Impacts of COVID-19 on Global Value Chains.

We investigate the impacts of COVID-19 on global value chains by examining bilateral trade in finished machinery products from January to June in both 2019 and 2020. We use the numbers of COVID-19 cases and deaths as measures of the impact of the pandemic. Specifically, we investigate how these impacts affect value chains in three scenarios-countries that import finished machinery products, countries that export finished machinery products, and countries that export machinery parts to countries exporting finished machinery products-to assess the impacts on demand, output, and supply chain, respectively. In our analysis, the largest negative impacts were from supply chain effects, followed by output effects. In contrast, we did not find significant impacts from demand effects. We also found that output effects are not so strong in intra-Asian trade compared with trade in other regions.

INTRODUCTION

The ongoing coronavirus pandemic (hereinafter, COVID‐19) has disrupted global value chains (GVCs) worldwide since its beginning. The decrease and delay in materials exported from China has resulted in decreases in production or changes of input sources in many countries. For example, according to an interview survey by the Japan External Trade Organization (JETRO), a Mexican affiliate of a Japanese firm was forced to switch its input sources from China to South Korea. Moreover, the decrease or delay in exports from China has decreased production in ASEAN countries, resulting in a reduction of their exports to Japan. According to the results of a questionnaire by the Japan Institute of Logistics Systems, as a result of the difficulty in importing goods from these countries, a Japanese firm switched its input source to domestic suppliers. These induced switches of input sources negatively affect firms' operations and reduce their exports of final goods.

This study examines the extent to which countries' exports of final goods have been disrupted since COVID‐19 began affecting supplier countries. Conceptually, a supplier country that is affected by the spread of COVID‐19 experiences a drop in the production of inputs. The resulting decrease in the amount of exported inputs and their increased price raises importers' costs and reduces their productivity in terms of final‐good production. This, in turn, reduces the export of the final goods from the input‐importing countries to other countries. Thus, the trade‐reducing effect of COVID‐19 spreads throughout GVCs. We also consider the effects of COVID‐19 on countries that export and import final goods. On the supply side, the decreased workforce and diminished productivity in exporting countries reduce the supply of final products. On the demand side, decreased earnings and lockdown measures in importing countries reduce the demand for final products. We take into account these direct effects of COVID‐19 on trade.

We empirically investigate the impacts of COVID‐19 on GVCs by examining bilateral trade in finished machinery products from January to June in both 2019 and 2020. Between these two periods, the world exports of finished machinery products decreased from US$1,551 billion to US$1,287 billion.1 Our dataset includes exports from 35 countries to 185 countries. To measure the prevalence of COVID‐19, we use the numbers of COVID‐19 cases and deaths collected by the European Centre for Disease Prevention and Control. Using these data, we investigate three country scenarios to capture pandemic‐related economic damage to the GVC: (1) a country that imports finished machinery products; (2) a country that exports finished machinery products; and (3) a country that exports machinery parts to the country exporting finished machinery products. These three country scenarios capture demand, output, and supply chain effects, respectively. We empirically investigate which of these effects has the largest impact on international trade during the COVID‐19 pandemic period.

Our findings can be summarized as follows. First, COVID‐19 did not have a significant effect on demand for finished machinery products in importing countries, whereas the finished machinery trade is significantly hurt by higher rates of COVID‐19 infection in both countries exporting finished machinery products and countries exporting machinery parts to those countries. In short, the impacts of COVID‐19 are primarily on the supply side, affecting both the outputs and inputs that play a crucial role in machinery trade; the impacts on demand play a less significant role. In particular, we found that the supply chain effect has greater impact on the finished machinery trade compared with the output effect. The insignificant result in importing countries is consistent with the finding by Hayakawa and Mukunoki (2020), which find that the pandemic in importing countries had no significant effect on total trade. Second, the output effect is not so strong when looking specifically at intra‐Asian trade. This may be because some Asian countries exempt the machinery industry from workplace‐closure orders due to its importance in the economy.2

The literature on GVCs is growing rapidly.3 Among many studies in this area, the following studies have examined the impacts of COVID‐19 on GVCs, as we do in this study. Inoue and Todo (2020) simulate the economic effect of a possible lockdown of Tokyo on production not only in Tokyo but also in other parts of Japan, through supply chains. They demonstrate that were Tokyo to be locked down for a month, the indirect effect on other regions would be twice as large as the direct effect on Tokyo, leading to a total production loss of 5.3% of the annual GDP. A similar analysis was conducted for the United Kingdom by Pichler et al. (2020), which sheds light on input–output linkages across sectors. George et al. (2020) focus on the epidemiological dynamics, that is, the transmission of diseases across countries and industries, through supply chains. Bonadio et al. (2020) estimate that the COVID‐19 shock is expected to decrease real GDP by 29.6% on average, and a quarter of the total is due to transmission of the shocks through GVCs. These recent studies performed simulation analyses of the effects on GVCs, whereas we examine these effects directly by using worldwide trade data collected during the COVID‐19 pandemic.4

The remainder of this study is organized as follows. Section 2 provides a theoretical background of the possible effects of COVID‐19 on GVCs. After explaining our empirical framework in Section 3, we report our estimation results in Section 4. Lastly, Section 5 presents the conclusion and discusses policy implications.

