Is Big Data a True Source of Market Power?
In this Q&A, Catherine E. Tucker, Sloan Distinguished Professor of Management and Professor of Marketing at MIT's Sloan School of Management, and an Analysis Group affiliate, discusses whether big data should be considered a source of market power, and the various ways in which big data is impacting antitrust issues and data breach-related litigation.
Q: What do you think are the most pertinent issues related to big data in antitrust?
Professor Tucker: Policy makers often argue data itself has become a source of market power. This contrasts with our experience as researchers, since we often focus on the headaches associated with trying to process and grapple with big data. It also doesn’t gel with standard strategic managerial models of what can confer market power. To establish this we borrowed a framework from the strategy literature, where we analyzed whether or not big data meets the four traditional criteria for being a barrier to entry or a source of sustainable competitive advantage (inimitability, rarity, value, and non-substitutability). At least at the moment, the kind of big data that digital firms have access to doesn’t make the cut. Big data’s advantages are relatively easily to mimic and by itself big data is often not that valuable. It is not clear that having two terabytes of data is much more valuable than having one megabyte of well-organized and easy-to-use data. Instead, competitive advantage stems from having the right personnel with the right training to make sense of the swathes of data.
Q: Other than thinking of data as a potential source of market power, are there other antitrust topics where big data has played a role?
Professor Tucker: Big data is also being discussed as a potential reason to disallow mergers, due to concerns related to consolidating market power. I must admit to inwardly smiling at this though, because my research suggests the opposite. Specifically, we studied what happened in hospitals when you have a merger and need to combine two sets of hospital data. The results of the study went a bit against the tide; we found that having a lot of data hurt the productivity of the merger, because of the lack of interoperability between the different data systems. Having two sets of big data that you are trying to merge can in fact increase the difficulty of a merger, and mitigate some of the benefits.
Q: Why does big data not turn out to be as helpful as we think?
Professor Tucker: I worked on another study related to this topic which was telling. In this study, we looked at what happened to search engine performance when the amount of data used to generate the query was limited due to a restraint imposed by the EU. This restraint led to a convenient natural experiment. We found that, when searches were based on six months of data rather than 18 months, the changes in the apparent quality of search results were minimal. We were surprised by this result; we thought if you are going to find returns from big data it would be in search, since the industry is notorious for using large amounts of data. However, when you think about it, this result makes a lot of sense. The proportion of our search queries that are unique is relatively large, and the searches that are most relevant to predicting behavior and results are often the most recent. Often, we collect volumes of data about a consumer and only a small sliver of it ends up being useful. I think this goes back to the idea that big data, when you start measuring in petabytes rather than terabytes, doesn’t make as big of a difference to relative competitive performance as is often assumed.
Q: We’ve touched upon antitrust in the form of market power and mergers; do you have any additional insights about how the world of big data is affecting data breach class action lawsuits?
Professor Tucker: What we have seen is that the explosion of digital data has been matched by an increase in litigation surrounding all the things that can happen with digital data. This litigation activity includes data breaches, alleged privacy intrusions, and many other things that can happen when companies collect detailed data about an individual customer. This increase in litigation is not surprising; in fact, I have a paper which shows that as you digitalize your data, you increase the likelihood that you will have a data breach. However, what I think is interesting, and is not said enough, is that parallel to this increase in data and resulting increase in data-related litigation is the opportunity to get more information about how the consequences of these different data breaches or privacy issues vary so much across individuals. For example, my own research on privacy emphasizes how idiosyncratic and varying people’s attitude toward privacy tends to be. For example, in my research, I have shown that while many users of social networking sites can respond negatively to intrusive use of their data in advertising, there is a subset of users who appreciate the personalization of their advertising and respond positively to it.
The other thing that I find fascinating is that, for even a single person, it is not clear that the effects of a privacy intrusion or a data breach are going to be constant across time or instance. I have two examples of this; first, I have done a study that shows that people’s privacy preferences evolve over time, and considerably change between the ages of 18 and 30. Second, we have observed that people’s individual preferences vary based on how they are financially compensated for the use of their data. People can be protective of their data, but for a small financial incentive they are willing to behave differently. We refer to this phenomenon as the privacy paradox. All of these inconsistencies and idiosyncrasies are important when thinking through the implications of class certification cases related to data breaches and privacy. ■