Impacts of Vaccination on International Trade During the Pandemic Era

Abstract

This paper examines how COVID‐19 vaccinations change international trade. We analyze monthly level trade data from January 2020 to March 2022 that cover the bilateral exports from 40 reporting countries to 220 partner countries. Our findings can be summarized as follows. On average, the effects of vaccination rates in importing and exporting countries on exports were found to be insignificant. When considering the income level, we also did not find significant effects of vaccination rates in high‐ and low‐income importing countries on exports. In contrast, the rise of vaccination rates in low‐income exporting countries significantly increased their exports though no significant increase in exports was detected when vaccination rates rose in high‐income exporting countries. These results imply that since low‐income countries are mainly engaged in labor‐intensive industries, the relaxation of lockdown orders (i.e., movement and gathering restrictions) driven by the rise of vaccination rates plays a crucial role in production activities in low‐income countries.

1. INTRODUCTION

The coronavirus (COVID‐19) pandemic has negatively affected trade on a global scale. To prevent the spread of the COVID‐19 virus, most countries implemented various kinds of nonpharmaceutical interventions, especially those to restrict people's movement. The stay‐at‐home orders decreased consumption opportunities, shrinking the imports of goods. Similarly, work‐from‐home impositions decreased production in factories, decreasing the export of goods. Thus, the severity of COVID‐19 damages led to a decrease in both exports and imports. According to the press release (PRESS/876, March 31, 2021) issued by the World Trade Organization (WTO), the global merchandise trade in 2020 decreased by 5.3%, proving that the COVID‐19 pandemic has contributed significantly to the decreasing global trade.

Against this backdrop, vaccinations are expected to play a key role in increasing international trade. International trade has experienced many external shocks, including various natural disasters and financial crises. These shocks are mostly temporary in nature. For example, although serious natural disasters may destroy factories, the economy gradually recovers by rebuilding newer factories. In the case of the ongoing COVID‐19 pandemic, the economy cannot fully recover until this virus is contained. To this end, the development of antiviral drugs and vaccines is indispensable. As of May 2022, various drugs and vaccines are available, although their effects are not necessarily clear. In particular, since the end of 2020, many countries have started vaccination drives, which are expected to contribute to economic recovery.

In this study, we examine how vaccinations change international trade during the pandemic era by using bilateral export data from 40 countries to 220 countries between the period of January 2020 and March 2022. Since COVID‐19 situations change from moment to moment, we use monthly trade data. In our empirical analysis, we assume that international trade depends on the restrictiveness of lockdown orders (i.e., movement and gathering restrictions) enforced in exporting and importing countries, as mentioned above. Then, we further assume that governments determine the order restrictiveness based on the number of newly confirmed cases and deaths and the vaccination rates. As a result, we regress bilateral trade values on these elements in exporting and importing countries. To control for unobservable factors, furthermore, we introduce various fixed effects. We estimate this model by the Poisson pseudo maximum likelihood (PPML) method.

Our main findings are summarized as follows. First, we empirically confirmed that lockdown restrictions decreased with an increase in vaccination rates. Second, on average, the effects of vaccination rates in importing and exporting countries on trade were found to be insignificant. Third, we did not find significant effects of vaccination rates in both high‐ and low‐income importing countries on trade. In contrast, the rise of vaccination rates in low‐income exporting countries significantly increased their exports though no significant increase in exports was detected when vaccination rates rose in high‐income exporting countries. Since low‐income countries are mainly engaged in labor‐intensive industries, the relaxation of lockdown orders driven by the rise of vaccination rates plays a key role in production activities in low‐income countries.

Our study is related to two strands of literature on the COVID‐19 pandemic. One is the literature on vaccination for COVID‐19, which is rapidly growing. Several studies have analyzed the optimal vaccination policy in terms of allocation (Babus, Das, and Lee 2020; Luyten, Tubeuf, and Kessels 2020; Agarwal et al. 2021; Vellodi and Weiss 2021; Forslid and Herzing 2021), subsidy (Bosi, Camacho, and Desmarchelier 2020; Goodkin‐Gold et al. 2020), speed (Gros and Gros 2021), distribution sites (Chevalier et al. 2021), and willingness to pay (Cerda and García 2021; Dong et al. 2020; García and Cerda 2020; Sarasty et al. 2020; Wong et al. 2020). For example, Gollier (2021) shows that vaccinating low‐risk people in vaccine‐rich countries before high‐risk people in vaccine‐poor countries worsens the global welfare consequences of the pandemic. Similarly, Więcek et al. (2021) examine the nexus among supply speed, efficacy, amount per dose, the interval between the first and second doses, and mortality. Some studies investigate vaccine hesitancy to seek the policy measure to increase its demand (Alsan and Eichmeyer 2021; Gans 2021; Thunström et al. 2020; Duquette 2020; Neumann‐Böhme et al. 2020; Bughin et al. 2021; Lazarus et al. 2021).

