Data Localisation: The Arrakis Argument

Nikita Melashchenko · NZCIEL · (9 March 2024) · opinion

Data localisation regimes create artificial monopolies over a resource whose value grows the more freely it circulates. In this respect they resemble the imperial control of spice on Arrakis, a strategy that concentrates power in the short term but leaves every participant worse off as retaliation spreads. A simple game-theory model shows that the states most exposed to this dynamic are open developing economies, the very markets that both liberal and authoritarian blocs are competing to absorb. Breaking out of this binary requires rethinking data governance from first principles.

Dune’s spice and Earth’s data

In Frank Herbert’s Dune, the desert planet Arrakis is the only source of a valuable substance called “spice”, or “melange”. The spice is the most valuable substance in the universe, and its control is a major factor in the political and economic power struggles of the characters.

Power over spice is power over all!

It was long known that the spice is a metaphor for oil, and the story of “Dune” is often seen as a reflection of the real-world politics and economics of the Middle East. However, I could not help but draw a parallel between the concept of the spice and data in our world, particularly in the context of trade policy issues related to data localisation.

Consider that like spice (as long as there are sandworms, of course), data (as long there are us, of course) is not a finite resource. It is constantly being generated and consumed. Moreover, its value is not just in its quantity, but in its quality and the insights that can be derived from it. Like spice, which was used for many purposes across the universe including space travel, data is used for many purposes across industries and borders. Finally, data forms datasets unique to each state, industry, and company, and the value of these datasets is enhanced by the ability to combine them with others. In this regard, national data domains are indeed like the fiefdoms of Arrakis, each with its own unique spice reserves, which result in scarcity and competition for control.

To protect them states impose data localisation regimes or requirements that data about a country’s citizens or residents be collected, processed, and/or stored within that country’s borders. This requirement is often motivated by concerns about national security, privacy, and economic development. However, it can also be seen as a form of protectionism, as it can limit the ability of foreign companies to access and use data from a particular country. This contributes even more to scarcity and competition for control, as it creates an artificial situation where data is not freely available to all, but is instead subject to the control of the state and its affiliated entities.

Perhaps, the analogy between the spice and data is not perfect, but it is a useful way to think about the value of data and the challenges of data governance. Arrakis was a natural monopoly for the spice production, which was economically and culturally significant for the entire universe. It was enforced by the Emperor’s decision to rotate the control over the planet between the noble houses. Data domains due to data localisation are artificial monopolies maintained by sovereign states, which prevent the global pool of data from being accessed and used by all, and thus limit the potential of data to contribute to the development of new products and services domestically and globally.

The paradox of spice–data analogy

The paradox in this situation lies in the dual nature of data as both a strategic asset and a commodity that increases in value when shared. On the one hand, like the spice on Arrakis, data within a state’s borders is a strategic asset that can grant significant advantages—economic, technological, and political. By guarding it diligently, a state aims to maintain control and security, much as the rulers of Arrakis would with their spice reserves. On the other hand, data, like spice, derives additional value from being shared, but also combined with other datasets. The insights and innovations gained from a global pool of data can far exceed what can be derived from any single source and significantly contribute to the development of new products and services to the benefit of all.

Developing this paradox further, I would argue that monopolising spice or data is thin ice. On the one hand, both are the source of power and wealth, but on the other hand, they are also mediums of exchange and cooperation. And both are subject to the same ambivalence. Both are valuable because they are scarce and controlled, but their value is also derived from their circulation and application. This inherent contradiction divides stakeholders into two camps: those who seek to monopolise and those who seek to share. The former are driven by the desire to maintain control and security, while the latter are driven by the desire to maximise the value of data by sharing and utilising it.

Now let’s consider the implications of this paradox for international trade and data governance. In my view, while a state may benefit in the short term from exercising an exclusive access to its territorial data, it would at certain point face the problem of the lack of access to the global pool of data. Such states would have to participate in open data markets that strive for technology transfer and accept the risk of sharing their data with foreign companies. Alternatively, such states may benefit from open data markets until the latter retaliate with similar measures, and the former lose access to the global pool of data. In either case, this asymmetric approach will undermine the very basis of data’s value, which is enhanced by aggregation effect and network effect.

