The Effect of Non-Tariff Measures on Trade in Medical Products During the COVID-19 Pandemic

Abstract

Countries have responded to the COVID-19 pandemic with different combinations of export controls and import liberalization measures, using non-tariff measures (NTMs). This study aims to understand the effects of NTMs on trade in medical products during the Covid-19 pandemic by obtaining their ad valorem equivalents (AVEs), comparing AVEs for pre-pandemic and pandemic years, and comparing AVEs with tariffs. We consider 59 key Covid-19 medical products listed by the World Customs Organization at the HS six-digit level for 15 countries selected based on their trade volume in Covid-19 medical supplies. We use a panel data set of 15 countries over the period 2002 to 2019 to estimate import demand elasticities for 59 commodities using 3 alternative methods: fixed effects, fixed effects instrumental variables, and system GMM. After estimating import demand elasticities, we estimate two sets of gravity-type import models for each commodity, one for the pre-pandemic year of 2019 and one for the pandemic year of 2020. We use the coefficients to obtain ad valorem equivalents of NTMs for each commodity and country in 2019 and 2020. We then compare the AVEs for the 2 years and with tariffs. We find that NTM AVEs increased during the pandemic, surpassing average tariffs in most countries and across most medical products. Our results show that NTMs were more heavily used during the pandemic and were more effective than tariffs on medical commodities during this period.

JEL Classification: F01, F13, F14, F19

Introduction

NTMs are a diverse set of practices, including import quotas, export restraints, licensing, domestic content requirements, government procurement policies, subsidies, technical regulations/standards, and sanitary/health measures, etc., formally classified as in UNCTAD (2024). NTMs have become the dominant instruments shaping international trade since tariffs declined globally (Ederington & Ruta, 2016; UNCTAD, 2017). The COVID-19 pandemic accelerated this trend as countries scrambled to secure access to essential goods. During the COVID-19 pandemic, technical measures in the form of Technical Barriers to Trade or TBTs and Sanitary and Phytosanitary Standards or SPSs, as well as anti-dumping and countervailing measures, were heavily implemented on essential agri-food and medical goods to secure the availability of critical supplies in the domestic market (S. J. Evenett, 2020). The pandemic caused unprecedented demand surges and supply chain disruptions for medical goods, revealing vulnerabilities in global medical supply chains (OECD, 2024). In this context, medical and personal protective products accounted for 73% of all NTMs imposed during the COVID-19 pandemic (UNCTAD, 2021). In fact, for 31% of all export controls on medical equipment and consumables implemented in 2020, there is no evidence to suggest they have been removed (S. J. Evenett, 2024).

Given the power of NTMs to shape trade patterns for medical products during the pandemic, it is essential to quantify their impact. However, the methodological challenges posed by the heterogeneous structure of NTMs, endogeneity of trade policy, and aggregation problems (Goldberg & Pavcnik, 2016) make it difficult to find corresponding ad valorem equivalents. In this paper, we aim to understand the impact of non-tariff measures on medical products listed in the World Customs Organization classification reference – Covid-19 medical supplies, Ed. 2, as updated on 9 April 2020. To this end, we proceed with the cumulative methodology initiated by Kohli (1991) and Harrigan (1997), then developed by Kee et al. (2008). We compile the most disaggregated data on medical products under the six-digit Harmonized System (HS) for 2002 to 2019, focusing on 15 countries selected based on their trade volumes of Covid-19 medical supplies. First, import demand elasticities are estimated; ad-valorem equivalents (AVEs) are calculated following Looi Kee et al. (2009) and then compared with tariffs.

This import demand (quantity) based approach to measuring NTMs is technically more involved and data-intensive than frequency or coverage type methods, price or unit value-based methods, quantity impact methods, or residual methods, but it is better grounded in economic theory. The originality of our study lies in bringing together several threads of literature: namely, the estimation of import elasticities at the commodity-country level, the estimation of quantity-based AVEs to facilitate comparison with tariffs, and the widespread use of NTMs on medical goods during the Covid-19 pandemic. In this study, we estimate the relative price elasticity of imports for 59 HS six-digit Covid-19 medical commodities using 3 econometric methods to improve robustness, rather than relying on existing literature estimates. We calculate AVEs for each commodity, country, and year (2019, pre-pandemic, and 2020, pandemic) using our own elasticity estimates. Then compare them to the respective tariffs. These efforts set our study apart from the bulk of the existing literature, which discusses the price and volume effects of NTMs in the context of all goods, with medical goods mentioned only in passing. The next section presents a brief overview of the literature, followed by our methodology, data, and empirical results. Conclusions follow.