The other is the literature that analyzes the effect of COVID‐19 on international trade. There are several empirical studies on goods trade (e.g., Evenett et al. 2022; Friedt and Zhang 2020; Hayakawa and Mukunoki 2021a, 2021b, 2021c; Kejzar and Velic 2020; Meier and Pinto 2020). These studies find that the severity of COVID‐19 damages in both exporting and importing countries leads to a decrease in the goods trade. Additionally, some studies show the propagation of such negative effects throughout supply chains. In particular, some studies demonstrate the role of various tools/strategies in mitigating the negative effects of COVID‐19 damages on international trade. For example, Ando and Hayakawa (2022) find that the negative effects on supply chains are mitigated when import sources are diversified. Moreover, Hayakawa, Mukunoki, and Urata (2022) show that the negative effect of COVID‐19 on importing countries decreases with developing e‐commerce. Here, we shed light on the role of vaccinations in international trade.1

The rest of this paper is organized as follows. The next section overviews the time‐series changes of COVID‐19 cases, vaccination doses, and international trade, followed by our empirical framework in Section 3. Our estimation results and conclusions are present in Sections 4 and 5, respectively.

2. OVERVIEW: CONFIRMED CASES, VACCINES, AND TRADE

This section gives an overview of the severity of COVID‐19 damages and the progress of vaccinations in addition to global trade. The number of cases has risen explosively since the World Health Organization (WHO) declared COVID‐19 a pandemic on March 11, 2020. At the same time, the virus has continued to mutate, and several variants have emerged. The expert group convened by the WHO has recommended using letters of the Greek alphabet for specific variants.2 The Beta variant was first documented in South Africa in May 2020, followed by the Alpha variant in the United Kingdom in September that year. Subsequently, the Delta and Gamma variants were documented in India in October and Brazil in November 2020, respectively. The Omicron variant was documented in November 2021. As of September 2022, these variants are designated as variants of concern. Furthermore, several other variants, which are designated as variants of interest, have been identified.

Figure 1 shows the aggregated monthly number of newly confirmed COVID‐19 cases by region. The data on this number are obtained from the COVID‐19 Data Repository maintained by the Center for Systems Science and Engineering at Johns Hopkins University.3 After the WHO's declaration in March 2020, the number of confirmed cases experienced an explosive rise. Until the third quarter of 2020, South Asia and Latin America & Caribbean were the main regions with the highest number of cases. After the Alpha variant was documented in the United Kingdom, the number of cases increased sharply in Europe & Central Asia, and North America. The global number hit its first peak in January 2021 and decreased in February. Indeed, since February 2021, the number of cases gradually decreased in Europe & Central Asia, and North America. However, the global number hit a second peak, which was higher than the first, in April 2021 because of the explosive rise in South Asia. The third peak was found in August 2021 due to the rise in East Asia & Pacific, Europe & Central Asia, and North America. However, the number of cases in these peaks were trivial compared with the number in the peak found in January 2022, which shows explosive increases in all regions except for sub‐Saharan Africa.

Figure 1.

The Number of Newly Confirmed Cases of COVID‐19 (1,000 persons) Source: The COVID‐19 Data Repository by the Center for Systems Science and Engineering (CSSE) at Johns Hopkins University.Note: This figure shows the aggregated number of monthly confirmed new cases of COVID‐19 by region. [Colour figure can be viewed at wileyonlinelibrary.com]

Scientists have developed potential vaccines for COVID‐19. According to the WHO website,4 at least 13 different vaccines have been globally administered, several types of which include inactivated or weakened virus vaccines, protein‐based vaccines, viral vector vaccines, and RNA and DNA vaccines. Evenett et al. (2021) show the high concentration and self‐reliance in COVID‐19 vaccine production among a group of 13 countries,5 which they refer to as the “Vaccine Club.” Many countries have made deals with vaccine manufacturers or pharmaceutical companies for vaccination doses. Furthermore, dozens of countries have obtained vaccines through COVAX, a consortium backed by the WHO to guarantee fair and equitable access to every country in the world. According to the WHO website,6 “[b]ased on what we know so far, vaccines are proving effective against existing variants, especially at preventing severe disease, hospitalization, and death.”

Figure 2 depicts the cumulative number of COVID‐19 vaccine doses administered at the beginning of each month. The data are obtained from the COVID‐19 Data Repository. The first mass vaccination program started in early December 2020, and the number has risen explosively during the study period. As of February 2021, vaccinations are concentrated in three regions including East Asia & Pacific, Europe & Central Asia, and North America. Other regions, especially South Asia, have gradually increased the number of vaccinations, with the largest number of doses credited to East Asia & Pacific, attributed mainly to the active vaccinations in China. Nevertheless, because of the global supply shortage, vaccination doses have been biased in rich countries. Indeed, Evenett et al. (2021) show that the Vaccine Club accounted for 60% of the total confirmed advance purchasing agreements with pharmaceutical companies for vaccination doses.

Figure 2.

The Cumulative Number of Vaccine Doses (1,000 doses) Source: The COVID‐19 Data Repository by the Center for Systems Science and Engineering (CSSE) at Johns Hopkins University.Note: This figure shows the cumulative number of COVID‐19 vaccine doses administered as of the beginning of each month. [Colour figure can be viewed at wileyonlinelibrary.com]

Last, we take an overview of global trade by employing monthly trade data drawn from the Global Trade Atlas by IHS Markit.7 Figure 3 shows the changes of global exports in 37 countries from January 2020 to January 2022.8 We use export statistics to avoid possible time‐lag issues on monthly import data, particularly for long‐distance trade via sea freight/sea cargo. This issue is relevant since our analysis is conducted at a monthly level, not an annual level. We aggregate these exports by months and normalize monthly global exports so that those values in January 2020 become a value of 1. As found in the figure, although exports decreased in April 2020, world exports have experienced a rapid recovery since June 2020.