On Arrakis, the problem of exclusive access to the spice is resolved by the Emperor’s decision to give the planet to this or that family house. By exchanging the control over the planet, the Emperor ensures that the spice is available to all the houses, and the planet is not destroyed by the struggle for control. Apart, of course, from the situation when the Emperor decides to use the power struggle over Arrakis to eliminate the house Atreides that has become too powerful and popular in his view.

In the real world, the solution to the problem of data localisation is not as simple as the Emperor’s decision, because it involves a multitude of stakeholders and hence a multitude of levels of control. However, my point is that states imposing data localisation are akin the “Dune”'s Emperor in their digital domains. Instead of maintaining a free market of data, they are creating an artificial situation, where they assign exclusive rights to data to certain entities mostly domestic and state affiliated companies. At the same time, they incentivise their companies to seek access to foreign data, which is not subject to the same restrictions, or the transaction costs of which are lower than the potential benefits, and sell their products and services abroad.

By doing so data produced within such states becomes inelastic, i.e. its supply to the global market is limited and unstable. Such states usually strive for innovation and economic growth, but they either need to rely on foreign services providers and share their data with them, or they need to invest in the development of their own data processing and storage infrastructure. The latter is a costly and time-consuming process, and it is not always possible to achieve the same level of technological advancement as foreign companies.

This is not least because foreign companies from open data markets have access to a global pool of data and can develop their products and services based on the insights derived from it, while data localisation states are limited to the data available within their borders and do not have access to the global pool of data if other states retaliate with similar measures. However, if other states do not retaliate, the states imposing data localisation will have a competitive advantage, as they will have access to the global pool of data, while the others will not have access to their data.

In addition, both groups of states seek to expand their markets and require data pockets for that. Since they are not likely to find them in each other, they must find them elsewhere. Let’s consider this situation from the perspective of basic game theory.

The game of data localisation

To create a non-zero-sum game scenario, we need to define the payoffs for each type of state when interacting with one another. We will have a classic payoff matrix, where each state’s decision to engage in either free data markets or data localisation—with the added layer of reciprocity—affects their own utility and that of others.

Let’s define four state types:

Here are some payoff assumptions we can make:

Data markets increase economic welfare, so states that engage in free data markets will generally have higher payoffs.^[1]

Data localisation reduces economic welfare due to inefficiencies but may increase internal data security and control, resulting in moderate payoffs.^[2]

Reciprocity adds a layer of trust and could potentially increase payoffs for states that engage in it.

States that practice data localisation when interacting with states that maintain free data markets might acquire economic benefits, but may miss out on them if those states retaliate with similar measures, receiving lower payoffs.

States that practice free data markets when interacting with states that practice data localisation might gain from the latter providing products and services, but may miss out on data sovereignty and the benefit from data exchange because of the lack of reciprocity and thus experiences the least efficiency and welfare.

Now, let’s set a payoff matrix. Each entry (i, j) in the matrix represents the payoff for state type i when interacting with state type j.

The matrix shows that the most significant benefits come from interaction between open data economies and the least from interaction between open data economies based on reciprocity and data localisation states due to isolation and economic inefficiency. The interaction between data localisation states is also not very beneficial due to inefficiencies, but it is still better than the interaction between open data economies based on reciprocity and data localisation states. In addition, the interaction between data localisation states and open data economies brings the most benefit to the former, while the latter receive less due to the costs of data localisation externally or data sovereignty risks domestically.

The same payoffs can be visualised as a game tree. Player 1 chooses a strategy (A, B, C, or D), then Player 2 responds. The leaf nodes show the resulting payoffs for each combination.

Here is the breakdown of the payoffs:

A–A and B–B (2,2): States with mutual open data markets have high economic welfare due to the free flow of data, leading to efficient markets and collaboration.^[3]

A–B and B–A (2,2): Similarly, even with State B’s reciprocity requirement, the interaction is optimal as both support open markets.

