Should Internet Companies Be Taxed for Profiting from Private Data?
The growing number of California-based technology companies exploiting internet communications to reap billions in revenue recently led Gavin Newsom, the new state governor, to propose a seemingly attractive idea: Why not compensate consumers for providing the data that has made these companies rich? Governor Newsom announced plans in February for the introduction of a so-called digital dividend, which he applauded as a way for consumers to “share in the wealth […] created from their data” by technology companies that “make billions of dollars collecting, curating and monetizing” user information.
Implementation details have not been defined yet, but the governor’s office reportedly noted proposals discussed elsewhere for taxes on data-driven corporate income and set up a team to work on designing the dividend. They should bear in mind that although a data tax is a good starting point, it is not enough to redress the power imbalance between firms and consumers that permeates the digital economy. It must be accompanied by measures that reinforce the agency of individuals.
The idea that consumers should be compensated for providing companies with data is not new. Pioneered by technologist Jaron Lanier in 2013, it has been discussed in academia and advocacy groups for some time. Legal scholar Eric Posner and economist Glen Weyl depict ordinary citizens as “technoserfs” who routinely give big corporations information on everything they do without adequate reward.
Technology firms counter that they offer widely popular services at no charge, providing significant value to consumers and to the economy as a whole. Indeed, leading internet platforms offer content that users are interested in, yielding efficiency gains in both business and private activities. Moreover, at least some individuals feel they receive psychological rewards from getting attention in return for parting with personal information.
These arguments are not convincing. Most consumers are unaware of how much data is collected from them and how firms are monetizing it. Upon learning more, some change their behavior—they leave the platforms entirely, even if it comes with a cost, or stop clicking on ads and buying promoted products. Not everyone realizes how targeted advertising can enable racial discrimination by showing job and housing opportunities only to some groups, or how shoppers browsing for the same goods from different locations may be shown different prices depending on how affluent their neighborhood is.
It is not realistic to contend that everyone is making an informed choice when relinquishing their privacy in exchange for platform services, even more so in markets that tend toward monopolization because of network effects. As argued by the German antitrust authority, platform users may subject themselves to mass data harvesting only because they have no alternative if they want to maintain access to certain categories of services. The combination of information asymmetries and limited competition makes for a supercharged market failure. Public intervention is justified.
In 2017, the European Commission attempted to introduce a 3 percent tax on corporate revenues from “activities where users play a major role in value creation.” As no agreement could be found at the EU level, some European countries made plans to introduce their own taxes. There is precedent in the United States, too: Washington state legislators proposed a 3.3 percent tax on revenues from data sales. These measures are conceptually different from taxes on sales of goods and services via ecommerce channels, as they aim at capturing the revenue produced by the availability of user data itself.
At first glance, a tax is an appealing option for governments seeking to give consumers compensation for data. It is easy to levy, despite the many challenges posed by the measurement of value added in the digital economy. It does not require a direct evaluation of how much each piece of information is worth, a notoriously difficult task. Once a few key choices are made—who should pay, what exactly is a data-driven activity—enforcement and monitoring are straightforward.
Policymakers, however, should not underestimate the challenge that comes with revenue allocation. The word “dividend” brings to mind monetary payments, but redistribution via direct transfers to consumers is not an option here. A back-of-the-envelope calculation suggests that a 3 percent tax on the US-based revenue of leading technology companies would have yielded about $32 per US resident in 2018., At roughly 0.1 percent of median per capita income, this sum would hardly make a difference for the vast majority of recipients. It is not an accurate representation of the role played by user data in digital innovation.
The dollar value is not the only issue. A transfer—no matter how large—does not work towards solving the problem of information asymmetries, limited individual agency over data, and generalized risks to privacy. The revenue of a data tax should be put to uses that tackle these problems directly. When aggregated over residents, the $32 become $1.3 billion for California, and more than $10 billion in the case of a nationwide scheme. The money should go toward funding three endeavors:
- Development of technology that makes consumers more aware of what information they provide to companies and what it is used for. Lengthy explanations in privacy policies, written in complex legal language, do not perform well in raising awareness. An app that visually tracks what is shared with whom and provides users with an intuitive way to limit circulation of their data is bound to do better.
- Experimentation with regulated data markets. Today, private markets for micro-level information are underdeveloped, exploitative, and dangerous. Consumers have no bargaining power vis-à-vis corporations, so attempts to sell data individually result in meager rewards. Criminals and spies can pose as legitimate buyers to harvest sensitive information. As pointed out by the Electronic Frontier Foundation, a digital rights group, an unfettered money-for-privacy model is ethically questionable. Data markets have the potential to empower individuals, but they need collective bargaining, regulation that sets limits on what can be sold, and transparent pricing mechanisms. The blockchain community is making headway when it comes to technical market architecture, but the regulatory part is missing.
- Research on privacy-preserving ways of extracting economic value from data, such as the generation of synthetic datasets and differentially private algorithms for statistical analysis. In the face of increasingly strict privacy legislation in many countries, with significant fines for noncompliance, companies already have an incentive to avoid invasive collection and use of personal data. If equally profitable alternatives were widely available and adopted, the pressure on consumers would decrease.
1. The European Commission clarified that targeting revenues instead of profits would have been a temporary measure, only in place until rules were developed to prevent multinational digital corporations from artificially apportioning all their profit to low-tax jurisdictions. The revenue tax would also have applied to companies that operate at a loss while growing their user base, a commonplace occurrence for internet-based businesses.
2. Calculation based on the combined 2018 revenue of Apple, Amazon, Facebook, and Google and its geographical distribution as provided by the companies in their financial statements.
4. The amount would increase as the digital economy expands, but for the transfer to reach $1,000, user-provided data would have to be responsible for about 50 percent of US GDP. This is not an impossible scenario further down the line: The IMF estimates that the digital sector alone, which accounts for online platforms and platform-based services but excludes many other uses of data throughout the economy, already accounted for 8.3 percent of GDP in 2015 (5.5 percent when excluding information and communications technology equipment, semiconductors, and software), before the machine learning boom. In such an economy, however, $1,000 would look as inadequate as $32 does now.