.--- layout: default ---
One of the most interesting talks that I attended last week was Abowd’s discussion of his economic framework for the analysis of the protection of privacy in the context of $\varepsilon$-differential privacy. From their 2015 paper:
The relatively new concept of differential privacy allows a natural interpretation of privacy protection as a commodity over which individuals might have preferences. In many important contexts, privacy protection and data accuracy are not purely private commodities. We show that it is feasible, at least in principle, to determine the optimal trade-off between privacy protection and data accuracy when the public-good aspects are important.
We develop a complete model of the technology associated with data publication constrained by privacy protection. Both the quality of the published data and the level of the formal privacy protection are public goods. We solve the full social planning problem with interdependent preferences, which are necessary in order to generate demand for the output of government statistical agencies.
Their theoretical model deserves much more effort that I put in my first reading, so I am going to refrain for discussing its relative merits. Also, I do not know how it fits in the literature because this is the first time I am exposed to, or even think about, an economic treatment of the problem. However, simple as it sounds, the overall notion of treating $\varepsilon$ as a social parameter and how its current value can be recovered from data is very illuminating and probably worth the time spent working through the details.