CONCEPTUAL FRAMEWORK

In this section, we discuss the theoretical background to the effects of COVID‐19 on the trade of machinery goods when exporters procure inputs through GVCs. An exporting country and an importing country of finished machinery products are denoted by country i and country j, respectively. The prevalence of COVID‐19 in country i and in country j affect trade between them. Impacts from COVID‐19 in countries that supply inputs to country i also have an influence on trade. Let country k denote a third country exporting inputs used for the production of the finished machinery products in country i. We summarize the trade effects of COVID‐19 in these countries separately.

We start by theoretically examining the impacts of COVID‐19 in exporting countries. An increase in the number of COVID‐19 cases/deaths in country i reduces the supply of goods because the COVID‐19 infections and deaths reduce labor participation. In addition, lockdown measures and the resulting decrease in mobility within workplaces further decrease productivity, unless remote work is sufficiently effective to maintain production activities. Given that the production of finished machinery products is relatively difficult to do at home or other remote location, COVID‐19 decreases the production of outputs. For instance, Dingel and Neiman (2020) estimate that the share of jobs that can be done at home is 22% in the manufacturing sector. A decrease in outputs, ceteris paribus, reduces exports of these products. We call this negative effect on trade the “output effect.” One countervailing effect is that finished machinery products might be exported more frequently, rather than supplied domestically, if domestic demand also decreases in the exporting country. Thus, although the net effect on trade in exporting countries is not straightforward, the impact of COVID‐19 is expected to decrease exports if the output effect is sufficiently large.

Next, we discuss the impacts of COVID‐19 on importing countries. Higher rates of COVID‐19 in country j can decrease workers' earnings in that country due to a decrease in work hours or the loss of jobs due to lockdown measures. Decreased earnings directly reduce the aggregate demand of the country. For instance, Bekaert, Engstrom, and Ermolov (2020) calculate that a decline in real GDP growth in the United States in the first quarter of 2020 was due mainly to an aggregate demand shock. Lockdown measures also decrease demand for products by restricting people's mobility and access to the retail market. These demand shocks have negative impacts on trade. COVID‐19 may also reduce trade by changing the composition of demands. Although the demands for some essential products, such as medical supplies and food, may increase, the demands for finished machinery products are likely to decrease because they are durable goods that are not purchased on a daily basis. Eaton et al. (2016) suggest that spending on durable goods decreases more than that on nondurable goods when a country is hit by negative demand shocks. Carvalho et al. (2020) use individual transaction‐level data from Spain to show that the COVID‐19 lockdown measures there decreased the market share of durables such as automobiles, computers, and furniture, while the market share of food increased. The demand shift from durable to nondurable goods has a negative effect on the imports of finished machinery products. We sum up these negative effects on trade as the “demand effect.”5

In addition to the direct effects on importing and exporting countries, COVID‐19 impacts in country k, which supplies the inputs to country i that are used to produce finished machinery products, may reduce trade from country i to country j through GVCs. A disruption in input production in country k caused by COVID‐19 decreases the volume of inputs supplied to country i. It also increases the price of those inputs.6 The resulting increase in input costs raises the production cost and decreases the productivity of the exporters in country i, as is suggested by Halpern, Koren, and Szeidl (2015). Blaum, Lelarge, and Peters (2018) also show that increased input costs raise the prices of final products, thereby reducing their demand. These effects through the output–input linkage reduce the output and export of final products. Indeed, Barrot and Sauvagnat (2016) show that, if suppliers are hit by a natural disaster, their customers experience a substantial drop in the sales of their products. Boehm, Flaaen, and Pandalai‐Nayar (2019) show that US firms that relied on Japanese inputs experienced large drops in production after the 2011 earthquake in Japan. Acemoglu and Tahbaz‐Salehi (2020) theoretically investigate how disruptions to supply chains magnify negative shocks. We denote the negative effect on trade of downstream products caused by a negative supply shock in the supplier countries of upstream inputs as the “supply chain effect.”

EMPIRICAL FRAMEWORK

This section presents our empirical framework for investigating the impacts of COVID‐19 on GVCs. Our simple model is as follows: where  is the export value of finished machinery products from countries i to j in time t. As a time‐variant country‐pair element, we introduce a regional trade agreement (RTA) dummy variable that takes a value of one if the two countries are members of the same RTA, and a value of zero otherwise ( ). It controls the trade effect of RTA that became effective during the sample period. The time‐variant exporter/importer characteristics include the respective country's logged GDP (ln and ln ). In this study, we furthermore assume that time‐variant exporter/importer characteristics include the extent of the damage of COVID‐19 in three countries (called COVID variables), including for the importing country ( ), the exporting country ( ), and countries exporting machinery parts to country i ( ). The variable is the country‐pair fixed effects, which control for the time‐invariant country‐pair characteristics such as the geographical distance between the two countries. Macro shocks are captured by year fixed effects, , and is a disturbance term.