Literature Review

The prevalence of non-tariff measures (NTMs) in international trade has been steadily increasing, to the extent that over 95% of globally traded products are now subject to at least one such measure (UNCTAD & World Bank, 2018). NTMs are a double-edged sword: they can simultaneously protect domestic supply while reducing welfare (Dhingra et al., 2023; Ederington & Ruta, 2016; Fell & Duver, 2024). Shi et al. (2025) analyzed the multidimensional impacts of NTMs on international trade by converting them into bilateral ad valorem equivalents. Utilizing an extensive country-by-product dataset, the study accounted for heterogeneity across sectors, countries, and NTM types. The authors distinguished between intensive and extensive margins of trade and found that NTMs exert statistically significant adverse effects on both margins. Sectoral and income-level differences among trading partners were shown to play a critical role in shaping these outcomes. A comparable study by Cadot and Gourdon (2016) employed the price-gap methodology to quantify the impact of NTMs on import prices, with effects assessed across sectors. Building upon the foundational price-gap methodology used in earlier analyses, a subsequent study by Cadot et al. (2018) integrated the impact of NTMs on import prices and trade flows, offering a more comprehensive assessment of their ad valorem equivalents and the broader implications for international trade policy.

NTMs pose challenges to international competitiveness because they can have an exclusionary effect, especially for developing or underdeveloped countries, and can harm sustainable development due to the inadequate institutional capacity of such countries (Taglioni & Kee, 2025). Moreover, as countries face structural or regulatory challenges, they tend to redirect their commercial pathways, thereby fragmenting global trade flows and ultimately contributing to the polarization of international trade (Chakraborty & Dey, 2024).

Although a significant part of the literature focuses on the economic effects of NTMs, they can also serve as strategic policy tools, for example, as Leonardi and Meschi (2026) discuss in their study examining the impact of non-tariff practices on the labor market in the United States. They find that an increasing implementation of protective trade measures targeting sectors with significant electoral influence can shape the trajectory of upcoming elections. Moreover, the presence of non-tariff measures in sectors critical to national security also underlines their strategic role (Wu, 2025).

During the COVID-19 pandemic, non-tariff measures were frequently applied to critical goods to address both immediate trade disruptions and long-term health policy implications. The early surge in restrictive NTMs (export bans, licensing) was followed by partial liberalization as governments sought to restore supply flows (APEC, 2021; Bown, 2021; S. Evenett et al., 2022). Empirical studies estimate that these NTMs raised trade costs and reduced trade volumes (Ahn & Steinbach, 2022, 2026; Dolabella, 2020). Effects varied by measure, product, and income group. Export restrictions constrained availability; generated global welfare losses and regressive outcomes, especially for small and low-income importers (Grassia et al., 2022; Lee & Parabhakar, 2021), while facilitative import measures (duty exemptions, fast-track procedures) mitigated shortages (APEC, 2021; Petrović, 2022). Compliance costs, testing requirements, and divergent regulations increased trade friction (Nicita & Koloskova, 2025; Gourdon, 2014; OECD/WTO, 2019), and inconsistent product standards delayed approvals (Baldwin & Evenett, 2020). Poor information sharing and delayed notification further amplified uncertainty (Miteva-Kacarski et al., 2021). Transparency and predictability proved key to minimizing harm (Lee & Parabhakar, 2021). OECD (2024) stresses the critical importance of data sharing and transparency in this context, emphasizing their role in strengthening future resilience mechanisms.

The increasing, multifaceted impact of NTMs on world trade, in general, and on trade in medical supplies in particular, has generated a growing body of literature with significant policy implications for the future. In this context, quantifying the pattern and magnitude of the impact is critical. Several empirical studies have attempted to quantify the price and trade-flow effects of NTMs. Still, none specifically focus on the import-demand-based ad valorem equivalents of NTMs on medical goods during the COVID-19 pandemic. Our study aims to fill this gap.

Methodology

Our methodology closely follows that of Kee et al. (2008) and Ghodsi et al. (2016). They use a semi-flexible trans log GDP function approach to formally derive import demands and their elasticities, which are estimated with data on import values, prices, GDPs, and endowments. The estimated import demand elasticities are defined as the percentage change in the quantity of an imported good when its price increases by 1%, holding constant the prices of all other goods, productivity, and the economy’s endowments.

