This working paper – which can be found here – attempts “to better understand data-driven markets: i.e. markets where the cost of quality production is decreasing in the amount of machine-generated data about user preferences or characteristics (henceforth: user information), which is an inseparable by-product of using services offered in such markets”.
The authors start from the assumption that data-driven markets are characterized by imperfect competition and subject to indirect network effects. In the light of this, they try to determine: (a) under which conditions a duopoly can be a stable market structure in a data-driven market; (b) when the propensity to market tipping (i.e. to monopolization) becomes overpowering; (c) the conditions that allow a dominant company in one data-driven market to leverage its position into another market.
The paper begins, at least implicitly, by distinguishing between indirect network effects (which mix supply and demand effects. They occur where increased demand leads to decreasing costs in obtaining a product input – in this case, user information – which allows firms to improve their perceived product quality and, in turn, reduces their innovation costs), direct network effects (where the consumption utility of one network consumer increases in proportion to the amount of consumers on that same network – and which are, hence, completely demand-driven) and dynamic economies of scale (also known as learning-curve effects, which are completely supply-driven effects). The relevance of these distinctions, according to the authors, is that: “In contrast to these [two latter] mechanisms, data-driven indirect network effects cannot easily be copied by competitors or destroyed by the arrival of a new technology”. According to the paper, this distinction is quite important, since:
- Regarding the Stability of a Duopoly and the Propensity of a Market to Tip
Starting from small quality differences in obtained data (which influence the quality of the product), indirect network effects may mean that “the market will eventually tip and one firm will dominate the market. Moreover, we show that such dominance is persistent, in the sense that, once the market has tipped, the weaker firm will never acquire more than a negligible market share in the future. The market is even tipping if it requires continuous, small investments in innovation to keep consumers’ perceived quality constant, which appears to be a reasonable description of dynamic, high-tech markets.”
Another important consequence of indirect network effects is that: “An important feature of a tipped market is that there are very little incentives for both the dominant firm and the ousted firm to further invest in innovation. [The dominant firm has] significantly lower marginal costs of innovation, due to its larger stock of user information. The latter characteristic enables the dominant firm to match any innovative activities of the ousted firm at lower marginal innovation cost and hence keep its quality advantage. As demand follows quality differences in our model, the smaller firm gives up innovating if its quality lags behind the larger firm’s too much. Knowing this, the dominant firm’s best response is to also save on investing in innovation and reap the monopoly profit.”.
- Regarding the Leveraging of Market Power across Markets
Following this, the authors develop the concept of “connected markets”: “which captures situations where user information gained in one market is a valuable input to improve one’s perceived product quality in another market. We show that user information in connected markets is two-way complementary, such that incentives to acquire user information in one market can justify market entry in another market, and vice versa”. The same results regarding market tipping and market stability identified for the original markets also apply to connected markets. This suggests that domino effects can occur, with the dominant firm in market A leveraging its advantage on user information in this market to make market B tip in its favour as well, even if market B is already served by traditional incumbent firms.
- Normative Implications
Because a tipped market provides low incentives for firms to innovate further, market tipping may be negative for consumers. It also deters market entry by new firms, even if they have or are able to develop revolutionary technologies. In order to address these concerns, the authors consider whether it would make sense to force data-driven business models to share their (anonymized) data about user preferences or characteristics with their competitors – and conclude that data sharing (voluntary, or not) eliminates the mechanisms causing data-driven markets to tip. However, while data sharing can prevent market tipping without negatively affecting the dominant firm’s incentives to innovate, its aggregate welfare consequences are ambiguous because of the possible duplication of innovation costs. The authors also hold that their model allows them to identify: “the characteristics of industries that may be prone to entry of data-driven firms, which has wide-ranging implications for suppliers, buyers, antitrust and regulation authorities in many industries, including some traditional sectors that are not thought of as data-driven today.”
As with most economic papers, I am not qualified to comment on the model developed by the authors (which is discussed in detail in Sections 3 and 4). It does seem to be based on an awful number of assumptions, and a cynical observer (which I am not) may think it was developed to understand Google’s business model / elaborate a theory of harm supporting the intuition that something is wrong with Google’s dominance. Further, it seems to assume that the only relevant variable for market competition is data, which limits the scope of the model a bit: in effect, the only businesses that the authors review are search engines and digital maps. I also have some issues with a number of the concepts used, which definition is crucial to the model. For example: (a) a data-driven firm is described as “a potential entrant, who uses a data-based business model and, hence, is harvesting indirect network effects” . This definitions is not only somewhat circular, it could theoretically be applied to any company which relies on user-data to improve their product; (b)“connected markets” are not defined; (c) what “innovation” may mean in this scenario (or how to measure it) lacks of clarity, even as it is presumed to lead to eternal market dominance.
Despite these reservations, if the model accurately reflects some of the phenomena underlying the digital economy, the paper may end up being seen as an important piece of work. First, it is able to explain how successful technological firms (“moligopolies”, in N. Petit’s formulation) are not only extremely dominant in their core market, but also find themselves regularly competing with other digital giants in related markets where they have displaced (or are displacing) traditional incumbents. Secondly, it provides a model that can be used to explain and assess the validity of a concern which has been (implicitly) articulated about data dominance: that data incumbents will be able to anticipate the creation of new markets and pre-empt competition in them.
In short, a stimulating paper, which may well lead to more research on “indirect network effects”.