In Chapter 2 of the ebook Digital Platforms and Concentration, forthcoming ahead of the Stigler Center’s annual antitrust conference on April 19 and 20, Justus Haucap of Heinrich-Heine University of Duesseldorf examines the distinctiveness of how Germany does antitrust for big tech, with specific looks at its current investigations of Google Shopping and Facebook and how useful data portability could really be.
Editors’ note: See also our 2017 conference volume Is There a Concentration Problem in America? Other chapters from this year’s ebook can be read at the following links:
The fast growth of digital platforms such as Amazon, Google, Facebook, and company, as well as their high market capitalization, is causing rising concerns for many policymakers around the globe, especially—but not exclusively—in Europe. Given that share prices reflect what markets expect regarding a firm’s future profitability, high share prices can also reflect expectations about (future) market power. As platforms are characterized by economies of scale and network effects, expectations or concerns about increasing market concentration are legitimate (see, e.g., Haucap & Heimeshoff 2014).
However, not every platform market is highly concentrated. Counterexamples include online retailers, digital real estate brokers, travel agents, and online dating sites. The presence of indirect network effects is not sufficient for a monopoly or high levels of market concentration. From a theoretical point of view, it is also not clear whether competition between several platforms is necessarily welfare-enhancing when compared to a monopolistic market structure. While, generally speaking, competition between several firms is almost always beneficial in standard markets (as long as the market is not characterized by natural monopoly conditions), this general wisdom does not always hold for multi-sided markets. Even if multiple platforms are not associated with additional fixed costs, the existence of multiple platforms may not be efficient due to the presence of direct and indirect network effects.
As Caillaud and Jullien (2003) and Jullien (2006) have shown, a monopoly platform can be efficient because network effects are maximized when all agents manage to coordinate on a single platform. Strong network effects tend to make monopoly structures efficient, while the risk of platform overload and lower participation rates and users’ so-called multi-homing opportunities suggest that competition is also efficient in digital markets. In fact, it is not only the welfare effects of a monopoly in such markets that are unclear, but also whether the market is quasi-naturally converging towards a monopoly structure.
Factors Determining Concentration in Platform Markets
Evans and Schmalensee (2008, 2015) have identified five factors that determine the concentration process in two-sided markets, as described in the following table:
It is relatively straightforward and immediately plausible that indirect network effects and economies of scale lead to increasing concentration. The strength of these indirect network effects will differ from platform to platform. In general, digital platforms are typically characterized by a cost structure with a relatively high proportion of fixed costs and relatively low variable costs (see, e.g., Jullien 2006). For example, for eBay, Expedia, Booking.com, etc., most of the costs arise for managing the respective databases, while additional transactions within the capacity of the databases usually do not cause additional cost. Increasing returns to scale are therefore typical for digital platforms.
The most important countervailing force is multi-homing opportunities. How easy it is for users to multi-home depends, among other things, on (a) switching costs (if they exist) and (b) the structure and height of platform charges. To switch, for example, from one online travel agency to another is usually associated with relatively low switching costs. Users can also switch away from Google to another search engine without major costs if a switch appears attractive. Similarly, it is typically relatively easy for sellers to open a second, third, or fourth Internet store, especially when compared to opening more brick-and-mortar stores. In contrast, switching costs between social networks are generally higher because of strong direct network effects and the effort needed to coordinate user groups. While for Google no significant direct network effects exist—i.e., it does not directly matter how many other people use Google—this is not true for social networks such as Facebook where the number of users is an important determinant.
Another form of switching costs can be found on auction platforms such as eBay, where apart from indirect network effects individual reputations are also highly relevant. As the reputation is built up as a function of the number of transactions already conducted over the platform and is therefore eBay- or platform-specific, changing platforms involves some switching costs if reputations are not portable across platforms.
Another countervailing force is capacity constraints. While for physical platforms such as shopping centers, fairs, and nightclubs, space is physically limited,1 this does not hold for digital markets. However, with regard to online markets advertising, space is often restricted since too much advertising can be perceived as a nuisance by users and can therefore decrease the platform’s value in the recipients’ eyes (Becker and Murphy 1993; Bagwell 2007).
In electronic markets like auction platforms or dating sites capacity limits can also emerge as a result of negative externalities caused by additional users. If additional users lead to a more heterogeneous group search costs may increase. In contrast, the more homogeneous the users are, the higher the value of a given platform can be for the demand side. If for example only certain people visit a particular platform (as some platforms are, for example, mainly visited by women, golf players, academics, etc.), targeted advertising is easier for advertising companies. Also note that some dating sites advertise that they only represent a certain group of clients (for example, only academics). This reduces the search costs for all visitors involved. Additional users would make the group of users more heterogeneous and not necessarily add value as they increase the search cost for other users.
