Big Data and Competition Policy

Size: px
Start display at page:

Download "Big Data and Competition Policy"

Transcription

1 Big Data and Competition Policy Alexandre de Streel University of Namur, CRIDS, CERRE 12 th Annual Conference of the Global Competition Law Centre Brussels, 26 January

2 OUTLINE 1. Big data value chain 2. Market power in Big data 3. Abuse of dominant position with big data 4. Data merger 5. Take Away 2

3 1. Big data 3 main parts in the value chain Data collection: direct or indirect (data brokerage) Data storage Data analysis of very large datasets: to reveal patterns of information not visible from smaller datasets Not new but 4Vs: Volume, Velocity, Variety, Value Rely on personal and non-personal data There is no direct link between the personality of the data (and its regulation) and its economic value 3

4 4

5 Public intervention: Be cautious Data economy has positive welfare effects Innovation and efficiency Better products and better process McAfee: companies that make the most of their data are 5% more productive and 6% more profitable than their competitors McKinsey: possible savings of up to 300 billion a year in the public services in the EU Free services Customer data are monetized on the other side of the two-sided market Payment with data, now recognised explicitly by EU law More targeted products and dynamic pricing Reduce information costs Welfare effect of perfect discrimination 5

6 2. Market Power in data collection Data is non-rivalrous, short-lived, ubiquitous, inexpensive and easy to collect BUT data collection may be limited by contractual restriction and/or hard to get (e.g. health data) Data exchange is less prevalent in EU than in the US, partly because of more stringent data protection rules Replicability analysis depends on the type and the use of the data Data on users preferences are often a by-product of another services 6

7 Feedback loops between data collection and data analysis User Feedback loops (Lerner, 2014) Monetisation feedback loop 7

8 Market Power in data processing Volume of data: economies of scale Variety of data: economies of scope Velocity of data: economies of time Experience curve may be very steep esp. with artificial intelligence and deep learning algorithms 8

9 3. Exploitative Abuses Excessively low privacy protection (German Facebook case) Facebook could be dominant on the market for social networks Facebook privacy terms of use can violate data protection rules and that could represent an abusive imposition of unfair conditions Perfect price discrimination Complex and multiple welfare effects Can increase total welfare when the market expansion effect is bigger than the firm appropriate effect Transfer between class of customers 9

10 Exclusionary Abuses Discrimination in the algorithm against competitors and leverage Google shopping case What could be the remedy? Transparency, public audit of algorithms Leverage of data base (e.g. customer list) French GDF Suez Case in 2014: Obligation of sharing data with competitors Belgian lottery case in 2015: Prohibition of data cross-subsidisation + Leverage from data-driven markets to non data-driven markets (Prufer and Schottmuller, 2016) Google conglomerate strategy: Search engines maps self-driving cars Mandatory sharing of raw data on user preferences 10

11 4. Merger: Easy Cases: data as a market Thomson/Reuters (2008): approved in phase II with remedies Horizontal effects on markets for some financial information: earning estimates, fundamental data of enterprises, time series of economic data Compulsory data sharing: divest copies of the database 11

12 More difficult Cases: data as an asset/input Google/DoubleClick (2008): approved in phase II without remedies Combination of data on search and web-browsing is replicable, data are available elsewhere Microsoft/Yahoo (2010): approved in phase I without remedies Merger will allow a more rapid improvement of MS search engine to compete with Google Facebook/WhatsApp (2014): approved in phase I without remedies Online advertising: Many alternative and valuable user data are not within Facebook exclusive control Facebook fallacy of defining static market Microsoft/LinkedIn (2016): approved in phase I with remedies Online advertising: large amount of data continue to be available and merger don t reduce the amount of data available to third-party 12

13 5. Take Away Identify the possible sources of market power in the big data value chain Abuses Possibly more exploitative abuses Exclusionary abuses focused on discrimination and leverage Merger Data as a market Data as an asset/input: Fallacy of market definition, Data as an infrastructure for innovation Competition law is complement, and not substitute, to other legal instruments (such as data protection or consumer protection laws) 13