Introduction
The first two articles of the series made the same argument in two different settings: Amazon in the United States and the platforms that dominate the developing world. Amazon did not become rich by extracting value from consumers or from its workers. The company got rich by making a surplus that did not exist before, and the losers in the process were first competitors and, to a lesser extent, the third-party sellers on its platforms. The same pattern holds in the developing world: Companies such as Jumia, Grab, and MercadoLibre deliver real surplus to consumers and small merchants, while the problems they face are data sovereignty and gig worker precarity.
A wrong diagnosis produces a wrong treatment; naming consumers and workers as the injured parties when the losses actually landed on competitors and third-party sellers, and every policy built on that reading will miss. Since the extraction narrative is flawed, the policies based on it should fail, and they do.
Governments have converged on three standard policy responses to Big Tech: digital services taxes, mandatory data localization, and antitrust action. The problem is that all of them share the assumption that platform wealth is extracted rather than created and thus needs to be retrieved or contained. The evidence on these policies now confirms that they tend to miss platforms and land on the consumers and small businesses they were meant to protect. A better agenda exists, built around data rights, worker classification, and competitive market structure; however, it remains mostly unbuilt and sits mainly in the EU, the UK, and a handful of other high-income states.
Section II: The Digital Services Tax Lands on the Wrong People
A digital services tax charges a flat percentage of the gross revenue a platform earns inside a country, regardless of physical presence or declared profit. Governments have aimed squarely at firms like Amazon and Google, and the appeal is political: the platform finally pays the treasury of the country whose users it profits from. However, the economic rationale for this move has proven to be less coherent.
The most direct evidence is provided by the very company under threat, Amazon itself. A recent investigation into the company’s response found that Amazon passes the tax almost entirely onto its sellers through higher fees and that most sellers then pass those fees straight to consumers. This is not an individual analyst’s impression. Economists Dominika Langenmayr and Rohit Reddy Muddasani discovered that “in most countries, Amazon increased its fees by roughly the amount of the digital service tax” and “firms using Amazon as a platform have largely passed these increased fees on to consumers.” Large digital firms thus bear only a smallpart of the tax burden. Instead, the tax burden falls not on the platform but on its domestic consumers. By their estimate, consumers in the markets they studied bore between 1.50 and 2.50 in local currency for every unit the tax actually raised. The ratio runs above one-for-one because the tax is levied once on gross revenue but passed on as a per-transaction fee, and each seller then adds its own margin on top of the higher fee, so consumers pay the tax plus the markup it triggers, while the treasury only collects the tax. The platforms’ own conduct confirms it. When the United Kingdom’s tax took effect, Amazon raised marketplace fees for UK sellers by 2 percent, Google raised advertising charges by 2 percent, and Apple adjusted App Store commissions, each of the three explicitly citing the DST. For a small business advertising on Google or selling through Amazon on thin margins, a 2 percent fee increase driven by that pass-through can wipe out a meaningful share of profit, a cost the small business did not set and cannot avoid.
Here the extraction framework proves itself ineffective. A government sets out to tax a platform it believes is bleeding its citizens, and the platform routes the tax back onto those same citizens. The diagnosis produced a cure that spread the disease.
Data Localization Damages the Economies It Claims to Protect
If digital service taxes are considered to be the application of the extraction narrative to money, data localization is the same applied to information. The reasoning is simple: the platforms take data from a country and monetize it abroad, so the country should require that data be stored and processed at home. From reading the last article, this looks like a data sovereignty measure. It is not. It mixes up where the data is stored with who profits from it.
The Information Technology and Innovation Foundation (ITIF) used the data-restrictiveness index based on OECD market-regulation data to “find that a 1-point increase in a nation’s data restrictiveness cuts its gross trade output 7 percent, slows its productivity 2.9 percent, and hikes downstream prices 1.5 percent over five years.” The specific country figures are sharper: South Africa is projected to see a 9.1 percent drop in gross output, a 3.7 percent drop in productivity, and a 1.9 percent increase in downstream prices between 2013 and 2018, tracking with the country’s rise in data restrictiveness in those years. Similarly, restrictive data policies are expected to “reduce Bangladesh’s volume of trade by 6 percent, Hong Kong’s by 5.7 percent, Indonesia’s by 5.8 percent, Pakistan’s by 3.7 percent, and Vietnam’s by 9 percent.”
None of this happens by chance. Data localization raises the cost of imported digital services that domestic firms rely on as inputs, and that cost travels downstream into higher prices and lower productivity across the whole economy. The value of data lies in its use, not in storage; localization moves the cost and leaves the value behind. Localization constrains use because the value comes from combining records and running them on the best available tooling. A mandate to keep data inside one jurisdiction splits a firm’s records into national silos, forces it to use whichever domestic cloud exists rather than the cheapest or most capable one, and blocks the cross-border pooling that makes fraud detection, credit scoring, and demand forecasting work at all. The servers move, but the analysis that made the data worth anything does not follow.