Our data sources are as follows. The study examines data from two time periods: January–June in 2019 and January–June in 2020. We obtain monthly data on exports of finished machinery products in reporting countries from the Global Trade Atlas maintained by IHS Markit.7 The 35 reporting countries and their 185 partner countries (i.e., importing countries) in our dataset are listed in Appendix 1.8 Machinery products are defined as those in the general machinery (HS84), electric machinery (HS85), transport equipment (HS86–89), and precision machinery sectors (HS90–92). Kimura and Obashi (2010) carefully classify HS six‐digit codes in these industries into finished products and intermediate products. By using this list, we restrict the study products only to the HS codes that are categorized into finished products and aggregate their exports at a country‐pair level.

As mentioned in Section 1, we use the numbers of COVID‐19 cases and deaths as measures of the impact of the pandemic, with data obtained from the European Centre for Disease Prevention and Control.9 These data were collected on a daily basis from reports from health authorities worldwide. We use the total number of cases and deaths from January to June in 2020.10 The numbers are set to zero for the same January–June period in 2019. Although we use both the numbers of cases and deaths, these two kinds of numbers are not necessarily perfectly correlated because the mortality rate differs widely across countries. Nevertheless, an increase in either number induces the government to implement measures to suppress transmission, which has an effect on people (e.g., stay‐at‐home orders) and companies (e.g., workplace‐closure orders). Furthermore, the strictness of such policies (e.g., recommended vs. required) also differs depending on the number of cases and deaths. Thus, to obtain robust results, we model using both numbers.

Specifically, our COVID variables are constructed as follows. and are the number of cases/deaths in an exporting country and an importing country, respectively. As mentioned above, captures the extent of COVID‐19 impacts in countries supplying machinery parts to the export country, country i. is calculated as the weighted average of the number of cases/deaths among those suppliers to country i. Specifically, this variable is constructed as follows: represents imports of machinery parts from country k in country i in year 2019. We use the share of imports of machinery parts from each country from January to June in 2019 (i.e., the pre‐COVID‐19 period) out of the total imports of machinery parts as weights. Data on imports of machinery parts in the 35 reporting countries are obtained from the Global Trade Atlas. We add a value of one to these three variables and then take their logs.

The data sources for other variables are as follows. We obtain the data on GDP from the World Economic Outlook Database on the IMF website. We use 2018 GDP figures for January–June 2019, and 2019 figures for January–June 2020 because GDP for 2020 has not yet been realized.11 Also, we intend to avoid GDP variables from containing the impacts of COVID‐19 because we capture those impacts solely by our COVID variables. Because we focus on the trade in the first half of each year, we can interpret our inclusion of GDP in the previous year as controlling for the demand/production conditions just before the first half of each year. We obtain the RTA dummy variable from Egger and Larch (2008) and its 2020 update by using the RTA information available on the World Trade Organization website.12 We estimate our equation by the Poisson pseudo maximum likelihood (PPML) method. The basic statistics are reported in Table 1.

Table 1

Basic Statistics

Variable	Obs.	Mean	SD	Min	Max
RTA dummy	11,232	0.354	0.478	0	1
ln importer's GDP	11,232	3.889	2.290	−3.170	9.973
ln exporter's GDP	11,232	6.741	1.284	3.762	9.973
ln (1 + importer's cases)	11,232	3.999	4.576	0	14.767
ln (1 + exporter's cases)	11,232	5.152	5.468	0	14.767
ln (1 + supplier's cases)	11,232	6.370	6.381	0	14.221
ln (1 + importer's deaths)	11,232	2.231	3.035	0	11.745
ln (1 + exporter's deaths)	11,232	3.563	4.036	0	11.745
ln (1 + supplier's deaths)	11,232	4.949	4.960	0	11.236

Source: Authors' calculation.

Before reporting our estimation results, we provide an overview of worldwide COVID‐19 impacts and data on machinery exports. Figure 1 depicts the daily numbers of worldwide COVID‐19 cases and deaths. In March 2020, both the cases and deaths increased exponentially. Therefore, in April, to slow the spread of COVID‐19, many countries imposed some form of restriction on the movement of people and business activities. Several countries declared citywide or nationwide lockdowns. Also, many countries imposed an entry ban on foreign travelers. As a result, the magnitude of increase in cases and deaths plateaued in April. Afterward, however, the numbers of cases and deaths move differently. Although cases started to increase considerably, deaths gradually decreased. That is, we observe a dramatic decline in mortality rates.

Figure 1

The Daily Numbers of COVID‐19 Cases and Deaths in the WorldSource: European Centre for Disease Prevention and Control.

Figure 2 shows the change in world machinery trade in 2020 relative to that in 2019. Since the trade data in China are not available for January and February separately, we examine the sum of trade values in those two months. In January–February, trade in both finished goods and parts was slightly smaller in 2020 compared with that in 2019. As shown in Figure 1, COVID‐19 spread across the globe in March. Consequently, the machinery trade started to dramatically decrease around the world starting in March. This decrease was larger for trade in finished machinery goods than for trade in machinery parts. In May 2020, trade in finished machinery goods and machinery parts was respectively more than 30% and 20% lower compared with 2019 levels. However, machinery trade started to recover in June, perhaps because most of the countries gradually started to lift their lockdown policies.

Figure 2

Global Machnery Trade in 2020 Relative to Trade in 2019Source: Authors' calculation.

