May 30, 2022
In a 2018 paper, Casella and Formenti rely on work undertaken by the United Nations Conference on Trade and Development (UNCTAD) to illustrate the differences between the FDI patterns observed among large multinational enterprises (MNEs) depending on their ‘internet intensity’. They map UNCTAD’s digital framework into a conceptual matrix positioning digital categories on the basis of their internet intensity (the internet intensity matrix or IIM), along two dimensions: production and operations, on the one hand, and commercialization and sales, on the other. The IIM distinguishes between purely digital multinational enterprises, non-digital MNEs and a group of ‘mixed model’ MNEs which falls somewhere between the two extremes. Their subsequent analysis and findings is where things get interesting: as it turns out, digital MNEs have a share of foreign sales that is more than 2.5 times the share of foreign assets compared to traditional, non-digital MNEs. In other words, digital firms do not tend to invest a great deal in markets abroad in order to secure foreign sales. This is despite the fact that many of the world’s largest digital MNEs often make in excess of half of their sales abroad.
Against this backdrop, it is fairly easy, at least from a South African perspective, to imagine the temptation to apply data localisation measures with a view to forcing digital MNEs to engage more extensively in foreign direct investment (FDI): if firms heavily reliant on South African data to make sales to South Africans profit heavily from those sales without investing as much as MNEs ordinarily invest, why not use data localisation measures as one way to force the issue? Some version of this question appears to be on the mind of certain South African policymakers, with the South African Department of Communications and Digital Technologies having published a Draft National Data and Cloud Policy (Draft NDCP) for comment in April 2021, which in its current form suggests that a number of policy interventions be implemented, including that all data which form part of South Africa’s ‘critical information infrastructure’ shall be processed and stored within the borders of South Africa, that a copy of all personal data transferred out of South Africa must be stored in the country for the purposes of law enforcement and that ‘[t]o ensure ownership and control … [d]ata generated in South Africa shall be the property of South Africa, regardless of where the technology company is domiciled’.
In turning the Draft NDCP into a final policy, it is perfectly conceivable that the South African government will be contemplating the extent to which using these types of policy levers, especially in the form of mandatory data localisation requirements, are likely to facilitate FDI flowing into the country. In this essay, I broadly contemplate the question of whether this could work – and at what cost, to whom.
South African Distributional Challenges as Context
It is difficult to dispute that the most pressing challenges in contemporary South Africa are those surrounding economic distributions. South African income and wealth inequality is notoriously high, with many a publication referring to South Africa as being one of the most unequal countries on the planet (if not the most). Accordingly to World Inequality Database (WID.world) data, the top 10% of income recipients in the country will receive approximately 66.5% of total pre-tax income in 2021 (compared to 47.3% in 1994), with the top 1% receiving as much as 21.9% (compared to 11.3% in 1994). While the fiscal system has done a fair deal of work to redistribute income, post-tax inequality has also been rising significantly during the country’s transition to democracy. WID.world further tells us that when it comes to total net personal wealth in 2021, an estimated 85.7% of it will accrue to the wealthiest 10% of South Africans, whereas the wealthiest 1% will account for 55% of total net wealth. South Africa’s economy is also characterised by high levels of economic concentration, which has often been converted into disproportionately high mark-ups by global standards. Moreover, according to one study, small and medium enterprises (SMEs) account for more than 98% of businesses in South Africa and between 50% and 60% of the country’s workforce across all sectors. All of these factors should be front and centre in the mind of any South African economic policymaker – they serve as critical context. After all, economic inequality at South African levels is a drag on growth and slows down the rate at which the growth that is realised reduces poverty.
Costs and Benefits: Contemplating the Potential Value of Forcing FDI through Localisation
It should, of course, also be noted that many digital MNEs operate in the South African market without facing much (if any) competition from local firms. Against this backdrop, the South African desire to benefit from data generated in South Africa, from South Africans is most understandable. But is the deployment of data localisation measures to force digital MNEs to invest in the country really the best way to achieve this? The Draft NDCP – with its reliance on the ‘critical information infrastructure’ concept and the clear attempt at assertion of sovereignty over the data generated by South Africans – appears to be looking at the Chinese data localisation model for inspiration. So, in answering this question, let us start by looking at parts of the Chinese experience.
