The COVID-19 shock and services trade decline: potential for digitalization matters

ABSTRACT

Global services trade declined by 20% during 2020 with significant heterogeneity across countries, geographical regions and sectors. We present stylized facts and provide hypotheses and empirical analysis seeking to explain this heterogeneity. The decline is found to be correlated with COVID-19 case and mortality rates; stringency of imposed lockdowns; the decline in merchandise trade; and with different ways of transacting services trade. The latter depends on the sectoral composition of services trade across countries, which in turn emanates from more fundamental determinants of comparative advantage in services, generating testable hypotheses to explain the observed heterogeneity in services trade decline. Focusing on attributes of digitalization and the role of value-chains, we find that human-capital-intensive countries with favourable digital-trade policies and greater ability to leverage ICT infrastructure were associated with relatively smaller declines. Moreover, the expected role of GVC-integration in accentuating the services trade decline finds little support in empirical results across sectors providing evidence instead for the GVC-resilience narrative.

KEYWORDS:

• Services trade
• COVID-19
• heterogeneity
• modes of supply
• digitalization
• gravity model

JEL CLASSIFICATION:

• F1
• F14

Disclosure statement

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

Data citation

Bilateral services trade data: ADB MRIO database (Available at https://mrio.adbx.online/)

COVID-19 cases and mortalities during 2020: European Centre for Disease Prevention and Control (Available at https://www.ecdc.europa.eu/en/publications-data/download-todays-data-geographic-distribution-covid-19-cases-worldwide)

Digital trade restrictiveness indices (DTRI): ECIPE (Available at https://ecipe.org/dte/database/)

Geographical remoteness: Baldwin, R. and J. Harrigan (2011). Zeros, quality, and space: Trade theory and trade evidence. American Economic Journal: Microeconomics, 3, 60-88.

Head, K., T. Mayer and J. Ries (2010). The erosion of colonial trade linkages after independence. Journal of international Economics, 81(1): 1-14.

GDP, PCGDP and human capital: World Development Indicators. World Bank. (Available at https://databank.worldbank.org/source/world-development-indicators#)

Governance indicators: World Governance Indicators. World Bank.Kaufmann, D., A. Kraay and M. Mastruzzi (2010). The worldwide governance indicators: methodology and analytical issues. Policy Research Working Paper Series 5430, The World Bank.

GVC participation: EORA MRIO database.Lenzen, M., K. Kanemoto, D. Moran and A. Geschke. 2012. Mapping the structure of the world economy. Environmental Science & Technology 46(15): 8374–8381.

Lenzen, M., D. Moran, K. Kanemoto and A. Geschke. 2013. Building Eora: A Global Multi-regional Input-Output Database at High Country and Sector Resolution. Economic Systems Research 25(1):20-49.

Modal shares in total services exports/imports: WTO TiSMoS (Available at https://www.wto.org/english/res_e/statis_e/trade_datasets_e.htm)

Networked Readiness Index (NRI): World Economic Forum (Available at https://reports.weforum.org/global-information-technology-report-2016/networked-readiness-index)

PTA-intensiveness: Hofmann et al. 2019 and OECD-WTO BaTiS

Hofmann, C., A. Osnago and M. Ruta (2019). The Content of Preferential Trade Agreements. World Trade Review, 18(3), 365-398. doi:10.1017/S1474745618000071.

(Available at https://www.wto.org/english/res_e/statis_e/trade_datasets_e.htm)

Quarterly merchandise and services trade data in 2019 and 2020: WTO (Available at https://data.wto.org/)

Stringency of government response: Oxford Stringency Index

Hale, T., A. Petherick, T. Phillips and S. Webster (2020). Variation in government responses to COVID-19. Blavatnik school of government working paper, 31, 2020-11.

(Available at https://www.bsg.ox.ac.uk/research/research-projects/covid-19-government-response-tracker)

Notes

¹ Commercial services are all services (broadly goods-related, travel, transport, construction, distribution, insurance and finance, ICT, business and personal) except for government services..

² These include construction; distribution; insurance; financial; telecommunications; computer and information; charges for the use of intellectual property; personal, cultural and recreational (audio-visual, health and education); and other business services (a diverse category that includes, inter alia, accounting, legal, engineering, research & development, management consulting, and technical services).

³ In WTO GATS parlance, services trade is transacted via four ‘modes of supply’. Mode 1 or ‘cross-border trade’ is the whole range of services that are delivered remotely e.g. BPO services. Mode 2 or ‘consumption abroad’ is the service transacted when the consumer travels to the economy of the supplier e.g. tourism. Mode 3 or ‘commercial presence’ is foreign affiliate activities in the host economy e.g. activties of foreign banks. Mode 4 or ‘movement of natural persons’ is the service delivered by the supplier in the economy of the consumer e.g. onsite software programmers.

⁴ Dingel and Neiman, (2020) find some service activities to be more amenable to WFH, including educational; professional, scientific and technical; management; insurance and finance; and information services in particular, but also wholesale trade; and real estate, rental and leasing services.