Quite a few parties that advocate platform regulation also have a vivid interest in limiting competition in the concerned markets.
In summary, while there are tendencies that foster concentration in digital markets, there are also countervailing forces. Moreover, one should note that many digital platforms drive competition in their respective product markets. Amazon, eBay, and other online retail platforms have intensified competition in retail markets, Uber and company have injected competition into the taxi cab industry, AirBnB and similar platforms facilitate competition within the short-term accommodation market, and so on. Not surprisingly, quite a few parties that advocate platform regulation also have a vivid interest in limiting competition in the concerned markets.
Antitrust Laws for Platform Markets
In response to the growing relevance of digital platforms, Germany—probably as the first jurisdiction—has, in 2017, introduced a number of additional criteria into its competition law that the German competition authority and the courts now need to consider when assessing market power in platform markets. The newly introduced §18 no. 3a of the Act against Restraints of Competition now requires that, when assessing an undertaking’s position in a market, the following five factors need to be also considered for multi-sided markets and networks:
(1) Direct and indirect network effects
(2) The parallel use of several services and users’ switching costs
(3) Economies of scale in relation with network effects
(4) Access to data relevant for competition
(5) Competitive forces of innovation
While these five factors certainly help to prevent especially the courts from focusing too narrowly on markets shares and other concentration measures only, it is rather unclear how these factors can be measured in practice. While the ease and practice of multi-homing may be relatively easily measured empirically, the strength of network effects is more difficult to measure, as are the remaining three factors. While competition authorities have so far heavily relied on factors that are reasonably easy to determine (market shares, concentration rates, price-cost margins, profitability measures, patterns of substitution, and so on), thereby economizing on law enforcement costs, the newly introduced factors are more difficult to operationalize. Hence, the additional factors also introduce more leeway for the competition authority as well as greater legal uncertainty for firms.
A second issue addressed in last year’s competition policy reform has been a clarification with respect to market definition. The fact that some services are not offered in exchange for money in itself no longer prevents competition authorities from defining a separate market for them. Hence, social networks or Internet search engines may constitute relevant markets not only in advertising, but also for users. Even though this clarification may be helpful for the courts, the practical difficulties in defining markets in the absence of sales and price variations remain. While theoretically an adjusted SSNIP test may be defined (see Filistrucchi 2008; Filistrucchi et al. 2014), it is in practice extremely difficult to measure a percentage decrease in quality or privacy or a percentage increase in data requirements. The practical difficulties in defining markets without sales can be illustrated by the approach the European Commission has taken in the Google shopping case.
On the European Commission’s Google Shopping Case
The European Commission holds the view that “Google has systematically given prominent placement to its own comparison shopping service” and it has “demoted rival comparison shopping services in its search results.” As Google Shopping is much more visible to consumers than typical comparison shopping services, the European Commission is of the view that “Google’s practices amount to an abuse of Google’s dominant position in general internet search by stifling competition in comparison shopping markets.” As will be outlined below, from an empirical perspective it is unclear, however, whether (a) there is a market for general internet search and, if so, who is active in that market and (b) whether Google Shopping is part of a comparison shopping market.
To be more precise, the European Commission assumes that there is a distinct market for comparison shopping services where consumers do not actually shop, but only compare offers. In the Commission’s eyes, Google Shopping, foundem, idealo.de, etc., are active in this particular market, while market places such as Amazon and eBay or online retailers such as Zalando are not.
To put it differently, the Commission assumes that most consumers do not choose between, say, Amazon and Google when they want to compare offers for new sports shoes, electronic consumer goods, or other products. In the Commission’s view, eBay or Zalando are not relevant alternatives for consumers, either, but only true product comparison sites where consumers cannot shop are in the same market. This assumption is, however, not based on any study or evidence of consumer behavior, but on the mere insight that Google Shopping does not offer products for sale, but only links to other webpages that offer products for sale. This also implies that the market definition would change should Google Shopping introduce one-click shopping, or vertically integrate into retailing products itself, or develop into a marketplace in its own. To ignore Amazon, eBay, Zalando, and the like as competitors for Google Shopping without studying actual consumer behavior is risky, at best. If I ask my students, for example, where they start shopping for products, many start immediately at Amazon and it is not implausible that many other consumers also shop this way. Neglecting true consumer behavior, however, is rather problematic.
