Scarcity nationalism during COVID-19: Identifying the impact on trade costs

During the COVID-19 pandemic, many countries used export and import policy as a tool to expand the availability of scarce critical medical products in the domestic market (scarcity nationalism). This paper assesses the direct and indirect (via trade in intermediates) increases in trade costs of critical medical goods resulting from these uncooperative policies. The results show that scarcity nationalism led to substantial increases in trade costs between February 2020 and December 2021 for most COVID-19 critical medical products, particularly garments (for example, face masks) and ventilators. The exception is vaccines, which saw a reduction in trade costs, which, however, was driven by the reduction in indirect trade costs for high-income countries, consistent with the view of a COVID-19 vaccine production club.


Introduction
During the COVID-19 pandemic, we witnessed a reemergence of trade protectionism. While traditionally protectionism is pursued by importers to shield domestic producers from foreign competition, the protectionism applied during the pandemic aimed at maximizing the availability of sensitive, foremost medical, products in the domestic market through a mix of import facilitation and export controls (Evenett et al., 2022). We refer to this policy mix as ''scarcity nationalism'' to distinguish it from traditional trade protectionism. We assess the impact of these policies on the trade costs of critical medical goods.
The analysis relies on Global Trade Alert (GTA) data on policy measures that have been adopted by countries at the HS6-tradedproduct level for medical products that have been critical during the COVID-19 pandemic, such as face masks and ventilators, and vaccines and vaccine inputs. After merging these data with bilateral product-level trade data, we measure the impact of these * Corresponding author at: ETH Zurich, Leonhardstrasse 21, 8092 Zurich,
1 We gratefully acknowledge the useful comments by an anonymous reviewer.
trade policies on trade margins. We disentangle the effect of trade measures (restricting and facilitating) on medical products using monthly data between January 2018 (prior to the pandemic, when no such measures were used) and December 2021. We also look into product, time and geographic heterogeneity and provide a portrait of the impact scarcity nationalism during the COVID-19 pandemic on trade costs. In doing so, we separate direct effects from indirect ones, as some affected products are intermediates. Interest in the effects of trade protectionism has risen in recent years as tensions in the world trade system have emerged. This paper relates to recent work on the economic effects of trade policy (Goldberg and Pavcnik, 2016;Handley et al., 2020;Amiti et al., 2019;Fajgelbaum et al., 2020). Differently from these studies, the focus here is not on the consequences of tariffs but on a broader set of trade-policy measures targeting medical products during the COVID-19 pandemic. Moreover, we study both the direct and indirect effects of trade policy through the intermediate trade channel.

Data
We use data from three sources. First, we construct binary indicators reflecting whether a country utilizes import (quotas, tariffs, monitoring, etc.)  etc.) policy measures or not. The source of these data is the Global Trade Alert (GTA), and the data vary by country j, product k, and month t. As both import and export measures can be restricting (if the GTA evaluation is red or amber) or facilitating (if the GTA evaluation is green), there are four indicators. We use them at monthly frequency between January 2018 and December 2021 and employ export measures in a way so that they are on the exports of a product to country j. The policy measures do not vary much across partner countries so that we make them importercountry-product-month-specific for both imports and exports (to the importer) upfront. We call them d m jk,t for import policies, a scalar for the mth (restraining and facilitating) import measures. Similarly, we refer to export measures towards importer j in k by d x jk,t , a scalar for the mth (restraining and facilitating) export measures. Stacking these measures for all importers J and products K Second, we use bilateral export data from country i to j of product k, x ijk,t for 180 products indexed by k on a monthly basis between January 2018 and December 2021. The monthly trade data are available for the European Union, the United States, and China, and we treat those three as partner countries (exporters or importers) of J = 247 countries, considering all possible bilateral relations between these countries in the exporting and importing directions. These data are used to estimate parameters revealing the ad-valorem-equivalent trade costs associated with the policy variables in ∆ t .
Third, we use cross-sectional bilateral export data for 180 + 1 products and J(J − 1) country pairs from UNCTAD. These are obtained from averaging annual export data for the period 2017-2019. Then, we construct a (J × J)(K + 1) matrix of bilateral trade flows. 2 We combine these data with the S × S (sector-by-sector) input-output matrix for the United States in 2012 (available from the BEA) to impute a K × K matrix by using product-sector concordance tables from UNCTAD. The obtained matrix will serve to compute input-output-repercussion indirect effects of trade policies on sensitive products.

