Trade restrictions on digital services and the impact on manufacturing exports

Abstract

In the digital age, digital services not only provide important intermediate inputs for manufacturing but they also affect the availability of other potentially ICT-enabled high-quality services as intermediate inputs. Therefore, it is particularly important and meaningful to explore the relationship between the liberalization of trade in digital services and manufacturing exports. This paper uses a panel dataset of bilateral exports of 15 manufacturing sectors in 54 countries, from 2014 to 2018, to examine the impact of digital services trade policy restrictions on exports of manufactured goods. The main results suggest that the impact of trade policy restrictions on manufacturing exports is mixed but mainly negative. Moreover, regulatory differences in digital services industries between bilateral country pairs also have a significant negative impact on the export performance of manufacturing industries. In addition, we further examine the heterogeneity of this impact mechanism across different policy areas, manufacturing sectors, and bilateral country-pair groups.

KEYWORDS:

• Digital services
• trade policy restrictions
• manufacturing
• exports
• regulatory differences

JEL Classifications:

• F13
• F14
• F60

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Fixed and variable trade costs include developing export markets, building distribution networks, maintaining foreign trade relations, insurance and transportation, etc.

2 WTO (2019) summarizes the benefits of trade liberalization in services into the following three aspects: (i) competition brought by liberalization increases the productivity of domestic services industries (ii) indirectly increases the productivity of manufacturing and services industries through input-output linkages (iii) promotes product innovation and differentiation and enhances the competitiveness of products in the international market, thereby stimulating demand. Of course, this is also a benefit of the liberalization of trade in digital services.

3 Ferencz (2019) suggests that regulation will be a disincentive to trade due to its restrictive, inconsistent or unpredictable characteristics.

4 For details, see ‘Measuring trade in ICT-enabled services: main findings and recommendations stemming from first pilot surveys’ published by UNCTAD in 2018.

5 Direct channels include the impact on the productivity of existing manufacturing firms and the availability of complementary services. Indirect channels include technology spillovers, discouraging firms/entrepreneurs from being involved in cost discovery processes, increasing the cost of manufacturing and hindering participation in global value chains (Su et al. 2020).

6 Ferencz (2019) states that some of the regulatory data on the DSTRI come from the existing STRI database, while the data for new regulatory measures comes from publicly available laws and regulations.

7 For the detailed framework and development methodology of the DSTRI, see Ferencz (2019).

8 There are 12 digital services sectors in total, and Table A2 in the Appendix lists the detailed ISIC Rev.4 2-digit code and sector name for each digital services sector.

9 Detailed classifications and ISIC Rev.4 2-digit codes of 15 manufacturing sectors are shown in the Appendix Table A3.

10 We are grateful to an anonymous reviewer for highlighting this point.

11 We used population-weighted distances since it can be used to consistently calculate both bilateral distances and internal distances.

12 Indirectly obtained through World Integrated Trade Solutions (WITS). We used the simple average most-favored-nation tariff rates. Since the tariffs on goods trade between EU members have been eliminated, and the UNCTAD TRAINS database does not provide tariff data between EU members, we used 0 instead.

13 Gouma and MP (2013) point out that most sector classifications in the ISIC Rev.3 and ISIC Rev.4 are very close, even with one-to-one matching. However, there are a number of newly defined activities at the 4-digit level in the ISIC Rev. 4 that are spread across many 2-digit industry aggregates in the ISIC Rev. 3. But since the trade data and input-output data are provided at the 2-digit level, a detailed and perfectly matched correspondence table does not yet exist. So we had to compromise and only roughly construct the ISIC Rev.3-ISIC Rev.4 correspondence table at the 2-digit level.

14 We did not use the tariffs weighted by trade value because the calculation of the tariffs weighted by trade value needs to rely on the trade value of individual sectors at ISIC Rev.3 2-digit level provided by the UNCTAD Comtrade Database. This is not only different from the ISIC Rev.4 sector classification version of the trade data provided by the OECD TiVA database, but also the recorded trade values will differ due to the different statistical benchmarks between the two databases. Therefore, calculated tariffs weighted by trade values will be highly biased.

15 According to the WITS statistics, the total export value of the 54 countries used in the empirical analysis of this paper accounts for about 81.46% of the world’s total export value for 2018. This means that the empirical analysis of this paper includes most of the world’s major exporting countries and the findings are representative and general.

16 We are grateful to an anonymous reviewer for highlighting this point.

17 Van der Marel and Ferracane (2021) pointed out that data localization measures force multinational enterprises to establish local servers to retain data, which may be a costly investment. In addition, some strictly restrictive data policies will also bring additional compliance costs to multinational enterprises, which will undoubtedly increase the fixed costs of enterprises entering foreign markets.

18 The percentage change in manufacturing exports from a decrease in the DSTRI of exporting and importing countries by one percentage point is calculated as follows: $[\exp (-1\times \mathrm{coefficient})-1]\times 100$.

19 The correlation coefficient between the MFN tariffs and RTA is -0.408, while the correlation coefficient between the effectively applied tariffs and RTA is -0.635.

20 See Appendix Table A3 for details of the industry classifications.

21 Finally, the percentage change in manufacturing exports from a decrease in the $\mathrm{Heterogeneityanswe}{r}_{\mathrm{ijt}}$ by 0.05 points is calculated as follows: $[\text{exp}({0.05}^{\ast }(-1.121+{0.408}^{\ast }\mathrm{DSTR}{I}_{\mathrm{ikt}}+{0.153}^{\ast }\mathrm{DSTR}{I}_{\mathrm{jkt}}))-1{]}^{\ast }100$.

22 Due to collinearity, we could not obtain estimates for the DSTRI of exporters.