Figure 3.

Evolution of the World Exports (January 2020 = 1) Source: Global Trade Atlas. [Colour figure can be viewed at wileyonlinelibrary.com]

3. EMPIRICAL FRAMEWORK

This section presents our empirical framework to examine how vaccinations change international trade. As discussed in Section 1, COVID‐19 has negative effects on international trade partly due to the imposed lockdown. In importing countries, the pandemic decreased consumption opportunity through stay‐at‐home orders, which further worsens business performance and thus lowers revenues and income. Such decreases in both consumption opportunity and income result in the shrinking of the demand size and imports of goods. Similarly, work‐from‐home orders in exporting countries decrease factory production (Dingel and Neiman 2020). Furthermore, the infection control measures in factories, such as social distancing, may lower productivity (Dutcher 2012; Etheridge, Tang, and Wang 2020), which in turn results in reducing production sizes and thus exports of goods.

In sum, the severity of COVID‐19 damages leads to a decrease in both exports and imports of goods through lockdown orders in exporting and importing countries. We formalize this structure simply as follows:

$${\mathit{\text{Trade}}}_{\mathit{\text{ijym}}}=\exp \left({\alpha }_1{\mathit{\text{Stringency}}}_{\mathit{iym}}+{\alpha }_2{\mathit{\text{Stringency}}}_{\mathit{jym}}+{\delta }_{\mathit{ijy}}+{\delta }_{\mathit{ijm}}+{\delta }_{\mathit{ym}}\right)\bullet {\epsilon }_{\mathit{\text{ijym}}},$$ (1)

where ${\mathit{\text{Trade}}}_{\mathit{\text{ijym}}}$ indicates export values from country i to country j in month m of year y. ${\mathit{\text{Stringency}}}_{\mathit{iym}}$ is the degree of lockdown orders' strictness in country i in month m of year y. Three kinds of fixed effects are denoted by ${\delta }_{\mathit{ijy}}$, ${\delta }_{\mathit{ijm}}$, and ${\delta }_{\mathit{ym}}$, which include other elements that explain bilateral trade and are discussed below. The term ${\epsilon }_{\mathit{\text{ijym}}}$ is a disturbance term. The stricter the lockdown orders in exporting and importing countries are, the smaller the bilateral trade values are. Thus, the coefficients for the two stringency variables are expected to be negative.

The government determines the restrictiveness of lockdown orders based on various elements, including the number of confirmed cases, that of fatalities, or their growth rates. After the vaccination drives, vaccination rates would be one of the key factors in this choice. Therefore, we formalize this choice as follows:

$${\mathit{\text{Stringency}}}_{\mathit{iym}}={\beta }_1\ln \left(1+{\mathit{\text{Cases}}}_{\mathit{iym}}\right)+{\beta }_2\ln \left(1+{\mathit{\text{Deaths}}}_{\mathit{iym}}\right)+{\beta }_3{\mathit{\text{Vaccines}}}_{\mathit{iym}}+{\rho }_i+{\rho }_{\mathit{ym}}+{\omega }_{\mathit{iym}}.$$ (2)

${\mathit{\text{Cases}}}_{\mathit{iym}}$ and ${\mathit{\text{Deaths}}}_{\mathit{iym}}$ are the number of newly confirmed cases and deaths, respectively. ${\mathit{\text{Vaccines}}}_{\mathit{iym}}$ is the vaccination rate (the total number of people who received at least one vaccine dose or all doses prescribed by the initial vaccination protocol per 100 people in the total population). The equation also includes two kinds of fixed effect (${\rho }_i$ and ${\rho }_{\mathit{ym}}$) and an error term (${\omega }_{\mathit{iym}}$).

Since the increase in confirmed cases or deaths induces governments to introduce more restrictive lockdown orders, their coefficients are expected to be positive. Furthermore, as discussed in the previous section, vaccines are believed to be effective in the prevention of severe disease, hospitalization, and death. Thus, even if the number of confirmed cases increases, restrictive orders may not be imposed if the number of vaccinated persons is considerably large. Indeed, some countries relaxed the restrictiveness of lockdown orders depending on the share of vaccinated persons (e.g., 70% of the total population in the United States). In sum, the increase in vaccination rates is expected to lower the restrictiveness of lockdown orders. Since the total number of populations does not change much in the short run, as long as controlling for country fixed effect (${\rho }_i$), the coefficients for the number of cases and deaths can also be interpreted as indicating the effects of their numbers per population. The year‐month fixed effect may capture the major type of variants in the world.