Kee et al. (2008) assume that $G^t(p^t,v^t)$ follows a flexible translog functional form with respect to good prices, p, and factor endowments, v, with n and k index goods m and l index factors:

$$\begin{array}{ll}\multicolumn{1}{c}{ln{G}^t\left(p^t,v^t\right)=}& \multicolumn{1}{c}{a_{00}^t+\sum _{n=1}^N{a}_{0n}^{t}ln{p}_n^t+\frac{1}{2}\sum _{n=1}^N\sum _{k=1}^N{a}_{nk}^{t}ln{p}_n^{t}ln{p}_k^t}\\ \multicolumn{1}{c}{}& \multicolumn{1}{c}{+\frac{1}{2}\sum _{m=1}^M\sum _{l=1}^M{b}_{ml}^{t}ln{v}_m^{t}ln{v}_l^t+\sum _{m=1}^M{b}_{0m}^{t}ln{v}_m^t}\\ \multicolumn{1}{c}{}& \multicolumn{1}{c}{+\sum _{n=1}^N\sum _{m=1}^M{c}_{nm}^{t}ln{p}_n^{t}ln{v}_m^t}\end{array}$$

(1)

The derivative of with respect to the $\ln G^t(p^t,v^t)$ gives the lnp_n^t equilibrium share of good n in GDP at period t:

$$s_n^t(p^t,v^t)\equiv \frac{p_n^t{q}_n^t(p^t,v^t)}{G^t(p^t,v^t)}$$

(2)

Which, after imposing restrictions on the functional form of the trans log GDP function to ensure that it is homogeneous of degree one with respect to prices and factor endowments and satisfies the symmetry property, results in:

$$\begin{array}{cc}\multicolumn{1}{c}{s_n^t(p^t,v^t)}& \multicolumn{1}{c}{=a_{0n}^t+\sum _{k=1}^N{a}_{\mathrm{nk}}^t{\mathrm{lnp}}_k^t+\sum _{m=1}^M{c}_{\mathrm{nm}}^t{\mathrm{lnv}}_m^t}\\ \multicolumn{1}{c}{}& =a_{0n}^t+a_{\mathrm{nn}}^t{\mathrm{lnp}}_n^t+\sum _{k\ne n}^N{a}_{\mathrm{nk}}^t{\mathrm{lnp}}_k^t+\sum _{m=1}^M{c}_{\mathrm{nm}}^t{\mathrm{lnv}}_m^t\hfill \end{array}$$

(3)

Where, $s_n^t$ is the share of good n in GDP at time t. Import demand elasticity of good n obtained from the GDP maximizing demand function is as follows:

$${\epsilon }_{\mathrm{nn}}^t\equiv \frac{\partial q_n^t(p^t,v^t)}{\partial p_n^t}\frac{p_n^t}{q_n^t}=\frac{a_{\mathrm{nn}}^t}{s_n^t}+s_n^t-1$$

(4)

We can calculate the import demand elasticities once $a_{\mathrm{nn}}^t$ is estimated based on equation 3. Note that the size of the import elasticity, ${\epsilon }_{\mathrm{nn}}^t$ , depends on the sign of $a_{\mathrm{nn}}^t$ , which captures the changes in the share of good n in GDP when the price of good n increases by 1%:

$${\epsilon }_{\mathrm{nn}}^t\{\begin{array}{c}\multicolumn{1}{c}{<-1\mathrm{if}{a}_{\mathrm{nn}}^t>0}\\ \multicolumn{1}{c}{=-1\mathrm{if}{a}_{\mathrm{nn}}^t=0}\\ \multicolumn{1}{c}{>-1\mathrm{if}{a}_{\mathrm{nn}}^t<0}\end{array}$$

(5)

Kee et al. (2008) obtain an estimable share equation by parameterizing a fully flexible trans log function as a semi-flexible trans log function following Diewert and Wales (1988) and by restricting all trans log parameters to be time-invariant. The resulting share equation is

$$s_n^t(p^t,v^t)=a_{0n}+a_{\mathrm{nn}}\ln \frac{p_n^t}{p_{-n}^t}+\sum _{m\ne l,m=1}^M{c}_{\mathrm{nm}}\ln \frac{v_m^t}{v_l^t}+{\kappa }_n^t$$

(6)

where the non-n goods price is calculated using the observed Tornqvist price index ${\mathrm{lnp}}_{-n}^t$ of all non-n goods using the GDP deflator $p^t$ , as in equation 7.

$${\mathrm{lnp}}_{-n}^t=(\ln p^t-{\overline{s}}_n^t{\mathrm{lnp}}_n^t)/(1-{\overline{s}}_n^t)$$

(7)

with ${\overline{s}}_n^t=0.5(s_n^t+s_n^{t-1})$ .

Equation 6 can be estimated using importer-specific product shares in GDP, the GDP deflator, product unit values, and information on factor endowments. Pooling data across countries and years for each good, while employing country and year fixed effects, the final share equation estimated by Kee et al. (2008) for each good takes the following form:

$$\begin{array}{cc}\multicolumn{1}{c}{s_n^t(p_{\mathrm{nc}}^t,p_{-\mathrm{nc}}^t,v_c^t)=}& \multicolumn{1}{c}{a_{0n}+a_{\mathrm{nc}}+a_n^t+a_{\mathrm{nn}}\ln \frac{p_{\mathrm{nc}}^t}{p_{-\mathrm{nc}}^t}}\\ \multicolumn{1}{c}{}& \multicolumn{1}{c}{+\sum _{m\ne l,m=1}^M{c}_{\mathrm{nm}}\ln \frac{v_{\mathrm{mc}}^t}{v_{\mathrm{lc}}^t}+u_{\mathrm{nc}}^t}\end{array}$$