Given that Asian countries are major suppliers of machinery products, we examine changes in intra‐Asian trade in machinery industries. In this paper, we define Asia as the 16 countries that have negotiated the regional comprehensive economic partnership; namely, the 10 ASEAN countries as well as Australia, China, India, Japan, South Korea, and New Zealand. These countries have developed sophisticated international production networks. The results are depicted in Figure 3, which shows a slightly different change from Figure 2. The trade of machinery parts in Asia shows a similar trend to that in the world average. Machinery‐parts trade started to decrease, particularly in April, and to recover in June. In contrast, the trade of finished machinery goods has remained at a low but relatively stable level compared with that in 2019, though it decreased considerably in May. Nevertheless, the magnitude of the decrease in machinery trade is smaller in the intra‐Asian trade than in the world average.

Figure 3

Intra‐Asian Machinery Trade in 2020 Relative to Trade in 2019Source: Authors' calculation.

EMPIRICAL RESULTS

This section reports our estimation results. We cluster standard errors by country pairs. Table 2 shows our baseline results.13 In this table, we introduce our COVID variables one by one. The RTA dummy variable and importer's GDP have insignificant coefficients, whereas the coefficients for exporter's GDP are significantly positive. Among the COVID variables, importer country has an insignificant effect. In contrast, both the number of cases and deaths in exporting countries have significantly negative effects on finished machinery trade. Thus, decreases in workforce size and productivity in exporting countries result in decreased trade. These results are consistent with those obtained for total trade in Hayakawa and Mukunoki (2020). Another noteworthy result is that the damage to suppliers as a result of the pandemic has a significantly negative effect on trade. These results imply that the impacts on the supply side, in terms of both outputs and inputs, play a crucial role in machinery trade, rather than affecting trade on the demand side.

Table 2

Baseline Estimation Results

	(1)	(2)	(3)	(4)	(5)	(6)
RTA dummy	0.039	0.031	0.022	−0.009	0.047	0.048
	(0.110)	(0.112)	(0.111)	(0.103)	(0.111)	(0.112)
ln importer's GDP	−0.085	−0.081	−0.181	−0.254	0.004	0.031
	(0.251)	(0.251)	(0.246)	(0.240)	(0.243)	(0.238)
ln exporter's GDP	1.726***	1.710***	1.767***	1.785***	1.457***	1.170***
	(0.343)	(0.345)	(0.322)	(0.311)	(0.350)	(0.392)
ln (1 + importer's COVID)	−0.002	−0.003
	(0.003)	(0.004)
ln (1 + exporter's COVID)			−0.014***	−0.020***
			(0.004)	(0.004)
ln (1 + supplier's COVID)					−0.053**	−0.099***
					(0.026)	(0.029)
COVID measure †	Case	Death	Case	Death	Case	Death
Log pseudolikelihood	−3E+10	−3E+10	−3E+10	−3E+10	−3E+10	−3E+10
Pseudo R‐squared	0.9965	0.9965	0.9966	0.9966	0.9965	0.9966
No. of observations	11,232	11,232	11,232	11,232	11,232	11,232

Note: This table reports the estimation results by the PPML method. The dependent variable is the export value of finished machinery products. The standard errors reported in parentheses are those clustered by country pairs. In all specifications, we control for country‐pair fixed effects and time fixed effects.

COVID measure indicates the measure of COVID‐19 variables. Case and Death represent the numbers of confirmed cases and deaths, respectively.

***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively.

Next, in columns (1) and (2) in Table 3, we introduce our COVID variables simultaneously. The results are qualitatively similar to those in Table 2. Although the demand effect is insignificant for both the cases and deaths, the output and supply chain effects are significantly negative. The average logged‐numbers of importing‐country deaths, exporting‐country deaths, and supplier‐country deaths in 2020 are 4.46, 7.13, and 9.90, respectively.14 Thus, using the estimates in column (2), we can see that importing‐country deaths increase trade by 0.2%, while exporting‐country deaths and supplier‐country deaths decrease trade by 12% and 76%, respectively.15 Thus, on average, the largest negative impact on trade can be found in the supply chain effect, followed by the output effect. One reason for the insignificant result of the demand effect might be that the rapid increase in online shopping and teleworking raised the import demands for electric machinery products, such as personal computers, smartphones, microphones, and cameras.

Table 3

Robustness Checks

	(1)	(2)	(3)	(4)	(5)	(6)
RTA dummy	0.029	0.006	0.027	−0.022	0.029	0.004
	(0.113)	(0.106)	(0.256)	(0.242)	(0.112)	(0.106)
ln importer's GDP	−0.087	−0.131	−0.524	−0.601	−0.092	−0.153
	(0.231)	(0.224)	(0.512)	(0.463)	(0.233)	(0.226)
ln exporter's GDP	1.505***	1.334***	3.397***	2.478***	1.465***	1.287***
	(0.348)	(0.388)	(0.807)	(0.665)	(0.351)	(0.384)
ln (1 + importer's COVID)	0.000	0.000	−0.003	0.003	0.000	0.001
	(0.003)	(0.003)	(0.008)	(0.008)	(0.003)	(0.003)
ln (1 + exporter's COVID)	−0.013***	−0.017***	−0.038***	−0.040***	−0.014***	−0.018***
	(0.004)	(0.005)	(0.010)	(0.011)	(0.004)	(0.005)
ln (1 + supplier's COVID)	−0.049*	−0.077**	−0.203***	−0.343***	−0.051**	−0.073**
	(0.025)	(0.031)	(0.071)	(0.061)	(0.026)	(0.029)
COVID measure †	Case	Death	Case	Death	Case	Death
Trade period	Jan–June	Jan–June	June	June	Jan–June	Jan–June
Covid period	Jan–June	Jan–June	Jan–June	Jan–June	Jan–May	Jan–May
Log pseudolikelihood	−2.6E+10	−2.6E+10	−1.2E+10	−1.1E+10	−2.6E+10	−2.6E+10
Pseudo R‐squared	0.9966	0.9966	0.9905	0.9912	0.9966	0.9966
No. of observations	11,232	11,232	10,184	10,184	11,232	11,232