Substituting equation (2) into equation (1), we obtain

$${\mathit{\text{Trade}}}_{\mathit{\text{ijym}}}=\begin{array}{l}\exp ({\gamma }_1\ln \left(1+{\mathit{\text{Cases}}}_{\mathit{iym}}\right)+{\gamma }_2\ln \left(1+{\mathit{\text{Deaths}}}_{\mathit{iym}}\right)+{\gamma }_3{\mathit{\text{Vaccines}}}_{\mathit{iym}}\\ +{\gamma }_4\ln \left(1+{\mathit{\text{Cases}}}_{\mathit{jym}}\right)+{\gamma }_5\ln \left(1+{\mathit{\text{Deaths}}}_{\mathit{jym}}\right)+{\gamma }_6{\mathit{\text{Vaccines}}}_{\mathit{jym}}\\ +{\vartheta }_{\mathit{ijy}}+{\vartheta }_{\mathit{ijm}}+{\vartheta }_{\mathit{ym}})\bullet {ϵ}_{\mathit{\text{ijym}}}.\end{array}$$ (3)

Based on the discussion above, the coefficients for the number of cases and deaths in both exporting and importing countries are expected to be negative, while their vaccination rates will have positive coefficients. Three kinds of fixed effects are denoted by ${\vartheta }_{\mathit{ijy}}$, ${\vartheta }_{\mathit{ijm}}$, and ${\vartheta }_{\mathit{ym}}$. The term ${ϵ}_{\mathit{\text{ijym}}}$ is a disturbance term. This equation is our main equation to be estimated by the PPML method. Specifically, we estimate it for monthly bilateral exports from 40 countries to 220 countries from January 2020 to March 2022.

Fixed effects in equation (3) control for various elements, with country pair–year fixed effects (${\vartheta }_{\mathit{ijy}}$) capturing the effects of standard gravity variables such as geographical distance in addition to importer's annual demand sizes and exporter's annual factor prices (e.g., wages). Additionally, this type of fixed effect works to control for the total population. Country pair–month fixed effects (${\vartheta }_{\mathit{ijm}}$) control for the seasonality of trade between the two countries. The country pair component in these fixed effects addresses an endogeneity concern on the variable of vaccination doses. When a country is a manufacturer of COVID‐19 vaccines, more doses of vaccines may be supplied to its partner country if the latter is a closer trade partner. Such an inherent trade link is controlled by the country pair component in the fixed effects above. Moreover, the number of confirmed cases may contain an error depending on the country's ability to detect the COVID‐19 cases. If the magnitude of such an error is related to the country's economic development level, it is controlled by the country–year component. Last, ${\vartheta }_{\mathit{ym}}$ represents the year–month fixed effects. As mentioned in equation (2), this type of fixed effect may capture the major type of variants in the world.

It is worth discussing three empirical issues. First, in equation (1), we consider that COVID‐19 damages affect trade only through changing the strictness of lockdown orders. However, the number of cases and deaths will also have direct effects on trade. For example, regardless of the order strictness, their large numbers imply that the number of healthy persons who can go out shopping or to work is small. Also, knowing their huge numbers, people may hesitate to go out and stay voluntarily at home. Therefore, the coefficients for these numbers in equation (3) include not only the indirect effect via lockdown orders but also such direct effects. Second, we also examine the effects of vaccination rates in high‐ and low‐income countries separately because of their possible differences. Last, although we avoid omitted variable bias by introducing detailed fixed effects, we also control for the existence of regional trade agreements (RTAs) by introducing a dummy variable RTA defined at a monthly level. This variable is constructed using the database updated by Egger and Larch (2008) and the website of the WTO.

The data are obtained from the same sources as in the previous section. Namely, we obtain the data on COVID‐19‐related variables from the COVID‐19 Data Repository, while the trade data are drawn from the Global Trade Atlas. There are four data issues. First, we restrict the study period to the COVID‐19 pandemic era (i.e., since 2020) because our interest lies in the effects of vaccinations rather than those of coronavirus. Second, depending on the trade data availability, our dataset for estimation is the unbalanced panel data. The export data particularly in March 2022 are available only for 14 countries. Third, in this trade database, China reported only aggregated trade in January and February of 2020. Therefore, we simply drop observations for these months that are reported by China. Fourth, most types of vaccines require two doses. In the choice of restrictiveness in lockdown orders (i.e., equation 2), the number of people who received at least one dose may have different effects from that of people who completed two doses. Thus, we examine vaccination rates in terms of both two numbers.9

4. EMPIRICAL RESULTS

This section presents our estimation results.10 Before estimating the main equation (i.e., equation 3), we estimate equation (2) to see how the restrictiveness of lockdown orders is associated with COVID‐19‐related variables. Notice that the more restrictive orders contribute to lowering the number of cases and deaths. Namely, the estimators in equation (2) suffer from the endogeneity bias sourced from the reverse causality. Specifically, the error term in equation (2) is negatively correlated with the number of cases and deaths. Thus, when we estimate equation (2) by the ordinary least square (OLS) method, the estimators for these numbers suffer from downward bias, and the positive coefficients are underestimated. However, we do not find suitable instruments for these two variables. Furthermore, the direction of bias is underestimation. Thus, we estimate this equation by the OLS method though the magnitude of the coefficients no longer indicates the causal effects.

As the dependent variable, we examine an index on the strictness of lockdown‐style policies that primarily restrict people's behavior. Specifically, we use the Government Response Stringency Index drawn from Hale et al. (2021), which is a composite measure based on nine response indicators including school closures, workplace closures, and travel bans. We rescale this index to a value from 0 to 1. The higher index indicates that more restricted measures are effective. Although the original data are available at a daily frequency, we compute the simple average for each month. We estimate this equation for 185 countries from December 2020 to March 2022, when vaccination programs were ongoing.

