...] we expect that advertising-funded search engines will be inherently biased towards the advertisers and away from the needs of the consumers. The Anatomy of a Large-Scale Hypertextual Web Sergey Brin & Larry Page Search Engine http://infolab.stanford.edu/~backrub/google.html 67 Distortion of search engines/recommendations Motivations Favour offers leading to Self-preferencing Discourage search to higher commissions or (in-house production or reduce competition purchase probabilities partial integration) among sellers Trade-off Larger Lower revenues participation (short term) (long term) 68 Behavioural biases and obfuscation § Context o Behavioural “biases”/ limited cognition may lead to suboptimal decisions. Example. Consumers not aware of extra charges after paying subscription fee o Consumers with misperceptions are highly valuable to sellers. § Main issue o Platform’s incentives to regulate sellers’ conduct? § Trade-off o Laissez-faire Sellers can exploit buyers’ behavioral biases. Profitable for the platform if it can extract part of the extra sellers’ Balance surplus. is unclear o Restrictive design (Example. Impose transparency of sellers’ offers) Sellers cannot exploit buyers → More buyers participate. Profitable for the platform if it can extract part of buyers’ surplus. 69 Insufficient control of negative network effects § Dominant platforms attract more “nuisance” o Poor quality Unreliable sellers, counterfeited products, AI-generated content, … o Malevolent behaviour Spams, trolls, fake reviews, click farms, hateful content, … o Cyber criminality § Main issue o Dominant platforms tend to underinvest in combating negative network effects (compared to social optimum) Private costs < Social costs ü Platforms may underestimate systemic risks Potential private benefits ü Commissions on sales of counterfeited products ü Advertising revenues from AI-generated content Locked-in users 70 Cory Doctorow may be right… “How platforms die: 1. They are good to their users; 3. They abuse those business 2. They abuse their users to make customers to claw back all the things better for their business value for themselves.” customers; If a platform and become by letting But this can launch dominant in service quality doesn’t mean successfully… its market. degrade… that it will die! 1 2 3 4 5 6 7 It can grow It may then or by larger and abuse its degrading it larger… market power… purposely. Quote: https://pluralistic.net/2023/01/21/potemkin-ai/ 71 European regulation. A move in the right direction → Targets large digital platforms, → Focuses on regulating online identified as "gatekeepers," to… intermediaries and platforms to… Ensure fair competition Protect users' fundamental rights Prevent unfair practices Promote a safer online environment Enhance contestability Mitigate systemic risks Promote innovation and consumer Enhance transparency and accountability choice Combat disinformation and illegal content 72