Etienne Pfister, "Data, Competition and Merger Control in the Digital Age"

Presented at the Legal Challenges of the Data Economy conference, March 22, 2019.


ETIENNE PFISTER: Thank you. And thanks to the organizers for inviting me to speak about data and competition. The organizer told me the same thing as they told Randall. They told me, they asked me to make a sexy title, and it looks like Randall and I have the same notion about what sexy is. But at least a common point. Crossing the American-Europe divide.

AUDIENCE: Apparently not.

ETIENNE PFISTER: Anyway, so I'm the chief economist of the French competition authority, and I will be speaking on my own behalf only. And as the title says, I'm going to speak only about data-related mergers. I'm not really going to talk about antitrust except maybe at the end if I have the time.

We have several merger decisions when we are dealing with data-related issues. The names are on the slides, and you all, you have heard about them. Some of them are quite a bit old now, like Google-DoubleClick that was in 2008. We have one last one, of our most recent one is Apple-Shazam, which was last year. And we have [INAUDIBLE] in France, which was also last year.

We're going to talk about the issues that have been raised in these decisions and are related to data. We have two issues there. The first one is what I have called the data advantage scenario. I'm going to describe it a little bit later on. The second one is the data increase scenario.

And keep in mind that I'm only going to talk about data-related subjects. These decisions raise many other points, but I'm only going to talk about the data-related ones-- first caveat. The second caveat is that due to time constraints, I'm going to oversimplify the reasoning. Bear this in mind also.

If you look at the decisions, they are all based on several arguments-- not on a single one, but on several arguments. And it's a compilation of these arguments that leads to the conclusion. What's interesting is that in all these decisions, data-- all of the scenarios had been identified-- data has never been rated as a competitive risk once all the facts have been carefully looked at.

So the main point is to say why? Why is it that when we're thinking about these operations, we all think about data risk. And why is it that? Well, the risk didn't turn out to be so high, at least at the time of examination. And when, if time allows, we will weigh some possible discussion points.

So first issue, what I have called the data advantage scenario. It's fairly simple. We have three companies, and they produce services. And all of them rely on some data. OK? Two of these companies merge, and when they merge, they may also merge their data set. And thanks to this merging of data sets, they're going to be-- they're going to have more data. They're going to produce more services or more high-quality services, like targeted ads for instance.

And of course, because they become more efficient thanks to this data, they grow. And not only do they grow, but their competitors, due for instance to network effects, their competitor will see that its market share is decreasing. Because compared to AB, the new entity, it gets more and more, less and less attractive.

But also, if you think about entry barriers, now in order to compete with AB, you need much more data than before. And this is an entry barrier that has been created thanks to the merger. So that's the first setting, very simple.

Another one, which is as simple but slightly different is that for instance, you have a company A which owns some data, and you have another which has no data at all. And this company B is not competing with A. B is only competing with C. When we have a merger between A and B, that's what we may call a vertical merger or conglomerate merger because A and B are not competing with each other.

Then we have the same story as before. B we'll get A's data, and thanks to this data, the new entity AB is going to get bigger and bigger, more efficient, more consumers, more data, more consumers, and so on. And C, of course, is going to go smaller. OK. So this is a basic scenario. It looks very simple, and probably this is what we all have in mind when we're thinking about some data-related mergers.

So now let's have a look at some real cases and what has happened. So this is the case with the Microsoft merger, well or the exclusive contract between Microsoft Bing and Yahoo regarding the search engines. Of course, we have a data advantage story there. Microsoft will get the data from Yahoo, and thus they will get more data and they're going to be more efficient. So we have clearly the data advantage scenario.

Is that a problem? Not really, because Microsoft-Yahoo is competing with a much larger competitor, which is Google, which has much more data. So the data advantage risk can be turned around. It's a data advantage opportunity. And this is what the commission has taken into account when it finally decided not to prohibit this merger. So that's fairly simple.

Another case. Let us assume that we have a company A which has some data. It is not competing with company B, which also has some data. B is competing with C. A and B merge as usual, and of course when we have the fear that A is going to transfer its data to B, and B will get bigger while C will get smaller. That's the data advantage risk. Is that so simple? Well, are the data owned by A so useful in order to sell B's products? This may not be the case.

If you look at the Apple-Shazam decision, this was one of the risks the commission, the European Commission studied. When Apple bought Shazam, were the data owned by Shazam useful to send music stream? The European Commission asked many operators about the potential risk, and the conclusion was there is maybe a very, very, very, very small advantage but not significant. And so we don't have a data advantage risk, because of the accumulation of data of A's data in B will not produce a substantial advantage. So the market structure will stay the same.

