International Trade Spillovers from Domestic COVID-19 Lockdowns

Shekhar Aiyar; Davide Malacrino; Adil Mohommad; Andrea F Presbitero

doi:10.5089/9798400212178.001.A001

Jump to Content Jump to Main Navigation

International Trade Spillovers from Domestic COVID-19 Lockdowns

Author:

Mr. Shekhar Aiyar

Mr. Shekhar Aiyar
Search for other papers by Mr. Shekhar Aiyar in
Current site
Google Scholar
PubMed

Mr. Davide Malacrino

Mr. Davide Malacrino
Search for other papers by Mr. Davide Malacrino in
Current site
Google Scholar
PubMed

Mr. Adil Mohommad

Mr. Adil Mohommad
Search for other papers by Mr. Adil Mohommad in
Current site
Google Scholar
PubMed

, and

Mr. Andrea F Presbitero

Mr. Andrea F Presbitero
Search for other papers by Mr. Andrea F Presbitero in
Current site
Google Scholar
PubMed

https://orcid.org/0000-0003-4622-941X

Publication Date:: 17 Jun 2022

eISBN:: 9798400212178

ISBN:: 9798400212178

Language:: English

Keywords:: containment policy; trade partner; goods import; goods trade; demand model; Imports; Global

Download PDF (1.4 MB)

While standard demand factors perform well in predicting historical trade patterns, they fail conspicuously in 2020, when pandemic-specific factors played a key role above and beyond demand. Prediction errors from a multilateral import demand model in 2020 vary systematically with the health preparedness of trade partners, suggesting that pandemic-response policies have international spillovers. Bilateral product-level data covering about 95 percent of global goods trade reveals sizable negative international spillovers to trade from supply disruptions due to domestic lockdowns. These international spillovers accounted for up to 60 percent of the observed decline in trade in the early phase of the pandemic, but their effect was shortlived, concentrated among goods produced in key global value chains, and mitigated by the availability of remote working and the size of the fiscal response to the pandemic.

Abstract

1 Introduction

The COVID-19 pandemic constituted both a demand and a supply shock and the twin nature of the shock was expected to lead to a sharp decline in trade (Baldwin and Tomiura, 2020; Buchel et al., 2020; Liu et al., 2022). For an economy experiencing COVID-19, the economic contraction resulting from domestic pandemic containment policies, such as restrictions on mobility (and social distancing), was expected to lead to a large contraction in demand, and thus a decline in imports. From the supply side, the sudden halt in production following strict lockdowns and outright closings of establishments would also imply a contraction in the exports of a COVID-19 hit country to its trading partners—or equivalently—the bilateral imports of trading partners (Baldwin and Freeman, 2022; Cerdeiro and Komaromi, 2020; Espitia et al., 2022; Lafrogne-Joussier et al., 2022). The former channel may be thought of as the direct impact of the pandemic on trade. The latter channel constitutes a spillover effect on global trade, resulting from domestic measures to contain the pandemic.

Trade indeed fell sharply with the onset of the pandemic, declining by more than 15 percent in the aggregate and by even more in services. However, despite the expectations of a prolonged collapse in global trade, there was a rapid and vigorous rebound, in striking contrast to previous global recessions (Figure 1, panel A). Another distinguishing feature of the pandemic trade shock is that the V-shaped recovery was driven by the strong performance of goods imports, which in 2021 overshot pre-pandemic levels, while services trade continued to languish well below pre-pandemic levels even two years after the start of the pandemic (Figure 1, panel B).

These patterns in global trade are usually explained by a constellation of factors, including the collapse and subsequent strong rebound in aggregate demand and the impact of travel restrictions and social distancing on key in-person traded services. In this paper, however, we focus primarily on the supply side spillover effects of containment policies. In particular, we use detailed product-level bilateral goods trade data to estimate the international spillover effect of lockdowns in a gravity setting, from the onset of the pandemic until mid-2021.

While the bulk of the analysis presented in this paper relies on granular bilateral trade data, a natural starting point to guage the role of factors other than domestic demand is to estimate a standard import demand model using multilateral data, and evaluate how well it can explain the observed trade patterns. Deviations of observed import growth from model predictions would suggest that drivers other than those related to demand also mattered. In particular, we focus on pandemic intensity indicators (number of COVID-19 cases and deaths, and measures of lockdown intensity and mobility). We find that while a standard demand model fits rather well observed historical goods and services import growth, it performs poorly in predicting what happened in 2020. Imports of goods fell by much less than predicted (and those of services much more than predicted) by the demand model. Moreover, countries whose trade partners had better health preparedness experienced larger prediction errors (goods imports fell by less than predicted), suggesting that policies implemented in partner countries in response to COVID-19 mattered.

Figure 2 shows that the fall in goods imports during the first half of 2020 was larger for countries whose trading partners implemented more severe containment policies.^¹ To explore this relationship further, we quantify the supply shock spillover channel of the crisis-induced lockdowns using bilateral product-level trade data in a gravity model. This setting makes it possible to absorb demand factors through a set of time-varying fixed effects defined at the country-industry level. In practice, we look at a country (e.g., the U.S.) and we compare the imports of a product—for instance, vehicles—in each month from countries with different containment policies. In this way, assuming that changes in the U.S. demand for vehicles follow the same pattern across partners, demand is controlled for, and the variation in the severity of containment policies across trade partners captures spillovers operating throgh supply. We find that international spillovers from lockdowns were economically large, but they were short-lived and concentrated in the early phase of the pandemic. Compared to a counterfactual with no lockdown restrictions, trade partners’ lockdowns explain up to 60 percent of the observed decline in imports in the first five months of 2020. But moving past the initial phase of the pandemic, the elasticity of trade to partners’ lockdown intensity becomes insignificant, as trade begins to recover despite the stringency of lockdowns being relaxed only marginally.

The average spillover effect masks several sources of heterogeneity. First, we consider the size of the fiscal response to the COVID-19 pandemic in trade partners and find that spillovers are stronger from countries that were less able to deploy large discretionary fiscal measures to mitigate the effects of the pandemic. Second, we exploit cross country heterogeneity in teleworkability—as measured by Dingel and Neiman (2020)—to show that the spillover effect of lockdowns is more than twice as strong from partner countries which are less able to rely on remote working compared to those that have a higher share of jobs that can be done from home. Third, we decompose the effect of the containment policies across GVC-intensive industries, pooling all the others into a residual category, and find that the effect of lockdowns is stronger in GVC-intensive industries, and especially in electronics, than in non GVC-intensive ones.^² Fourth, we exploit the fact that the effect of containment policies is likely to differ across products depending on their position along the value chain, with more downstream industries, for which output is closer to the end user, relatively more exposed to GVCs and, therefore, to the restrictions imposed by lockdowns. In addition, by interacting the stringency index with a measure of upstreamness proposed by Antràs et al. (2012), we can fully control for multilateral resistance as in standard gravity models including exporter-time fixed effects. While in this setting we cannot estimate the average effect of the lockdown, we can better identify its differential effect across industries at different points along the value chain. The results show that the negative effect of stringency measures is dampened for industries which are further upstream (like metals and minerals products), while it is stronger for those downstream (like transportation and textiles).

Finally, our findings support the notion that the pandemic was accompanied by a rotation in demand from services and towards goods. Indeed, the greater the intensity of the pandemic in a given country, the smaller the decline in goods imports relative to the predictions of the standard import demand model. In other words, it appears that domestic containment policies boosted demand for imported goods—as distinct from the negative spillover effect on imports due to partner country lockdowns. In addition, services imports fell by much more than would be predicted by demand alone in countries for which tourism services comprised a large share of total imports.

Related literature. Our paper contributes to a burgeoning literature on the propagation of the COVID-19 pandemic shock through global trade (e.g., Bonadio et al., 2021; Espitia et al., 2022; Lafrogne-Joussier et al., 2022; Liu et al., 2022).

First, we build on contemporaneous analyses looking at the effect of domestic lockdowns on trade partners. A set of papers zooms in on the experience of individual countries. Heise (2020) shows that imports of U.S. firms that sourced from China before the pandemic fell by 15% more than those of comparable firms that were importing from other countries. Similarly, Lafrogne-Joussier et al. (2022) consider French firms engaged in GVCs and show that those relying on Chinese inputs experienced a 7% greater decline in their overall imports following the Chinese lockdown relative to firms that were not exposed to Chinese inputs.^³ Also using data on French firms, Brussevich et al. (2022) explore trade patterns vis-a-vis several countries finding evidence that the pandemic shock propagated through supply chains and that more restrictive lockdown policies in trading partners resulted in larger drops in exports and imports by French firms, with the adjustment borne mostly at the intensive margin and by smaller and less automated firms. Evidence from Portuguese firms shows that the severity of the health crisis and mobility restrictions are associated with a large contraction in trade flows, although there is evidence of a progressive adaptation of trading firms to new circumstances, as the effect of lockdowns turns insignificant in the second half of 2020 (Pimenta et al., 2021).^⁴

Closer to our approach, Cerdeiro and Komaromi (2020) disentangle the supply and demand channels in global trade and quantify the effect of supply spillovers from lockdowns using daily data on bilateral seaborne trade and tracking the transmission of supply disruptions across space. Their analysis suggests a downstream propagation of countries’ lockdowns through global supply chains and points to short-lived supply spillovers of lockdowns through international trade. Espitia et al. (2022) are among the first to use product-level bilateral trade data to look at the heterogeneous trade effects of the pandemic across sectors. Their analysis, based on data covering the trade of 28 countries in the first six months of the pandemic, shows that a decline in mobility in trade partners is associated with a reduction of imports of durable goods. Also, the negative effect of lower mobility is stronger for sectors which rely more on imported inputs. Contemporaneous work by Berthou and Stumpner (2022) is based on 31 reporting countries and documents a significant effect of lockdowns put in place by exporting and importing countries on bilateral trade.^⁵

With respect to this literature, we show evidence of the international spillover effects of lock-downs based on a large product-level dataset covering 98 reporting countries and 95% of global trade in goods over 18 months. With these data we can fully control for time-varying factors which could affect product-specific demand in different importing countries. We also investigate in detail the (lack of) persistence in lockdown spillovers and identify industries more exposed to containment policies. Exploting differences across industries in their position along the value chain, we are also able to quantify the differential effect of lockdowns on trade partners controlling for all unobserved time-varying factors that differ across exporters and that could confound the spillover effect of lockdowns.