It is true that China has to a large degree, largely on sovereignty grounds, cordoned off its citizens data to the benefit of both state owned and private Chinese firms. Through a host of measures – mandatory localisation being a prominent one among these – the Chinese government has been able to leverage its large population and the size of its internal market to build its own national tech giants such as Baidu and Tencent. This approach has undoubtedly facilitated certain types of investment, with many foreign firms purchasing large stakes in Chinese tech firms. Naspers – one of South Africa’s own largest multinationals – is perhaps a prime example: through Prosus, its Dutch-listed international asset division, it is the single largest shareholder in Tencent, which has consistently been one of the world’s ten largest publicly traded firms since at least 2017.
This does not mean, however, that the Chinese approach is heavily reliant on attracting FDI per se. There are still several examples, of course, where China has used its localisation laws to pressure large foreign multinationals into investing directly into its economy. In 2017, for example, Apple announced that it would invest more than $1 billion in order to build its first data centre in China. The data centre, which was subsequently built (and is operational as of May 2021) in partnership with the Chinese state-owned firm Guizhou Cloud Big Data Industry in Guizhou – a relatively poor province in southern China –in order to comply with data localisation laws. Similarly, in April 2021, after initial resistance, Tesla announced that it, too, would be building a data centre in China. The data centre, Tesla’s first outside of the United States, was recently opened in Shanghai.
That said, it would be very difficult for South Africa to imitate the Chinese model – which according to a number of studies has also had significant costs – for a broad range of reasons. Most important among these, perhaps, is that internal market size matters: according to calculations based on WID.world data, China’s aggregate national income (adjusted for differences in purchasing power) is at least 36 times larger than South Africa’s national income. This leaves South Africa with far less economic clout of the kind necessary to strongarm some of the world’s largest multinationals. There are various other factors that come into play too: for example, South Africa and China’s political regimes are completely different, an upshot of which is that the South African government is far less capable of forcing firms – domestic and foreign alike – to submit to its will (the fact that South Africa is a constitutional democracy with incredibly high levels of economic inequality makes it even for difficult for it to successfully do so).
Additionally, the success of data localisation strategies used by the South African government to attract FDI will necessarily depend on a multiplicity of other factors, too, including the extent of the costs imposed by data localisation requirements – especially for South African SMEs, which, should data localisation measures be taken, would be poorly poised to handle what are likely to be significant additional costs for business across a majority of sectors of the economy in the short to medium-term – how firms operating in the digital economy are taxed, the extent of competition between foreign and local firms and the availability of various types of skilled labour. While detailed economic studies on the exact costs of potential data localisation measures in South Africa are currently lacking (the kinds of studies that would be necessary to thoroughly assess the costs and benefits of prospective localisation), what is quite clear is that even if data localisation measures were implemented and resulted in increased FDI, FDI flows are no panacea when it comes to addressing South Africa’s economic woes.
This is especially true if one has regard to the fact that the focus of efforts to attract FDI through data localisation laws would predominantly be geared at forcing digital MNEs with considerable bargaining power to invest in data centres. This has a number of ramifications. For example, while data centres tend to be profitable for their owners, they do not ordinarily create a great deal of jobs, especially given that much of what goes on inside the typical data centre is automatable (if not already automated). All things being equal, attracting FDI and adding some jobs would naturally be preferable to nothing. But all things would not be equal: data localisation would, as noted above, mean additional costs for businesses across the economy, including those that are in no way associated with large digital MNEs. Moreover, as other contributions to this colloquium, there are a meaningful number of other considerations when adopting data localisation requirements that extend well beyond FDI and costs. These may entail trade offs which South African policymakers will have to consider very carefully.
Whatever policies are adopted in relation to data localisation in South Africa, then – which policies should have regard to economic inequality as context and should not contribute to its further exacerbation – must be carefully thought through in a holistic fashion; they must also allow for flexibility and closely monitored experimentation aimed at continuously gauging whether the benefits of an adopted policy outweigh the costs, especially given that data localisation requirements are a relatively new development in South Africa. Finally, it should be mentioned that there are other policy avenues for creating a more thriving and inclusive digital economy in South Africa. Done right, a number of these, such as those with the goal of fostering more equal access to quality education and digital infrastructure, will bear fruit with little risk of downsides – these policy goals should accordingly be prioritised.
 This essay is largely based on research undertaken for a policy brief produced as part of a Mandela Institute project titled 'Africa's Digital Economy: Protectionism, Development and Democracy'. The policy brief, titled 'Data Localisation in Kenya, Nigeria and South Africa: Regulatory Frameworks, Economic Implications and Foreign Direct Investment', is available here.