Effects of pandemic-related trade policy measures on trade costs and imports
We distinguish between direct and indirect effects (through imported inputs) of the trade policy measures of interest during the time period January 2018-December 2021. For this, use g jk,t to denote the import-value equivalent for importer j, product k, and month t. We establish later how to estimate it and postulate that g jk,t is linearly composed of the direct effects associated with the four import-and (partner-)export-related policy measures which are collected in ∆ jk,t , and with as many corresponding indirect effects.
of expenditure shares for each product k (Eaton and Kortum, 2002;Caliendo and Parro, 2014). For this, start with M, a JK × J imports matrix. The diagonal elements (expenditure on domestic produce) can be predicted from exporter and importer fixed effects in a gravity equation per product, assuming domestic sales frictions are zero. The matrix M is then row-normalized (dividing by aggregate expenditure by product) to obtain where O is the S×S input-output matrix of the United States with sector number S < K , and C is a K × S conversion matrix which 2 We consider K + 1 products here, in order to account for the bulk of products (a residual category) beyond the 180 sensitive ones in focus here. associates products with sectors. 3 Assuming that an estimate of the (global country-product-pair) Leontief-type (pseudo-)inverse The latter is nothing else where the first term captures the indirect relevance of the application of policy measures in ∆ t for each importer j and product k from other units (the same or other importers and the same or other products) in month t. As ∆ t is JK × 4, so is Λ t , and the latter has typical row Λ jk,t .
Using η k and ξ k for the 4 × 1 parameter vectors on the columns of ∆ jk,t and Λ jk,t , respectively, we can define Conditional on the direct effects ∆ jk,t η k , Λ jk,t ξ k measures the indirect effects of the policy measures in ∆ jk,t η k on g jk,t .
We can estimate g jk,t from bilateral product-level data on exports or imports x ijk,t , assuming a customary level-multiplicative or log-additive gravity model of product-level sales of country i to j in month t: where α ik,t , β ijk , and γ jk,t are parameters capturing sales/supplypotential factors of k in i, the trade freedom of k between i and j, and the demand/consumption potential of k in j, all in logs and at time/month t. When only trade but no domestic sales data are available, the unilateral frictions on k enter γ jk,t only.
We proceed in two steps to identify the parameters η k and ξ k of interest: (i) estimate γ jk,t as importer-product-month fixed effects from monthly bilateral product-level export data x ijk,t on 180 sensitive products using Eq. (3) with the aforementioned fixed effects; (ii) estimate the impact of unilateral trade frictions on product k from γ jk,t = g jk,t +γ jk + λ where g jk,t is defined as in (2), andγ jk , λ γ t and u γ ik,t are importerproduct-fixed, month-fixed, and importer-product-month residual effects, respectively.
As all ∆ jk,t are binary indicators, all parameters η measure (direct) semi-elasticities of imports. Note that under the assumption of a constant-elasticity-of-substitution model with (direct) trade elasticity (1 − σ k ), 4 one can scaleγ jk,t by (1 − σ k ) to obtain the iceberg-trade-cost increment associated with pandemicrelated policies alone by product, importer, and month. E.g., τ jk,t = 0.02 then means that trade costs are (directly plus indirectly) raised by pandemic-related policies for product k and country j in month t by about two percentage points. Apart from reporting on τ jk,t , we (trade-weighted) aggregate these trade costs effects for each of the 247 countries and each month t to obtain τ j,t . Finally, we aggregate the same trade-cost effects across countries j to obtain τ t for the average country and sensitive product.