Second, by examining the trade dynamics of goods and services, we document the large shift of imports from services to goods and the sharp increase in goods imports, which cannot be fully explained by demand factors. In this respect, our evidence is in line with previous research looking at domestic consumption patterns in several economies—see Bounie et al. (2020) for France, Andersen et al. (2022) for Denmark, Carvalho et al. (2021a) for Spain, Baker et al. (2020) for the United States, and Chronopoulos et al. (2020) for the United Kingdom. Our analysis suggests that these trends are common across countries, driven by pandemic-specific factors, such a shift in preferences across sectors (e.g., Guerrieri et al., 2022) and pent-up demand, and partly reflected in trade flows.

Our paper also contributes to a large literature which analyzes the (potential) health-economy trade-offs associated with lockdowns and pandemic-related containment policies (Baqaee et al., 2020; International Monetary Fund, 2020; Aum et al., 2021). The decision to impose a lockdown, and to mandate how severe the lockdown should be, is challenging given the need to strike a balance between protecting lives and minimizing the harm to economic activity. While most of the economic literature focuses on the potential domestic costs of lockdowns in terms of reduced mobility and lower economic activity (Chen et al., 2020; Deb et al., 2021), our analysis tries to quantify the international costs borne by other countries via the trade channel. In this respect, by showing a large but relatively short-lived fall in trade due to lockdowns, our results provide additional evidence to inform policy makers about the potential costs of containment policies.

2 Were the effects of the pandemic on imports “unprecedented”?

Many consequences of the pandemic have been assessed as unprecendented. To assess whether the same can be said of its impact on trade – and especially imports – we perform the following exercise. First, we estimate a simple demand model based on the work of Bussière, Callegari, Ghironi, Sestieri and Yamano (2013), who used it in the context of the Great Trade Collapse in 2008–2009, on a sample of 127 countries over the period 1985–2019.^⁶ Second, we use the estimated model to predict import growth in 2020 and obtain country-level prediction errors.^⁷ Third, we aggregate the errors across countries and compare the aggregate prediction error in 2020 with similarly aggregated errors in the previous years, to establish whether the change in imports observed during the pandemic year can be explained by standard economic forces. Fourth, we correlate the errors in the 2020 cross-section with country level factors to understand their drivers.

2.1 A simple empirical import demand model

As in Bussière et al. (2013), we begin by estimating country-level regressions of importer country m’s log change in real imports in year t (M_mt) on a country specific constant (π_m), the log change in a measure of import-intensity adjusted demand (D_mt) and import prices relative to the domestic GDP deflator (P_mt):

\begin{matrix} Δ \ln M_{m t}^{τ} = π_{m} + β_{D m}^{τ} Δ \ln D_{m t}^{τ} + β_{P m}^{τ} Δ \ln P_{m t} + ε_{m t,}^{τ} \forall τ \in {g, s} & (1) \end{matrix}

The regressions are estimated separately for imports of goods (τ = g) and services (τ = s).

Following Bussière et al. (2013), import-intensity adjusted demand is a harmonic weighted average of different domestic demand components (private consumption C, government consumption G, investment I and exports X to take into account the import content of export demand). It is defined as:

\begin{matrix} D_{m t}^{τ} = C_{m t}^{ω_{τ} C_{m}} G_{m t}^{ω_{τ} G_{m}} I_{m t}^{ω_{τ} l_{m}} X_{m t}^{ω_{τ} X_{m}}, \forall τ \in {g, s}, & (2) \end{matrix}

where the weights (ω's) represent the total import content of each component and are computed from the Eora Global Supply Chain Database following the method in Bussière et al. (2013). The import relative prices are computed using data from the World Economic Outlook database of the International Monetary Fund.

In Table 1, we report various statistics from the distribution of estimated coefficients from the country-by-country specifications in equation 1. The average estimated coefficient on demand is about 1.3 and the median estimate is quite close to that for goods while it is slightly lower (1.1) for services. Most estimated coefficients on demand are positive. The coefficients on relative prices are mostly negative, with the mean (median) estimates at -0.2 (-0.14) for goods and -0.30 (-0.23) for services. While in most cases import demand covaries negatively with prices, in some countries the estimated coefficient on prices is positive. This is possibly due to standard endogeneity (prices increasing with demand) which would bias the coefficient upwards. Since the main goal of our exercise is to predict import growth, we do not address this issue; the model turns out to perform well in sample as we highlight in the next subsection.^⁸

2.2 Model performance

As shown in Figure 3 (top panels), combining the estimates from the country-by-country regressions (weighted by shares in world imports) yields accurate predictions-i.e. fitted values-of import growth up to 2019. But for 2020 the model fails at predicting the large observed fall in services trade (the model predicts a growth rate of about -8%, while in 2020 trade fell by 25%) and slightly overpredicts the fall in goods trade (10% predicted vs 6% observed fall). The bottom panels of Figure 3 report the prediction errors series: the error for services in 2020 is 0.2 log-points, quite literally “off-the-charts” with respect to previous years.^⁹

To understand what drove the poor performance of the model in 2020, we proceed by linking the prediction errors for 2020 to various pandemic related variables and other country features.

2.3 The drivers of pandemic prediction errors: going beyond demand

To fix ideas concerning this analysis, notice that equation 3 can be derived from the following expression for import demand (the index τ for the type of import has been omitted for simplicity):

\begin{matrix} M_{m t} = D_{m t}^{β_{D}} {(\frac{{\tilde{P}}_{M m t}}{{\tilde{P}}_{m t}})}^{β_{P}} \exp {α_{m} t + c_{t} + η_{m t}} \equiv D_{m t}^{β_{D}} P_{m t}^{β_{P}} \exp {α_{m} \times t + c_{t} + η_{m t}} & (4) \end{matrix}

In words, import demand by country m at time t can be though of as a function of a measure of domestic demand (D_mt, in our case the import-intensity adjusted measure), prices of imports relative to domestic prices $\frac{{\tilde{P}}_{M m t}}{P_{m t}}$ , a country- specific linear time trend α_mt, aggregate shocks captured by c_t and other time varying specific factors captured by η_mt (e.g., preferences, trade costs not subsumed in the price indexes, the impact of demand on imports not captured by the measure of demand or supply factors faced by country i and different from aggregate supply shocks that are not immediately priced in). The estimated equation 3 can be obtained by taking logs, differencing over time, adopting the definitions γ_t ≡ ∆c_t and ε_mt ≡ ∆η_mt, and—concerning the trend—noticing that α_m × t — α_m × (t — 1) = α_m.

Hence the residual in the equation captures elements such as changes in preferences, or supply shocks having an impact on imports not immediately captured by standard price indexes. The pandemic likely produced various shocks of this sort and the following results confirm this intuition.

Results are reported in Table 2. The pandemic induced higher than expected goods imports. Countries that experienced a more severe pandemic (more cases, more stringent measures or less mobility) show better than expected good import growth. Consistent with the previous discussion, it is possible that the pandemic induced a shift in preferences away from services (domestic like restaurants, and imported like travel) towards goods. The insignificant coefficients on imported services, however, suggest that possibly the shift away from services mostly affected domestic rather than imported services.^¹⁰

Importantly for what follows, trade partners’ health preparedness was associated with more goods imports. The ability of countries to increase their goods import above the expected amount was associated with their partners’ health preparedness as captured by the Global Health Security Index.^¹¹ These results are shown in Table 3, where the relevant variable is an import-weighted average of the index. These results suggest that containment policies in partner countries likely played a role beyond those countries’ borders, spilling over internationally through trade connections. To investigate this possibility more formally, we turn next to an analysis of country-to-country trade linkages.

3 International spillover effects of lockdowns

3.1 Modeling international spillovers

Having assessed the role of pandemic-related factors and policies on the exceptional fall in imports, we turn now to the spillover effects that lockdowns could have on trade partners. To quantify the international spillover effects of containment policies, we estimate the following standard gravity equation at the country pair-industry-month level (see Yotov, 2022, for a recent review of gravity models), in which goods imports are a function of trade partners’ containment policies:

\begin{matrix} M_{m e i t} = \exp [β S t r i n g e n c y I n d e x_{e t} + δ C o n t r o l s_{m e t} + α_{m e i} + γ_{m i t}] + ε_{m e i t} & (5) \end{matrix}

The bilateral imports of products in industry i (M_meit) by importer country m from exporter country e in month t is regressed on: i) the time-varying index of lockdown intensity in the exporter country e (Stringency Index_e,t), measured using the monthly average values of the Oxford COVID-19 Government Response Stringency Index (Hale et al., 2021); ii) a set of variables that vary across country pairs and time (Controls); and iii) a set of fixed effects (α_mei,γ_mit).

The main dataset covers bilateral goods imports disaggregated at the 6-digit product level in the Harmonized System (HS6). These data are available at a monthly frequency from Trade Data Monitor (TDM) from January 2020 to June 2021.^¹² The original sample includes 99 importing countries which trade with 196 exporting countries. These data have been used recently to look at the recent effects of trade tensions and of the COVID-19 pandemic on trade flows (Berthou and Stumpner, 2022; Grant et al., 2021; Arita et al., 2022). Figure A3 shows that TDM covers almost the universe of goods trade, as the aggregation of all import flows account for about 95 percent of goods trade, as reported by the IMF Direction of Trade Statistics. When looking at a common sample made by 99 reporting countries, the coverage is very close to 100 percent.

We aggregate bilateral monthly data on goods imports at the HS6 codes over about 300 industries, to be able to match import flows with a measure of insustry upstreamness, as computed by Antràs et al. (2012). The aggregation is done using the concordance between the 6-digit HS codes and I-O commodity codes, as published by the Bureau of Economic Analysis.^¹³

The trade data are also matched with a set of other variables at the exporting country level. First, we merge the TDM data with a monthly measure of the stringency of pandemic containment policies in exporting countries (the Oxford COVID-19 Government Response Stringency Index), as computed by Hale et al. (2021), and with the number of new COVID-19 cases and death per capita, as reported by Our World in Data (Ritchie et al., 2020). Second, we use the Global Trade Alert (GTA) data (Evenett, 2019) to construct a measure of export restriction at the country-pair level, by counting, at the quarter level, the number of new export interventions (e.g., bans, quotas, non-tariff measures, tariffs, etc.) implemented by the exporter country e versus the importing country m. For completeness, the model also includes the number of export barriers which have been removed.^¹⁴

This dataset, which is the basis for our main analysis, includes 98 importing countries and 163 exporting ones, for a total of 15,880 country pairs and 4,652,840 unique industry-exporter-importer trade corridors.