The problem mentioned in the previous article, that data from the Global South is monetized in the North with very little return, does not go away here. Localization misses this problem entirely and rather moves servers without moving returns and charges the local economy for it.
The Remedies that Would Work, and Where They Are
The argument running through this series is that every harm must be met with its proper remedy. Currently, policies that address data sovereignty and labor precarity already exist; however, they are implemented unevenly and mostly in high-income states.
Take labor precarity first. The most ambitious legislative response to it is the Platform Work Directive, adopted by the European Union in October 2024 and entering into force in December 2024, with member states obligated to incorporate it into their national legislation by December 2026. As with any good labor remedy, this one includes the legal presumption of employment that shifts the burden of proof from the worker to the platform. This singular shift turns precocity into a claim a worker can actually make: a driver no longer has to prove he is an employee to reach sick pay, notice, and a wage floor. Now, the platform has to prove he is not. The harm was misclassification, and the remedy is aimed at misclassification, which is precisely what a revenue tax cannot do. Moreover, it requires the platforms to make the decisions on termination and suspension of employment by a human reviewer, not by an algorithm. Instead of taxing platform revenue, the directive reclassifies the worker, which is where the harm actually is.
The logic applies elsewhere as well. Argentina and Mexico have introduced guidelines for the gig economy, while Chile passed a labor law regulating the classification of platform workers. Judicial precedent is also easily transferred across borders. After the United Kingdom Supreme Court in 2021 decided that Uber drivers were workers and not independent contractors, due to the control Uber has over them, South African drivers took a class action lawsuit against Uber’s local and Dutch subsidiaries in the Johannesburg Labor Court to seek the same recognition. The tension is worth naming: these remedies originate in high-income states, yet Argentina, Mexico, Chile, and South Africa have already adopted or litigated versions of them. That is the point. Classification reform travels on judicial reasoning and statutory language which cross borders cheaply. On the other hand, a revenue tax depends on enforcement leverage over a foreign firm and does not travel at all.
The point is not that the Global South should copy the European text word-for-word, considering how different the labor markets are there; however, the remedy addressing the actual harm exists in a usable format, while the remedies aimed at the imagined harm, the taxes and localization policies, are the tools that the developing-world governments have reached for first.
The Counter-Argument
There is a practical argument supporting the current policies: they may be misdirected, but they do generate some revenue and put pressure on companies that otherwise pay virtually nothing for their operations. France and the United Kingdom alone gathered over 400 million euros from their digital service taxes in 2021. That is less than a thousandth of either country’s annual tax take, and it goes into general revenue—not the merchants and consumers on whose behalf the tax is defended. Alas, some pressure is better than none.
The revenue is real; the question is who paid it. Section II demonstrated that the tax burden will actually fall on the shoulders of domestic consumers and merchants selling through the platforms and on the platform itself. The government that imposes the tax and congratulates itself for taxing Amazon has done nothing but tax its citizens while stamping Amazon’s name on the receipt. That is not a partial success, as it mislabels the cost.
Moreover, the argument that some pressure is better than none does not hold, because the pressure has produced blowback instead of leverage. These taxes drew repeated US trade retaliation across the Trump and Biden administrations, and the OECD’s Pillar One initiative, which is meant to replace them by reallocating multinational profits to the countries where users are located, was stalled in 2024. A developing country that imposes such a tax as a bargaining tool ends up with trade complaints filed against it.
The “better than nothing” approach works only if there is indeed nothing else, yet that is not the case. There is another agenda in Section IV: policies that address the actual harm, avoid the pass-through issue, and do not allow the trading partner to file a complaint. The choice was never between these taxes and inaction. It is between these taxes and a policy that works.
The Diagnosis Determines the Cure
Let us return to where the series started. The extraction narrative has been satisfying and flawed; it looked for the victims among the consumers and workers, who actually gained from the operation, and missed the actual losers. This article has shown the consequences of this diagnosis when it turns into policy. The extraction narrative produces extraction-shaped policies as remedies that miss the platforms and punish the people whom it was supposed to protect.
There is a better agenda, and it is narrow. The data sovereignty regulations should regulate the returns to data from the developing world rather than localization. Worker classification reform should follow the logic of the EU’s Platform Work Directive, adapted to the labor market where most work is informal. Market structure regulations should prevent monopolies (the genuine competitive harm this series has granted from the start) rather than impose taxes.
Some of the framework already exists. The International Labor Organization has been developing its standards on decent work in the platform economy and introduced them at its conferences in 2025 and 2026. The opportunities are there. What is missing is the willingness to use them instead of unilateral taxes that, from the evidence above, cost their own citizens more than they ever cost the firms they were aimed at.
Getting the diagnosis right was never an academic point. It is the condition for a cure that treats the disease instead of the patient.





Leave a Reply