Note: This table reports the estimation results by the PPML method. The dependent variable is the export value of finished machinery products. The standard errors reported in parentheses are those clustered by country pairs. In all specifications, we control for country‐pair fixed effects and time fixed effects.

COVID measure indicates the measure of COVID‐19 variables. Case and Death represent the numbers of confirmed cases and deaths, respectively.

***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively.

The large contribution of the supply chain effect suggests that the negative effects of COVID‐19 on trade are larger for firms that have developed GVCs with countries more seriously hit by COVID‐19. It also implies that firms in supplier countries did not have sufficient inventory of inputs. From another point of view, however, the large supply chain effect indicates that trade of finished machinery products will recover sharply if the COVID‐19 cases/deaths decline in supplier countries. For instance, Meijerink, Hendriks, and van Bergeijk (2020) suggest that world merchandise trade after the COVID‐19 pandemic seems to follow a sharp “V‐shaped” pattern and recovery from the trade collapse appears to be taking place much more quickly than those of the Global Financial Crisis in 2008–9. Restored supply chains in machinery products might drive a faster recovery.

In the remaining columns, we account for the gap in months between shipments and contracts to some extent. Specifically, in columns (3) and (4), we replace the dependent variable with the trade values only in June, and we replace the variables for COVID‐19 with those from January to May in columns (5) and (6).16 The results of the COVID variables in both robustness checks show similar results to those in columns (1) and (2).

The results of some more robustness checks are available in Appendix 2. First, to take logs, we added a value of one to the COVID variables. Our results do not qualitatively change even if adding a very small number other than one (Appendix Table 1). Second, in our study sample, the numbers of exporting and importing countries are 35 and 185, respectively. To check if the insignificant results for importer's COVID‐19 are not driven by an asymmetric study sample, we estimate our model for the symmetric study sample by restricting importing countries only to the 35 reporting countries. The results are reported in Appendix Table 2 but again show the insignificant coefficients for importer's COVID‐19. Finally, in the previous tables, we used the number of cases or deaths as a measure of COVID. As its alternative measure, we also use their ratio to total population, which is named COVID per capita. The results are reported in Appendix Table 3 and again show the robust results of the significantly negative coefficients for supplier's COVID‐19.

App. Table 1

Adding a Small Number to COVID‐19 Variables

	(1)	(2)	(3)	(4)	(5)	(6)
RTA dummy	0.036	0.032	0.05	0.049	0.036	0.029
	(0.112)	(0.111)	(0.255)	(0.253)	(0.112)	(0.110)
ln importer's GDP	−0.045	−0.028	−0.426	−0.346	−0.049	−0.053
	(0.241)	(0.238)	(0.535)	(0.493)	(0.242)	(0.241)
ln exporter's GDP	1.489***	1.250***	3.331***	2.243***	1.467***	1.258***
	(0.351)	(0.387)	(0.849)	(0.680)	(0.356)	(0.386)
ln (1.E‐06 + importer's COVID)	0.001	0.001	−0.001	0.001	0.001	0.001
	(0.001)	(0.002)	(0.004)	(0.004)	(0.001)	(0.002)
ln (1.E‐06 + exporter's COVID)	−0.004**	−0.004*	−0.013***	−0.010*	−0.004**	−0.005**
	(0.002)	(0.002)	(0.005)	(0.005)	(0.002)	(0.002)
ln (1.E‐06 + supplier's COVID)	−0.047*	−0.086***	−0.197**	−0.362***	−0.047*	−0.072**
	(0.027)	(0.031)	(0.082)	(0.065)	(0.028)	(0.031)
COVID measure †	Case	Death	Case	Death	Case	Death
Trade period	Jan–June	Jan–June	June	June	Jan–June	Jan–June
Covid period	Jan–June	Jan–June	Jan–June	Jan–June	Jan–May	Jan–May
Log pseudolikelihood	−2.7E+10	−2.6E+10	−1.2E+10	−1.1E+10	−2.7E+10	−2.6E+10
Pseudo R‐squared	0.9965	0.9966	0.9902	0.9908	0.9965	0.9966
No. of observations	11,232	11,232	10,184	10,184	11,232	11,232

Note: This table reports the estimation results by the PPML method. The dependent variable is the export value of finished machinery products. The standard errors reported in parentheses are those clustered by country pairs. In all specifications, we control for country‐pair fixed effects and time fixed effects.

COVID measure indicates the measure of COVID‐19 variables. Case and Death represent the numbers of confirmed cases and deaths, respectively.