The key parameter of interest β measures the effect of trade partners containment policies on imports. Figure 1 (panel B) illustrates that the increase in mobility restrictions at the outbreak of the pandemic coincided with the sharp collapse in goods imports in the first two quarters of 2020. However, the stringency index could be correlated with other simultaneous changes in the exporter country. In particular, an important element to consider is the response of trade policy to the pandemic, as several governments imposed new export restrictions on specific goods (see, for instance Evenett, 2020; Evenett et ah, 2021, on export restrictions on medical supplies and food products). To account for the role of trade restrictions, we include the number of new export restrictions and the number of removed export barriers at the country-pair level, constructed from the GTA data. To further mitigate the omitted variable bias, the set of controls includes the number of new COVID-19 cases and deaths per month (per million inhabitants) measured in the exporter country^¹⁵

Country-pair-industry fixed effects (α^mei) control for differences in industry-specific trade flows between each pair of importer and exporter countries, such as time-invariant natural, cultural and geographical factors, as well as the presense of trade agreements (as long as they are present throughout the sample period). The importer-industry-time fixed effects (γ_mit) absorb unobserved time-varying heterogeneity across both importers and industries. In other words, all unobserved changes in demand for goods in a given industry, including those coming from domestic lock-downs, should be absorbed by the fixed effects.

Conditional on this rich set of controls and fixed effects, the coefficient β captures the impact of lockdowns on imports via the supply channel. A negative coefficient on the stringency index indicates that the size of the decline in imports for a given country is related to the severity of the pandemic restrictions imposed by its trade partners, due to the corresponding reduction in the supply of goods by those trade partners.

Before turning to the results it is worth considering a final caveat when interpreting the coefficient β as a measure of a supply channel. Import demand is controlled for under the assumption that the proportional change in country-specific demand for products in a given industry (in a given month) is the same across countries. For instance, we assume that the proportional change in demand for vehicles by U.S. consumers in April 2020 was the same for both Japanese and German cars. As the analysis looks at monthly changes and focuses on a period of high uncertainty, it is plausible to assume that consumers did not adjust their demand differentially across producers in different countries.

Since the truncation of import flows at zero biases the standard log-linear OLS approach and leads to inconsistent coefficient estimates, we follow an extensive trade literature on gravity models (Silva and Tenreyro, 2006, 2022; Anderson and Yotov, 2016; Espitia et al., 2022; Arita et al., 2022; Yotov, 2022) and estimate equation 5 by Poisson pseudo-maximum likelihood (PPML), as implemented by Correia et al. (2020). Standard errors are clustered at the exporter level, which is the source of variation in the stringency index.

3.2 Baseline Results

The main results are shown in Table 4 and reported in Figure 4. The first five columns show the negative and significant association between the stringency of partners’ containment policies and domestic imports. Moving from a model with time varying importer fixed effects (column 1) to one with time varying importer-industry fixed effects (column 2) shows that the point estimate of coefficient of the stringency index is stable and suggests that the model captures most of the variation from the demand side. The spillover effect is robust to controlling for the extent of the health crisis (measured by the number of COVID-19 cases and deaths per capita) and changes in export restrictions put in place by trade partners, including when controlled for jointly (columns 3–5). The effect is also economically meaningful. The semielasticity is about -0.15 and implies that that a one percentage point increase in the stringency index is associated with a 0.15 percent reduction in imports.^¹⁶

The pandemic was a novel shock, so it is plausible that over time there was adaptive behaviour, with the implication that the size of spillover effects could be changing from month to month. To examine this possibility we split the coefficient β over time and estimate the spillover effects of trade partner containment policies in each month. Figure 4 shows the dynamics of the spillover effect of lockdowns. The effects were strongest in the first 5 months of 2020, gaining in strength in February and March when COVID-19 evolved from a regional crisis to a pandemic, but then declining and becoming insignificant in June, when goods imports started to rebound. Interestingly, there is a smaller but significant effect in the Spring of 2021, coincident with the spread of the Delta variant. As containment policies persisted throughout the period—the stringency index does not show any visible decline (Figure 4)—this would suggest that countries started adjusting to the presence of lockdowns and pandemic-related restrictions, consistent with evidence found by Heise (2020), Cerdeiro and Komaromi (2020), Lafrogne-Joussier et al. (2022), Berthou and Stumpner (2022), Pimenta et al. (2021), and Arita et al. (2022) in different settings.^¹⁷

Since the impact of lockdowns on imports is found to be large but short-lived, we estimate the baseline model over the first half of 2020 to better gauge the economic effect during the first phase of the crisis. The results, reported in column 6, indicate that the semielasticity is more than twice that estimated for the whole sample. This point estimate is used to generate the evolution of goods imports under a counterfactual with no containment policies in trade partners. Comparing this series with the actual evolution of imports indicates that containment policies accounted for about 60 percent of the observed fall in imports (Figure 5), the headline quantification of the spillover effect discussed in the chapter.^¹⁸

3.3 Heterogeneous Effects of Lockdowns

The spillover effects of containment policies could depend on the capacity of countries to mitigate them and adapt to the new prevailing circumstances. A first aspect is the size of the fiscal policy response to the COVID-19 pandemic. We use data collected by the IMF on COVID-19 related fiscal measures taken or announced between January and June 2020. These data cover government discretionary measures that supplement existing automatic stabilizers and include both above-the-line (additional spending and forgone revenue) and below-the-line (equity, loans, and guarantees) fiscal operations. All numbers are scaled by GDP.^¹⁹ We estimate the semi-elasticity of imports to the stringency index in countries which implemented a relatively small fiscal response to the crisis (the bottom quartile of the distribution) and in those which implemented a relatively large fiscal reponse. The results, reported in Table 4 (column 7) and at the top of Figure 6, indicate that the international spillovers from lockdowns are significantly larger for countries whose trade partners have been less able to combat the effects of the pandemic with discretionary fiscal measures.

A second key aspect is the capacity to rely on remote working. The results shown in columns 8 and 9 in Table 4 exploit cross country heterogeneity in the proportion of jobs which can be done at home to test whether the supply effect due to the lockdown is stronger for countries which import more from countries where jobs are less likely to be done remotely. Teleworkability is measured using the cross-country data computed by Dingel and Neiman (2020) and the sample of trade partners is split between those with a low share of jobs that can be done remotely (the bottom quartile of the distribution) and those with a high share of teleworking. As the use of the teleworkability measure reduces the sample size, the baseline model is estimated on the restricted sample (column 8). Even in this case there is a negative (albeit smaller) and significant spillover effect. What is more interesting is that the spillover effect of lockdowns is more than twice stronger for countries which are less able to rely on remote working compared to those that have a higher share of jobs that can be done from home (column 9). This finding, shown also in Figure 6, is consistent with existing evidence showing that the feasibility of remote work mitigated the negative effects of reduced worker mobility (Pei et al., 2021), while suggesting that the benefits extended beyond national boudaries to trade partners.

A third dimension of heterogeneity is across industries. Column 10 in Table 4 and Figure 6 report the results obtained from decomposing the effect of containment policies across four GVC-intensive goods, defined to include inputs and finished goods in automotive industries, electronics, textiles and garments, and medical goods. Together these goods account for about 24 percent of global goods trade, and are typically considered to be at the forefront of GVCs (Sturgeon and Memedovic, 2010). All other products are pooled into a residual category (non-GVC-intensive industries). The results indicate that the effect of lockdowns is stronger in GVC-intensive industries, and especially in electronics, than in non GVC-intensive ones. This finding is consistent with some recent evidence based on firm-level data (Pimenta et al., 2021) as well as product-level flows (Berthou and Stumpner, 2022). It supports the intuition that imports in GVC-intensive industries are relatively more exposed to disruptions in the supply chain (in this case due to lockdowns).

3.4 A Full-Fledged Gravity Model

The baseline analysis does not fully control for multilateral resistance as in standard gravity models since it does not include time-varying exporter fixed effects. Adding this term makes it impossible to identify the semielasticity of the stringency index, given that its source of variation is also at the exporter-time level. However, the richness of the product-level data allows to go one step further and better identify the supply channel of lockdowns exploiting the fact that the effect of the lockdown is likely to differ across industries.

The sensitivity of imports could depend on the industry’s reliance on the sourcing of inputs, as measured by the industry’s upstreamness (i.e., the average distance from final use). Using a Bartik (1991)-style approach, the stringency index is interacted with a measure of GVC upstreamness computed by Antràs et al. (2012) from U.S. input-output tables.^²⁰ This leads to an augmented version of equation 5:

\begin{matrix} ​ M_{m e i t} = δ (β S t r i n g e n c y I n d e x_{e t} \times U p s t r e a m_{t} + δ C o n t r o l s_{e i t} + α_{m e i} + γ_{m i t} + μ_{e t} + ε_{m e i t}), & (6) \end{matrix}

which includes both multilateral resistance terms (γ_mit and µ_et) and identifies the differential effect of the stringency index across industries. In other words, the (time-invariant) upstreamness of the industry is a measure of its exposure to the (time-varying) lockdown supply shock. The intuition is that more downstream industries, for which output will go to the end user (e.g., automobile, electronics), would be relatively more exposed to GVCs and sourcing inputs and, therefore, to the restrictions imposed by lockdowns.

Table 5 shows the results. When the exporter-time fixed effects are not included, the results show that the negative effect of stringency measures is dampened in industries which are very upstream (like metals and minerals products), while it is stronger for those downstream (like transportation and textiles). A one standard deviation of the upstream index (SD = 0.82) reduces the supply effect of the lockdown by about 20 percent (column 1).^²¹ More importantly, once fully controlling for unobserved (time-varying) heterogeneity across exporters including the multilateral resistance term (column 2), the differential effect of the lockdown across industries with different degree of upstreamness remains statistically significant and similar in size. This evidence is in line with the results of Brussevich et al. (2022) using data from French firms.

