A lot of work has been done over the past two years on laying out systematic foundations for our understanding of some key characteristics of large digital platforms. The Furman, Stigler, and Cremer Reports (HM Treasury 2019, Scott Morton 2019, Crémer et al. 2019), as well as multiple agency studies by the UK Competition and Markets Authority (CMA 2019), the Australian Competition and Consumer Commission (ACCC 2019), and others (e.g. Autorité de la concurrence and Bundeskartellamt 2019), have set out how internet aggregators, online marketplaces and social networks benefit from direct or indirect network effects and scale; how these forces have led to high concentration and winner-takes-all dynamics; and how the ability to collect and exploit data at scale plays a critical role. This has all been good background.
One area where issues are well understood is dynamic competition for the platform. From the original network effects models of the 1980s to the Microsoft era, we have theories (most notably, Carlton and Waldman 2002 and Choi and Stefanidis 2001) which rationalise why a platform might want to undermine certain complementary services because they constitute potential dynamic threats to the platform’s core monopoly. These models rationalise well the historic Microsoft cases and subsequent enforcement against Google in Europe. Further literature (Choi and Jeon 2020, Etro and Caffarra 2017) also helped us understand how pricing constraints – e.g. the need to set a zero price on certain consumer-facing services – can generate inefficiencies which incentivise colonisation of adjacent markets on the part of platforms.
But there are gaps in our understanding of how platforms behave and what this means for welfare. The main issue to be cracked for the forthcoming major regulation effort in Europe1 is competition on the platform. How does the owner of a proprietary platform set the ‘rules of the game’ for third parties that operate and create joint value on their ecosystem? What about cases where the platform owner also supplies services in competition with these third parties? While control by the platform owner might generate efficiencies (e.g. if it internalises negative spillover effects on its quality, for example, in a way that other platform participants do not), it is also right to worry about harmful effects. This ‘dual role’ has come under major scrutiny in Europe, with investigations inter alia into the commissions Apple sets for in-app purchases of digital goods and the operation of Amazon’s online marketplace. Concerns around these ‘umpire and player'2 issues extend also to Google’s role operating the dominant search engine while being also a major player in online advertising and operating its own specialised search services (e.g. in comparison shopping, local, travel).
There is an economic literature which has started to look at the economics of ‘competition on the platform’ and is expanding. Relevant references here are Hagiu and Wright (2015), De Corniere and Taylor (2019) and Hagiu et al. (2020).3 But more needs to be done to make these models speak to our current policy questions around how to conceptualise and measure the welfare effects of platform rules and structures. These insights are necessary both for effective antitrust cases and any regulation of platforms. Tech observers and bloggers like Ben Thompson of Stratechery4 and Ben Evans5 often remark on how the regulatory discourse around large platforms is naive and uninformed: “We have locked in the tech giants for a generation by not understanding how their businesses work”; “What I see again and again is an insistence on a framework of understanding that is obsolete. And then people wonder why regulation is so ineffective, or why things like GDPR make these companies stronger. This outcome was knowable!”; and “Aggregators and Platforms are fundamentally different, but by treating everything as a platform we are not simply failing to regulate Aggregators, we are strengthening their hand at every turn. Yet conference after conference, regulation after regulation, it’s the same thing. Pretending these are platforms when they are something different — and different not because they are bad actors, but because the economics are so fundamentally different. Yet we keep pretending it is the same”.
What new thinking do we need?
What extensions do we need to the existing literature? What are the right models to think about these issues and how can they be interrogated to inform the current debate on regulating ‘gatekeeper platforms’?
One major source of differentiation we need to take on board is distinctions in the business models the various ecosystems operate, and how these different strategies for monetising the surplus created by their platforms influence their incentives (Caffarra 2019). Unlike, for instance, telecoms operators that were subjected to industry regulation that applied fairly uniformly because they all operated roughly the same business model, digital tech giants operate a spectrum of business models.