***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively.

App. Table 2

Estimation Results with the Symmetric Dataset

	(1)	(2)	(3)	(4)	(5)	(6)
RTA dummy	0.01	−0.019	−0.101	−0.165	0.01	−0.021
	(0.118)	(0.112)	(0.263)	(0.248)	(0.118)	(0.111)
ln importer's GDP	0.075	0.014	−0.277	−0.383	0.063	−0.022
	(0.354)	(0.340)	(0.839)	(0.748)	(0.356)	(0.342)
ln exporter's GDP	1.293***	1.157**	3.192***	2.354***	1.252***	1.103**
	(0.422)	(0.459)	(0.995)	(0.807)	(0.425)	(0.454)
ln (1 + importer's COVID)	−0.002	−0.002	−0.009	−0.005	−0.002	−0.002
	(0.003)	(0.004)	(0.009)	(0.009)	(0.003)	(0.004)
ln (1 + exporter's COVID)	−0.014***	−0.018***	−0.041***	−0.044***	−0.015***	−0.020***
	(0.005)	(0.005)	(0.013)	(0.013)	(0.005)	(0.005)
ln (1 + supplier's COVID)	−0.041	−0.065*	−0.192**	−0.323***	−0.044	−0.063**
	(0.029)	(0.034)	(0.084)	(0.073)	(0.029)	(0.032)
COVID measure †	Case	Death	Case	Death	Case	Death
Trade period	Jan–June	Jan–June	June	June	Jan–June	Jan–June
Covid period	Jan–June	Jan–June	Jan–June	Jan–June	Jan–May	Jan–May
Log pseudolikelihood	−1.4E+10	−1.3E+10	−6.7E+09	−6.0E+09	−1.4E+10	−1.3E+10
Pseudo R‐squared	0.9967	0.9968	0.9900	0.9910	0.9967	0.9968
No. of observations	2,290	2,290	2,234	2,234	2,290	2,290

Note: This table reports the estimation results by the PPML method. The dependent variable is the export value of finished machinery products. The standard errors reported in parentheses are those clustered by country pairs. In all specifications, we control for country‐pair fixed effects and time fixed effects. In this estimation, we restrict importing countries only to the same countries as exporting countries.

COVID measure indicates the measure of COVID‐19 variables. Case and Death represent the numbers of confirmed cases and deaths, respectively.

***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively.

App. Table 3

Estimation Results for COVID Per Capita

	(1)	(2)	(3)	(4)	(5)	(6)
RTA dummy	0.055	0.016	0.101	−0.018	0.053	0.019
	(0.111)	(0.106)	(0.250)	(0.232)	(0.110)	(0.105)
ln importer's GDP	−0.116	−0.29	−0.734*	−1.106**	−0.181	−0.302
	(0.231)	(0.241)	(0.440)	(0.523)	(0.234)	(0.245)
ln exporter's GDP	0.923**	0.462	1.383**	0.538	0.744*	0.415
	(0.398)	(0.414)	(0.644)	(0.866)	(0.409)	(0.434)
Importer's COVID per capita	−4.602	−55.531	−11.969	−59.18	−6.082	−45.438
	(3.927)	(47.637)	(8.469)	(126.260)	(5.840)	(49.840)
Exporter's COVID per capita	−11.317**	−23.957	−38.703***	−130.892	−13.018*	−33.722
	(4.915)	(62.395)	(11.595)	(175.248)	(7.134)	(69.977)
Supplier's COVID per capita	−54.5***	−1,011.1***	−194.5***	−2,955.9***	−74.4***	−1,062.3***
	(15.5)	(236.5)	(28.2)	(654.0)	(22.2)	(282.7)
COVID measure †	Case	Death	Case	Death	Case	Death
Trade period	Jan–June	Jan–June	June	June	Jan–June	Jan–June
Covid period	Jan–June	Jan–June	Jan–June	Jan–June	Jan–May	Jan–May
Log pseudolikelihood	−2.6E+10	−2.6E+10	−1.1E+10	−1.1E+10	−2.6E+10	−2.6E+10
Pseudo R‐squared	0.9966	0.9966	0.9911	0.9907	0.9966	0.9966
No. of observations	11,232	11,232	10,184	10,184	11,232	11,232

Note: This table reports the estimation results by the PPML method. The dependent variable is the export value of finished machinery products. The standard errors reported in parentheses are those clustered by country pairs. In all specifications, we control for country‐pair fixed effects and time fixed effects.

COVID measure indicates the measure of COVID‐19 variables. Case and Death represent the numbers of confirmed cases and deaths, respectively.

***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively.

In Figure 3, we found a slightly different trend for machinery trade in Asia, compared with the world average. To investigate differences in the effects of COVID‐19 on trade, we introduce the interaction terms of our COVID variables with an Asian dummy (Asia), which takes a value of one for intra‐Asian trade. The results are reported in Table 4. As for Table 3, we examine the dependent variable and COVID variables measured by different time periods. Non‐interacted COVID variables have similar results to those in Table 3. Although the coefficients for importing‐country impacts are insignificant, exporting‐ and supplier‐country cases and deaths have significantly negative coefficients. The coefficients for the interaction terms with the Asian dummy are significant only in the case of exporters.17 The coefficients for the interaction terms between exporting‐country impacts and the Asian dummy are estimated to be significantly positive, indicating that exporting‐country COVID‐19 cases and deaths do not have strong negative effects on the intra‐Asian trade of finished machinery products.