In addition, the Commission’s analysis also ignores that Google Shopping is very clearly labelled as advertising. Hence, it is questionable whether many consumers expect Google Shopping to be an encompassing comparison shopping site. Moreover, it is even explained that Google is paid for these advertisements, which is by far less clear on other comparison shopping sites. Hence, it is not clear how many consumers really expect a “neutral” listing of Google Shopping results, given that it is labelled as advertising. Put differently, Google Shopping does not portray itself as a neutral metasearch engine, as some competing comparison shopping sites do. Quite possibly advertising platforms (such as Google Shopping) and metasearch engines are regarded as substitutes by consumers, but it still remains speculative without any evidence on consumer behavior.
Similarly, the Commission simply assumes that a market for general search exists. Whether such a distinct product market exists is rather unclear, however. Consumers typically have specific rather than general questions. They look for information on books, people, the weather, sports results, hotels, flights, share prices, restaurants, sports shoes, and so on. Most of this information can be searched for on Google, but it can also be searched at Amazon (books, sports shoes), LinkedIn and Wikipedia (people), specialized weather and sports sites, Booking (hotels), Yelp and Foursquare (restaurants), and so on. While it is true that many of these sites do not directly provide links to third-party webpages (even though many also do), people typically do not search for links, but for information. For almost every specific question that Internet users have there are more options than searching on Google. Taking books as an example, one may assume that Amazon is the world’s leading search engine for books where people want to find new books or information about certain books. Is Amazon, therefore, the dominant book search engine that may not favor its own offers?
Given that Google does not charge users who search and, hence, Google’s price (of zero) on this side of its market does not vary, we know close to nothing about potential consumer responses to potential price increases. This also implies that we do not know whether a market for general search in fact exists at all or whether there are many, many markets for specific searches and it just happens that Google is active in all of these markets. Again, completely neglecting consumer behavior does not strengthen the European Commission’s case.
The Commission has decided to take a normative rather than a positive approach to market definition, which leaves for economics the role of identifying effects that remedies may have, at best.
The Commission has decided to take a normative rather than a positive approach to market definition, which leaves for economics the role of identifying effects that remedies may have, at best.
On the German Facebook Case
Since many platforms do not charge both sides of their market, but only one of them (e.g., advertisers, but not users), new forms of exploitative abuse may (theoretically) emerge, such as demanding “too much data.” Interestingly enough, Germany’s Federal Cartel Office (FCO) is currently conducting an investigation into Facebook’s behavior vis-à-vis its users. More precisely, the FCO is investigating whether Facebook has a dominant position in the market for social networks and whether Facebook’s general terms and conditions are inadequate and constitute an exploitative abuse of market power. While the FCO’s theory of harm has not been laid out in writing yet and is therefore not entirely clear as to the details, it appears that, as a supposedly dominant player, Facebook has responsibilities that go beyond the responsibilities of non-dominant parties with respect to privacy standards and data usage.
Such a requirement, however, may not safeguard but even jeopardize competition in the concerned markets. The reasoning leading to this conclusion can be briefly explained as follows: if one assumes that social network users do not receive disutility from sharing personal data and having data sets combined, collecting and combining data from users can obviously not be an exploitative abuse, as consumers cannot be exploited if they do not mind providing the data that is collected. Put differently, there cannot be any harm inflicted onto users if they do not receive any disutility from having their data combined. On the contrary, as combining data facilitates the development of better matching technologies to rank news and other information to match user interests, the prohibition to do so would lead to a deterioration of the services offered (as the matching technology would deteriorate).
At the same time, Facebook would become less competitive in advertising markets vis-à-vis Google and other market participants. As data is used to develop and offer better services, preventing Facebook from collecting, combining, and using the data is equivalent to requiring Facebook to be less innovative and to offer inferior services—both would harm competition.
In contrast, Facebook users and advertisers tend to benefit from the use and combination of “on-Facebook” and “off-Facebook” data, as the usage and combination of different data sources facilitates the improvement of matching algorithms to rank information and news for users. In addition, it is difficult to conceive how users can be exploited by using their data as their data resources are not depleted when used. Hence, any analogy with data as a form of money or payment is misleading, as monetary resources cannot be used multiple times. Finally, empirical evidence suggests that (most) people do not feel exploited when their data is used. Quite in contrast, people tend to willingly share data in order to obtain benefits such as improved services.
If, however, we assume that sufficiently many consumers do receive disutility from data being combined, requiring Facebook to use or combine less data or only data from certain sources and to offer higher privacy standards than competitors would be equivalent to requesting Facebook by law to offer superior products than rivals (and, in the extreme case, to foreclose the market), which would also harm competition.
Hence, requiring Facebook to use or combine less data or only data from certain sources would stifle competition. Either Facebook would be required to become less innovative and to deteriorate their service (in the likely case that most users do not receive disutility from having data sets combined) or, alternatively, Facebook would be required to outperform its rivals (if most users did receive disutility from having data sets combined).