At one end we have device-funded ecosystems like Apple’s (or, similarly, license fee or subscription funded ecosystems like Microsoft’s), whose value is generated by the presence of a desirable set of complementary services (e.g. Apple’s App Store or the ecosystem of applications around Microsoft’s OS and cloud products) but which are primarily monetised via consumer-side ‘access fees’ (whether this be device prices or subscriptions). Then at the other end, there are ad-funded ‘aggregators’ who provide ‘free’ services to consumers and monetise through ads or data collection. Google and Facebook are the leading examples in this category, with services (Social Networking, Search, Gmail, YouTube) provided at a zero price to consumers and monetised via advertising, or by gathering data which can be used and monetised elsewhere.
Of course, the distinction is not new; different business models have co-existed elsewhere, including in direct competition with each other. Subscription-based pay TV and ad-funded TV have been around for decades. The market for news involved a variety of business models well before the advent of online news. But any regulatory framework needs to account for these differences and, in the case of digital platforms, the choice of business model determines how each platform will react to the evolution of the ecosystem, and how strategies for interacting with third party complementors will affect consumers.
Modelling ‘umpire and player’ across business models: Some preliminary insights
How might a platform’s business model affect our specific thinking on conduct? Some incipient, ongoing research can provide some insights.
A first example considers competition between a platform that is funded mostly through the sale of a physical product (say, a device) and one monetised indirectly via advertising.6 The obvious direct analogy is to the respective ecosystems of Apple (which is able to monetise its OS and app-store via sales of the iPhone and other devices) and Google Android (which distributes its OS through independent manufacturers without charging a license fee, monetising its operations indirectly through its broader ad-funded ecosystem). However, the results provide more general insights into how monetisation strategies impact the respective incentives for investment, and the pricing of platforms’ app ecosystems.
Both platforms want a vibrant and rich app store because this enhances the attractiveness of their devices, and hence, the number of users on their platform. Both the ad-funded and device-funded platform can monetise via commissions on sales of third-party apps, but the device-funded platform can set the price of its devices whereas the ad-funded platform cannot, instead generating revenue based on the total number of users and ‘eyeballs’ it can monetise.
Consider the incentives to invest in the quality of the platform’s ecosystem (e.g. by better curating third-party apps, introducing more stringent quality standards, and so on). Such investments are likely to generate benefits for users, potentially drawing more of them onto the platform, but this quality will be costly. The internalisation of these benefits will differ across business models.
A device-funded platform is likely to internalise the benefit of an ecosystem that consumers value more highly because it can monetise that ecosystem directly by charging a higher price for its devices. By contrast, the degree of ‘internalisation’ by an ad-funded platform is likely to be different. Under this business model the platform cannot directly alter the price of devices and so cannot as easily alter the ‘access fee’ for the platform in response to changes in the quality of the overall ecosystem. This means that its revenues and profits are less closely tied to the quality of the ecosystem: a higher-quality ecosystem is helpful of course, insomuch as it leads to more users. But the increase in value per user cannot be as easily extracted with advertising. This difference will tend to push the ad-funded ecosystem towards fostering cheaper devices which bring in more users rather than a high-quality ecosystem that can be monetised via higher-end prices.
There is of course a broader range of conducts of concern around app ecosystems (from levels of commissions to alleged foreclosing behaviour) which are also being explored, but the key overall insight is that the incentives generated by the underlying monetisation strategy matter for competition and for welfare outcomes. To give another example, the ‘incidence’ of a commission on sales of third-party apps is likely to be very different for a product-funded ecosystem (which might be incentivised by such a revenue stream to set lower prices for its main products relative to a counterfactual with no commissions) versus an ad-funded operator (where pass-through will tend to be more diffuse).7
Take a second example. Consider a business model like Amazon’s, in which sales by third-party sellers are monetised via commissions while Amazon’s own retail business sells potentially the same products – thus monetising via direct sales to consumers. Multiple critical questions have arisen in the public and regulatory debate on the interaction between Amazon and third-party sellers hosted on its platform. Does Amazon have incentives to ‘steer’ consumers towards its own products (‘self-preferencing’)? Does it have incentives to enter as a retailer (or even launch own label products) in ways that disadvantage third-party sellers?