Table 4

Asia Effects

	(1)	(2)	(3)	(4)	(5)	(6)
RTA dummy	0.045	0.015	0.103	0.019	0.045	0.015
	(0.112)]	(0.105)	(0.252)	(0.234)	(0.111)	(0.104)
ln importer's GDP	−0.181	−0.184	−0.679	−0.688	−0.184	−0.206
	(0.238)	(0.233)	(0.498)	(0.467)	(0.240)	(0.235)
ln exporter's GDP	1.247***	1.195***	2.553***	1.873***	1.195***	1.116***
	(0.381)	(0.413)	(0.790)	(0.691)	(0.382)	(0.413)
ln (1 + importer's COVID)	0.005	0.002	0.014	0.013	0.006	0.003
	(0.005)	(0.005)	(0.014)	(0.010)	(0.005)	(0.004)
ln (1 + importer's COVID) × Asia	0.000	0.005	−0.009	−0.004	0.000	0.006
	(0.007)	(0.008)	(0.016)	(0.017)	(0.007)	(0.008)
ln (1 + exporter's COVID)	−0.015***	−0.017***	−0.051***	−0.049***	−0.015***	−0.019***
	(0.006)	(0.006)	(0.012)	(0.013)	(0.006)	(0.006)
ln (1 + exporter's COVID) × Asia	0.016**	0.013	0.064***	0.056***	0.016**	0.015*
	(0.007)	(0.009)	(0.014)	(0.019)	(0.007)	(0.009)
ln (1 + supplier's COVID)	−0.058**	−0.077**	−0.244***	−0.349***	−0.061**	−0.075**
	(0.026)	(0.032)	(0.064)	(0.061)	(0.026)	(0.029)
ln (1 + supplier's COVID) × Asia	−0.006	−0.007	−0.031*	−0.025	−0.007	−0.009
	(0.009)	(0.009)	(0.018)	(0.018)	(0.009)	(0.009)
COVID measure †	Case	Death	Case	Death	Case	Death
Trade period	Jan–June	Jan–June	June	June	Jan–June	Jan–June
Covid period	Jan–June	Jan–June	Jan–June	Jan–June	Jan–May	Jan–May
Log pseudolikelihood	−2.6E+10	−2.6E+10	−1.1E+10	−1.1E+10	−2.6E+10	−2.6E+10
Pseudo R‐squared	0.9966	0.9967	0.9909	0.9914	0.9966	0.9967
No. of observations	11,232	11,232	10,184	10,184	11,232	11,232

Note: This table reports the estimation results by the PPML method. The dependent variable is the export value of finished machinery products. The standard errors reported in parentheses are those clustered by country pairs. In all specifications, we control for country‐pair fixed effects and time fixed effects.

COVID measure indicates the measure of COVID‐19 variables. Case and Death represent the numbers of confirmed cases and deaths, respectively.

***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively.

It is worth discussing these results on the interaction terms. The insignificant results in the interaction with supplier‐country impacts may be because firms in Asia tend to use just‐in‐time (JIT) production systems in machinery industries. Thus, because they do not have enough inventory for machinery parts, COVID‐19 impacts in supplier countries negatively affect the exports of finished machinery products, as in the case of other regions.18 In contrast, the nonnegative impact of COVID‐19 in exporting country on intra‐Asian trade might be because firms in Asia retained some finished machinery products in their inventories. That is, their inventory adjustment may enable manufacturing firms in Asia to continue to export, even when their production operations are decreased or delayed. Indeed, according to questionnaires conducted by JETRO and some organizations in Malaysia, Indonesia, and the Philippines, many Japanese manufacturing firms reacted to workplace‐closure orders by adjusting the amount of final products held in their inventories.19

Another important reason for the nonnegative effect of COVID‐19 impacts in exporting countries involved in intra‐Asian trade is that factory operations were often exempt from lockdown and workplace‐closure orders. In many countries, factory operations were banned by workplace‐closure orders. In Asian countries, however, some industries were permitted to operate if adequate infection control measures (e.g., social distancing) were taken. This typically includes those that produce essential products, such as medical supplies and equipment and food products. Some Asian countries also allow operations to continue in export‐oriented companies, firms in Special Economic Zones, or industries that require production to maintain the supply chain. Permission for these industries to operate can be found in some Asian countries, including China, India, Malaysia, and the Philippines, though it is also observed in other regions (e.g., Mexico). Given that such exemptions generally apply to the machinery industry, the negative effects of COVID‐19 on exporting countries involved in machinery trade may have been mitigated.20

CONCLUDING REMARKS

This paper investigates the supply chain effect of COVID‐19 on trade by using monthly trade data of finished machinery products. Although there have been some simulation analyses examining the impacts of COVID‐19 on GVCs, this is the first paper to use observed trade data to examine how the COVID‐19 pandemic has disrupted trade of final goods through input–output linkage. Our empirical findings indicate that exports of final goods decrease if an exporting country imports inputs from countries more seriously affected by COVID‐19. We have also confirmed that COVID‐19 impacts on exporting countries have a significantly negative effect on their exports. From a quantitative viewpoint, the former supply chain effect was found to be larger than the latter output effect. In contrast, the COVID‐19 impacts in importing countries do not have significant effects on trade. These results indicate that both the output effect and the supply chain effect play a key role in identifying the impacts of COVID‐19 on trade in a world with increasing connectedness through GVCs.