(8)

Once ${\hat{a}}_{\mathrm{nn}}$ is obtained from equation 8, the consistent estimate of import demand elasticity of good n in country i, ${\hat{\epsilon }}_{\mathrm{ni}}$ , is constructed using the average import shares of each country c, ${\overline{s}}_{\mathrm{nc}}$ , and ${\hat{a}}_{\mathrm{nn}}$ as:

$${\hat{\epsilon }}_{\mathrm{nc}}=\frac{{\hat{a}}_{\mathrm{nn}}}{{\overline{s}}_{\mathrm{nc}}}+{\overline{s}}_{\mathrm{nc}}-1$$

(9)

Kee et al. (2008) mention three econometric considerations that require additional steps. First are the endogeneity and measurement errors in prices. The second is selection bias due to zero imports, and the third is the partial adjustment of imports. In this study, the first is addressed using an instrumental variables estimator. The second consideration is not relevant to the present study because only 59 specific COVID-19 products and 15 countries are focused on, and there are no zero import values (apart from missing observations) in the sample over the 2002 to 2019 period. The third consideration was addressed using a system GMM estimator and robust standard errors.

Instrumental variables estimation was performed using the instruments suggested in Kee et al. (2008): the simple and inverse-distance-weighted averages of the unit values of the rest of the world, as well as the trade-weighted average distance of country i to all the exporting countries of good n. So, for $z_1$ and $z_2$ we used the following formulation:

$$\begin{array}{cc}\multicolumn{1}{c}{z_{\mathrm{ni}}^t=}& \multicolumn{1}{c}{\ln \frac{{\overline{p}}_{n-c}^t}{{\overline{p}}_{-n-c}^t}=\ln \left(\sum _{k\ne c}{w}_k{p}_{\mathrm{nk}}^t\right)}\\ \multicolumn{1}{c}{}& \multicolumn{1}{c}{-(\frac{\ln \sum _{k\ne c}{w}_k{\overline{p}}_k^t-(\sum _{k\ne c}{w}_k{\overline{S}}_{\mathrm{nk}}^t)\ln (\sum _{k\ne c}{w}_k{p}_{\mathrm{nk}}^t)}{1-\sum _{k\ne c}{w}_k{\overline{S}}_{\mathrm{nk}}^t}}\end{array}$$

(10)

$$w_k=\{\begin{array}{cc}\multicolumn{1}{c}{\frac{1}{C-1},}& \multicolumn{1}{c}{\mathrm{for}\mathrm{the}\mathrm{simple}\mathrm{average}}\\ \multicolumn{1}{c}{\frac{1}{\mathrm{distance}\mathrm{in}\mathrm{km}\prime \mathrm{s}\mathrm{btw}\mathrm{country}\mathrm{c}\mathrm{and}\mathrm{k}},}& \multicolumn{1}{c}{\mathrm{for}\mathrm{the}\mathrm{weighted}\mathrm{average}}\end{array}$$

Where C refers to the total number of countries. The third instrument, $z_3$ , the trade-weighted distance, is calculated as follows:

$$z_{\mathrm{nc}}^t=\sum _k{w_k}^t\mathrm{distanc}{e}_{\mathrm{kc}}$$

(11)

Where k denotes all exporting countries, the distance between country k and c is measured in kilometers, ${w_k}^t$ is the share of k in world exports of good n in year t.

After estimating import demand elasticities, we use them to obtain ad valorem equivalents of NTMs in effect across our 15 countries for the 59 Covid-19 goods in 2019 and 2020. Here, we follow the methodology outlined in Looi Kee et al. (2009). Derived from the n-good equilibrium model with log-linear utilities and log-linear constant returns technologies as in Leamer (1988, 1990), they use the following specification:

$$\begin{array}{cc}\multicolumn{1}{c}{\ln m_{n,c}=}& \multicolumn{1}{c}{{\alpha }_n+\sum _k{\alpha }_{n,k}{C}_c^k+{\beta }_{n,c}^{\mathrm{ntm}}\mathrm{nt}{m}_{n,c}}\\ \multicolumn{1}{c}{}& \multicolumn{1}{c}{+{\epsilon }_{n,c}\ln (1+t_{n,c})+{\mu }_{n,c}}\end{array}$$

(12)

where $m_{n,c}$ is the import value of good n in country c deflated by standard unit values so that $m_{n,c}$ represents imported quantities, ${\alpha }_n$ are tariff line dummies that capture any good-specific effect; $C_c^k$ are k variables representing country characteristics. Relative factor endowments (agricultural land over GDP, labor over GDP, and capital over GDP) are used to represent country characteristics. ${\alpha }_{n,k}$ are parameters in front of the variables that capture country characteristics; $\mathrm{nt}{m}_{n,c}$ is a dummy variable indicating the presence of a NTM on good n in country c. ${\beta }_{n,c}^{\mathrm{ntm}}$ is the parameter that captures the impact that a NTM imposed on good n in country c has on imports of good n in country c. $t_{n,c}$ is the ad-valorem tariff on good n in country c; ${\epsilon }_{n,c}$ is the import demand elasticity and ${\mu }_{n,c}$ is an i.i.d. error term.