Ongoing research (Etro 2020b) considers an e-commerce platform whose main choice variables are deciding what terms to set for third-party sellers, what commissions to set, what prices to set for the products it retails, and how to organise logistics. While the platform has a comparative advantage in logistics, rival suppliers often have more information about how to create a niche product and may have a brand or useful channel with which to reach customers.
In deciding whether to ‘enter’ as a direct retailer in competition with third-party sellers, the platform owner should compare commission revenues from sellers with net profits from direct sales of its own version; and these entry decisions should be based on profit-maximisation objectives. How far are the platform owner’s choices from anything that maximises consumer welfare? Well, the business model matters again.8 The analysis first considers a uniform product sold by many external providers competing to make a sale on the platform. If the e-commerce platform sets the commissions on third-party sales appropriately, at a level that maximises its own profits, it will then decide to enter as a direct retailer only if its efficiency relative to alternative sellers in terms of logistics and marketing will lead to prices that are lower enough to increase aggregate consumer welfare.9 In this set up, there is no systematic bias in the direction of the e-commerce platform ‘overproviding’ its own products, i.e. entering as a first-party or a private label seller to crowd out competitive third-party sellers. But this is why the business model matters: critically, closer alignment with the interest of consumers might not hold if the platform was monetising sales in a different way. As we know from the Google Shopping case and other equivalent settings, self-preferencing will emerge much more naturally for an ad-funded business model.
There are a variety of corollaries of this analysis. The first emerges when third-party sellers have at least some market power. In that case, the e-commerce platform has greater incentives to enter with a private label version of the product to avoid double marginalisation on sales. Instead, it will have fewer incentives to enter as a direct retailer in the case of products where it faces higher wholesale prices (as for brands for luxury, beauty and apparel). Yet, in both these cases, market power by third-party sellers leads, if anything, to a bias toward underprovision of private label and first-party products relative to what would be in the best interest of consumers – because consumers tend to gain more from the price reductions than the profit Amazon can gain from its sales. A second analysis involves introducing competition for customers with other platforms. Adding this dimension tends to reduce both the commissions on third-party sellers and the prices of products retailed directly, while decisions to ‘enter’ as a retailer are even more closely in line with the interest of consumers (again in this setting).
We need to extend and challenge this type of work
What does all this mean? The research we have briefly mentioned here suggests that business models matter fundamentally when one is evaluating the dual role of platform owners as umpires and players in their ecosystem. At a general level, platforms that monetise on the consumer side (through the sale of a product, or consumer-side commissions on transactions) are those that best internalise the interest of consumers in their interactions with third-party players, relative to ad-funded platforms. This fundamental distinction is important to explain why we cannot just assume that, if internet businesses/aggregators like Google were found to have strong incentives to engage in self-preferencing, other platforms that operate a very different business model also inevitably have the same incentives – i.e. we cannot ‘map’ the exact same concerns merrily across all internet businesses and platforms. Rather, economists will have to create models that capture the incentives of each platform and rigorously evaluate the concerns.
More importantly, we need this type of work to be extended, challenged and tested on observations and actual data. To make policy recommendations and plan the forthcoming ex-ante ‘gatekeeper regulation’, on which an important public consultation has begun, we need a better understanding of the magnitude of the problems businesses are respectively trying to solve; how the different players respond to changes in prices or rules; how consumers differ and how many are of which type; how to measure overall welfare effects, not just partial analyses; and therefore how to determine which model best fits the data and which factors are most important.
For economists, what other models do we need? Do we have models that describe these situations well? What insights and empirical predictions can be drawn? For companies, one approach is to keep heads below the parapet and resist scrutiny as much and for as long as possible. Yet scrutiny comes with size, and power, and if one has nothing to hide it is better to show one’s hand and allow for empirical independent academic research to proceed.