Another data advantage scenario. Let's look at WhatsApp-Facebook for instance, and that decision. So we have Facebook which buys WhatsApp, and, of course, the risk what is it? Facebook will grab hold of WhatsApp data. Is that so simple? The commission showed at that time, no. Why so?

First, it said Facebook wouldn't be allowed to get hold of WhatsApp data because of a GDPR. Most people who use Facebook have only given their consent to use data for WhatsApp, not for Facebook. OK. We have seen earlier that things are a bit more ambiguous once we take Facebook behavior into account.

Another problem is that at that time in 2014, the commission forward that WhatsApp didn't have enough market power to increase its data and to send it to Apple. One of the fact that was noted by the commission at that time was that as soon as Apple announced that it was buying WhatsApp, WhatsApp lost 20% of its subscribers for fear of data advances.

And another problem is technical issues. We think that merging data is very, very easy. This may not be the case. At least, this is what Apple told the commission. This is what Facebook told the commission at that time. He said, I wouldn't be able to merge the data set of WhatsApp on Facebook. And two years later, the commission realized that this was a lie, because at that time Facebook was already able to merge this data, and Facebook has been condemned for this lie.

There are many other reasons why the merger between Facebook and WhatsApp has been authorized. It was not only this technical issue. But same kind of reasoning is used in Google-DoubleClick and in Microsoft in [INAUDIBLE] but I'm not going to divulge it.

Let's move on. Another obstacle to the data advantage scenario. So we have Facebook and WhatsApp, and we have many other platforms, communication services, social network services, and so they're merging together. So we may fear that it's going to be a huge data advantage, but what the commission says is no way. There won't be any data advantage because there is so much data everywhere available for everyone, not only for Facebook and WhatsApp. But all other kind of platforms can use these data through third parties, and so on.

So this is one peculiarity of the digital world. Data is everywhere, so it's very hard to get a data advantage when there is so much data everywhere. That's what the commission thought at least in 2014.

And so the market structure is not changed, and the operation is authorized. Again, this was one argument on top of several others. Also see Google-DoubleClick and Verizon-Yahoo, same reasoning. You can see the slides. Google-DoubleClick for instance, a combination of data about searches. Google data. We have data but uses web surfing behavior. DoubleClick data is already available to a number of Google's competitors today. Which competitors? Microsoft, Yahoo, third parties, internet service providers, and so on.

Let us come to the SeLoger-Logic Immo decision. SeLoger-Logic Immo are real estate portals for non French audience. These are portals on which real estate agencies go to publish their ads for the goods for houses, for flats that they sell. And this is the portals where consumers go when they want to rent or to buy a house.

They have some competitors like LeBon Coin, Bienici, and several others. And they all have end consumers, internet viewers, and they also have real estate agencies. Of course, when SeLoger buys Logic Immo, we have a data advantage story in mind. SeLoger-Logic Immo will get more data so they will have an advantage over their competitors, and they will get bigger. And because we are on a market with network effects, this will be bad for competition. But again, not so simple. Why?

First, end consumers and estate agencies, they use on average four real estate portals, our. Which means that when SeLoger has some data about a particular consumer, there are free other portals on average that have the same data. And this is true for end consumers, and it is true for real estate agencies. First issue.

Second issue. Because end consumers use several real estate portals, when SeLoger buys Logic Immo, will it get much more data? No. Because most of the consumers of Logic Immo, they already use SeLoger. So SeLoger already has the data.

And finally, when you look at data, what is the relevant metrics of data? It's not marketing share and value. It would rather be market share in volumes, number of consumers, number of ads. Who is the stronger when you look at the number of ads or number of viewers? It is LeBon Coin. Much bigger than the new company of SeLoger-Logic Immo. And also Bienici is probably going to be much bigger because Bienici is a subsidiary of all the real estate agencies, so the real estate agencies are going to give possibly much more data to Bienci than to other portals.

OK time is running, so I may skip the Microsoft-LinkedIn but it's kind of the same story as before. So that was the first scenario about data advantage. The important thing to see is that a scenario has been spotted, and that's the first part of a story. The second part is, is this scenario realistic given all the facts that we know about the market.

Second scenario is a data increase scenario. We are more into data collection here. It's like a price increase. You know, what is the fear when two companies merge? The fear is that price is going to increase. If we are on a market on which for services free, the fear is that data collection is going to increase. Is this fear justified?

Let us see a fairly simple sketch here. We have three companies, A, B, and C. They are competing. And we have consumers and customers. Each of these companies collect some data. We may think that the amount of data collection is the result of an equilibrium. Why?

Because if A chooses to collect more data, the risk is that some consumers or customers are going to move. They're going to move to B, or they're going to move to C. And so, of course, has a trade-off. I want to collect more data, but if I do so, I'm going to lose some consumers. So there is a trade-off to be made.

Now let us assume that we have a merger between A and B. What's going to happen? Well, A will be able to collect more data than before. Because if a consumer goes to B, if a consumer is unhappy with the increased data collection by A, this consumer will go to B. But this consumer is no longer lost for A, because A and B are the same company. So this is a risk of increased data collection.

But again, the scenario is not that simple. Because first, when you look at all the past cases that we had, you have only few cases where firms are really competing one with the other. Most of the times, we have vertical mergers or we have conglomerate mergers. So they're not direct competitors, so this ethic does not come into play.

The second is that we have a GDPR and business sets a limit to data collection. And the third limit is that as Randall said, consumers not much informed about data collection. What does that mean? It means that actually, many companies, they already set the amount of data collection to the top, and it does not matter whether they acquire B or not. They're already at the top, because the consumer either does not care or is not informed.

The more tricky scenario is that one. We have two companies, A and B. They're not competing one with the other. And A is competing with C, and B is competing with D. B does not collect any data, but it is bought out but by A. And A needs some data. So possibly what A will do is that A will instruct B into collecting data, and then A will grab the data.

This is something that happened with Facebook-WhatsApp. This is something that may have happened with Facebook-WhatsApp. When WhatsApp was bought, it was not collecting any data because it has a very strict privacy policy. Its model was not about selling ads and so on. Its model was something else. No one knew what the model was at that time, actually.

But once it had been bought by Apple, then the data transfer find motivation. But at that time, WhatsApp was only a very, very small market player. And as I said, the commission thought that many consumers are going to flee away from WhatsApp if they see that some information is going to be taken.

I have some time left for possible discussion. Yeah.

When we think about merger control related to data, there are two things that can be difficult. There won't be any slides here. I'm going to shorten my discussion because of time constraints. The first one is that these are markets with zero sales. Sometimes you target a company. It has no sales at all because it sells free services. It is not present on the advertising market yet. So its turnover can be either zero or it can be very low. It can be below the notification thresholds.

For instance, in the last 17 years, Google has bought something like 270 companies but only one of these acquisitions was notified to the European Commission. Fortunately, there is a simple solution for this. And many countries are working towards it. It's to lower the notification threshold, or it's to change the definition of a threshold. Instead of using sales, use the transaction value.

And a third solution is to use ex-post control. That is, I do not intervene yet, but I have a certain time to intervene once the merger has been done in order to disentangle the merger if I see that there are some anti-competitive effects. So that's the first difficulty. But I think that there are some solutions.

A second difficulty when we are dealing about merger control in these markets. The second difficulty is that these markets are nascent, recent markets, and they keep evolving all the time. And when we are in merger control, we have to think ahead. We have to think what will be the outcome equilibrium in five years time? What will be the conduct of companies in five years time? When we see that these markets are changing so rapidly, this exercise can be really difficult.

And this is combined with another difficulty is that if you look at merger control. In order to prohibit a merger, you have a standard of proof to respect. You have to bring some evidence that the risk of an anti-competitive conduct is high. And if you are much uncertainty, first demonstration is much harder to make.

This is one of the questions about which many economists and lawyers are thinking about. I'm thinking about Tommaso Valletti, the chief economist from the European Commission. I am thinking about the Furman report to which Randall also referred to early on. I'm thinking also about the [INAUDIBLE] report on digital conglomerates.

What these reports say is that maybe when we are looking at acquisition made by super dominant companies, maybe we should think about not only taking for probabilities of having an anti-competitive result, but also think about the cost of making the bad decision. What is the more costly? Is it to prohibit Google or Facebook from buying a company, or is it let Google buy a company and see that competition is reduced even though the probability may be low?

Because we know that there is a risk of mistake, but if we make that mistake the cost can be very important. This is one of the reports which is made sometimes towards the Facebook-WhatsApp decision. Because at that time in 2014, WhatsApp didn't really look like a competitor at least on the advertising market. And possibly it may have been a competitor five years later. We will never know. But possibly the cost of prohibiting is too high. But again, this means shifting a little bit, modifying the burden of proof.

I think I'm going to make just one last point if I have one minute. I'm talking more about general antitrust. It is often said that antitrust comes sometimes too late. I don't think that is so true if you look at all the recent decisions by the commission. Of course, we have the Google Shopping decision which has been waited for quite a long time, but this was because of an exceptional circumstance which was a failed commitment procedures. If you look at the other cases, Google cases, they took a fairly short amount of time. That's the first thing.

And the second thing, if we are reasonably certain that there is a competitive risk with a given conduct, we have interim measures that suspend the conduct. And this has been used recently by the French competition authority no later than two months ago against Google against the refinancing criteria by Google. Thank you very much.

AUDIENCE: Thank you.

Big data