3.5 Robustness

Our results are robust to additional exercises aimed at testing the sensitivity of the findings to the choice of variables, sample and methodology.

Measuring containment policies. The main results are robust to measuring containment policies with an index measuring only the severity of workplace closures. This index, which assumes discrete values from 0 (no restrictions) to 3 (closing down or working from home for all-but-essential workplaces), is one of the 8 containment and closure policy indicators used to calculate the Oxford COVID-19 Government Response Stringency Index (Hale et al., 2021).^²² While its categorical nature compresses its variability over time, the index is the closest to the idea of measuring how lockdowns could affect production and spillover to international trade. The index of workplace closings and the stringency index are highly correlated, and they show a very similar evolution over time (Figure A4). In particular, the correlation in the pooled sample is equal to 0.82 and a regression of the stringency index against the workplace closings index with month and country fixed effects gives a coefficient equal to 13.3 (s.e. = 0.56). We replicate the baseline results using the measure of workplace closings and show that our findings are still valid. Table 6 shows that more stringent containment policies in workplaces put in place by trade partners are associated with a significant decline in imports (columns 1–3). This effect is significantly stronger for trade partners which implemented a smaller discretionary fiscal response to the COVID-19 pandemic (column 5) and for trade partners with a low capacity to work remotely (column 6), while it is larger for goods upstream in the value chain (column 8).

Robustness across different country groups. Because of the asynchronous dynamics of the COVID-19 pandemic and its differential intensity across countries, one could imagine that results are sensitive to specific countries or regions. To address this concern, we estimate the baseline model by dropping, one at a time, specific country groups classified by income and region from the set of exporting countries. Figure 7 shows that the significance of the spillover effect is robust to alternative samples. However, it also suggests that the semielasticity of the stringency index becomes smaller when emerging markets and Asian countries are excluded. This evidence is consistent with the impact of containment policies being larger in the first phase of the crisis, when the COVID-19 shock disproportionately affected Asian countries, shutting down production and disrupting global trade. By contrast, the semielasticity is higher when advanced economies and European countries are excluded, suggesting that containment policies in Europe (which occured later in the year) had weaker spillover effects. In a second exercise we zoom in on the role of China, given its importance for global trade and its prominent role in the early phase of the pandemic. We re-estimate the baseline model dropping China from the set of exporters and focusing on the heterogeneous effects of lockdowns over time. Figure A5 in the appendix, which replicates Figure 4, illustrates that our results on the international spillover of lockdown are not China-specific and remain economically meaningful in the sample excluding China. The smaller spillover effects in February and March and the larger ones in April and May could indicate that the initial spillover effect on global trade was driven by the lockdowns in China. By constrast, lockdowns in other countries matter more in the Spring of 2020, when China started easing restrictions, while other regions (e.g., Europe and the United States) were still under severe lockdowns.

Clustering. Table A3 in the appendix reports the main results discussed in the chapter estimated by clustering the standard error at the exporter-month level. The significance of the findings is not affected, and the estimated standard errors are—if anything—smaller.

4 Conclusions

Goods trade fell sharply in the early stages of the pandemic, but rebounded quickly. Trade in services, on the other side, has not yet recovered at the time of writing in mid-2022. We show that standard factors of import demand observed in 2020—components of domestic demand and relative import prices—are consistent with weaker than observed demand in goods imports and stronger than observed demand in services imports. Indeed, feeding these standard factors into a simple model of import demand estimated on historical data suggets that non-standard drivers of imports came to the fore during the pandemic. We show countries whose trade partners had better health preparedness experienced smaller declines in goods imports relative to model predictions in 2020, hinting at the possibility of spillovers from partner country pandemic responses on trade.

We then show that pandemic containment policies in partner countries played an important role beyond those countries’ borders. Using product-level bilateral trade data, we document that international spillovers from supply disruptions due to lockdowns were sizable and negative. Specifically, we estimate the extent to which more stringent pandemic containment policies in a given country affected imports from that country, after controlling for import demand. We find that a percentage point increase in the stringency index is associated with a 0.15 percent reduction in imports. This is a economically sizeable impact, implying that up to 60 percent of the observed fall in imports between January and May 2020 can be explained by lockdowns. We also document that such spillover effects were (i) short-lived, fading out starting in the second half of 2020, (ii) stronger for goods belonging to GVC-intensive industries, and (iii) mitigated if exporting countries were more conducive to teleworking and implemented large discretionary fiscal measures in response to the pandemic.

Finally, we show that where the pandemic was more intense, goods imports rebounded faster than predicted by the import demand model. This is consistent with the oft-made argument, (e.g., in Guerrieri et al., 2022), that the pandemic induced a shift in preferences away from services (domestic services, like restaurant meals, and imported services, like travel) towards goods. These findings are relevant to a better understanding of how the economic effects of the pandemic propagated across countries via trade linkages.

Figures

Tables

^{Table 1:}

Import demand model: Summary statistics from the country-by-country estimates

Notes: Results from regressions estimated country by country on a sample of 127 countries with at least 16 observations between 1985 and 2019.

^{Table 1:}

Import demand model: Summary statistics from the country-by-country estimates

		Total	Services	Goods
Demand coefficients	Mean	1.2955	1.3472	1.3305
	Median	1.3468	1.0932	1.3574
	Interquantile range	[0.95,1.54]	[0.68,1.76]	[0.99,1.69]
Price coefficients	Mean	-0.2230	-0.2931	-0.2030
	Median	-0.1913	-0.2266	-0.1390
	Interquantile range	[-0.39,0.00]	[-0.57,0.09]	[-0.43,0.10]
Number of countries		127	127	127

Notes: Results from regressions estimated country by country on a sample of 127 countries with at least 16 observations between 1985 and 2019.

^{Table 1:}

Import demand model: Summary statistics from the country-by-country estimates

		Total	Services	Goods
Demand coefficients	Mean	1.2955	1.3472	1.3305
	Median	1.3468	1.0932	1.3574
	Interquantile range	[0.95,1.54]	[0.68,1.76]	[0.99,1.69]
Price coefficients	Mean	-0.2230	-0.2931	-0.2030
	Median	-0.1913	-0.2266	-0.1390
	Interquantile range	[-0.39,0.00]	[-0.57,0.09]	[-0.43,0.10]
Number of countries		127	127	127

Notes: Results from regressions estimated country by country on a sample of 127 countries with at least 16 observations between 1985 and 2019.

^{Table 2:}

Residual analysis: Pandemic-relevant variables

Notes: Results from a regression of the import demand model prediction errors in 2020 on COVID-19 related variables. The total number of COVID-19 cases in 2020 (in logarithm), mobility and the stringency index are extracted from the Our World in Data database. The stringency index and mobility are computed as averages over 2020. Changes in the number of observations originate from the variable of interest being missing. Standardized coefficients represent the number of standard deviation changes in the dependent variables associated to one standard deviation change in the variable of interest. Standard errors, in parenthesis, are robust to heteroscedasticity ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

^{Table 2:}

Residual analysis: Pandemic-relevant variables

		(1) Total	(2) Services	(3) Goods	(4) Total	(5) Services	(6) Goods	(7) Total	(8) Services	(9) Goods
Total COVID-19 cases		0.00321	-0.00639	0.00812**
	Standardized coefficient	(0.00360) 0.0574	(0.0123) -0.0416	(0.00408) 0.121**
Stringency index					0.00172* (0.000895)	0.00177 (0.00231)	0.00229** (0.00103)
	Standardized coefficient				0.197*	0.0738	0.217**
Mobility								-0.00236** (0.000918)	-0.00183 (0.00330)	-0.00363*** (0.00117)
	Standardized coefficient							-0.239**	-0.0674	-0.305***
Observations		125	125	125	121	121	121	99	99	99
Adjusted R-squared		-0.004	-0.006	0.009	0.038	-0.003	0.049	0.074	-0.005	0.104

^{Table 2:}

Residual analysis: Pandemic-relevant variables

		(1) Total	(2) Services	(3) Goods	(4) Total	(5) Services	(6) Goods	(7) Total	(8) Services	(9) Goods
Total COVID-19 cases		0.00321	-0.00639	0.00812**
	Standardized coefficient	(0.00360) 0.0574	(0.0123) -0.0416	(0.00408) 0.121**
Stringency index					0.00172* (0.000895)	0.00177 (0.00231)	0.00229** (0.00103)
	Standardized coefficient				0.197*	0.0738	0.217**
Mobility								-0.00236** (0.000918)	-0.00183 (0.00330)	-0.00363*** (0.00117)
	Standardized coefficient							-0.239**	-0.0674	-0.305***
Observations		125	125	125	121	121	121	99	99	99
Adjusted R-squared		-0.004	-0.006	0.009	0.038	-0.003	0.049	0.074	-0.005	0.104

^{Table 3:}

Residual analysis: Trade partners’ health preparedness

Notes: Results from a regression of the import demand model prediction errors in 2020 on trade partners health preparedness, computed as the import-weighted average of the Global Health Security Index across all countries from which country i imports goods. Standardized coefficients represent the number of standard deviation changes in the dependent variables associated to one standard deviation change in the variable of interest. Standard errors, in parenthesis, are robust to heteroscedasticity. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

^{Table 3:}

Residual analysis: Trade partners’ health preparedness

		(1) Total	(2) Services	(3) Goods
Trade partners’ health preparedness		0.00213 (0.00212)	-0.00520 (0.00546)	0.00518*** (0.00183)
	Standardized coefficient	0.0964	-0.0854	0.195***
Observations		122	122	122
Adjusted R-squared		0.002	-0.001	0.036

^{Table 3:}

Residual analysis: Trade partners’ health preparedness

		(1) Total	(2) Services	(3) Goods
Trade partners’ health preparedness		0.00213 (0.00212)	-0.00520 (0.00546)	0.00518*** (0.00183)
	Standardized coefficient	0.0964	-0.0854	0.195***
Observations		122	122	122
Adjusted R-squared		0.002	-0.001	0.036

^{Table 4:}

The international spillover effect of lockdowns

Notes: The table reports the results of the estimation of equation 5 by Poisson pseudo-maximum likelihood. The stringency index is the Oxford COVID-19 Government Response Stringency Index, computed by Hale et al. (2021). The sample spans the period 2020:ml-2021:m6; in column 6 the sample is restricted to the first six months of 2020, while in columns 8–9 it is limited to exporting countries for which the measure of teleworkability computed by Dingel and Neiman (2020) is available. The coefficient of the stringency index is estimated separately for: i) trade partners which have announced and implemented high and low discrectionary fiscal measures in response to the pandemic between January and June 2020 (fiscal spending is scaled by GDP and the sample is split along the first quartile of the distribution of fiscal spending over GDP); ii) trade partners with high and low values of the teleworkability index (the sample is split along the first quartile of the distribution of the index); and iii) for goods in GVC-intensive industries (Automotive, electronics, medical, textiles) and in non-GVC-intensive ones. Standard errors in parenthesis are clustered at the exporter level. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. TW = teleworkability.