Authors’ note: The authors have been involved to different degrees in advisory work both for and against tech platforms, including Apple, Amazon, Microsoft, Uber and others.
ACCC – Australian Competition and Consumer Commission (2019), Digital platforms inquiry - final report.
Autorité de la concurrence and Bundeskartellamt (2019), Algorithms and Competition.
Bertoletti, P and F Etro (2016), “Preferences, entry and market structure”, The RAND Journal of Economics 47(4): 792-821.
Bertoletti, P and F Etro (2017), “Monopolistic competition when income matters”, The Economic Journal 127(603): 1217-1243.
Buehler, B (2015), “Do international roaming alliances harm consumers?”, Journal of Industrial Economics 63(4): 642-672.
Caffarra, C (2019), “Follow the Money - Mapping issues with digital platforms into actionable theories of harm”, Antitrust Case Laws e-Bulletin, e-Competitions Special Issue Platforms, August.
Carlton, D W and M Waldman (2002), “The strategic use of tying to preserve and create market power in evolving industries”, The RAND Journal of Economics 33(2): 194.
Choi, J P and C Stefanadis (2001), “Tying, investment, and the dynamic leverage theory”, The RAND Journal of Economics, 52-71.
Choi, J P and D-S Jeon (2020), “A leverage theory of tying in two-sided markets with non-negative price constraints”, American Economic Journal: Microeconomics, in press
CMA – Competition and Markets Authority (2019), Online platforms and digital advertising. Market study interim report.
De Corniere, A and G Taylor (2019), “A model of biased intermediation”, The RAND Journal of Economics 50(4): 854-882.
Crémer, J, Y-A de Montjoye and H Schweitzer (2019), Competition policy for the digital era, European Commission.
Etro, F (2020a), “Device-funded vs Ad-funded Platforms”, mimeo, University of Florence,
Etro, F (2020b), “Product Selection in Online Marketplaces”, mimeo, University of Florence.
Hagiu, A and J Wright (2015), “Marketplace or reseller?”, Management Science 61(1): 184-203.
Hagiu, A, T-H Teh and J Wright (2020), “Should Amazon be allowed to sell on its own marketplace?”, mimeo, Boston University
HM Treasury (2019), Unlocking digital competition, Report of the Digital Competition Expert Panel.
Khan, L M (2016), “Amazon’s Antitrust Paradox”, Yale Law Journal 126: 710-805
Scott Morton, F (Chair) (2019), Report of the Committee for the Study of Digital Platforms, Stigler Center for the Study of the Economy and the State.
Spence, M (1976), “Product selection, fixed costs, and monopolistic competition”, Review of Economic Studies 43(2): 217-235.
2 See US Senator Elizabeth Warren’s comment that “you don’t get to be the umpire and have a team” (shorturl.at/yF038); see also Khan (2016).
3 See also Hagiu and Julian Wright (2015).
6 A formal treatment is in Etro (2020a), analysing competition between a device-funded platform and an ad-funded platform in a model with heterogeneous consumers and monopolistic competition between apps on each platform (based on the formalisation of Bertoletti and Etro 2017).
7 In particular, the formal analysis shows that a platform that monetises mostly through the sale of a product has incentives to set both the commission and the price of the product in ways that internalise the impact on the total surplus of consumers, while an ad-funded platform does not have these same incentives. More precisely, consumers benefit from a positive commission on sales of third-party apps in the case of a product-funded platform because this allows for a profit-maximising price for the product that is lower than it would have been in the absence of such a commission.
8 For a formal treatment see Etro (2020b), whose entry analysis relies on the literature on entry (Spence 1976, Bertoletti and Etro, 2016).
9 Formally, such alignment between the interest of the platform and its consumers requires conditions on the demand system, which hold under common demand functions (as linear, loglinear and isoelastic demand functions). More generally, either overprovision or underprovision of entry may emerge for a product under the assumption of competitive 3P sellers, but without a bias in either direction.