A–C and A–D (0.5,1.5): State A, with its open policy, only sees limited benefit (0.5) when dealing with C or D states that localise data, as A cannot fully tap into C and D’s markets. However, C- and D-type states can access State A’s market, receiving greater benefits (1.5) and leaving A with (0.5) benefits.

B–C and B–D (0,0): Since State B requires reciprocity to trade and C- and D-type states are closed, there is no benefit for any of them.

C–A and D–A (1.5,0.5): Here we see the benefits (1.5) reaped by C and D from A’s open market while projecting their own policy costs (0.5) onto A but leaving A with (0.5) by participating in A’s market.

C–C (0.5,0.5): When two data-localising states interact, economic welfare is constrained by inefficiencies. Each side manages some limited engagement (0.5) within the constraints of mutual data localisation, but neither gains the advantage that comes from asymmetric access or open markets.

C–D and D–C (1.5,1) and (1,1.5): These interactions account for a bit more benefit as they both have data localisation but may have some alignment, leading to more efficacy than C with C but less than D with D because of lack of reciprocity in their restrictive practices.

D–D (1.5,1.5): States with aligned data localisation and reciprocity policies find a balance (1.5) that suggests a functional, though not optimal, economic interaction.

The Arrakis argument

The payoff matrix captures the economic logic, but it does not capture the politics. In the real world, states are not isolated actors choosing strategies in a vacuum — they form alliances, exert pressure, and compete for influence over uncommitted partners. This is where the analogy with Arrakis becomes most instructive.

Let’s also consider the parallels between the game tree players and states in the real world. A- and B-type states are likely liberal and developed economies that are open to data exchange and reciprocity, while C- and D-type states are likely authoritarian and developing states that maintain data localisation regimes.

However, there is also an argument that in fact A-type states are developing states that are open to data exchange without reciprocity because they do not have the capability to maintain data localisation regimes, they suffer from the lack of access to technology and accept the risk of sharing their data with foreign companies in the name of economic growth.

B-type states, on the other hand, are developed economies and technology exporters that are happy to trade with alike states and developing states from category A, but not with authoritarian states from category C and D. C- and D-type states are likely authoritarian states that maintain data localisation regimes and seek to retain absolute control over their resources, with group D seeking to expand their web of reciprocal data localisation markets.

Hence, Arrakis in this game is the group A of developing states that are open to data exchange to all, but because B-type states have a doctrinal incompatibility with C- and D-type states, especially with regards to asymmetry in data exchange, both groups will likely try to pull A-type states into category B or D, respectively. There are viable arguments for A-type states to join either group, as they will benefit from the open data markets with B-type states, but they will also benefit from the reciprocal data localisation markets with D-type states.

In fact, I would argue that there is little incentive for A-type states to remain in category A as they progress in their development, because they will have to face the problem of the divided markets between B- and D-type states, and to some extent C, and they will have to choose between the two. Remaining in category A will only make them vulnerable to the asymmetry in data exchange. They get little benefit from C- and D-type states, and by moving to category D they get more benefits. With moving to category B they won’t get more benefits from B-type states, but they will cut off the benefits to C- and D-type states.

Here is the final consideration. Fremen or no Fremen in the game, the strategic goal for B-type states is to expand their web of reciprocal open data markets, thereby enhancing collective economic welfare and innovation within their circle, and reducing advantages to authoritarian states by not providing unilateral market access. The strategic goal for D-type states is to expand their web of reciprocal data localisation markets, thereby enhancing collective economic welfare and security within their circle, and reducing advantages for B-type states. In this struggle, A-type states, or in fact their markets, are the prize for either group unless they figure out the way to provide an alternative to the asymmetry in data exchange. For now, in the near complete absence of results in the field of global data governance, including within the MC13 in Abu Dhabi, the binary B-D paradigm seems to be reinforced. To come up with an innovative solution to this problem, we finally need to start thinking outside the box.