Fuqua Insights Podcast: Why Is Your Data Worth So Little?

Tanner Morgan  0:00  
Welcome to Duke Fuqua insights, a podcast where we explore faculty research and the actionable takeaways for business leaders at every level

(music)

Tanner Morgan  0:13  
Whether we're scrolling, shopping or streaming, our data is constantly being collected, analyzed and traded. We're often told that this data is incredibly valuable and that sharing it leads to better products, more personalization and smarter decisions. But what are the trade offs of today's data economy? My name is Tanner Morgan, a recent graduate of Fuqua daytime MBA program. I'm joined by Professor Ali Makdoumi, an associate professor of Decision Sciences at Fuqua, to help answer these essential questions. His research explores how information flows through markets, shaping everything from pricing to privacy. He co-authored “Too much data,” a paper examining the unintended consequences of how personal data is bought and sold. Ali, thanks for joining us today.

Ali Makhdoumi: Thank you, Tanner, for having me, pleasure to be here.

Tanner Morgan  1:08  
So to start off, I have a couple of questions for you. There's no shortage of analysis on how valuable data can be, but you focus instead on when there's too much of it. What motivated your research?

Ali Makhdoumi: 1:18  
So I guess the conventional wisdom is that data is valuable for, you know, personalized advertisement, you know development of AI, and perhaps you know personalized AI tools, or you know, personalized recommendation and so on.

Ali Makhdoumi  1:36  
And you know more data is obviously better for these purposes. But what we argue in this work is, you know, where does data come from? And in most cases, the answer to this question is, you know, it's coming from people. And then the other sort of the question is, you know, are we having the right amounts of data for what the tech companies or, you know, platforms are compensating users for. And that's where, you know, we argue data is under priced, and you know, we're having too much of it. In other words, you know, sort of, there are two ways to look at it. So one is, you know, for what the platforms are offering, or, you know, paying, we have too much personalized data collected from users. And you know, the other way to look at it is, you know, I guess, and this is closer to policy design, is, you know, for the amount of user data that they're collecting, they should be paying more or compensating users more.

Tanner Morgan  2:43  
So it sounds like a key point in your research is that the effect of sharing personal data extends far beyond the individual. Why is that? And what are some of the economic consequences of that?

Ali Makhdoumi  2:54  
Great question. Again, the sort of broad intuition is quite simple, and so is, you know, I guess, in short, it is because of data externality. So let me explain this. So imagine that, you know, I want to be very private, so I go to the basement of my house, you know, not sharing any data. So does that mean that my personal data, and, you know, personal taste will be, will be protected? Not really. So why is that? Because the fact that my, you know, friends, family and colleagues who have, you know, perhaps similar taste to me, are sharing the data reveals something about me as well. So this implies, you know, personal data and privacy is more of a public good, and giving control to individuals, even though you know necessary is not going to resolve the whole issue, so we need, we need more than that. So in fact, the issue remains, you know, to exist, even if individuals are willing to share their data, even if, you know, I don't want to be too private, so I want to get the fair price for my data and share it. Then, you know, the issue still remains. Why is that? So imagine that you know, you and I want to share our data and get a fair price for this, get a reasonable price for our, you know, personal data. So now the data buyer will come to you and say, I'm willing to purchase part of your data at a certain price, and if the price is good, you will share your data. Now the data buyer will come to me and say, well, now that Tanner's data has been shared, given the correlation between your data, for instance, because you're both affiliated with Fuqua, something about you has been revealed as well. So therefore the sort of marginal value of my data has decreased, and I will be willing to share at the lower price than what I could get. It, you know, in isolation. So again, the data buyer will come to you and say, Now that Ali's data has been revealed, the marginal value of your remaining data is lower, and I will be buying it, you know, at the lower price, and so on. So once you sort of create this market for personal data, you and I as users will end up competing with each other. And you know, we will end up sharing our data at the very marginal, marginal price, so we will not get the fair price for our for our data. So I guess the bottom line is, you know, when one person sells or gives away their data, others’ privacy is compromised even if they don't want to share, even if they're not sharing their data. And the economic consequence, I would say, is inefficient oversharing. Meaning platforms acquire lots of lots of data at, you know, very low marginal prices, and individuals are going to end up undervaluing their own privacy, because if much of it, you know, has already been leaked by others,

Tanner Morgan  6:14  
You make a really interesting point on competition. Competition between platforms is often something that we believe can solve market problems. Can you explain the role of competition and its impact in this space?