This model allows for both tariffs and NTMs to deter trade with the importing country, with effects that vary by country and by good. Since we have estimated the import elasticities, we can substitute our estimates and the tariffs into the equation and move the fourth term on the right-hand side to the left.

$$\begin{array}{c}\multicolumn{1}{c}{\ln m_{n,c}-{\epsilon }_{n,c}\ln (1+t_{n,c})}\\ \multicolumn{1}{c}{={\alpha }_n+\sum _k{\alpha }_{n,k}{C}_c^k+{\beta }_{n,c}^{\mathrm{ntm}}\mathrm{nt}{m}_{n,c}+{\mu }_{n,c}}\end{array}$$

(13)

As in Looi Kee et al. (2009), we allow the β parameters to vary across tariff lines and countries such that they have a product and a country-specific impact, where each country’s factor endowments will capture the latter as in Leamer’s (1988, 1990) comparative-advantage approach:

$${\beta }_{n,c}^{\mathrm{ntm}}={\beta }_n^{\mathrm{ntm}}+\sum _k{\beta }_{n,k}^{\mathrm{ntm}}{C}_c^k$$

(14)

Where all ${\beta }_{n,k}^{\mathrm{ntm}}$ s are product-specific parameters. The country variation comes from the interaction with the comparative advantage variables. Substituting equations 13 and 14 we get:

$$\begin{array}{c}\multicolumn{1}{c}{\ln m_{n,c}-{\epsilon }_{n,c}\ln (1+t_{n,c})}\\ \multicolumn{1}{c}{={\alpha }_n+\sum _k{\alpha }_{n,k}{C}_c^k+({\beta }_n^{\mathrm{ntm}}+\sum _k{\beta }_{n,k}^{\mathrm{ntm}}{C}_c^k)\mathrm{nt}{m}_{n,c}+{\mu }_{n,c}}\end{array}$$

(15)

Once the product-specific and the country-specific components of the ${\beta }_{n,c}^{\mathrm{ntm}}$ are estimated, equation 14 is motivated.

The ad-valorem equivalent of an NTM is the change in domestic price as a response to NTM, that is:

$$\mathrm{ave}=\frac{\partial p_{n,c}^d}{\partial \mathrm{nt}{m}_{n,c}}$$

(16)

To obtain the AVE, we differentiate equation 12 with respect to NTM to get:

$$\frac{\partial \ln m_{n,c}}{\partial \mathrm{nt}{m}_{n,c}}=\frac{\partial \ln m_{n,c}}{\partial p_{n,c}^d}\frac{\partial p_{n,c}^d}{\partial \mathrm{nt}{m}_{n,c}}={\epsilon }_{n,c}{\mathrm{ave}}_{n,c}^{\mathrm{ntm}}$$

(17)

Thus, we get:

$${\mathrm{ave}}_{n,c}^{\mathrm{ntm}}=\frac{1}{{\epsilon }_{n,c}}\frac{\partial \ln m_{n,c}}{\partial \mathrm{nt}{m}_{n,c}}=\frac{{\beta }_{n,c}^{\mathrm{ntm}}}{{\epsilon }_{n,c}}$$

(18)

AVEs are calculated in each country at the HS-six-digit product level. Here, since we could not constrain the estimated ${\beta }_{n,c}^{\mathrm{ntm}}$ to be non-positive, the estimated AVEs are not constrained to be non-negative.

Data and Estimation

We use a panel of 15 countries and 18 years from 2002 to 2019 for the 59 medical products listed at the HS-six-digit level in World Customs Organization classification reference – Covid-19 medical supplies Ed. 2, as updated on 9 April 2020. We obtain estimates of 59 import demand elasticities using equations 8–11 and use these elasticities to estimate AVEs of NTMs for each of the 59 goods in each of the 15 countries in 2019 and 2020 using equations 15 and 18.

Countries are selected based on their trade volume of Covid-19 medical supplies and include Belgium, Canada, China, France, Germany, India, Italy, Japan, the Netherlands, Russia, Spain, Switzerland, Turkey, the United Kingdom, and the USA. Missing observations in standard unit values are interpolated using the average trends in the data. Table 1 presents the definitions and sources of the data we used.