While privacy issues may need to be newly addressed in digital markets, antitrust laws do not appear to be the most effective instrument for safeguarding privacy.
Data Portability and Data Access as a Salutary Remedy?
A remedy regularly proposed to facilitate competition in digital markets is regulated access to data resources that dominant—or even all—firms use (see, e.g., Argenton & Prüfer 2012). While there are obvious privacy issues, access may be granted to anonymized or pseudonomized data. It is unclear, however, whether access to a dominant firm’s anonymized or pseudonomized data can really mitigate major competition problems.
To provide an example, consider the case of Google search where access to Google’s historical search and click data has been proposed as a promising remedy. Since Google search is becoming more and more personalized, access to pseudonomized data is of limited usefulness. Google’s search results are, among other things, very good because many people also use Google calendar, Gmail, Google Maps, etc. Given that Google can use personal information from my mails, my calendar and so on, it can produce very good, personalized search results. Hence, it is access to very personal, non-anonymous data that is decisive for obtaining ideal search results, while anonymous data is of limited usefulness. Forcing Google or Google users, however, to also provide private data to competitors would conflict with privacy concerns. Similarly, access to anonymized or pseudonomized data from social networks is likely to be of limited usefulness for competitors. Simply assuming that competitors can offer services of equal quality by accessing a dominant’s firm data in pseudonomized form may only yield limited insights.
What may be more helpful is empowering users to voluntarily port their data themselves to competing service providers. One should note, however, that difficult questions remain for data that is generated through interaction (as, for example, in social networks) so that individual rights of more than one party are concerned.
The concern that digital platforms may become dominant in certain markets is clearly legitimate. How to react and which remedies to impose is a much more difficult question, though. A number of proposals for fiercer platform regulation would also limit competition in the concerned markets such as retailing, transport, accommodation, etc., and should, accordingly, be digested with caution.
In addition, competition authorities sometimes even over-enforce nondiscrimination rules, as, for example, in most so-called dual pricing cases (see Haucap & Stühmeier 2016), thereby even promoting market concentration.
Justus Haucap is the director of the Duesseldorf Institute for Competition Economics (DICE) at Heinrich-Heine University of Duesseldorf.
Argenton, C. & J. Prüfer (2012), Search Engine Competition with Network Externalities, Journal of Competition Law and Economics 8, 73-105.
Bagwell, K. (2007): The Economic Analysis of Advertising, in: Mark Armstrong and Robert Porter (Eds.): Handbook of Industrial Organization 3, 1701-1844, Elsevier, Amsterdam.
Becker, G.S. & K.M. Murphy (1993), A Simple Theory of Advertising as a Good, Quarterly Journal of Economics, 108, 941-964.
Caillaud, B. & B. Jullien (2003), Chicken & Egg: Competition among Intermediation Service Providers, Rand Journal of Economics, Vol. 34, 309-328.
Evans, D. S. & R. Schmalensee (2008), Markets with Two-Sided Platforms, Issues in Competition Law and Policy, Vol. 1, 667-693.
Evans, David S. & Richard Schmalensee (2015), The Antitrust Analysis of Multi-Sided-Platform Businesses, in: R. Blair and D. Sokol (eds.), Oxford Handbook on International Antitrust Economics, Vol. 1, Oxford University Press, 404-449.
Filistrucchi, Lapo (2008), A SSNIP Test for Two-Sided Markets: The Case of Media, NET Institute Working Paper No. 2008-34.
Filistrucchi, L., D. Geradin, E. van Damme & P. Affeldt (2014), Market Definition in Two-Sided Markets: Theory and Practice, Journal of Competition Law and Economics 10, 293-339.
Jullien, B. (2006), Two-Sided Markets and Electronic Intermediaries. In: G. Illing and M. Peitz (Eds.) Industrial Organization and the Digital Economy, MIT-Press: Cambridge, 272-303.
Haucap, J. & U. Heimeshoff (2014), Google, Facebook, Amazon, eBay: Is the Internet Driving Competition or Market Monopolization?, International Economics and Economic Policy, 11, 49-61.
Haucap, J. & T. Stühmeier (2015): Competition and Antitrust in Internet Markets, in: J. Bauer and M. Latzer (eds.), Handbook on the Economics of the Internet, Edward Elgar: Cheltenham, 183-210.
Disclaimer: The ProMarket blog is dedicated to discussing how competition tends to be subverted by special interests. The posts represent the opinions of their writers, not necessarily those of the University of Chicago, the Booth School of Business, or its faculty. For more information, please visit ProMarket Blog Policy.
- The capacity on one side of the market may be more limited than on the other. For example, the number of stands may be more limited in a trade show than the space for potential visitors.[↩]