The estimation results are shown in column (1) in Table 1. The standard errors are clustered by country. In column (1), we use vaccination rates in terms of the number of people who received at least one dose. As is consistent with our expectation, the coefficients for the number of cases and deaths are significantly positive, while that for vaccination rates is significantly negative. These results are unchanged when we use vaccination rates in terms of the number of people who completed two doses, as shown in column (2). One interesting difference between the two columns is that the absolute magnitude of the coefficient for vaccination rates is larger in column (2). This result may indicate that the rate of completely vaccinated persons is more crucial in the government decision on the restrictiveness of lockdown orders. In columns (3) and (4), we introduce the interaction terms of vaccination rates with a dummy variable on high‐income countries.11 However, their coefficients are insignificantly estimated, indicating no significant differences in the lockdown stringency–vaccination rates nexus between high‐ and low‐income countries.

Table 1.

Regression of Stringency Index

	(1)	(2)	(3)	(4)
ln(1 + cases)	1.166** (0.495)	1.320*** (0.491)	1.283** (0.506)	1.394*** (0.502)
ln(1 + deaths)	2.003*** (0.412)	1.871*** (0.409)	1.890*** (0.426)	1.812*** (0.417)
Vaccines	−12.249*** (3.122)	−16.126*** (3.112)	−8.659** (4.206)	−12.779*** (4.656)
Vaccines  × high income			−5.380 (4.694)	−4.464 (5.082)
Vaccine measure	One	Full	One	Full
No. of observations	2,936	2,936	2,936	2,936
Adjusted R 2	0.649	0.655	0.65	0.655

Note: The dependent variable is a stringency index. This table reports the estimation results obtained using the OLS method. The standard errors reported in parentheses are those clustered by country. In all specifications, we control for country fixed effects and year‐month fixed effects. The measure of vaccines for “one” is the total number of people who received at least one vaccine dose per 100 people in the total population, while that for “full” is the total number of people who received all doses prescribed by the initial vaccination protocol per 100 people in the total population.

*** and ** indicate the 1% and 5% levels of statistical significance, respectively.

We also estimate equation (1) by the PPML to see the overall effect of lockdown order restrictiveness on trade. The results are shown in column (1) in Table 2. The standard errors are clustered by country pair. The coefficients for both the importer's and exporter's stringency are estimated to be significantly negative. Thus, the more restrictive lockdown order in importing or exporting countries decreases international trade significantly. The coefficient for the RTA dummy is also significantly positive, indicating the trade creation effect of RTAs even on a monthly basis (notice that we control for country pair–year fixed effects). In the previous section, we pointed out the direct effects of COVID‐19 cases and deaths on trade. To examine these effects, we introduce the number of cases in column (2), that of deaths in column (3), and both numbers in column (4). While we can find the significant direct effects of cases reported in exporting and importing countries, the direct effects of deaths, especially those in exporting countries, are insignificant. After introducing these numbers, importer's stringency still has significantly negative coefficients, while the coefficients for exporter's stringency turn out to be insignificant. The former result indicates that vaccination rates may be a remaining part that explains importer's stringency and has significant effects on trade. On the other hand, vaccination rates in exporting countries may not account for such a part.

Table 2.

PPML Estimation Results

	(1)	(2)	(3)	(4)
Importer's stringency	−0.1305*** (0.0318)	−0.0524* (0.0274)	−0.0644** (0.0287)	−0.0525* (0.0277)
Exporter's stringency	−0.0703*** (0.0271)	−0.0269 (0.0259)	−0.0348 (0.0238)	−0.0294 (0.0247)
ln(1 + importer's cases)		−0.0083*** (0.0019)		−0.0079** (0.0034)
ln(1 + importer's deaths)			−0.0068*** (0.0018)	−0.0004 (0.0032)
ln(1 + exporter's cases)		−0.0047** (0.0023)		−0.0060* (0.0035)
ln(1 + exporter's deaths)			−0.0032 (0.0026)	0.0015 (0.0038)
RTA	0.0513* (0.0281)	0.0412* (0.0241)	0.0398 (0.0249)	0.0423* (0.0243)
No. of observations	186,884	186,884	186,884	186,884
Pseudo R 2	0.998	0.998	0.998	0.998

Note: The dependent variable is trade values. This table reports the estimation results obtained using the PPML method. The standard errors reported in parentheses are those clustered by country pair. In all specifications, we control for country pair–year fixed effects, country pair–month fixed effects, and year‐month fixed effects.

***, **, and * indicate the 1%, 5%, and 10% levels of statistical significance, respectively.

The PPML estimation results for equation (3) are reported in Table 3. The standard errors are clustered by country pair. In this table, we use vaccination rates in terms of the number of people who received at least one dose. Due to the high correlation between the number of cases and deaths, we introduce these numbers separately in columns (1) and (2). Both numbers are simultaneously introduced in column (3). Column (1) shows that the number of cases in importing and exporting countries have significantly negative coefficients. In column (2), the number of deaths in importing countries has a significantly negative coefficient but the number in exporting countries has a less significant coefficient. In column (3), while the results on the number of cases are unchanged, those on the number of deaths turn out to be insignificant. Furthermore, in all the three columns, coefficients for vaccination rates in exporting and importing countries are insignificant, indicating that vaccination drives do not contribute to a significant trade increase on average. In Table 4, we use vaccination rates in terms of the number of people who completed two doses and obtained similar results.

