Thread by @GregorySCrawfor

A (long) thread complementing those by @Caffar3Cristina and @johnnyryan encouraging competition economists to make theories of harm based on (lack of) privacy and data protection core features of antitrust investigations from our @VoxEU article last week. voxeu.org/content/antitr…

Side note: my views have been strongly influenced by the fight alongside @Caffar3Cristina & others to have the types of harms we identify in the piece be taken seriously in the Google-Fitbit merger. The specifics of our opposition here:

https://twitter.com/GregorySCrawfor/status/1341332072879800321

2/

The focus is mostly digital platform deals and conduct. DPs are well understood by this point: they offer something consumers like (search results, email, maps, etc.) “for free.” This is the first bit of bad framing that limits effective antitrust around data. 3/

No money may change hands, but of course platforms do get paid – with our data. Usually some combination of demographic, location, and/or interest data, which they then (mostly) monetize (so far) in online advertising markets. 4/

As such, it’s time to call a firm’s policies regarding the data they collect, how they combine it with other data, how they use it, and how they sell it what it is: the *price* of using digital platform services. 5/

Framing (lack of) privacy and/or data protection as a price is far more useful than the more typical economic approach of thinking of it as a quality attribute. 6/

Why? Because in standard economics the relationship between quality and market power is generally ambiguous (e.g. see our intro here: aeaweb.org/articles?id=10…, with @MatthewShum2 & Alex Shcherbakov). 7/

But there is little ambiguity about the direction of privacy protection (down) and data collection practices (up) as ad-funded digital platforms (G&F) have become dominant (see, e.g., @DinaSrinivasan here: papers.ssrn.com/sol3/papers.cf…). 8/

This is no surprise: we pay with data and prices rise with market power (see the UK’s CMA’s impressive online advertising market study). gov.uk/cma-cases/onli… 9/

Further complicating matters is that consumers often don’t know how their data are being used and therefore don’t even know the prices they pay (e.g. Appendix G in the CMA’s online advertising market study linked to above). 10/

The literature on “unobserved prices” finds that this can further reduce welfare as competition won’t improve outcomes on attributes consumers can’t see (See Paul Heidhues and Botond Koszegi’s survey in the Handbook of Behavioral Econ sciencedirect.com/science/articl…). 11/

Seeing (lack of) privacy and data protection as a price means mergers like Google-Fitbit that combine data are particularly bad. Not only is this an ex-post price increase for Fitbit users (whose health/wellness data can now be monetized by Google)… 12/

…..but also it is an ex-post price increase for users of Google’s services (whose demographic/location/interest data can now be monetized in health markets). 13/

Who knew that this is what you were signing up for when you chose to use Gmail, download Chrome, or bought an Android phone (or bought a Fitbit)??? 14/

It’s time to treat data collection and use as the price increases that they are and make them a core part of the harms to be prevented in investigations of deals or dominant firm conduct. How? 15/

First, learn what data is collected and how it used/sold now and, if part of a deal, how data sources will be combined in future. Internal documents/investor presentations/similar can be helpful here. 16/

Second, measure consumer tastes for these uses (e.g. privpapers.ssrn.com/sol3/papers.cf…). Yes, survey methods have their flaws, but conventional methods don’t work when prices are unobserved. 17/

These are the prices/price changes from the conduct/transaction. If enforcers/their economists are willing to push beyond the conventional wisdom, they can serve as a basis for merger conditions involving purpose limitations or a block. 18/

A second thread that we argue competition economists should pick up is that lack of privacy and data protection can facilitate exploitation and foreclosure. 19/

Data often creates markets with increasing returns to scale, causing them to tip to a dominant firm. It also allows personalized offers, and if paired with dominance this creates discriminatory power and thus the potential for exploitation. 20/

Data is also general-purpose input, inducing complementarities that can be exploited by all the regular “tricks” of dominant firms seeking to extend their market power and foreclose rivals (e.g. tying, bundling, restrictive contracting, etc.). 21/

In Google-Fitbit, both exploitation and foreclosure concerns were present. Because Google is dominant in non-health data, allowing them to combine it with health and wellness data implies a significant risk of exploitation in insurance and other “health-tech” markets. 22/

Similarly, @ProfFionasm and @DavidDinielli (from p25) argue Facebook foreclosed competitors by deceptively tracking users while claiming in its privacy policies that it wasn’t (based on @DinaSrinivasan's work above). 23/ omidyar.com/news/roadmap-f…

Motivated by Google-Fitbit, Chen, Choo, Cong, and Matsushima (2020) analyze “data-driven mergers,” showing that personalization can lead to exploitation and the leveraging of market power.
ideas.repec.org/p/dpr/wpaper/1…. 24/

Indeed, there is a growing academic literature on harms from data, e.g. “privacy policy tying,” “privacy externalities,” and more; it’s time for competition economists to take these on board and also make them a central part of investigations.
papers.ssrn.com/sol3/papers.cf…
25/

Privacy experts have been way ahead of competition economists on consumer harms from digital platform data collection and exploitation; it’s time we listened and worked to bring their concerns into our models and data-related cases. END/