Results
In the following, we first display the evolution of τ k,t for selected sensitive products -namely, garments, ventilators, vaccines and tests. The results are summarized in Fig. 1. Fig. 1 illustrates the evolution of import trade costs expressed as iceberg-surcharge costs over mill prices. For garments such as face masks ( Fig. 1(a)) those appeared to be very low (in the 1% range) for the period before the pandemic. We observe a slight increase in trade costs already in January of 2020, and by April of 2020 the costs peaked at more than 140%. They came down again only to levels below 5% after June of 2021. Interestingly, indirect effects on trade costs account for most of the increase during March and April 2021, but later become negative, perhaps suggesting a change in the production process of garments.
We see a sharp facilitation of the trade of the COVID-19 vaccines from the time that such vaccines became available in the first quarter of 2021 both in terms of direct and indirect costs ( Fig. 1(b)). This result is not in contrast with the ''vaccine nationalism'' view. It simply reflects the fact that these products were invented and introduced during the pandemic and they required trading in inputs that were not much traded before. Much of this trade took place within the set of economies that were producing vaccines, which were mostly high-income countries, with limited trade with non-producing countries (Evenett et al., 2021).
For ventilators (Fig. 1(c)), trade costs increased from virtually zero to about 10% even as early as the beginning of 2019 -obviously unrelated to the COVID-19 pandemic. Frictions increased sharply with a slight delay relative to garments but similarly peaking in the middle of 2020, at somewhat higher than 10%. Unlike garments, frictions on ventilators came down at the end of the sample period to the level they had prior to the COVID-19 pandemic.
Policy-induced trade frictions on tests ( Fig. 1(d)) increased gradually prior to the outbreak of the pandemic, but there was a sharp increase from the beginning of the pandemic, especially in terms of indirect trade costs. The total costs experienced a more moderate increase thanks to the negative contribution of the direct trade costs.  Upon aggregating trade-cost increases for the average month during the period after February 2020 until the end of 2021 for the (trade-weighted) average product, we obtain the following insights. First, import frictions were increasing at high levels in the average month in particular for the United States, Türkiye, India, Japan, China and the Russian Federation followed by other countries, mostly in the northern hemisphere and in South America. Second, when considering changes between the period from February 2020 until the end of our sample coverage relative to the period from the beginning of 2018 to January 2018, we see that the change in the uptake of restraining policies was quite significant and widespread with important peaks in Russia, Indonesia, China and Argentina for the average sensitive product.
Finally, in Table 1, we aggregate the information on changing trade frictions into country groups by income bracket and report the change in percentage points between the period after February 2020 and the period before up until January 2020 in the average month. 5 The table is organized in three parts vertically: combined direct and indirect components of trade-cost levels and their policy-induced changes are at the top; the two panels underneath pertain to the direct (own-cost) and indirect components only. The numbers in the same cell of the two panels at the bottom sum up to the corresponding one at the top. The numbers suggest that the biggest trade cost increases happened for upper-middleincome countries (through partner export or own import policies) for the imports of ventilators and for high-income countries for the ones of garments. The biggest trade cost reductions occurred for vaccines in high-income countries (consistent with the findings in Fig. 1). The indirect effects tend to soften the direct ones for tests, ventilators, and garments, whereas they reinforce the (negative) direct effects for vaccines. The indirect effects are particularly large and dominant for upper-middle-and high-income countries in the case of vaccines, whereas the direct effects tend to dominate in other cells. Interestingly, the indirect effects are of relatively large importance in comparison to the direct effects throughout the table pointing to the fact that the indirect costs of trade policy are key. 5 We list the associations of countries with country groups in the table footnote.
Overall, we identify trade-cost increases specifically for sensitive medical products during the Covid-19-pandemic protectionist wave. We document a substantial trade-cost surge between February 2020 and December 2021 for medical products. Particularly, trade costs for garments and ventilators peaked at the beginning of the pandemic. Indirect, input-output-transmitted costs played a crucial role with the policy effects.

Data availability
Data will be made available on request.