Table 1.

Data Definitions and Sources.

Variable	Definition	Source
For equation 19
s	share	UNCOMTRADE
lnrelp	natural logarithm of the price ratio SUV over the price index defined by equation 7	UNCOMTRADE
lnvkn	natural logarithm of the capital-labor ratio	WDI and PWT 11
lnvnl	natural logarithm of the labor-land ratio	WDI
lnvkl	natural logarithm of the capital-land ratio	WDI and PWT 11
lnm	natural logarithm of import quantity	UNCOMTRADE
z1	instrumental variable defined by equation 10	CEPII Geo-Dist and WDI
z2	instrumental variable defined by equation 10	CEPII Geo-Dist and WDI
z3	instrumental variable defined by equation 11	CEPII Geo-Dist and WDI
For equation 20
lnmm	natural logarithm of import quantity modified as in equation 15	UNCOMTRADE
lnvny	natural logarithm of the labor-RGDP ratio	WDI
lnvly	natural logarithm of the land-RGDP ratio	WDI
lnvky	natural logarithm of the capital-RGDP ratio	WDI and PWT 11
lnx	natural logarithm of export quantity	UNCOMTRADE
dm18	change in import quantity in 2018	UNCOMTRADE
dm19	change in import quantity in 2019	UNCOMTRADE
ntm	dummy = 1 if NTM is imposed, =0 otherwise	WTO I-TIP

The empirical form equation 8 is:

$$\begin{array}{c}\multicolumn{1}{c}{s_{\mathrm{nc}}^t=a_{n0}+a_{\mathrm{nc}}+a_n^t+a_{\mathrm{nn}}{\mathrm{lnrelp}}_{\mathrm{nc}}^t+c_{n1}{\mathrm{lnvkn}}_{\mathrm{nc}}^t}\\ \multicolumn{1}{c}{+c_{n2}{\mathrm{lnvkl}}_{\mathrm{nc}}^t+c_{n3}{\mathrm{lnvnl}}_{\mathrm{nc}}^t+u_{\mathrm{nc}}^t,}\\ \multicolumn{1}{c}{n=1,\dots .,59;c=1,\dots 15;t=2002,\dots .,2019}\end{array}$$

(19)

Here, $s_{\mathrm{nc}}^t$ is the share of good n in the GDP of country c at time t; ${\mathrm{lnrelp}}_{\mathrm{nc}}^t$ is the natural logarithm of the relative price of good n with respect to other goods as calculated by equation 7 in country c at time t; ${\mathrm{lnvkn}}_{\mathrm{nc}}^t$ is the natural logarithm of the capital-labor ratio in country c at time t, ${\mathrm{lnvkl}}_{\mathrm{nc}}^t$ is the natural logarithm of the capital-land ratio in country c at time t, ${\mathrm{lnvnl}}_{\mathrm{nc}}^t$ is the natural logarithm of the labor-land ratio in country c at time t.

We estimated the 59 models 3 times using different techniques to capture the best ${\hat{a}}_{\mathrm{nn}}$ s to motivate equation 9. First, a simple two-way fixed-effects (FE) model with country and year fixed effects was estimated. Then, to control for potential endogeneity of relative prices, the two-way fixed effects instrumental variable (FEIV) model was estimated using the instruments $z_1$ , $z_2$ and $z_3$ . Lastly, to control for serial correlation and partial adjustment in import shares, an Arellano and Bover (1995) system GMM estimator with the lagged dependent variable as an additional regressor was used. Country-wise clustering and robust standard errors were used in all estimations.

We selected the best estimate for each commodity from the three models based on their diagnostics. We chose the GMM estimates only if the model passed the diagnostic tests such that the AR (1) p-value is less than .10, the AR (2) p-value is larger than .10, and the Hansen J test p-value is between .10 and .90. We chose the FEIV estimates based on instrument relevance and exogeneity diagnostics such that the Kleibergen–Paap LM (KP LM) statistic is greater than 10 and the Anderson–Rubin (AR) test p-value is less than .05. Whenever these conditions were not met, we used the FE estimates. In the end, the FEIV estimates were selected for 37 commodities, the FE estimates for 14, and the system GMM estimates for only 8. The resulting set of ${\hat{a}}_{\mathrm{nn}}$ s was used to calculate the import elasticities according to equation 9.

The kernel density plots of the estimated import elasticities by country are given in Figure 1, and a complete list of estimated values is provided in Appendix Table A1. The import demand for this set of goods is relatively inelastic on average, as expected, given that these are essential medical goods. The distributions in Figure 1 are trimmed to exclude extreme values. Overall, the distributions are consistent. The elasticities have the expected sign and magnitude for the most part. However, we observe anomalous values for China, the Russian Federation, and India, which may be due to data quality issues and/or atypical trade regimes.

