“Academic engagement” of a Big Tech firm: that job should exist. But it has not happened, and it probably won’t happen, unless we get a new mindset among regulators and researchers. Tommaso Valletti explains why.
This piece started with a joke. April 1 was a school holiday in the UK, and I had promised to take my kids for a bicycle ride in the morning. As I was waiting for them to get ready, I sat lazily on the couch reading the latest news. I had already been subjected to their pranks since they had woken up, so I decided to write the following tweet for April Fool’s day:
Some *personal news* – after many amazing years at @imperialcollege I’m leaving to take up a *joint* position at #Facebook and #Google as the new Global Head of Academic Engagement and Research for the Common Good.— Tommaso Valletti (@TomValletti) April 1, 2021
Then we left for our ride. Everyone that knows me, my vocal history of being against the power of Big Tech, and hopefully my sense of humor, knew that it was a joke, and I was expecting a few chuckles. By the time we got back a few hours later, the tweet had gone a bit viral (ah, the power of platforms!). Many people did not get it—perhaps a testimony that it was a good joke. I had received several requests from journalists for an interview, strangers had emailed me offering their services (and I apologize for raising their expectations), and angry followers were upset at me for having sold my soul for corporate money. Even friends wrote or called, fantasizing about amazing research questions they could answer using data from Facebook or Google.
In fact, I realized that in tweet that I had accidentally thrown out, there were several things to unpack. “Academic engagement” of a Big Tech firm: that job could exist, that job should exist. But it has not happened, and probably it won’t happen.
I am talking only about my field, economics, and competition economics in particular. In the past 20 years, we have witnessed the incredible success of Google, Amazon, Facebook, Apple, and Microsoft (GAFAM). They are the top companies in the world by market capitalization. They affect every aspect of our lives. They are worth several trillion dollars collectively and sit on an unprecedented trove of data that could potentially allow researchers and policymakers to shed light on extremely relevant questions, and in a very detailed fashion. Given that empirical work is the gold standard for economic studies, one would expect a large interest from economists worldwide. You’d expect a lot of “academic engagement,” and useful interactions between academia and those corporations, producing a whole body of scholarly research that would advance our knowledge. Instead, there are almost zero published papers in the competition arena using the relevant data from these platforms.
Imagine how many questions related to digital platforms are out there. We have understood a few aspects in theory: network effects, multi-sidedness with “free” services on one side, the significance of data for the business model, scale and scope economies, and so on. The missing but crucial piece is quantification: Can we quantify these effects, and measure their practical importance? Or what about this question: When are digital markets going to tip, or when does a nascent entrant become a competitive threat? Or Senator Elizabeth Warren’s (D-MA) “Umpire and Player” story, which is relevant to Apple’s relationship with App developers as well as to Amazon’s relationship with sellers: When does integration imply that competition and consumers are harmed? The papers exploring these empirical questions should have already been written a long time ago, but it has not happened.
Giant tech companies have too much power. My plan to #BreakUpBigTech prevents corporations like Amazon from knocking out the rest of the competition. You can be an umpire, or you can be a player—but you can’t be both. #WarrenTownHall pic.twitter.com/73y1002QVv— Elizabeth Warren (@ewarren) April 23, 2019
It’s not that GAFAM are not interested in economics. On the contrary, some Big Tech companies have large internal economics departments and hire many PhD economists, whom they use for business optimization. Very little of the work or activities of these groups is generally visible to the outside world. I am sure they actually produce a lot of potentially interesting results: Big Tech companies engage in continuous A/B experimental testing to uncover behavioral biases and adjust their strategy to what they learn. But we never get to learn what they see. Try to look at the programs of major conferences, or the publications of articles: You won’t find employees of GAFAM presenting original research. Microsoft Research is an exception, as they have had for many years a well-regarded “blue-sky thinking” group of economists and data scientists. But again, there are no published empirical papers in competition economics over recent years.
Instead, these companies have well-known chief economists or public affairs staff who are regularly present at roundtables—but their objective is to direct public discourse, not to promote or conduct unbiased research. GAFAM actually do give money to some researchers, some departments, and even more so to “think tanks” as I have previously written in ProMarket (often without a proper disclosure). But they never give data, which is the most valuable thing to an empirical researcher. Their funding is basically money to buy influence.