^{Table 4:}

The international spillover effect of lockdowns

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
Stringency index	-0.00141*** (0.000)	-0.00149*** (0.000)	-0.00183*** (0.000)	-0.00160*** (0.000)	-0.00182*** (0.000)	-0.00307*** (0.000)		-0.00062*** (0.000)
Covid cases per million, lagged			0.00002 (0.000)		0.00002 (0.000)
Covid deaths per million, lagged			-0.00056 (0.001)		-0.00051 (0.001)
Number of new export restrictions				0.01631 (0.011)	0.00917 (0.010)
Number of removed export restrictions				-0.00299 (0.002)	-0.00199 (0.002)
Stringency index x Small fiscal response							-0.00211*** (0.000)
Stringency index x Large fiscal response							-0.00146*** (0.000)
Stringency index x Low telework									-0.00126*** (0.000)
Stringency index x High telework									-0.00058*** (0.000)
Stringency index x Automotive										-0.00169** (0.001)
Stringency index x Electronics										-0.00312*** (0.001)
Stringency index x Medical										-0.00246*** (0.001)
Stringency index x Textiles										-0.00243*** (0.000)
Stringency index x non-GVC industries										-0.00118*** (0.000)
Observations	23,594,169	23,531,808	21,787,468	23,531,808	21,787,468	6,118,735	23,256,563	14,764,840	14,764,840	23,531,808
Exporter-importer-industry FE	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y
Importer-month FE	Y	-	-	-	-	-	-	-	-	-
Importer-industry-month FE	N	Y	Y	Y	Y	Y	Y	Y	Y	Y
Industry-month FE	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y
Sample	All	All	All	AU	AU	2020H1	All	TW	TW	AU
Semielasticity	-0.1413	-0.1494	-0.1828	-0.1594	-0.1818	-0.3068		-0.0616

^{Table 4:}

The international spillover effect of lockdowns

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
Stringency index	-0.00141*** (0.000)	-0.00149*** (0.000)	-0.00183*** (0.000)	-0.00160*** (0.000)	-0.00182*** (0.000)	-0.00307*** (0.000)		-0.00062*** (0.000)
Covid cases per million, lagged			0.00002 (0.000)		0.00002 (0.000)
Covid deaths per million, lagged			-0.00056 (0.001)		-0.00051 (0.001)
Number of new export restrictions				0.01631 (0.011)	0.00917 (0.010)
Number of removed export restrictions				-0.00299 (0.002)	-0.00199 (0.002)
Stringency index x Small fiscal response							-0.00211*** (0.000)
Stringency index x Large fiscal response							-0.00146*** (0.000)
Stringency index x Low telework									-0.00126*** (0.000)
Stringency index x High telework									-0.00058*** (0.000)
Stringency index x Automotive										-0.00169** (0.001)
Stringency index x Electronics										-0.00312*** (0.001)
Stringency index x Medical										-0.00246*** (0.001)
Stringency index x Textiles										-0.00243*** (0.000)
Stringency index x non-GVC industries										-0.00118*** (0.000)
Observations	23,594,169	23,531,808	21,787,468	23,531,808	21,787,468	6,118,735	23,256,563	14,764,840	14,764,840	23,531,808
Exporter-importer-industry FE	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y
Importer-month FE	Y	-	-	-	-	-	-	-	-	-
Importer-industry-month FE	N	Y	Y	Y	Y	Y	Y	Y	Y	Y
Industry-month FE	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y
Sample	All	All	All	AU	AU	2020H1	All	TW	TW	AU
Semielasticity	-0.1413	-0.1494	-0.1828	-0.1594	-0.1818	-0.3068		-0.0616

^{Table 5:}

The international spillover effect of lockdowns across industry upstreamness

Notes: The table reports the results of the estimation of equation 6 by Poisson pseudo-maximum likelihood. The stringency index is the Oxford COVID-19 Government Response Stringency Index, computed by Hale et al. (2021). The sample spans the period 2020:m1–2021:m6. Standard errors in parenthesis are clustered at the exporter level. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

^{Table 5:}

The international spillover effect of lockdowns across industry upstreamness

	(1)	(2)
Stringency index	-0.00234*** (0.001)
Stringency index x Upstreamness	0.00039* (0.000)	0.00057*** (0.000)
Observations	23,531,808	23,531,808
Exporter-importer-industry FE	Y	Y
Importer-industry-month FE	Y	Y
Exporter-month FE	N	Y

^{Table 5:}

The international spillover effect of lockdowns across industry upstreamness

	(1)	(2)
Stringency index	-0.00234*** (0.001)
Stringency index x Upstreamness	0.00039* (0.000)	0.00057*** (0.000)
Observations	23,531,808	23,531,808
Exporter-importer-industry FE	Y	Y
Importer-industry-month FE	Y	Y
Exporter-month FE	N	Y

^{Table 6:}

The international spillover effect of lockdowns: Workplace closings

Notes: The table reports the results of the estimation of equation 5 (columns 1–6) and equation 6 (columns 7–8) by Poisson pseudo-maximum likelihood. Rather the using the Oxford stringency index, the table uses the index of workplace closings, also computed by Hale et al. (2021). The sample spans the period 2020:ml-2021:m6; in column 3 the sample is restricted to the first six months of 2020, while in columns 5–6 it is limited to exporting countries for which the measure of teleworkability computed by Dingel and Neiman (2020) is available. The coefficient of the stringency index is estimated separately for: i) trade partners which have announced and implemented high and low discrectionary fiscal measures in response to the pandemic between January and June 2020 (fiscal spending is scaled by GDP and the sample is split along the first quartile of the distribution of fiscal spending over GDP); and ii) trade partners with high and low values of the teleworkability index (the sample is split along the first quartile of the distribution of the index). Standard errors in parenthesis are clustered at the exporter level. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. TW = teleworkability.

^{Table 6:}

The international spillover effect of lockdowns: Workplace closings

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Workplace closings	-0.02831*** (0.010)	-0.02840*** (0.010)	-0.07762*** (0.021)	-0.00885 (0.006)			-0.04202*** (0.016)
Covid cases per million, lagged		0.00001 (0.000)
Covid deaths per million, lagged		-0.00089 (0.001)
Number of new export restrictions		0.00797 (0.010)
Number of removed export restrictions		-0.00309* (0.002)
Stringency index x Small fiscal response					-0.04878*** (0.013)
Stringency index x Large fiscal response					-0.02696*** (0.010)
Workplace closings x Low telework						-0.02177* (0.012)
Workplace closings x High telework						-0.00814 (0.006)
Workplace closings x Upstreamness							0.00622 (0.004)	0.00801*** (0.003)
Observations	23,531,808	21,787,468	6,118,735	14,764,840	23,256,563	14,764,840	23,531,808	23,531,808
Exporter-importer-industry FE	Y	Y	Y	Y	Y	Y	Y	Y
Importer-industry-month FE	Y	Y	Y	Y	Y	Y	Y	Y
Exporter-month FE	N	N	N	N	N	N	N	Y
Industry-month FE	Y	Y	Y	Y	Y	Y	Y	Y
Sample	All	All	2020H1	TW	AU	TW	AU	All

^{Table 6:}

The international spillover effect of lockdowns: Workplace closings

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Workplace closings	-0.02831*** (0.010)	-0.02840*** (0.010)	-0.07762*** (0.021)	-0.00885 (0.006)			-0.04202*** (0.016)
Covid cases per million, lagged		0.00001 (0.000)
Covid deaths per million, lagged		-0.00089 (0.001)
Number of new export restrictions		0.00797 (0.010)
Number of removed export restrictions		-0.00309* (0.002)
Stringency index x Small fiscal response					-0.04878*** (0.013)
Stringency index x Large fiscal response					-0.02696*** (0.010)
Workplace closings x Low telework						-0.02177* (0.012)
Workplace closings x High telework						-0.00814 (0.006)
Workplace closings x Upstreamness							0.00622 (0.004)	0.00801*** (0.003)
Observations	23,531,808	21,787,468	6,118,735	14,764,840	23,256,563	14,764,840	23,531,808	23,531,808
Exporter-importer-industry FE	Y	Y	Y	Y	Y	Y	Y	Y
Importer-industry-month FE	Y	Y	Y	Y	Y	Y	Y	Y
Exporter-month FE	N	N	N	N	N	N	N	Y
Industry-month FE	Y	Y	Y	Y	Y	Y	Y	Y
Sample	All	All	2020H1	TW	AU	TW	AU	All

