Big Data, Privacy and Price Discrimination: A Behavioral Economics Perspective

By Oren Bar-Gill*

This is the era of Big Data. We live on the Internet. And (almost) everything about our lives – our choices, preferences, interests, location, friends, employment, income, and more – is potentially being tracked. Moreover, information that we do not reveal directly can often be inferred with the help of powerful computer algorithms. Many of us are happy to share this information in exchange for tailored ads and free services. Others bemoan the end of privacy.

The proper regulation of Big Data and the optimal protection of privacy are fiercely debated in this country and others. The FTC issued a 2012 report titled “Protecting Consumer Privacy in an Era of Rapid Change,” followed by a 2014 report on Data Brokers. This year, the White House circulated a draft “Consume Privacy Bill of Rights.” Similar initiatives, some much further advanced, are taking shape in the European Union and in other countries.

My goal in this post is to highlight one particular problem raised by the Big Data revolution – the problem of price discrimination. Price discrimination describes a situation where a seller, usually a monopolist (or, more generally, a seller with some degree of market power), charges different prices to different consumers for the very same product or service. The seller seeks to identify her customer’s willingness to pay, and then charges higher prices to those consumers who are willing to pay more.

Price discrimination has been around for a long time – long before Big Data. Business travelers pay more for flights than tourists (even when they both fly coach). Businesses pay more than individual consumers for the same product. Prices differ based on geographical areas, the timing of the purchase and more. But until recently sellers’ ability to price discriminate was relatively limited. Since a customer’s willingness to pay is not written on her forehead, sellers had to resort to observable characteristics, like business customers vs. individual consumers, city dwellers vs. suburban shoppers, whites vs. blacks, etc. The categories of price discrimination were rather coarse.

Enter Big Data. Think about the amount of information that Google or Facebook has about us. And, with the help of data brokers who aggregate and sell information, you don’t even have to be a Google or a Facebook to know almost everything about your customers. With this vast amount of information about preferences, perceptions and income constraints, it is almost as if each consumer has her willingness to pay written on her digital forehead. And this opens the door to much more fine-grained price discrimination. Sellers can practically charge a different price to each individual consumer. (Even with these vast amounts of information, sellers’ ability to price discriminate is not unlimited: fairness concerns and the possibility of arbitrage impose meaningful limits on price discrimination in many markets. My point is only that Big Data has the potential of pushing the frontier of price discrimination beyond what we have been accustomed to.)

Price discrimination is not necessarily bad. Consider a monopolist. If the monopolist cannot price discriminate, it will set a single price. This relatively high price will inefficiently prevent a potentially large number of consumers with a lower willingness to pay from purchasing the product, creating the infamous monopoly deadweight loss. But if the monopolist can price discriminate – charging a higher price to consumers with a higher willingness to pay and a lower price to consumers with a lower willingness to pay – this deadweight loss can be avoided. (This form of price discrimination raises important distributional concerns, as the monopolist appropriates the entire consumer surplus, but let’s focus on efficiency for now.)

This advantage of price discrimination depends on a crucial and often underappreciated assumption. It assumes that a consumer’s willingness to pay reflects the consumer’s preferences regarding the product or, simply, the benefit that the consumer will derive from the product. But willingness to pay is not only a function of preferences; it can also be a function of consumer misperceptions. Willingness to pay reflects the perceived benefit from the product, not the actual benefit.

When sellers price discriminate based on willingness to pay and willingness to pay is based on perceived benefit, there is no reason to believe that price discrimination will increase efficiency. In particular, if benefit is overestimated, price discrimination will increase the number of product units sold, but some of these units will be sold to consumers who purchase them only because of the misperception. In other words, price discrimination facilitates efficiency-reducing transactions.

The normative assessment of price discrimination, especially when driven by Big Data, depends on a deconstruction of the willingness to pay estimates that sellers use. When misperception plays a larger role in these estimates, the less likely it is that price discrimination will enhance efficiency. The challenge for policymakers – the challenge of deconstructing the willingness to pay data – is not a simple one. But without such deconstruction, we cannot determine if price discrimination is good or bad. Now, clearly, price discrimination is only one implication of Big Data and, therefore, only one consideration in the ongoing debates over the regulation of Big Data. Still, it is a potentially important consideration and one that requires careful analysis.

 

*Oren Bar-Gill is a professor at Harvard Law School whose scholarship focuses on the law and economics of contracts and contracting.