Figure 1.

Kernel density of import elasticity (trimmed 1%–99%) by countries.

The next step in the empirical strategy is to estimate equation 15, whose empirical form is:

$$\begin{array}{c}\multicolumn{1}{c}{\mathrm{lnm}{m}_{\mathrm{nc}}={\alpha }_n+{\alpha }_{n1}\mathrm{lnvl}{y}_c+{\alpha }_{n2}\mathrm{lnvn}{y}_c+{\alpha }_{n3}\mathrm{lnvk}{y}_c}\\ \multicolumn{1}{c}{+{\beta }_n^{\mathrm{ntm}}\mathrm{nt}{m}_{n,c}+{\beta }_{n,1}^{\mathrm{ntm}}\mathrm{lnvl}{y}_c\mathrm{nt}{m}_{n,c}}\\ +{\beta }_{n,2}^{\mathrm{ntm}}\mathrm{lnvn}{y}_c\mathrm{nt}{m}_{n,c}+{\beta }_{n,3}^{\mathrm{ntm}}\mathrm{lnvk}{y}_c\mathrm{nt}{m}_{n,c}+{\mu }_{n,c}\hfill \\ \multicolumn{1}{c}{n=1,\dots .,59;c=1,\dots 15;}\end{array}$$

(20)

We estimate equation 20 for 2019 and 2020 to obtain the coefficient estimates used in equation 14, which, in turn, motivate equation 18 for calculating the AVEs. While a two-way FE estimator with country and commodity effects appears to be the obvious choice, we opted to estimate 59 separate models due to the complexities posed by our sparse ntm dummy, collinearity issues, and limited degrees of freedom in the context of fixed effects and instruments. To address the potential endogeneity of NTMs, FEIV estimation is conducted using the instruments of the natural logarithms of exports and the past change in imports, as proposed by Looi Kee et al. (2009). All models passed the standard diagnostic tests. AVEs estimated using the relevant coefficient estimates from these regressions and equation 18 for 2019 and 2020 are presented in Table A2 and Table A3 in the Appendix. Theoretically, AVEs are expected to be non-negative. In our dataset, 170 of the 600 AVEs we estimated were non-negative in 2019, and 109 of the 496 were non-negative in 2020.

Comparison of AVEs of 2019 and 2020

Figure 2 illustrates the average non-negative AVE for each country in 2019 and 2020, with countries ordered vertically by their 2019 average. The figure also includes the number of commodities for which we can estimate non-negative AVEs for each year. We see that, across 10 of the 15 countries, the average AVE increased from 2019 to 2020, indicating a greater restrictive effect of NTMs on average. The remaining countries exhibit minor reductions in average AVE, suggesting facilitative effects of NTMs. These differences remain visible even after winsorizing AVE values at the 5 to 95 percentiles, indicating that the observed shifts represent broad policy changes rather than outliers. Variation in sample sizes suggests that some changes should be interpreted cautiously, particularly for countries with limited product coverage. The magnitude of the impact is particularly pronounced in large emerging markets such as Russia, China, and India, in which external shocks are likely to generate structural fragilities. In the case of major European economies, including Germany, France, and the United Kingdom, the extent of change has remained limited. This actually highlights cross-country heterogeneity in country responses to the COVID-19 crisis, as well as differences in country characteristics such as trade dependencies, product mixes, and legislative structures.

Figure 2.

Average AVEs (winsorized 5%–95%) by country.

Figure 3 compares average AVE estimates between 2019 and 2020 for each commodity, with the commodities ordered vertically by their 2019 average value. These are the 19 commodities for which we obtained non-negative AVE estimates for both years. For most commodities, 2020 AVEs lie to the left of 2019 AVEs, indicating greater restrictiveness, including medical mobility devices (HS 87139099), medical diagnostic and test instruments (HS 902780), and liquid soap (HS 340120), goggles (900490) and syringes (901831). Only a handful of commodities exhibit stability over time, and none show a decline in restrictiveness. Although sample sizes are small for some items, the consistent rightward shifts suggest a broad tightening of NTMs on critical medical and chemical goods during the pandemic period.

Figure 3.

Average AVEs (winsorized 5%–95%) by commodity.

Comparison of AVEs and Tariffs

Figure 4 presents a set of intuitive scatter graphs to compare AVEs and tariffs. Each point gives the AVE and tariff values for a country–commodity pair. The solid curve shows a nonparametric Locally Estimated Scatterplot Smoothing (LOESS) fit that captures the average relationship in the data, without forcing it to be linear. At the same time, the dashed 45° line serves as a reference showing where tariffs and AVEs would be proportional. The number of observations above the 45° line increases from 22 to 26 in 2020. The LOESS curves show a noticeable upward shift in 2020, reflecting the increased AVEs of NTMs relative to tariffs. The widening AVE dispersion in the 2020 plot is consistent with the surge in regulatory controls. Many low-tariff essential goods saw significant increases in NTMs due to COVID-19 (export bans, quality controls, licensing requirements). Even zero-tariff products (e.g., medical goods) had high NTMs during the pandemic. Some higher-tariff sectors also saw increased NTM protection. Overall, NTMs had a larger restrictive impact than tariffs during the pandemic.