⁵ The pandemic also led to substitutability between modes of supply in service delivery. For instance, students switching to online classes instead of moving physically to the country where the educational institutes are located, are an illustration of a Mode 2 service delivered cross-border. Similarly, a professor teaching classes online without travelling to the country where the students are located, is an example of a Mode 4 service also delivered digitally. Thus, the potential for digital delivery is not unique to Mode 1 trade but also depends on other characteristics such as information intensity.

⁶ If the OECD and above-median groups were experiencing a higher YoY growth in services exports before the pandemic (say 5%) than the non-OECD and below-median groups (say 2.5%) and both groups experienced a comparable shock from COVID-19 (say -20%), the data would present a misleading result that the non-OECD and below-median groups experienced more acute YoY declines due to the pandemic.

⁷ We would like to thank an anonymous referee for suggesting these robustness checks. Consistent with these suggestions, the econometric analysis that follows also compares 2020 to the period average of 2017–2019, though the inclusion of time-varying exporter- and importer-fixed effects in those specifications obviates the need for scaling services exports with respect to GDP.

⁸ Data on all variables (except for GVC-participation and share of services exports in PTAs that pertain to 2015 and the Networked Readiness Index and digital-trade-restrictiveness that pertain to 2016), on the basis of which the total sample is ‘split’ above and below median to present stylized facts in this section, pertain to 2017. All variable descriptions and data sources are provided in Appendix A Table 1.

⁹ Bangladesh saw its imports of goods-related services increase over 13 times in 2020 relative to 2019, though most other countries reported a decline in their imports of services across sectors but especially in travel and transport. In a world of regional and global value-chains, the decline in services imports not only has an adverse effect on exports but also on economic activity in general, given increasing servicification, with additional implications for post-pandemic recovery (Shingal, 2020b, 2022).

¹⁰ The Oxford COVID-19 Government Response Tracker (OxCGRT) has been assembling publicly available information on different attributes of COVID-19 incidence (cases, deaths, tests) and policy indicators since the onset of this crisis. There are nine indicators, which record information on containment and closure policies, such as school and workplace closures and restrictions in internal and international movement. A stringency index is constructed on the basis of these indicators to measure the strictness of ‘lockdown style’ policies. The value of the stringency index ranges from 0 to 100, with higher values associated with more restrictions on individual behaviour and economic activities of a country. In the empirical analysis that follows, the stringency index is rescaled to lie between 0 and 1 for ease of interpretation of coefficient estimates (also see Liu et al. 2022).

¹¹ The relationship remains qualitatively similar if we use the average value of the stringency index during June or December 2020 or over the period from 16 March to 31 December 2020. This is not surprising given that a 10% increase in COVID-19 cases during 2020 is found to be associated with a 0.7% rise in the value of the stringency index at the mean.

¹² Foreign affiliate transactions were also affected by lockdowns and containment measures.

¹³ These include maintenance & repair and manufacturing services used as inputs.

¹⁴ At the same time, pandemic-induced problems with freight services, such as lack of truck drivers and ship crews or restrictions on their cross-border movement, adversely affected trade in goods..

¹⁵ This is the latest year for which these data are available in TiSMoS.

¹⁶ Note that TiSMoS, the source of services trade data by mode of supply, is a constructed database that relies on fixed shares for most countries (Wettstein et al. 2017). Thus, some of the export and import allocations by mode of supply in TiSMoS may not be completely accurate. For instance, banking and insurance services are also delivered via commercial presence though TiSMoS data suggest that these are completely delivered cross-border. That said, TiSMoS data provide a reasonable picture of how different services are traded internationally, besides being the only source of multi-sector cross-country services trade data by mode of supply that are publicly available.

¹⁷ These include a wide range of professional services (accounting, legal, R&D, etc.) as well as insurance and financial services.

¹⁸ We would like to thank an anonymous referee for suggesting that we examine the impact of the pandemic in 2020 relative to the trend growth of services exports over 2017–2019 instead of only one preceding year. That said, our overall findings remain qualitatively similar if we consider only two years (2019 and 2020) or organize the data in a panel over four years (2017 to 2020) instead of averaging the data over 2017–2019.

¹⁹ Unfortunately, we do not have a sample of bilateral services trade data for 2021 yet that also include data on intra-national flows. We need the latter to construct the $\mathit{INT}{L}_{\mathit{ij}}$ binary dummy, which is a crucial element of our identification strategy (see below for details). Moreover, WTO data show that several countries were already showing signs of recovery in 2021, which makes 2020 appropriate to examine the immediate impact of COVID-19-induced demand and supply shocks on services trade. Limiting the analysis to 2020 also obviates the need to control for the moderating effects of COVID-19 vaccinations, which nonetheless is an exercise worth-pursuing once more recent services trade data with intra-national flows are available.

²⁰ The data are available at https://mrio.adbx.online/. For the purpose of this analysis, the input-output structure of the data was converted to a panel at the exporter-importer-sector-year level. These data cover 63 reporting and partner countries (of which Rest of the World was excluded from analysis), and 17 disaggregated services sectors that were aggregated into 11 broad commercial services sectors (construction; distribution; maintenance & repair; hotels and restaurants; transport; post and telecommunications; financial intermediation; real estate and other business services; education; health and social work; and other community, social, and personal services).