Table 3.

PPML Estimation Results

	(1)	(2)	(3)
ln(1 + importer's cases)	−0.0101*** (0.0020)		−0.0090*** (0.0034)
ln(1 + importer's deaths)		−0.0090*** (0.0019)	−0.0013 (0.0031)
Importer's vaccines	0.0217 (0.0439)	0.0251 (0.0447)	0.0231 (0.0454)
ln(1 + exporter's cases)	−0.0057*** (0.0022)		−0.0060* (0.0035)
ln(1 + exporter's deaths)		−0.0046* (0.0025)	0.0003 (0.0039)
Exporter's vaccines	0.0739 (0.0523)	0.0743 (0.0530)	0.0723 (0.0528)
RTA	0.0356 (0.0232)	0.0314 (0.0233)	0.0353 (0.0233)
No. of observations	186,884	186,884	186,884
Pseudo R 2	0.998	0.998	0.998

Note: The dependent variable is trade values. This table reports the estimation results obtained using the PPML method. The standard errors reported in parentheses are those clustered by country pair. In all specifications, we control for country pair–year fixed effects, country pair–month fixed effects, and year–month fixed effects. The measure of vaccines is the total number of people who received at least one vaccine dose per 100 people in the total population.

*** and * indicate the 1% and 10% levels of statistical significance, respectively.

Table 4.

PPML Estimation Results: Full‐Vaccination Rates

	(1)	(2)	(3)
ln(1 + importer's cases)	−0.0103*** (0.0020)		−0.0089*** (0.0035)
ln(1 + importer's deaths)		−0.0092*** (0.0018)	−0.0015 (0.0031)
Importer's vaccines	−0.0390 (0.0388)	−0.0410 (0.0389)	−0.0391 (0.0396)
ln(1 + exporter's cases)	−0.0057*** (0.0022)		−0.0062* (0.0033)
ln(1 + exporter's deaths)		−0.0046* (0.0025)	0.0006 (0.0038)
Exporter's vaccines	0.0295 (0.0386)	0.0273 (0.0386)	0.0291 (0.0388)
RTA	0.0343 (0.0244)	0.0307 (0.0243)	0.0341 (0.0243)
No. of observations	186,884	186,884	186,884
Pseudo R 2	0.998	0.998	0.998

Note: The dependent variable is trade values. This table reports the estimation results obtained using the PPML method. The standard errors reported in parentheses are those clustered by country pair. In all specifications, we control for country pair–year fixed effects, country pair–month fixed effects, and year–month fixed effects. The measure of vaccines is the total number of people who received all doses prescribed by the initial vaccination protocol per 100 people in the total population.

*** and * indicate the 1% and 10% levels of statistical significance, respectively.

In Table 5, we introduce the interaction terms of vaccination rates with the high‐income country dummy. In columns (1)–(3), we use vaccination rates in terms of the number of people who received at least one dose, while the number of people who completed two doses is used for vaccination rates in column (4). The results indicate that the coefficients for importer's vaccination rates and their interaction terms are mostly insignificant. Thus, importer's vaccination rates do not have significant effects on international trade. On the other hand, we can see a clear contrast in the results for exporter's vaccination rates. The non‐interacted exporter's vaccination rates have significantly positive coefficients, while the coefficients for their interaction terms are significantly negative. Furthermore, the absolute magnitude is almost the same. These results imply that exporter's vaccination rates increase trade when exporters are low‐income countries but not when they are high‐income countries. Specifically, in low‐income countries, a 1 percentage point rise in vaccination rates increases their exports by around 0.17%.12

Table 5.

PPML Estimation Results: High‐Income versus Low‐Income Countries

	(1)	(2)	(3)	(4)
ln(1 + importer's cases)	−0.0094*** (0.0019)		−0.0096*** (0.0030)	−0.0092*** (0.0031)
ln(1 + importer's deaths)		−0.0079*** (0.0018)	0.0004 (0.0028)	−0.0007 (0.0028)
Importer's vaccines	0.0079 (0.0524)	0.0169 (0.0521)	0.0014 (0.0519)	−0.0393 (0.0502)
Importer's vaccines × high income	0.0567 (0.0352)	0.0467 (0.0344)	0.0599* (0.0342)	0.0405 (0.0385)
ln(1 + exporter's cases)	−0.0066*** (0.0023)		−0.0006 (0.0039)	−0.0033 (0.0035)
ln(1 + exporter's deaths)		−0.0076*** (0.0029)	−0.0067 (0.0048)	−0.0039 (0.0043)
Exporter's vaccines	0.1566*** (0.0466)	0.1741*** (0.0466)	0.1768*** (0.0464)	0.1219*** (0.0340)
Exporter's vaccines × high income	−0.1659*** (0.0336)	−0.1900*** (0.0381)	−0.1880*** (0.0414)	−0.1843*** (0.0365)
RTA	0.0552** (0.0255)	0.0460* (0.0250)	0.0521** (0.0254)	0.0388 (0.0251)
Vaccine measure	One	One	One	Full
No. of observations	186,884	186,884	186,884	186,884
Pseudo R 2	0.998	0.998	0.998	0.998

Note: The dependent variable is trade values. This table reports the estimation results obtained using the PPML method. The standard errors reported in parentheses are those clustered by country pair. In all specifications, we control for country pair–year fixed effects, country pair–month fixed effects, and year–month fixed effects. The measure of vaccines for “one” is the total number of people who received at least one vaccine dose per 100 people in the total population, while that for “full” is the total number of people who received all doses prescribed by the initial vaccination protocol per 100 people in the total population.