Figure 4.

Comparison of AVEs with tariffs for 2019 and 2020.

Conclusion

The literature indicates a growing body of research on the economic effects of NTMs. Studies show that NTMs can protect domestic supply but also adversely affect trade volumes and prices. In this context, quantifying the pattern and magnitude of the impact is critical. Several empirical studies have attempted to quantify the price and trade-flow effects of NTMs. Still, there is a lack of focused research on the import-demand-based AVEs of NTMs on medical goods during the COVID-19 pandemic.

Our study aims to fill this gap by estimating import demand elasticities and ad valorem equivalents (AVEs) of NTMs on the imports of medical products listed in the World Customs Organization (2020), comparing AVEs for the pre-pandemic and pandemic years, and comparing AVEs to tariffs. For this purpose, we estimate ad valorem equivalents of non-tariff measures imposed by 15 countries on 59 medical products at the most disaggregated HS level. In doing so, we also derive import demand elasticities for each commodity and country. Then we compare the 2019 and 2020 mean AVEs of NTMs on medical products. We also compare these with tariffs on the same goods in the same time frame. We find that NTM AVEs increased during the pandemic, surpassing average tariffs in most countries and across most medical products. NTMs were more restrictive and more frequently used than tariffs in 2020, with effects likely lingering after the pandemic. Patterns varied by country and product; some countries saw minor decreases in AVEs or facilitative NTM use, too. We see that, across 10 of the 15 countries, the average AVE increased from 2019 to 2020, indicating greater restrictiveness. The remaining countries exhibit minor reductions in average AVE, suggesting facilitative effects of NTMs. For most commodities, 2020 AVEs lie to the left of 2019 AVEs, indicating greater restrictiveness, including medical mobility devices (HS 87139099), medical diagnostic and test instruments (HS 902780), and liquid soap (HS 340120), goggles (900490), and syringes (901831). Only a handful of commodities exhibit stability over time, and none show a decline in restrictiveness. Comparing 2019 and 2020 in relation to AVEs and tariffs, we see increased AVEs of NTMs relative to tariffs. Overall, these findings are consistent with the literature and further support the claim that NTMs were more heavily used during the pandemic and more effective than tariffs on these commodities.

The findings contribute to trade theory in three ways. First, they show that NTMs function as flexible crisis instruments when tariff use is constrained by international commitments. Second, they show that trade policy can be endogenous because the relative effectiveness of instruments shifts in response to exogenous shocks. Third, the observed heterogeneity across countries and products highlights the role of domestic political economy factors—such as supply chain resilience, production capacity, healthcare systems, and institutional quality—in shaping trade policy outcomes.

Several limitations apply. First, the sample covers only 15 major traders in COVID-19 medical goods and the 59 COVID-19 medical goods; this limits generalizability, particularly to developing countries with different economic structures and institutional capacity. Second, import demand elasticities are estimated using pre-pandemic data (2002–2019), and structural breaks during the pandemic may undermine the precision of elasticity estimates and consequently the AVE estimates. In a similar vein, the gravity framework relies on functional-form and exogeneity assumptions that may not hold fully during a global crisis. Third, the AVEs of NTMs capture price-equivalent effects but not non-price barriers such as administrative delays, which may be important during the pandemic. Fourth, the analysis consists of a static comparison of 2019 and 2020 so it does not capture dynamic policy adjustments or lag effects. Finally, NTMs are not disaggregated by type or effect, potentially masking heterogeneous impacts of different classes of measures.

Future research should expand country coverage, use higher-frequency data to examine dynamic policy adjustments, and disaggregate NTMs (e.g., SPS, TBT, restrictive, facilitating, etc.). Welfare analysis could clarify the distributional effects across consumers and producers. Further work on the political-economy determinants of NTM adoption and their interactions with broader crisis policies would deepen understanding of trade policy choices. Research on reforming international rules to secure the international supply chain for medical products remains especially relevant.

Maintaining trade flows of medical supplies during the COVID-19 pandemic has been critical. The trade policy response should ensure that trade channels for these products are open, existing hurdles are removed, and unnecessary costs and delays are prevented, as the World Bank Trade and COVID-19 Guidance Note (April 2020) advised. Our results suggest that this advice has not necessarily been followed, and NTMs created additional cost-increasing effects. The need to remove unnecessary NTMs to avoid further disruptions and costs in future crises is a key issue that trade policy should address in the aftermath of the pandemic.