A few incredibly successful academics in the profession, such as Susan Athey, Pat Bajari, Steve Tadelis, or Hal Varian even managed to have positions with these companies, and in a couple of instances went back to academia after what must have been a tremendously interesting experience. But these are superstars, exceptions we should not confuse with the average (talented) economists entertaining a career with GAFAM.
This is not an academic curiosity, but it is the intentional outcome of data access choices made by the large digital players. The lack of access also implies that there is de facto no system of revolving doors between these companies and academia when it comes to young economists: if you are a young scholar and cannot publish, you are not going to come back to academia after an experience at a GAFAM company.1 Possibly even worse, generations of bright academic economists have avoided investigating the field empirically altogether, as they anticipated that their efforts would not lead to publishable work. Because of this, academics have stopped asking relevant questions about GAFAM for too long, and the external understanding of digital issues has not much advanced empirically, compared to the internal knowledge of GAFAM that are therefore several steps ahead of anyone else, including the regulators. This lack of academic scrutiny also plagues the interpretation of the very few available studies coming from cases or other data sources, so the debate remains at some vague abstract level, missing all the necessary empirical details.
Courageous researchers have tried to compensate for the lack of direct access to data in various ways, scraping Amazon’s data from the web, finding evidence of Amazon “steering” demand to its advantage, or devising experiments and getting some data from Google’s competitors, arguing that Google’s success does not seem to be matched by good reviews. Companies respond in various ways. Often, they say that these proxies are not good enough and results are “flawed” and should thus be dismissed, as “their” internal data and experiments show different results. But of course, we cannot see those; perhaps the enforcers can, but they have different problems in terms of resources and time pressure. In some cases, they even intimidate the researchers, which is what happened with a team at NYU that was studying targeted political advertising on Facebook, using data scraped from the web. Facebook sent their lawyers telling the team to shut down the project.
Notice the irony here: Facebook went publicly after a group of researchers that teamed up with Facebook users to hold Facebook to account on its own promises to fight misinformation, and Facebook did it in the name of protecting users’ privacy. This is the same company that decided to hide as long as possible, and not to prevent, the very same scraping that was at the basis of the 2015 Cambridge Analytica scandal (involving 80 million users) as well as the more recent leakage of 533 million users’ personal data and phone numbers.
Even when some data have been made available by other digital companies (e.g., Uber) the objectivity of some of the output is sometimes questioned: companies grant data for studies that tend to enhance their public image, so that academic research inadvertently becomes part of the platforms’ lobbying effort, a phenomenon that Luigi Zingales labelled academic capture.
23 years ago, economists Carl Shapiro and Hal Varian (who then became Google’s chief economist for the next two decades) published a book called Information Rules that became quite influential in Business Schools teaching the new economy. It had the right ingredients: network effects and positive feedbacks, multi-sided markets, lock-ins, competition for the market as opposed to competition in the market. A quarter of a century later, we are still repeating pretty much the same anecdotes, highlighting the same trade-offs—in theory—with little advancement as to their practical relevance. We have to acknowledge that our understanding has progressed little since then, especially when compared to the phenomenal success of digital platforms. We have been collectively exposed to the power and consequences of Big Tech. Several antitrust investigations have been launched in Europe and are finally underway in the US (after two decades of inaction). Important reports have come out highlighting the various questions and possible solutions that policymakers are thinking about. But we have to admit that little has come out from academic economists.
Possible Ways Forward
It won’t be easy to change this but I have a few suggestions below. Perhaps the biggest obstacle is to shift the priors of many economists and get a revamped interest in the topic. The digital environment—due to inherently strong network effects, high complexity, and low transparency—has quickly become a highly concentrated institution where a few big players have monopolized its core sectors. Having worked on antitrust cases myself for three years when I was chief competition economist of the European Commission, most of my priors have changed. People—by and large—don’t search beyond the first or second search result, they don’t change the defaults on their devices, they rarely multihome. But how can I convince my fellow economists that our mental defaults should change too, if I cannot show them the data?