²¹ Given the way in which multi-regional input-output (MRIO) databases are constructed, the effect of the COVID-19 shock at both exporting- and importing-country levels could not be estimated in a combined regression as one set of explanatory variables was dropped due to collinearity. We thus estimate the effects at the exporting- and importing-country levels in distinct regressions.

²² There is variation in this variable over the years included in the sample. For instance, EU’s agreements with Singapore and Vietnam were only negotiated in 2020, as were Australia’s agreements with Hong Kong, Indonesia and Peru.

²³ The share of the number of COVID-19 cases in population; the share of the number of COVID-19 deaths in COVID-19 cases; and the Oxford stringency index are all exporter-time and importer-time varying variables. Since our econometric specifications include time-varying exporter and importer fixed effects, these variables are completely collinear with the fixed effects and can thus not be estimated directly. However, since the dependent variable includes data on both cross-border ($X_{\mathit{ij}}$) and intra-national flows ($X_{\mathit{ii}}$), we construct a binary dummy, $\mathit{INT}{L}_{\mathit{ij}}$, that takes the value one for every observation in the dataset where the exporting country differs from the importing country. Interacting $\mathit{INT}{L}_{\mathit{ij}}$ with the $\mathit{Covi}{d}_{\mathit{it}}$ and $\mathit{Covi}{d}_{\mathit{jt}}$ variables makes the dimension of the interacted term $\mathit{ijt}$-specific and thus, no longer collinear with the time-varying exporter and importer fixed effects. The effects of the COVID-19 variables can thus be estimated directly in a fully-specified model.

²⁴ The data on the policy variables pertain to an early pre-pandemic period to assuage endogeneity-related concerns in estimation. This is consistent with established practice in this literature, for instance see Freund et al. (2022)..

²⁵ Given the growing servicification of economic activity in countries across the world (WTO, 2019), we prefer using GDP in constructing the measure for geographical remoteness. However, the role of geographical remoteness in explaining the services trade decline and our overall findings remain qualitatively similar if we use services value added instead of GDP in constructing this measure.

²⁶ We used principal component analysis to orthogonalize data on these three variables and then estimated one component as a measure of country ability to leverage ICT infrastructure. The estimated component explained 63% of the variation in the three variables.

²⁷ The Networked Readiness Index (NRI) provides an assessment of the factors, policies and institutions that enable a country to fully leverage ICT for increased competitiveness and well-being. The index maps the network readiness landscape of economies based on their performance in four areas: technology, people, governance, and impact. We use the governance dimension of the NRI to reflect the conduciveness of the policy environment around ICT and digital trade.

²⁸ GVC participation is defined as the sum of backward and forward participation; these terms were constructed using EORA MRIO data for the year 2015 as the share of foreign value added (FVA) and indirect value added (DVX) in gross exports (GX), that serve as measures of backward and forward participation, respectively (for instance see Aslam et al. 2017).

²⁹ To focus more on sector-specific determinants and exploit heterogeneity across sectors, the GVC variables were also constructed at the sector-level and used in sector-specific estimating equations in sensitivity analysis. Our overall findings remained qualitatively similar..

³⁰ Australia, Austria, Bangladesh, Belgium, Bhutan, Brazil, Brunei Darussalam, Bulgaria, Cambodia, Canada, China, Croatia, Cyprus, Czech Republic, Denmark, Estonia, Fiji, Finland, France, Germany, Greece, Hong Kong, Hungary, India, Indonesia, Ireland, Italy, Japan, Kazakhstan, Kyrgyz Republic, Laos, Latvia, Lithuania, Luxembourg, Malaysia, Maldives, Malta, Mexico, Mongolia, Nepal, Netherlands, Norway, Pakistan, Philippines, Poland, Portugal, Romania, Russia, Singapore, Slovak Republic, Slovenia, Spain, South Korea, Sri Lanka, Sweden, Switzerland, Taiwan, Thailand, Turkey, United Kingdom, United States, and Vietnam.

³¹ In the case of the stringency index for instance, this is calculated as $[\mathit{exp}(\mathit{Coefficien}{t}^{\mathit{OSI}\ast \mathit{INTL}}\ast \mathit{s}.\mathit{d}{.}^{\mathit{OSI}}\ast \mathit{Mea}{n}^{\mathit{INTL}})-1]\ast 100$.

³² The effect is computed as $[\mathit{exp}(\mathit{Coefficien}{t}^{\mathit{PTA}})-1]\ast 100.$.

³³ The results reported in Tables 5 and 6 remain qualitatively similar if services exports are destined to meet intermediate or final demand or if the data are organized in a panel comprising both intermediate and final demand in the partner countries.

³⁴ See Francois and Hoekman, (2010) for an excellent early review and Arnold et al. (2011), (2016); Lodefalk, (2014); Beverelli et al. (2017); Hoekman and Shepherd, (2017); Fiorini and Hoekman, (2018) for more recent analysis..