This paper – which can be found here – makes more use of the dismal science than the paper above, but makes a similar point: claims that online platforms have secured permanent monopolies protected by barriers to entry arising from network effects and stockpiles of data are inconsistent with economics, the technology literature, and the history of online competition.
The paper is structured as follows:
- Section I provides an Introduction. It notes that: “The record is replete with forecasts, soon proved wrong and then forgotten, that winners that took all, or most, were unbeatable’. Furthermore, there are four reasons why comparison to old staid corporate giants is unsuited to online platforms. First, turbulent waves of disruptive innovation have constantly shaken the business models of platform leaders and opened new avenues for entry and competition since the dawn of the digital age. Second, online platforms pegged as leaders in one area compete with each other in many other areas. They identify one another as competitors and commonly encroach on each other’s turf. Third, online platforms are more susceptible to attack by entrants than the network industries of a century ago. Multi-homing and low capital costs mean that the network effects that give rise to concerns about permanent dominance are reversible. Fourth, while many online platforms are “leaders” in particular categories, such as portals, these firms share a common business model: they compete for the scarce attention of people, and then compete to sell slices of that attention to advertisers.
- Section II delves into the economics and technology of online platforms. It provides a description of direct and indirect network effects, before explaining why they can be reversed in the digital economy.
First, indirect network effects do not necessarily lead to monopoly. Instead, indirect network effects may become exhausted at some point, as additional users stop adding value. Additionally, online platforms can soften the impact of indirect network effects by differentiating themselves on one or both sides of the platform in order to be appealing to different users. Looking at the reversal of indirect network effects, the author focuses on the difference between physical and online networks. Indirect network effects can be difficult to reverse for physical networks because users make specific capital investments in those networks that they would have to duplicate if they joined another network. By comparison, indirect effects are easier to reverse for online platforms. First, currently there are extremely low capital requirements to start a business – online platforms don’t have to sink capital to build the physical facilities needed to provide services, which means that, when online platforms become the incumbents, they are also much more vulnerable to entry and displacement than traditional network firms. Second, online platforms can rely on operating systems that sit on top of physical networks to achieve fast global distribution. Third, online platforms can easily add features because they are based on software, which also allows these features to be easily distributed to final users. This means that online platform providers often compete intensively with each other by adding features to improve their basic product offerings and to enter new markets, so as to avoid losing market position. Lastly, online platform users can multi-home and switch platforms at low cost. Online platform competition is dynamic and unpredictable because waves of disruptive innovation expand opportunities for entry and pose challenges to incumbents.
A last point is that different platforms compete for attention and advertising. Online attention platforms compete with each other, in various dimensions, even though, from the perspective of consumers, they seem to be doing different things. On the consumer side of the platform, they are all vying directly against each other to attract attention that can then be monetized through advertising. On the other side, they are also competing against each other for advertising spending. Even market winners are only winning the right to provide certain services for free – and compete with other market winners for advertising money. This seems to me to be a different point from the ones above – I assume the point here is that, regardless of all considerations about network effects and market power, there is competition between online platforms in the specific markets for consumers’ attention and advertising.
- Section III then provides an overview of dynamic competition among online platforms between 1995 and 2016. A first section is devoted to explaining how online platforms compete with each other across a range of products and services despite each having gotten a toehold in the digital economy by doing completely different things. This competition is both static (e.g. for advertising money) and dynamic (over new products and markets). As regards static competition, the main online platforms have posed, and continue to pose, a significant threat to at least another large online platform. If one looks at dynamic competition instead, one can see multiple entries, indicating that barriers to entry and switching costs to another platform are low. The author provides a number of examples – e.g. search engines, messaging apps, music downloads, lodging – to demonstrate that between the mid-2000s and early 2010s a number of platforms that looked like unstoppable leaders, protected by supposedly insurmountable network effects and first-mover advantages, fell from their perches. Knowledge that seemingly secure platform leaders have been toppled forces online platform leaders to engage in constant innovation to keep attracting users and holding on to their positions.
- Section IV deals with the argument that data results in a significant barrier to entry. The point is not really elaborated. Instead, it is merely held that: “One can make various conceptual arguments as to why data should or shouldn’t be important for online competition, and there is a thriving literature that does so. A fundamental point, however, is that as a general description of how online platform competition works the data-barrier to entry theory is inconsistent with the facts. The history described in the previous section doesn’t support the view that data acts either as a significant barrier to entry for online platforms or as an asset that protects incumbent platforms from competition”.
- Finally, Section V concludes. In short: “None of the firms act as if it has a moat around it, protected by data, network effects, or anything else. There don’t seem to be sleepy monopolies littering the digital economy. Few who participate in this sector can sleep well at night.” In an earlier segment of the paper, the author outlines the consequences of this for competition agencies: “Like all firms, online platforms may engage in unlawful collusion or monopolization. They may attempt mergers and acquisitions that could harm competition. Some online firms may have significant market power in particular lines of commerce and abuse that market power, through unilateral or vertical practices, to squelch competition. Evidence-based analysis, sharply focused on whether there is harm to consumers, and an understanding of the history and economics of online rivalry, however, should inform vigilance over the digital economy.”
Comment: This paper provides a good summary of the author’s work on the topic – which has long sought to argue that the platform and zero-cost economy is competitive. Personally, I would need to see stronger evidence of the reversibility of indirect network effects. Empirical support for multi-homing occurring in practice is relatively thin on the ground, so it would have been good to see a discussion of that. Furthermore, and as in the paper above, the implications of this for market power in the online platform’s core markets are somewhat limited – online platforms may be concerned with potential competition, and there may be competition pressures, which means that antitrust intervention is no so pressing as some claim. And yet, the question remains: when is such an intervention called for?
There is a basic argument uniting all authors of the “there is nothing to see here” persuasion: evidence-based analysis and empirical studies are needed before intervention occurs. I’m all for this, and this is a valuable precept of decision theory as applied to antitrust. However, another dimension of decision theory, which seems to be forgotten by these authors, is that decisions will have to be made under conditions of uncertainty, and that non-intervention may have its own costs. The argument that antitrust intervention is required in the digital economy is often premised on the assumption that the costs of non-intervention are extremely high (and, to be clear, people calling for antitrust intervention in the digital economy often have values other than short-term consumer welfare in mind, which is a different problem). The “no problem to see here” school implicitly argues that the costs of non-intervention are not that high at all, and will likely be a lot higher than the costs of mistaken intervention – but they rarely engage with the conditions under which the competition risks are such that they may require regulatory intervention. This may well be because they don’t know what these conditions are, which is something we should be working on.