← 1. In emerging fields of study such as this, questionnaires can be useful tools for obtaining background information to help guide research efforts. They are not, however, without caveats. For example, the questionnaire can suffer from biases arising from: i) the channels of distribution; ii) the modes of this distribution; and iii) self-selection. For inference to be made about the larger population of firms an appropriate sampling strategy, ensuring the representativeness of the sample, is needed. Obtaining such a representative sample, in this instance, would stretch beyond the resources available for this work and therefore the results of the questionnaire are to be seen as initial information on which observations or “rebuttable presumption” for deeper analysis can be made.
← 2. The aim of the questionnaire was to garner information about how business use data and the economic implications of data regulation. To ensure that data was comparable, it was important to undertake the analysis at the firm level rather than to undertake a separata questionnaire targeted to consumers or to business associations.
← 3. The distribution was somewhat skewed towards the OECD economies. Responses were received from business located across 32 economies, of which 18 were OECD economies. Japan was most represented with 25 responses, followed by the United States with 9 responses, Chinese Taipei with 6, Ireland (5), and the United Kingdom (3). In terms of sectors, coverage was more spread out with responses from businesses engaged in agriculture, manufacturing, and services. In terms of size distribution, the sample was heavily biased towards larger firms, which represented 75% of responses.
← 4. It is important to note that firms that deal with personal data might be more likely to have responded to the questionnaire as they are more likely to be aware of the issues because there is more regulation aimed at privacy/personal data and, as a result, might be more motivated to respond.
← 5. These are the weighted averages of the response by categories taking the middle of the range as the point of reference. That is, if 30% of respondents claim that costs increase by 6-10% then we assign a weight of 30% to the cost increase of 8% (middle of the distribution). We then take the sum of these multiplications across all categories.
← 6. The question asked was: “Do you think that regulations that require safeguarded transfers of data across borders can have a positive impact on trust and therefore sales?”
← 7. To obtain this number, the average cost increase in the given bracket was taken and multiplied by the share of responses that chose that bracket. For cost increases that led business to claim that they would stop their activities costs of 500% are assumed.
← 8. Trust is a complex phenomenon to capture, including in the context of personal data protection (Acquisti, Taylor and Wagman, 2016[38]). In this paper, trust is assumed to be demand enhancing. Ceteris paribus, a consumer will demand more from a firm that safeguards its information over one that does not.
← 9. For example, an economy that combines a stringent ad-hoc data transfer mechanisms (Category 3) with a prohibitive data localisation policy would not see a change in the (trade) costs and (trust) benefits of data policies.
← 10. Geoeconomic blocs are defined based on UN voting patterns and more specifically the difference in voting of economies relative to the United States and China. Following (Métivier et al., 2023[40])three groups are defined, a Western group, an Eastern group, and a group of non-aligned economies.
← 11. There are also different categories of localisation requirements with varying degrees of restrictiveness. They are distinct from data transfer policies and an economy can have any combination of them, although economies with restrictive data transfer policies tend to have restrictive localisation requirements as well. The four scenarios discussed above do not include any policy shocks related to data localisation. Instead, the potential economic impacts of localisation are explored with four separate scenarios.
← 12. In Scenario A, data management costs and trade costs can change for economies with prohibitive data localisation policies, since prohibitive localisation policies are no longer inhibiting data flows. In Scenario 1 only trade costs and WTP change. Data management costs are not projected to change, since there are no economies with ad hoc data flow authorisation policies nor an absence of data localisation policies.
← 13. These fixed effects help control for multilateral resistance which can lead to unobserved heterogeneity biasing coefficients (Annex B).
← 14. This can be done by multiplying the DSTRI coefficient in the gravity equation, , by the change in the DSTRI that would arise from a change in the data regulation, , divided by the trade elasticity, : . See also López González, Sorescu and Kaynak (2023[35]).
← 15. DSTRI scores are the same for economies in Category 1 and 2, respectively Open and Pre-authorised Safeguards. The reason is that DSTRI scores do not aim to capture policies into that much detail. Correspondingly, the counterfactual policy scenarios are necessarily stylised.
← 16. Since these costs reflect increased production costs, gravity estimates, which are related to trade costs specifically, cannot be employed for data localisation policies.
← 17. The reported costs are the projected additional data management costs, average over all sectors and economies. The values are larger than in Figure 3.7, which reports the increase in total costs in each of the sectors reported emerging from the need to procure more data management services.