Ali Makhdoumi  6:30  
Great question again. So competition may or may not help in the context of market for personalized data as we study it. Why is that? Because once you have multiple platforms that are competing, we're going to end up having a fragmented market, say, users will join one of the two platforms. Now there are fewer externalities on each of these platforms, and that can increase prices, so users will be better off in that sense. However, there's some, you know, benefits, you know, for instance, for service quality and so on that we only get if we have, like, a lot of users on the platforms. And you know, now that we have fewer users on a platform, those benefits are going to be less so overall this, you know, second force may dominate and competition may worsen the user's welfare.

Tanner Morgan  7:27  
So one of the solutions you propose involves transforming the data to reduce its impact on others. How would this kind of quote-unquote, decor decorrelation process work in a real world setting,

Ali Makhdoumi  7:40  
So I guess, if you think about it, the the issue is, you know, data externality or data correlation. So how can we fix it? We need to, you know, remove the correlation, so that when, when I decide as a user to share my data, I only share my data and nothing about others. So technically, this involves removing statistical correlation between individuals data. So my data will have a part that is orthogonal to others, and when I decide to share my data, I will be sharing only something about me, myself, and nothing about others. In practice? Well, it's a hard question because it's hard to achieve. So why is that? Because to decorrelate the data of users, we need to have access to the data of all of them. And you know what that means is we're going to need a trusted party that collects the data of all individuals, and somehow figure out how to decorrelate the data of each user from the data of others. But I guess we don't have a, you know, third party. We don't have a trusted third party. So it is hard to implement this in practice, unless, you know, we come up with some smart ways of, you know, implementing this by some blockchain ideas and so on. But I guess, like the bottom line is, you know, once we decorrelate the data of users to make sure that when I decide to share my data, I'm sharing only something about me, myself and nothing about others, then the issue will be, will be resolved, and we're going to have, you know efficiency in this , you know market for personal data.

Tanner Morgan  9:24  
So, many Fuqua grads go on to work in data heavy industries such as healthcare or tech. How should they or anyone else working in these fields be thinking about your research, and just as a quick follow on, you also talk a lot about trust, what does it mean to have a more responsible data strategy.

Ali Makhdoumi  9:43  
Great question again. So I would say they should be thinking not just about what data they collect, but about whose privacy is being affected by that data collection. So in healthcare, for instance, in healthcare, you mentioned that, for instance, sharing one person's genetic data can reveal something about their relatives, for instance. So here in this context, a responsible strategy would sort of account for these externalities, perhaps by securing a group consent or adjusting compensation to reflect broader privacy costs. I guess, like in any other context, a responsible strategy should take into account that privacy is a public good, and giving individuals control over their data, although it's necessary, is not really sufficient. For instance, you know, take the sort of a more recent regulations that have been implemented in Europe, for instance, like the GDPR, or, you know, the California one CDPR. So at a high level, they're giving individuals control over the data. What we argue here in this paper is that we do have to do even you know, more than that, because privacy or, you know, protecting privacy, requires more of a public effort.

Tanner Morgan  11:12  
So before we wrap up, I'd love to hear a little bit more about where you think we go from here. Do you see a path where we rethink the role of personal data entirely, or are we heading more towards an even deeper entanglement, unless something dramatic happens?

Ali Makhdoumi  11:27  
Well, I guess that's one of the questions that keep me awake at night. So we are clearly at a crossroads. If we continue down the current path -- unregulated data markets -- we'll see, you know, worsening privacy violations and increasingly sort of distorted market incentives. And you know, this is only going to benefit the platforms, and it's going to, you know, keep reducing the user's welfare. But on the positive side, there is a sort of a growing awareness, both among policymakers and also technologists, that data isn't just an individual asset, it's a social one. Moving forward, we need to sort of rethink data governance, and there have been some moves in Washington toward that, perhaps you know, by regulating data markets, you know, implementing taxation, or, you know, mediation mechanisms, or by redesigning data architectures. And there are, you know, many interesting ideas on, you know, how one can implement those, you know, regulations. And one thing I should highlight is that there is no doubt that data is valuable and necessary for innovation. So the goal is not to halt innovation, but to sort of ensure that it aligns with societal benefits and the user incentives. So I guess that's kind of important to keep in mind.

Tanner Morgan  13:00  
Absolutely. Well, thank you so much for joining us, professor, really fascinating research. 
Ali Makhdoumi:  Thanks again, Tanner.