References

Amiti, M. and Weinstein, D. E. (2011). Exports and Financial Shocks. The Quarterly Journal of Economics, 126 (4), 1841–1877.
- Search Google Scholar
- Export Citation
Andersen, A. L., Hansen, E. T., Johannesen, N. and Sheridan, A. (2022). Consumer Responses to the COVID-19 Crisis: Evidence from Bank Account Transaction Data. Scandinavian Journal of Economics, Forthcoming.
- Search Google Scholar
- Export Citation
Anderson, J. E. and Yotov, Y. V. (2016). Terms of trade and global effciency effects of free trade agreements, 1990–2002. Journal of International Economics, 99, 279–298.
- Search Google Scholar
- Export Citation
AntràS, P., Chor, D., Fally, T. and Hillberry, R. (2012). Measuring the Upstreamness of Production and Trade Flows. American Economic Review, 102 (3), 412–416.
- Search Google Scholar
- Export Citation
Arenas, G. C., Majune, S. and Montfaucon, A. F. (2022). The Impacts of Lockdown Policies on International Trade in the Philippines. Policy Research Working Paper 9911, The World Bank, Washington DC.
- Search Google Scholar
- Export Citation
Arita, S., Grant, J., Sydow, S. and Beckman, J. (2022). Has global agricultural trade been resilient under coronavirus (COVID-19)? Findings from an econometric assessment of 2020. Food Policy, 107, 102204.
- Search Google Scholar
- Export Citation
Aum, S., Lee, S. Y. T. and Shin, Y. (2021). Inequality of fear and self-quarantine: Is there a tradeoff between GDP and public health? Journal of Public Economics, 194, 104354.
- Search Google Scholar
- Export Citation
Baker, S. R., Farrokhnia, R. A., Meyer, S., Pagel, M. and Yannelis, C. (2020). How Does Household Spending Respond to an Epidemic? Consumption during the 2020 COVID-19 Pandemic. The Review of Asset Pricing Studies, 10 (4), 834–862.
- Search Google Scholar
- Export Citation
Baldwin, R. and Freeman, R. (2022). Risks and Global Supply Chains: What We Know and What We Need to Know. Annual Review of Economics,, 14 (1).
- Search Google Scholar
- Export Citation
Baldwin, R. and Tomiura, E. (2020). Thinking ahead about the trade impact of COVID-19. In R. Baldwin and B. Weder di Mauro (eds.), Economics in the time of COVID-19, London: CEPR, pp. 59–71.
- Search Google Scholar
- Export Citation
Baqaee, D., Farhi, E., Mina, M. and Stock, J. H. (2020). Policies for a Second Wave. Brookings Papers on Economic Activity, 2020 (2), 385–443.
- Search Google Scholar
- Export Citation
Barrot, J.-N. and Sauvagnat, J. (2016). Input Specificity and the Propagation of Idiosyncratic Shocks in Production Networks. The Quarterly Journal of Economics, 131 (3), 1543–1592.
- Search Google Scholar
- Export Citation
Bartik, T. (1991). Boon or Boondoggle? The Debate over State and Local Economic Development Policies. Vol. 1, W.E. Upjohn Institute for Employment Research, pp. 1–16.
- Search Google Scholar
- Export Citation
Bas, M., Fernandes, A. and Paunov, C. (2022). How Resilient Was Trade to COVID-19? Policy Research Working Paper 9975, The World Bank.
- Search Google Scholar
- Export Citation
Berthou, A. and Stumpner, S. (2022). Trade under lockdown. Working Paper 867, Banque de France, unpublished manuscript.
- Search Google Scholar
- Export Citation
Boehm, C. E., Flaaen, A. and Pandalai-nayar, N. (2019). Input Linkages and the Transmission of Shocks: Firm-Level Evidence from the 2011 To¯hoku Earthquake. The Review of Economics and Statistics, 101 (1), 60–75.
- Search Google Scholar
- Export Citation
Bonadio, B., Huo, Z., Levchenko, A. A. and Pandalai-nayar, N. (2021). Global supply chains in the pandemic. Journal of International Economics, 133, 103534.
- Search Google Scholar
- Export Citation
Bounie, D., Camara, Y., Fize, E., Galbraith, J., Landais, C., Lavest, C., Pazem, T. and Sa-vatier, B. (2020). Consumption dynamics in the COVID crisis: Real-time insights from French transaction and bank data. COVID Economics, 59, 1–39.
- Search Google Scholar
- Export Citation
Brussevich, M., Papageorgiou, C. and Wibaux, P. (2022). Trade and the COVID-19 Pandemic: Lessons from French Firms. Working Paper 2022/081, International Monetary Fund.
- Search Google Scholar
- Export Citation
Buchel, K., Legge, S., Pochon, V. and Wegmuller, P. (2020). Swiss trade during the COVID-19 pandemic: an early appraisal. Swiss Journal of Economics and Statistics, 156 (22), 1–15.
- Search Google Scholar
- Export Citation
BussièRe, M., Callegari, G., Ghironi, F., Sestieri, G. and Yamano, N. (2013). Estimating Trade Elasticities: Demand Composition and the Trade Collapse of 2008–2009. American Economic Journal: Macroeconomics, 5 (3), 118–151.
- Search Google Scholar
- Export Citation
Cameron, E., Nuzzo, J. B. and Bell, J. A. (2019). Global Health Security Index: Building Collective Actionand Accountability. Tech. rep., Johns Hopkins Bloomberg School of Public Health, Baltimore, MD: Center for HealthSecurity.
- Search Google Scholar
- Export Citation
Carvalho, V. M., Garcia, J. R., Hansen, S., Ortiz, Á., Rodrigo, T., Mora, J. V. R. and Ruiz, P. (2021a). Tracking the COVID-19 crisis with high-resolution transaction data. Royal Society Open Science,, 8 (8).
- Search Google Scholar
- Export Citation
Carvalho, V. M., Garcia, J. R., Hansen, S., Ortiz, Á., Rodrigo, T., Mora, J. V. R., Nirei, M., Saito, Y. U. and Tahbaz-salehi, A. (2021b). Supply Chain Disruptions: Evidence from the Great East Japan Earthquake. The Quarterly Journal of Economics, 136 (2), 1255–1321.
- Search Google Scholar
- Export Citation
Cerdeiro, D. and Komaromi, A. (2020). Supply Spillovers During the Pandemic: Evidence from High-Frequency Shipping Data. IMF Working Paper 20/284, International Monetary Fund.
- Search Google Scholar
- Export Citation
Chen, S., Igan, D., Pierri, N. and Presbitero, A. (2020). Tracking the Economic Impact of COVID-19 and Mitigation Policies in Europe and the United States. IMF Working Paper 20/125, International Monetary Fund.
- Search Google Scholar
- Export Citation
Chronopoulos, D. K., Lukas, M. and Wilson, J. O. S. (2020). Consumer Spending Responses to the Covid-19 Pandemic: An Assessment of Great Britain. COVID Economics, 34, 145–186.
- Search Google Scholar
- Export Citation
Correia, S., GuimarÃEs, P. and Zylkin, T. (2020). Fast Poisson estimation with high-dimensional fixed effects. The Stata Journal: Promoting communications on statistics and Stata, 20 (1), 95–115.
- Search Google Scholar
- Export Citation
Deb, P., Furceri, D., Ostry, J. D. and Tawk, N. (2021). The Economic Effects of COVID-19 Containment Measures. Open Economies Review, 33 (1), 1–32.
- Search Google Scholar
- Export Citation
Dingel, J. I. and Neiman, B. (2020). How many jobs can be done at home? Journal of Public Economics, 189, 104235.
- Search Google Scholar
- Export Citation
Espitia, A., Mattoo, A., Rocha, N., Ruta, M. and Winkler, D. (2022). Pandemic trade: COVID-19, remote work and global value chains. The World Economy, 45 (2), 561–589.
- Search Google Scholar
- Export Citation
Evenett, S., Fiorini, M., Fritz, J., Hoekman, B., Lukaszuk, P., Rocha, N., Ruta, M., Santi, F. and Shingal, A. (2021). Trade policy responses to the COVID-19 pandemic crisis: Evidence from a new data set. The World Economy, 45 (2), 342–364.
- Search Google Scholar
- Export Citation
Evenett, S. J. (2019). Protectionism, state discrimination, and international business since the onset of the Global Financial Crisis. Journal of International Business Policy, 2 (1), 9–36.
- Search Google Scholar
- Export Citation
Evenett, S. J. (2020). Sicken thy neighbour: The initial trade policy response to COVID-19. The World Economy, 43 (4), 828–839.
- Search Google Scholar
- Export Citation
Frederick, S. (2019). Global value chain mapping. In S. Ponte, G. Gereff and G. Raj-Reichert (eds.), Handbook on Global Value Chains, Northampton, MA: Edward Elgar Publishing.
- Search Google Scholar
- Export Citation
Frederick, S. and Lee, J. (2017). Korea and the Electronics Global Value Chain, Global Value Chains Center, Duke University, Durham, NC.
- Search Google Scholar
- Export Citation
Grant, J. H., Arita, S., Emlinger, C., Johansson, R. and Xie, C. (2021). Agricultural exports and retaliatory trade actions: An empirical assessment of the 2018/2019 trade conflict. Applied Economic Perspectives and Policy, 43 (2), 619–640.
- Search Google Scholar
- Export Citation
Guerrieri, V., Lorenzoni, G., Straub, L. and Werning, I. (2022). Macroeconomic Implications of COVID-19: Can Negative Supply Shocks Cause Demand Shortages? American Economic Review, 112 (5), 1437–1474.
- Search Google Scholar
- Export Citation
Hale, T., Angrist, N., Goldszmidt, R., Kira, B., Petherick, A., Phillips, T., Webster, S., Cameron-blake, E., Hallas, L., Majumdar, S. and Tatlow, H. (2021). A global panel database of pandemic policies (Oxford COVID-19 Government Response Tracker). Nature Human Behaviour, 5 (4), 529–538.
- Search Google Scholar
- Export Citation
Heise, S. (2020). How Did China’s COVID-19 Shutdown Affect U.S. Supply Chains?, Liberty Street Economics. Federal Reserve Bank of New York, Available at: https://libertystreeteconomics.newyorkfed.org/2020/05/how-did-chinas-covid-19-shutdown-affect-us-supply-chains/.
- Search Google Scholar
- Export Citation
International Monetary Fund (2020). World Economic Outlook, October 2020: A Long and Difficult Ascent. Washington, DC: International Monetary Fund.
- Search Google Scholar
- Export Citation
Kose, A., Sugawara, N. and Terrones, M. E. (2020). Global Recessions. Discussion Paper 14397, CEPR.
- Search Google Scholar
- Export Citation
Lafrogne-joussier, R., Martin, J. and Mejean, I. (2022). Supply shocks in supply chains: Evidence from the early lockdown in China. IMF Economic Review, Forthcoming.
- Search Google Scholar
- Export Citation
Liu, X., Ornelas, E. and Shi, H. (2022). The Trade Impact of the Covid-19 Pandemic. The World Economy, Forthcoming.
- Search Google Scholar
- Export Citation
Majune, S. and Addisu, L. (2021). The Effect of Lockdown Policies on International Trade:Evidence from Kenya. Global Working Paper 148, Brookings Institution, Washington DC.
- Search Google Scholar
- Export Citation
Pei, J., Vries, G. and Zhang, M. (2021). International trade and Covid-19: City-level evidence from China’s lockdown policy. Journal of Regional Science.
- Search Google Scholar
- Export Citation
Pimenta, A. C., Gouveia, C. M. and Amador, J. (2021). COVID-19, Lockdowns and International Trade: Evidence from Firm-Level Data. Working Paper 14, Bank of Portugal.
- Search Google Scholar
- Export Citation
Ritchie, H., Mathieu, E., Rodes-guirao, L., Appel, C., Giattino, C., Ortiz-ospina, E., Hasell, J., Macdonald, B., Beltekian, D. and Roser, M. (2020). Coronavirus Pandemic (COVID-19). Published online at https://ourworldindata.org/coronavirus.
- Search Google Scholar
- Export Citation
Silva, J. M. C. S. and Tenreyro, S. (2006). The Log of Gravity. The Review of Economics and Statistics, 88 (4), 641–658.
- Search Google Scholar
- Export Citation
Silva, J. M. C. S. and Tenreyro, S. (2022). The Log of Gravity at 15. Portuguese Economic Journal.
- Search Google Scholar
- Export Citation
Sturgeon, T., Daly, J., Frederick, S., Bamber, P. and Gereffi, G. (2016). The Philippines in the AutomotiveGlobal Value Chain. Durham, NC: Center on Globalization,Governance and Competitiveness, Duke University.
- Search Google Scholar
- Export Citation
Sturgeon, T., Daly, J., Frederick, S., Bamber, P. and Memedovic, O. (2010). Mapping global value chains: Intermediate goods trade and structural change in the world economy. Working Paper 05/2010, UNIDO, Vienna.
- Search Google Scholar
- Export Citation
Yotov, Y. (2022). Gravity at Sixty: The Bijou of Trade,. Working Paper 2022–1, LeBow College of Business, Drexel University.
- Search Google Scholar
- Export Citation