***, **, and * indicate the 1%, 5%, and 10% levels of statistical significance, respectively.

One may propose another indirect effect of vaccinations on trade, which is the one through lowering the number of cases and deaths. Therefore, in order to uncover the whole effect of vaccinations on trade, it may be interesting to entroduce only vaccination rates in exporting and importing countries in equation (3) by dropping their number of cases and deaths. However, this estimation is problematic. While vaccination rates affect the number of cases and deaths, countries with higher numbers try to raise vaccination rates. Namely, since there is reverse causality, omitting the number of cases and deaths from equation (3) yields endogeneity bias in the estimators of vaccination rates. In particular, since the error term in equation (3) without these numbers is positively correlated with vaccination rates, the estimators in vaccination rates suffer from upward bias, and their positive coefficients are overestimated. Therefore, although vaccination rates may have significant effects on trade via changing the number of cases and deaths, our estimation results suggest at least that vaccination rates do not have additional effects on trade once we control for those numbers. Such additional effects can be found only when low‐income countries are exporters.

Last, it is worth discussing such additional effects of vaccination rates. The insignificant effects of importer's vaccination rates indicate that the change of the lockdown stringency driven by importer's vaccination rates does not affect trade significantly. This result may suggest a change of lifestyle. As a consequence of the pandemic, society has gradually become familiar with online shopping as one of the major styles of purchase. According to eMarketer, retail e‐commerce sales in the world grew by 26.4% in 2020 and 16.3% in 2021. As a result, the relaxation of lockdown orders based on the vaccination rate may not significantly change people's consumption amount and thus imports. On the other hand, the rise of vaccination rates in low‐income exporters increases their exports, while trade does not change according to the vaccination rates in high‐income exporters. This difference may be because low‐income countries are mainly engaged in labor‐intensive industries. Thus, the relaxation of restrictive orders driven by the rise of vaccination rates plays a key role in production activities in low‐income countries. In high‐income countries, on the other hand, since the manufacturing process tends to be automated, production operations are relatively feasible with a small number of onsite workers even under the restrictive order.

5. CONCLUDING REMARKS

This paper empirically investigated how vaccinations change trade. To do that, we examined the bilateral trade data from January 2020 to March 2022. As a result, we found that the rise of vaccination rates in low‐income exporters significantly increases their exports. On the other hand, no significant increases of trade were detected when vaccination rates rose in importers or high‐income exporters. Vaccines were not available in most low‐income countries, especially in the initial period of vaccination drives (i.e., the former half of 2021). However, due to their industry structure, which has comparative advantages in labor‐intensive industries, low‐income countries require more vaccines to continue production activities than high‐income countries. As introduced in Section 1, a previous study demonstrated that vaccinating low‐risk people in vaccine‐rich countries before high‐risk people in vaccine‐poor countries worsens the global welfare. Similarly, our results may suggest that the global trade and welfare improve by vaccinating people in countries with comparative advantages in labor‐intensive industries before people in countries with comparative advantages in automated and modern industries.

APPENDIX 1. STUDY COUNTRIES

185 Countries in Table 1

ABW, AFG, AGO, ALB, AND, ARE, ARG, AUS, AUT, AZE, BDI, BEL, BEN, BFA, BGD, BGR, BHR, BHS, BIH, BLR, BLZ, BMU, BOL, BRA, BRB, BRN, BTN, BWA, CAF, CAN, CHE, CHL, CHN, CIV, CMR, COD, COG, COL, CPV, CRI, CUB, CYP, CZE, DEU, DJI, DMA, DNK, DOM, DZA, ECU, EGY, ERI, ESP, EST, ETH, FIN, FJI, FRA, FRO, GAB, GBR, GEO, GHA, GIN, GMB, GRC, GRD, GRL, GTM, GUM, GUY, HKG, HND, HRV, HTI, HUN, IDN, IND, IRL, IRN, IRQ, ISL, ISR, ITA, JAM, JOR, JPN, KAZ, KEN, KGZ, KHM, KIR, KOR, KWT, LAO, LBN, LBR, LBY, LIE, LKA, LSO, LTU, LUX, LVA, MAC, MAR, MCO, MDA, MDG, MEX, MLI, MLT, MMR, MNG, MOZ, MRT, MUS, MWI, MYS, NAM, NER, NGA, NIC, NLD, NOR, NPL, NZL, OMN, PAK, PAN, PER, PHL, PNG, POL, PRI, PRT, PRY, PSE, QAT, ROU, RUS, RWA, SAU, SDN, SEN, SGP, SLB, SLE, SLV, SMR, SOM, SRB, SSD, SUR, SVK, SVN, SWE, SWZ, SYC, SYR, TCD, TGO, THA, TJK, TKM, TLS, TON, TTO, TUN, TUR, TWN, TZA, UGA, UKR, URY, USA, UZB, VEN, VIR, VNM, VUT, YEM, ZAF, ZMB, ZWE.

40 Exporting Countries in Tables 2, 3, 4, 5

ARG, AUS*, AUT*, BEL*, BRA, CAN*, CHE*, CHN, CIV, DEU*, DNK*, ESP*, FIN*, FRA*, GBR*, GRC*, HKG*, IDN, IND, IRL*, ISR*, ITA*, JPN*, KEN, KOR*, LUX*, MEX, MYS, NLD*, NZL*, PHL, PRT*, RUS, SGP*, SWE*, THA, TWN*, USA*, VNM, ZAF (* indicates high‐income countries).

220 Importing Countries in Tables 2, 3, 4, 5

ABW*, AFG, AGO, AIA, ALB, AND*, ARE*, ARG, ARM, ATF, ATG*, AUS*, AUT*, AZE, BDI, BEL*, BEN, BFA, BGD, BGR, BHR*, BHS*, BIH, BLR, BLZ, BMU*, BOL, BRA, BRB*, BRN*, BTN, BWA, CAF, CAN*, CCK, CHE*, CHL*, CHN, CIV, CMR, COG, COK, COL, COM, CPV, CRI, CUB, CXR, CYM*, CYP*, CZE*, DEU*, DJI, DMA, DNK*, DOM, DZA, ECU, EGY, ERI, ESH, ESP*, EST*, ETH, FIN*, FJI, FLK, FRA*, FRO*, FSM, GAB, GBR*, GEO, GHA, GIB*, GIN, GLP, GMB, GNB, GNQ, GRC*, GRD, GRL*, GTM, GUF, GUY, HKG*, HND, HRV*, HTI, HUN*, IDN, IND, IRL*, IRN, IRQ, ISL*, ISR*, ITA*, JAM, JOR, JPN*, KAZ, KEN, KGZ, KHM, KIR, KNA*, KOR*, KWT*, LAO, LBN, LBR, LBY, LCA, LKA, LSO, LTU*, LUX*, LVA*, MAC*, MAR, MDA, MDG, MDV, MEX, MHL, MKD, MLI, MLT*, MMR, MNG, MNP*, MOZ, MRT, MSR, MTQ, MUS, MWI, MYS, NAM, NCL*, NER, NFK, NGA, NIC, NIU, NLD*, NOR*, NPL, NRU, NZL*, OMN*, PAK, PAL, PAN*, PCN, PER, PHL, PLW*, PNG, POL*, PRI*, PRK, PRT*, PRY, QAT*, REU, ROM, RUS, RWA, SAU*, SDN, SEN, SGP*, SHN, SLB, SLE, SLV, SMR*, SOM, SPM, STP, SUR, SVK*, SVN*, SWE*, SWZ, SYC*, SYR, TCA*, TCD, TGO, THA, TJK, TKL, TKM, TON, TTO*, TUN, TUR, TUV, TWN*, TZA, UGA, UKR, URY*, USA*, UZB, VCT, VEN, VGB*, VNM, VUT, WLF, WSM, YEM, ZAF, ZMB, ZWE (* indicates high‐income countries).

APPENDIX 2. OTHER TABLES

App. Table 1.

Basic Statistics in Equation (2)

Variable	Obs.	Mean	SD	Min	Max
Stringency	2,936	51.3795	17.9414	0.8968	94.7997
ln(1 + cases)	2,936	8.6567	3.2684	0	16.8271
ln(1 + deaths)	2,936	4.4127	2.7765	0	11.6956
Vaccines	2,936	0.2549	0.2909	0	1.0551

App. Table 2.

Correlation Matrix in Equation (2)

Variable		(1)	(2)	(3)
Stringency	(1)	1.00
ln(1 + cases)	(2)	0.35	1.00
ln(1 + deaths)	(3)	0.37	0.88	1.00
Vaccines	(4)	−0.05	0.30	0.11

App. Table 3.

Basic Statistics in Equation (3)

Variable	Obs.	Mean	SD	Min	Max
Trade	186,884	1.9E+08	1.2E+09	0	5.7E+10
ln(1 + importer's cases)	186,884	6.9250	4.0494	0	16.8243
ln(1 + importer's deaths)	186,884	3.4322	2.9577	0	11.6956
Importer's vaccines	186,884	0.1333	0.2439	0	1.2457
ln(1 + exporter's cases)	186,884	9.6561	3.5992	0	16.8243
ln(1 + exporter's deaths)	186,884	5.3207	3.2988	0	11.6956
Exporter's vaccines	186,884	0.2093	0.3036	0	0.9449
RTA	186,884	0.3565	0.4790	0	1

App. Table 4.

Correlation Matrix in Equation (3)

Variable		(1)	(2)	(3)	(4)	(5)	(6)	(7)
Trade	(1)	1.00
ln(1 + importer's cases)	(2)	0.13	1.00
ln(1 + importer's deaths)	(3)	0.12	0.90	1.00
Importer's vaccines	(4)	0.08	0.30	0.18	1.00
ln(1 + exporter's cases)	(5)	0.02	0.37	0.26	0.29	1.00
ln(1 + exporter's deaths)	(6)	0.00	0.25	0.18	0.17	0.90	1.00
Exporter's vaccines	(7)	0.03	0.25	0.17	0.69	0.37	0.18	1.00
RTA	(8)	0.08	0.14	0.14	0.07	0.05	0.05	0.03