At this stage one might ask: Companies are out there to make money, so why should they care about academic research? Aren’t all companies doing the same? Not really. First, the GAFAM continue with their narrative that their goal is to make a better world, and, given their size, the world should know what they do and the impact they produce on society. They have obligations that other companies do not have. But even if you do not accept the moral imperative, in other fairly big industries it is possible to conduct empirical research. Think of airlines, pharmaceuticals, telecommunications, or supermarkets, to name just a few. Datasets are available to analyze their competitive landscape. In many cases, regulators have stepped in, but also private initiatives exist to collect meaningful datasets where any individual data are anonymized following established protocols. Why should Big Tech be any different?
A cynic might say, “Nothing can be done, there’s just too much money at stake, and it’s too late to rein in the power of Big Tech.” This is quite close to the truth, and data is the very origin of their power, but I would hope we can actually move forward in different ways.
One is through enlightened leadership. Perhaps Jeff Bezos, or Sundar Pichai, or other CEOs could understand the value of engagement with unbiased academics, the good public image they could achieve, and push for it. Some of the Big Tech founders were idealists. Google founders Sergey Brin and Larry Page actually came directly from an academic environment. Perhaps the current leaders could revisit their priors as to whether the world today is in a better place compared to 10 years ago. But company lawyers (or shareholders) won’t allow it, precisely because there’s just too much money at stake. True academic engagement simply does not pass a cost-benefit analysis for each individual platform, because they only see the risk side. Sure, expect lots of “policy” engagement, but nothing more than this. I would not count on it.
So, what else can be done? Research is badly needed. The implications of digital systems for socio and economic outcomes cannot be grasped by a single discipline, be it economics, sociology, psychology, computer science, or data science. The ramifications of digital systems for the different facets of modern society call for truly multi-disciplinary research analysis. Economics tends to suffer from a silo mentality that has to change in order to be more open to contributions from other fields. There’s hope that this could happen, with some increasing evidence that the extramural influence of economics scholarship has been accelerated by the diversity of research. We need to help academic economists get out of their towers and translate policy questions into workable research questions. We also need the editors of journals to be open-minded, possibly including weaker data standards in order to inform first-order policy questions as the matters are so big and the proper data just does not (yet) exist. They also need to be much faster in their editorial decisions. Or else our findings, which are already rare, will come out simply too late to have any practical impact.
What about a market-based solution? I mentioned above some big industries where data are available. In many cases, data exist because of private data providers that enable academic research. And they exist also because the players in the industry themselves want the data, as they have access to their own but not those of their competitors. Those are fairly concentrated industries, but not as concentrated as Big Tech. But would you need access to third-party data when you already are the market? The answer is no. Dominance comes with many consequences, and this is another manifestation of it.
What ultimately remains is the regulatory solution. Authorities can oblige the release of data when companies reach high levels of market power. We can study airports and airlines because the Federal Aviation Authority has made data open to everyone. We can study spectrum auctions or broadband development because the Federal Communication Commission can help you do that. The Stigler Center Report suggests the creation of a Digital Authority with the powers to do this. Regulators in several parts of the world are going to come out with regulations to deal with digital platforms, and quite soon. Regulators must get data from digital platforms as part of their new responsibilities and, as a low threshold, publish their own reports—in which academics could be involved. Even better, they should make such data open and available for academic research. This is the point in time when academics should seriously think about pragmatic solutions that are actionable, and turn what is currently only an April’s Fool joke into a dream job for the next generation.
- I note that this bifurcation in careers right after the PhD is quite specific to economics. In Computer Science (CS), for instance, employees of Google and Microsoft publish a lot and sit on the program/organizing committee of the largest CS conferences. By some metrics, Microsoft is the most influential CS department in the world with Google being 6th (also a reflection of their size). There are examples of people that went from companies to tenured positions in academia. I conjecture that they publish much more than economists because what they do is very often not company-specific. The contribution of a machine learning paper, for instance, is usually methodological and quite general, and also appreciated in the academic community. Importantly though, the vast majority of these papers do not use internal data for anything, instead relying on standard benchmarks or publicly available datasets. That being said, there are critical views also on the impact that Big Tech have on CS researchers, most notably on Artificial Intelligence and AI & ethics.