Online Appendix—Not for Publication

Additional Figures

Additional Tables

^{Table A1:}

Import demand model: Panel estimates

Notes: Results from a panel regression estimated on a sample of 127 countries with at least 16 observations between 1985 and 2019. IAD = imported adjusted demand. Standard errors in parenthesis are clustered at the country level. Standard errors in parenthesis are clustered at the country level. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

^{Table A1:}

Import demand model: Panel estimates

	(1)	(2)	(3)	(4)	(5)	(6)
	Total	Total	Services	Services	Goods	Goods
Relative Price	-0.29*** (0.099)	-0.30*** (0.10)	-0.24** (0.11)	-0.25** (0.11)	-0.31** (0.13)	-0.32** (0.13)
IAD – Total	0.99*** (0.088)	0.94*** (0.090)
IAD – Services			1.11*** (0.15)	1.09*** (0.16)
IAD – Goods					0.96*** (0.096)	0.91*** (0.096)
Adj. R²	0.51	0.53	0.086	0.088	0.39	0.41
Country FE	Y	Y	Y	Y	Y	Y
Year FE	N	Y	N	Y	N	Y
Number of countries	127	127	127	127	127	127

^{Table A1:}

Import demand model: Panel estimates

	(1)	(2)	(3)	(4)	(5)	(6)
	Total	Total	Services	Services	Goods	Goods
Relative Price	-0.29*** (0.099)	-0.30*** (0.10)	-0.24** (0.11)	-0.25** (0.11)	-0.31** (0.13)	-0.32** (0.13)
IAD – Total	0.99*** (0.088)	0.94*** (0.090)
IAD – Services			1.11*** (0.15)	1.09*** (0.16)
IAD – Goods					0.96*** (0.096)	0.91*** (0.096)
Adj. R²	0.51	0.53	0.086	0.088	0.39	0.41
Country FE	Y	Y	Y	Y	Y	Y
Year FE	N	Y	N	Y	N	Y
Number of countries	127	127	127	127	127	127

^{Table A2:}

Residual analysis: Travel imports as a share of total service imports

Notes: Results from a regression of the import demand model prediction errors in 2020 on the share of travel import over total service import, computed from the WTO service import database and averaged over the period 2016–2019. Standardized coefficients represent the number of standard deviation changes in the dependent variables associated to one standard deviation change in the variable of interest. Standard errors, in parenthesis, are robust to heteroscedasticity. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

^{Table A2:}

Residual analysis: Travel imports as a share of total service imports

		(1) Total	(2) Services	(3) Goods
Travel imports over total service imports		-0.00158 (0.00101)	-0.00760*** (0.00235)	-0.000426 (0.000966)
	Standardized coefficient	-0.189	-0.331***	-0.0423
Observations		105	105	105
Adjusted R-squared		0.024	0.105	-0.008

^{Table A2:}

Residual analysis: Travel imports as a share of total service imports

		(1) Total	(2) Services	(3) Goods
Travel imports over total service imports		-0.00158 (0.00101)	-0.00760*** (0.00235)	-0.000426 (0.000966)
	Standardized coefficient	-0.189	-0.331***	-0.0423
Observations		105	105	105
Adjusted R-squared		0.024	0.105	-0.008

^{Table A3:}

The international spillover effect of lockdowns: Alternative clustering

Notes: The table reports the results of the estimation of equation 5 (columns 1–6) and equation 6 (columns 7–8) by Poisson pseudo-maximum likelihood. The stringency index is the Oxford COVID-19 Government Response Stringency Index, computed by Hale et al. (2021). The sample spans the period 2020:ml-2021 :m6; in column 3 the sample is restricted to the first six months of 2020, while in columns 5–6 it is limited to exporting countries for which the measure of teleworkability computed by Dingel and Neiman (2020) is available. The coefficient of the stringency index is estimated separately for: i) trade partners which have announced and implemented high and low discrectionary fiscal measures in response to the pandemic between January and June 2020 (fiscal spending is scaled by GDP and the sample is split along the first quartile of the distribution of fiscal spending over GDP); and ii) trade partners with high and low values of the teleworkability index (the sample is split along the first quartile of the distribution of the index). Standard errors in parenthesis are clustered at the exporter-month level. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. TW = teleworkability.

^{Table A3:}

The international spillover effect of lockdowns: Alternative clustering

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Stringency index	-0.00149*** (0.000)	-0.00182*** (0.000)	-0.00307*** (0.001)	-0.00062*** (0.000)			-0.00234*** (0.001)
Covid cases per million, lagged		0.00002 (0.000)
Covid deaths per million, lagged		-0.00051 (0.001)
Number of new export restrictions		0.00917 (0.008)
Number of removed export restrictions		-0.00199 (0.002)
Stringency index x Small fiscal response					-0.00201*** (0.000)
Stringency index x Large fiscal response					-0.00146*** (0.000)
Stringency index x Low telework						-0.00126*** (0.000)
Stringency index x High telework						-0.00058*** (0.000)
Stringency index x Upstreamness							0.00039** (0.000)	0.00057*** (0.000)
Observations	23,531,808	21,787,468	6,118,735	14,764,840	23,256,563	14,764,840	23,531,808	23,531,808
Exporter-importer-industry FE	Y	Y	Y	Y	Y	Y	Y	Y
Importer-industry-month FE	Y	Y	Y	Y	Y	Y	Y	Y
Exporter-month FE	N	N	N	N	N	N	N	Y
Industry-month FE	Y	Y	Y	Y	Y	Y	Y	Y
Sample	All	All	2020H1	TW	AU	TW	AU	All

^{Table A3:}

The international spillover effect of lockdowns: Alternative clustering

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Stringency index	-0.00149*** (0.000)	-0.00182*** (0.000)	-0.00307*** (0.001)	-0.00062*** (0.000)			-0.00234*** (0.001)
Covid cases per million, lagged		0.00002 (0.000)
Covid deaths per million, lagged		-0.00051 (0.001)
Number of new export restrictions		0.00917 (0.008)
Number of removed export restrictions		-0.00199 (0.002)
Stringency index x Small fiscal response					-0.00201*** (0.000)
Stringency index x Large fiscal response					-0.00146*** (0.000)
Stringency index x Low telework						-0.00126*** (0.000)
Stringency index x High telework						-0.00058*** (0.000)
Stringency index x Upstreamness							0.00039** (0.000)	0.00057*** (0.000)
Observations	23,531,808	21,787,468	6,118,735	14,764,840	23,256,563	14,764,840	23,531,808	23,531,808
Exporter-importer-industry FE	Y	Y	Y	Y	Y	Y	Y	Y
Importer-industry-month FE	Y	Y	Y	Y	Y	Y	Y	Y
Exporter-month FE	N	N	N	N	N	N	N	Y
Industry-month FE	Y	Y	Y	Y	Y	Y	Y	Y
Sample	All	All	2020H1	TW	AU	TW	AU	All

Shan Chen and Bryan Zou provided excellent research assistance. We thank Oya Celasun, Romain Duval, Stacey Frederick, Miguel Fuentes, Gita Gopinath, Petya Koeva Brooks, Andrei Levchenko, Brad MacDonald, Sergii Meleshchuk, Timothy Sturgeon, Petia Topalova and Daria Taglioni for helpful comments and suggestions. We received excellent feedback from seminars at the World Trade Organization, Bruegel, Swiss National Bank, Universidad Adolfo Ibanez, Universidad de Chile, Central Bank of Chile, Central Bank of Colombia, Fedesarrollo (Bogotà), Asia School of Business in Kuala Lumpur, Asean +3 Macroeconomic Reasearch Office, and the Monetary Authority of Singapore. The views expressed here are those of the authors and should not be attributed to the International Monetary Fund, its Executive Board, or its management.

Unless specifed otherwise, throughout the paper we measure the severity of containment policies using the Oxford COVID-19 Government Response Stringency Index, as computed by Hale et al. (2021). For simplicity, we refer to it as the stringency index.