We also find that the negative effects of COVID‐19 on exporting countries are smaller in intra‐Asian trade. This result is perhaps because some Asian countries permit the machinery industry to operate even when workplace‐closure orders are in effect or because manufacturing firms in Asia can mitigate negative supply chain shocks by adjusting their existing inventory of finished goods. In short, our results suggest that addressing the supply chain effect is most important for ameliorating the impacts of the COVID‐19 pandemic on trade. It does not mean, however, that firms should shift their input sourcing from foreign suppliers to domestic suppliers. The localized sourcing of inputs instead exacerbate the negative impacts of COVID‐19 on the domestic economy. A diversification of input sourcing including both domestic and international markets and “dual sourcing” of the same inputs from suppliers in different countries should help prevent the spread of negative shocks through value chains. Improving inventory management and providing exceptions of lockdown policies for manufacturing of key inputs will also mitigate the negative effects. We hope this paper sheds new light on how to confront these challenges.

App. Table 4

Intra‐Asian Trade versus Asian Exports to Non‐Asia

	(1)	(2)	(3)	(4)	(5)	(6)
RTA dummy	0.028	0.026	0.073	0.055	0.031	0.032
	(0.109)	(0.108)	(0.254)	(0.252)	(0.108)	(0.108)
ln importer's GDP	−0.141	−0.175	−0.611	−0.721*	−0.135	−0.185
	(0.223)	(0.222)	(0.392)	(0.384)	(0.223)	(0.223)
ln exporter's GDP	0.799	0.890*	1.094	0.931	0.678	0.718
	(0.507)	(0.517)	(0.680)	(0.672)	(0.499)	(0.518)
ln (1 + importer's COVID)	0.002	0.000	−0.016*	−0.01	0.003	0.001
	(0.006)	(0.005)	(0.009)	(0.009)	(0.006)	(0.005)
ln (1 + importer's COVID) × ExAsia‐ImAsia	0.003	0.006	0.022*	0.02	0.004	0.007
	(0.008)	(0.008)	(0.012)	(0.015)	(0.008)	(0.008)
ln (1 + importer's COVID) × ExAsia‐ImNAsia	0.006	0.003	0.064***	0.051***	0.005	0.002
	(0.013)	(0.011)	(0.020)	(0.018)	(0.013)	(0.011)
ln (1 + exporter's COVID)	−0.032***	−0.026***	−0.126***	−0.104***	−0.038***	−0.031***
	(0.010)	(0.009)	(0.019)	(0.017)	(0.010)	(0.009)
ln (1 + exporter's COVID) × ExAsia‐ImAsia	0.035***	0.023**	0.147***	0.117***	0.042***	0.029***
	(0.011)	(0.010)	(0.022)	(0.021)	(0.012)	(0.011)
ln (1 + exporter's COVID) × ExAsia‐ImNAsia	0.040***	0.026**	0.143***	0.117***	0.046***	0.032**
	(0.013)	(0.012)	(0.023)	(0.022)	(0.013)	(0.013)
ln (1 + supplier's COVID)	−0.067**	−0.066*	−0.263***	−0.288***	−0.076**	−0.073**
	(0.031)	(0.035)	(0.055)	(0.064)	(0.031)	(0.033)
ln (1 + supplier's COVID) × ExAsia‐ImAsia	−0.025*	−0.016	−0.130***	−0.095***	−0.031**	−0.021*
	(0.013)	(0.011)	(0.024)	(0.022)	(0.013)	(0.011)
ln (1 + supplier's COVID) × ExAsia‐ImNAsia	−0.035**	−0.019	−0.171***	−0.133***	−0.040**	−0.023*
	(0.016)	(0.014)	(0.024)	(0.023)	(0.016)	(0.013)
COVID measure †	Case	Death	Case	Death	Case	Death
Trade period	Jan–June	Jan–June	June	June	Jan–June	Jan–June
Covid period	Jan–June	Jan–June	Jan–June	Jan–June	Jan–May	Jan–May
Log pseudolikelihood	−2.5E+10	−2.5E+10	−9.6E+09	−9.7E+09	−2.5E+10	−2.5E+10
Pseudo R‐squared	0.9967	0.9967	0.9921	0.9921	0.9967	0.9967
No. of observations	11,232	11,232	10,184	10,184	11,232	11,232

Note: This table reports the estimation results by the PPML method. The dependent variable is the export value of finished machinery products. The standard errors reported in parentheses are those clustered by country pairs. In all specifications, we control for country‐pair fixed effects and time fixed effects. ExAsia‐ImAsia takes a value of one if both exporting and importing countries are Asian countries and zero otherwise. ExAsia‐ImNAsia takes a value of one if an exporting country is an Asian country while an importing country is a non‐Asian country.

COVID measure indicates the measure of COVID‐19 variables. Case and Death represent the numbers of confirmed cases and deaths, respectively.

***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively.