2. Data taxes as a Pigouvian tax

Pigouvian taxes are an oft-proposed solution to a variety of harms, at least amongst academics.Footnote ¹⁰ Pigouvian taxes are named after Arthur Pigou, who over a century ago identified the phenomenon of differences between the social costs of industrial activities and the private costs to industrial actors.Footnote ¹¹ Pigouvian taxes aim to correct the market failure created by this difference in social and private costs by forcing private actors to internalise these social costs. The idea is that, when an actor is allowed to undertake harmful activities without bearing any of the costs of harmful activities themselves, they will undertake that harmful activity at too high of a rate. As a result, social welfare will not be maximised.Footnote ¹² A classic and straightforward example is pollution. Companies are expected to maximise their profits by setting their level of production such that the company’s marginal costs are equal to the marginal benefits received. Because pollution is a cost of production that is not borne by the company, but instead by society at large, the company will not take that cost into account when setting their level of production and will thus overproduce relative to the optimal level of production.Footnote ¹³

Pigouvian taxes come in to correct this market failure by forcing the company to take on the social cost of the activity, thus reducing the level of production to its socially optimal level.Footnote ¹⁴ The primary goal of Pigouvian taxes is, therefore, to change the taxpayer’s production behaviors, rather than raise revenues or redistribute income and wealth. As Victor Fleischer explains, these ‘[c]orrective taxes are taxes that are designed primarily to change behavior rather than raise revenues’.Footnote ¹⁵ While pollution is a classic example of a potential Pigouvian tax, Pigouvian taxes have been proposed or employed to solve a variety of social harms, ranging from the environmental to public health harms.Footnote ¹⁶

The data economy is, in many ways, an ideal candidate for such a Pigouvian tax.Footnote ¹⁷ The social harms of datafication have been well-documented by scholars.Footnote ¹⁸ Some of these harms stem from violations to individual privacy and dignity, with the mere collection of data causing harm, regardless of use.Footnote ¹⁹ Datafication can also limit autonomy and individual capacity for self-determination and allow for manipulation of human behavior.Footnote ²⁰ The exploitative harms of datafication have also been highlighted, as digital companies amass wealth and power by harvesting data about human behaviors and activities without compensating those individuals. These practices within the data economy have been described by scholars as technofeudalism,Footnote ²¹ a new form of colonialism,Footnote ²² and as comparable to slavery.Footnote ²³ Harms exist not only on the individual but also on the societal level,Footnote ²⁴ and the data economy has been cited as contributing to growing inequality.Footnote ²⁵ As Salomé Viljoen explains: ‘What makes datafication wrong is not (only) that it erodes the capacity for subject self-formation, but instead that it materialises unjust social relations: data relations that enact or amplify social inequality.’Footnote ²⁶ Not only are the harms of the data economy well-documented, but many argue that our existing legal regimes are failing to adequately address these harms.Footnote ²⁷

In the context of these harms and the struggle of legal regimes to adequately address them, a Pigouvian tax on data collection is an intriguing policy tool and for that matter a policy tool that links to the broader rights-protection and power-balancing aims of digital constitutionalism. Importantly, the choice to pursue a Pigouvian data tax implies a political view that datafication is harmful, and it is not beneficial for society to have the data economy continue unchecked. It implies a view that the government needs to regulate the new data economy to mitigate the societal harms of datafication.

The harms of the data economy are social costs that are not being borne by the private actors involved in an economy activity. The digital companies that are collecting and economically benefitting from users’ data are analogous to the industrial company that pollutes the environment while producing their products. Because digital companies are not required to bear this social cost, they will produce (or, in this case, collect data) at a rate that exceeds the socially optimal level. A Pigouvian tax would be a means to correct this market failure and force companies to bear the social costs of data production. As a result, digital companies should reduce their level of data collection to a socially optimal level where the costs of datafication, including societal harms, match the benefits. Reducing the level of data collection should, therefore, reduce the harms associated with the data economy.Footnote ²⁸

A Pigouvian data tax would also raise government revenue.Footnote ²⁹ It would, therefore, further revenue-raising goals to a certain extent. It would also serve a redistributive role by reallocating economic resources from the data-collecting digital companies to the public, thereby combatting economic and political power concentration. But revenue-raising and redistribution cannot serve as a key goal of a Pigouvian data tax. If the goal of the Pigouvian data tax is to reduce levels of data collection (and, in turn, the harms of datafication), the tax should be set at a rate that would cause digital companies to reduce the volume of data they collect, until data collection falls to a socially optimal level. Therefore, if designed effectively, imposing the data tax will erode the tax base. As a revenue-raising mechanism, Pigouvian taxes are, by design, self-defeating.Footnote ³⁰ So while a Pigouvian data tax could be an effective regulatory mechanism to reduce data collection harms, it can only serve an incidental revenue-raising and redistributive function. But, as the following section discusses, revenue-raising and redistribution are central concerns for many proponents of data taxes, and these concerns implicate a different motivation for data taxes from reducing the harms of datafication.

3. Data as a new tax base

Another possible motivation for data tax reforms stems from the insight that income no longer serves as an effective tax base in the context of the data economy. The failure to recognise the economic value of data, separate and apart from any monetary gains derived from it, has both limited governments’ revenue-raising capacities and encouraged the concentration of economic resources and power in the hands of data-collecting digital companies. Therefore, if tax is to serve its purpose of raising government revenues and redistributing economic resources in the data economy, a new (or at least complementary) tax base needs to be found. Data has been identified as that base.

The sentiment that Big Tech companies are not paying their ‘fair share of taxes’ is pervasive in political discourse surrounding international tax reform.Footnote ³¹ And various tax scholars have highlighted that corporate income no longer functions as an effective tax base within the information-driven data economy.Footnote ³² Why income no longer serves as an effective tax base within the data economy relates to both the nature of data as a value form and the unique business models that it triggers.Footnote ³³ The defining feature of the data economy is the collection and exploitation of data about people (or social data) in order to predict and manipulate human behavior.Footnote ³⁴ In the data economy, social data serves as a new factor of production creating prediction value – value that stems from the capacity to predict and manipulate human behavior.Footnote ³⁵ Prediction value is a subset of use value and that differs from (and often does not neatly translate into) exchange value.Footnote ³⁶ Exchange value conceptualises of the value of a thing as its monetary, fair market price.Footnote ³⁷

The central importance of social data and the prediction value it creates is problematic for tax law’s reliance on income as a base in a couple of ways. First, tax law measures income in terms of monetary exchange value.Footnote ³⁸ Because prediction value is distinct from monetary exchange value, it does not get captured in the income tax base. As a result, a valuable economic resource that is increasingly important to the economy – prediction value – slips through the cracks of the income tax.Footnote ³⁹ Second, even when companies leverage prediction value to create monetary exchange value, the specific business models and practices that dominate the data economy are in many cases distinct from those that have dominated industrial capitalism. Therefore, when existing income tax laws are applied in this new context, they often fail to achieve the law’s revenue-raising and redistributive goals.Footnote ⁴⁰

For example, many of the business practices associated with the data economy focus on longer-term company growth versus earning company income in the short or medium term.Footnote ⁴¹ This focus on growth over income is a result of the nature of social data as a factor of production. In order to amass prediction value, companies must have access to steady streams of social data; therefore, they will focus on growing large user and customer networks at the expense of current income.Footnote ⁴² To grow these user bases, digital companies may offer free or low-cost services, running losses while amassing user bases.Footnote ⁴³ Or they may focus on building up ecosystems of goods and services, thereby locking in users and guaranteeing access to future streams of social data from them.Footnote ⁴⁴ Digital companies who use these business strategies may leverage the social data and its resulting prediction value to earn monetary income at some point in the future. They might do this through direct means, such as using prediction value to sell targeted advertising services, or through indirect means, such as using prediction value to improve their products and services.Footnote ⁴⁵ Data collected about people may eventually lead to income for companies. But, the delayed nature of this income realisation in the data economy is a challenge for tax law because it allows companies to defer tax liabilities. This deferral deprives governments of revenues and delays redistribution while simultaneously allowing companies to amass a valuable economic resource – social data and its resulting prediction value.Footnote ⁴⁶ And the ability to amass a valuable economic resource brings with it a concentration of economic and political power in the hands of the companies that hold that resource,Footnote ⁴⁷ resulting in the type of power imbalance that digital constitutionalism aims to combat.

Additionally, the business practices and models of the data economy can result in a misalignment between the country from which social data is collected and the country in which the monetary income derived from that social data’s prediction value is realised. The problem of multisided markets has been frequently cited in debates over the appropriate tax reforms in response to the digitalisation of the global economy.Footnote ⁴⁸ This is part of a common criticism that the place of taxation no longer aligns with the place of value creation – the need to realign the locus of taxation with the locus of value creation has been a driver of many recent international tax reform efforts.Footnote ⁴⁹

A data tax could serve to fix the failings of the income tax in the data economy and thereby address the economic and political power imbalances that are a key concern within digital constitutionalism. By using data, which serves as a material store of prediction value, as its base, a data tax would recognise prediction value as a form of economic value distinct from monetary exchange value. Because it takes into account prediction value as a distinct value form, a tax system that includes data, and not just income, as part of its tax base would better allocate taxing burdens based on relative ability to pay. Using data as part of the tax base would also eliminate the tax deferral problems discussed above. Companies would be taxed when they collect data. The timing of taxation would align with time when they amass prediction value, rather than delaying taxation until they potentially convert that prediction value into monetary exchange value. Using data as a tax base would also address the misalignment of the locus of taxation and the locus of value creation in the digital economy. The country from which data is extracted would be the country that would be allocated taxing authority, rather than the country where profits derived from that data’s prediction value might eventually be realised. For these reasons, data taxes have the potential to make up for the failings of income as a tax base in the data economy. Data taxes could allow for substantial revenue to be raised from digital companies in their users’ home countries. And, by acknowledging prediction value as a distinct form of economic value creation, data taxes could also allow for a fairer and more effective redistribution of economic resources.

However, if the goal of data taxes is to correct for the failings of income as a tax base, the design considerations for policymakers must be distinct from the design considerations for a Pigouvian data tax. Policymakers would need to focus on setting data taxes at a rate that maximises revenue and fulfils redistributive goals. This contrasts with the Pigouvian data tax, in which the design considerations would instead by focused on setting the data tax at a rate that would reduce data collection to a level where the social harms of datafication match its benefits. Any data tax, regardless of design, will likely have some dampening effect on data collection behaviors of digital companies, but the extent of that effect will depend on whether it is designed as a Pigouvian or revenue-raising tax. Additionally, a base-building data tax suggests a political choice to not intervene in the new data economy but instead adapt the tax system to tax data value creation in the same way as other forms of economic value creation. While this choice certainly does not foreclose recognising datafication’s harms and regulating it via other areas of the law, it does stand in contrast to a Pigouvian data tax, which requires a clear normative view of the social harms and benefits of datafication and the government’s role in regulating them.

4. Path forward

Data taxes have the potential to serve as a powerful tool in efforts to respond to the unique challenges brought by datafication and the data economy – a powerful tool that can operate within the framework of digital constitutionalism. The data tax could serve as a Pigouvian tax with the aim of reducing data collection to a socially optimal level, thus balancing the social harms and benefits of datafication. Alternatively, data taxes could remedy the failings of income as a tax base in the data economy, raising needed government revenues and achieving redistributive goals. As scholars and policymakers continue to think about the role that a data tax may serve, it is essential to understand that designing an effective data tax requires a clear understanding and consensus around what problem a data tax is meant to solve.