GVC-intensive industries include automobiles, electronics, textiles and garments, and medical goods. The mapping of goods from the “six-digit” level under the Harmonized System to the inputs and final goods in these GVCs are compiled from Frederick and Lee (2017) (electronics), Sturgeon et al. (2016) (automobiles), and Frederick (2019) (textiles and garments, medical devices).

The result that the COVID-19 shock had a larger effect on trade for products that reply on intermediate inputs from China is also shown in a cross-country setting by Bas et al. (2022).

Similar evidence is shown also using trade data. Majune and Addisu (2021) find that the introduction of lockdown measures by Kenya’s trading partners had a negative effect on imports, which fell by 23% on average after the measures were put in place. Looking the the Philippines, Arenas et al. (2022) find that the drop in imports has been even larger (-57% compared to pre-pandemic levels).

A related literature investigates the disruptions in agricultural trade during the pandemic, showing that containment policies and reduced mobility are associated with the decline in bilateral trade, although the size of decline has been smaller than for non-agricultural trade and the effect of containment polices has weakened over time (Arita et al., 2022). More generally, our analysis builds on an extensive literature that measures the effects on trade flows of a variety of domestic shocks, natural disasters (Barrot and Sauvagnat, 2016; Boehm et al., 2019; Carvalho et al., 2021b) and financial sector shocks (Amiti and Weinstein, 2011).

The sample is selected based on data availability: a country is in the sample if we can observe at least 16 complete observations between 1985 and 2019.

In what follows we use the term prediction errors to refer both to in-sample and out-of-sample prediction errors. The former type of error is computed for years 1985 to 2019, the years included in our estimation sample. The latter type of error is computed for 2020, as data for that year are not included in the estimateion sample.

Table A1 also reports the results from a panel version of the equation in 1, including importer and time fixed effects (α_m and γ_t):

\begin{matrix} Δ \ln M_{m t}^{τ} = α_{m}^{τ} + γ_{t}^{τ} + β_{D}^{τ} Δ \ln D_{m t}^{τ} + β_{P}^{τ} Δ \ln P_{m t} + ε_{m t}^{τ}, \forall τ \in {g, s} & (3) \end{matrix}

Consistent with economic intuition, the estimated coefficients on relative prices are negative and significant across all specifications and close to 0.3. The coefficient on import-intensity adjusted demand is positive and close to 1 for goods, and around 1.1 for services. The adjusted R² suggests that the model for imports of goods explains about 40% of the variation in the data, while it is just shy of explaining 10% of the variation in the growth of service imports.

Figure A1 summarizes the distribution of prediction errors in 2019 and 2020. Concerning services, the distribution in 2020 is more disperse than the distribution in 2019, has a lower mean and a prominent negative skew. While the distributions of the errors coming from the model for the import growth of goods are closer to each other, the variance of errors in 2020 is still larger and the distribution features a prominent long right tail, consistent with the positive average prediction error observed in aggreagate. The mean square errors reported in Figure A2 in the appendix deliver a similar message, although the mean square errors for goods was higher in 2000 than 2020.

While imported services such as travel indeed declined, this was not the case for other categories such as communication. Relatedly, as countries closed their borders to contain the spread of the virus, tourism collapsed, explaining much of the fall in service imports. Large importers of tourism (as captured by the average value of travel imports as a share of GDP between 2016 Q1 and 2019 Q4) saw a much larger than expected drop in service import growth, as shown in Table A2 in the appendix.

For details on the index, see Cameron et al. (2019), as well as other material that can be found on the Global Health and Security Index website at https://www.ghsindex.org/about/.

Trade Data Monitor data are available by subscription at https://tradedatamonitor.com/ and cover also exports.

See: https://www.bea.gov/industry/historical-benchmark-input-output-tables.

See: https://www.globaltradealert.org/data_extraction for details on the coverage and the methodology of the GTA data.

Another potential variable to control for would be mobility, measured by the average of all the components of the Google mobility score excluding parks and residential. However, a decline in mobility is an inevitable consequence of a lockdown, making it strongly correlated with the stringency index. In the sample 2020:m1–2021:m6, the elasticity of mobility to the stringency index (computed by a simple regression controlling for time and country fixed effects) is -0.5. As the analysis focuses on the effects of the containment policy measures (e.g., lockdowns) rather than on actual behavior (which could also reflect individual choices), the empirical model incorporates the stringency index rather than mobility.

Interestingly, this finding is very close to the results of Pimenta et al. (2021) using firm-level Portuguese data.

An alternative interpretation is that, after the initial shock, the stringency index does not capture adequately the intensity of the lockdown measures relevant for production and trade. However, measuring containment policies exclusively by an index of workplace closings delivers similar results, mitigating concerns about measurement issues— see Section 3.5.

The effective fall in imports is equal to the value of the series in January (96.5) minus the value in May (72.5). In the same way, the fall in the counterfactual without containment policies is 100–90.3. Thus, lockdowns account for (24–9.6)/24 = 59.7 percent of the actual import decline.

See: https://www.imf.org/en/Topics/imf-and-covid19/Fiscal-Policies-Database-in-Response-to-COVID-19; we use the June 2020 vintage of the database to focus on the initial fiscal reponse as the international spillovers are concentrated in the first 6 months of 2020 (Figure 4).

Antràs et al. (2012) also compute the industry measure of upstreamness for other economies with I-O tables and show that this is generally stable across countries. Given the primary goal of keeping bilateral trade flows in the gravity model as large as possible, the US measure of upstreamness is applied to all exporter countries.

Given that average value of the upstreamness indicator is equal to 2.07, the coefficient on stringency (measured at the average level of upstreamness) is equal to -0.00234 + 0.00039 ∗ 2.07 =-0.0015. A one standard deviation increase in upstreamness increases the coefficient by 0.00039 ∗ .81 = 0.0003, which correspond to about one fifth of teh average effect (.0003/.0015 = 0.20).

See https://www.bsg.ox.ac.uk/research/research-projects/covid-19-government-response-tracker for further details on the Oxford stringency index and its individual components.

Other Publishers

Save
Cite
Email this content

Share Link

Copy this link, or click below to email it to a friend
Email this content
or copy the link directly:

https://www.elibrary.imf.org/view/journals/001/2022/120/article-A001-en.xml
The link was not copied. Your current browser may not support copying via this button.

Link copied successfully

Collapse
Expand

International Trade Spillovers from Domestic COVID-19 Lockdowns

Author:

Mr. Shekhar Aiyar

Mr. Davide Malacrino

Mr. Adil Mohommad

, and

Mr. Andrea F Presbitero

Volume/Issue:: Volume 2022: Issue 120

Publisher:: International Monetary Fund

ISBN:: 9798400212178

ISSN:: 1018-5941

Pages:: 48

DOI:: https://doi.org/10.5089/9798400212178.001

Keywords
1 Introduction
2 Were the effects of the pandemic on imports “unprecedented”?
3 International spillover effects of lockdowns
4 Conclusions
Figures
Tables
References
Online Appendix—Not for Publication
- Additional Figures
- Additional Tables

View raw image

^{Figure 1:}

Trade in the pandemic

View raw image

^{Figure 2:}

Change in imports and partner countries’ lockdown stringency

View raw image

^{Figure 3:}

Trade in the pandemic: Results from an import demand model

View raw image

^{Figure 4:}

The international spillover effect of lockdowns over time

View raw image

^{Figure 5:}

Quantifying the international spillover effect of lockdowns

View raw image

^{Figure 6:}

The heterogeneous international spillover effect of lockdowns

View raw image

^{Figure 7:}

Robustness: the international spillover effect of lockdowns dropping country groups

View raw image

^{Figure A1:}

Performance of the import demand model: 2019 vs 2020

View raw image

^{Figure A2:}

Performance of the import demand model: Mean square errors

View raw image

^{Figure A3:}

Representativeness of the TDM data

View raw image

^{Figure A4:}

The stringency index and workplace closings

View raw image

^{Figure A5:}

The international spillover effect of lockdowns over time, excluding China as an exporter

Foresee

^{Table A3:}

The international spillover effect of lockdowns: Alternative clustering

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Stringency index	-0.00149*** (0.000)	-0.00182*** (0.000)	-0.00307*** (0.001)	-0.00062*** (0.000)			-0.00234*** (0.001)
Covid cases per million, lagged		0.00002 (0.000)
Covid deaths per million, lagged		-0.00051 (0.001)
Number of new export restrictions		0.00917 (0.008)
Number of removed export restrictions		-0.00199 (0.002)
Stringency index x Small fiscal response					-0.00201*** (0.000)
Stringency index x Large fiscal response					-0.00146*** (0.000)
Stringency index x Low telework						-0.00126*** (0.000)
Stringency index x High telework						-0.00058*** (0.000)
Stringency index x Upstreamness							0.00039** (0.000)	0.00057*** (0.000)
Observations	23,531,808	21,787,468	6,118,735	14,764,840	23,256,563	14,764,840	23,531,808	23,531,808
Exporter-importer-industry FE	Y	Y	Y	Y	Y	Y	Y	Y
Importer-industry-month FE	Y	Y	Y	Y	Y	Y	Y	Y
Exporter-month FE	N	N	N	N	N	N	N	Y
Industry-month FE	Y	Y	Y	Y	Y	Y	Y	Y
Sample	All	All	2020H1	TW	AU	TW	AU	All

Topics

Countries

Series

About

Resources

Abstract

1 Introduction

2 Were the effects of the pandemic on imports “unprecedented”?

2.1 A simple empirical import demand model

2.2 Model performance

2.3 The drivers of pandemic prediction errors: going beyond demand

3 International spillover effects of lockdowns

3.1 Modeling international spillovers

3.2 Baseline Results

3.3 Heterogeneous Effects of Lockdowns

3.4 A Full-Fledged Gravity Model

3.5 Robustness

4 Conclusions

Figures

Tables

References

Online Appendix—Not for Publication

Additional Figures

Additional Tables

Same Series

Other IMF Content

Other Publishers

Asian Development Bank

Inter-American Development Bank

The World Bank

Share Link

Table of Contents

Headings

Figures

Metrics

Foresee