COVID-19, public procurement regimes and trade policy.

This paper analyses a prominent dimension of the initial policy response to the COVID-19 pandemic observed in many countries: the imposition of export restrictions and actions to facilitate imports. Using weekly data on the use of trade policy instruments during the first seven months of the COVID-19 pandemic (January-July, 2020), we assess the relationship between the use of trade policy instruments and attributes of pre-crisis public procurement regulation. Controlling for country size, government effectiveness and economic factors, we find that use of export restrictions targeting medical products is strongly positively correlated with the total number of steps and average time required to complete procurement processes in the pre-crisis period. Membership of trade agreements encompassing public procurement disciplines is associated with actions to facilitate trade in medical products. These findings suggest future empirical assessments of the drivers of trade policy during the pandemic should consider public procurement systems.

INTRODUCTION

One element of the policy response to the COVID‐19 pandemic by governments was to greatly expand public procurement (PP) of critical medical supplies, notably personal protective equipment (masks, gloves, face‐shields, respirators), ventilators, laboratory equipment (kits, reagents, swabs, laboratory consumables) and medicines. In some countries, this procurement response included requisitioning of available stocks of such products and a ban on their export. Many countries made active use of trade policy instruments to enhance access to essential supplies, involving a mix of measures to facilitate imports (lowering taxes and import tariffs and creating ‘green channels’ at borders to speed through imports) and export controls (Evenett et al., 2021). WTO rules permit the use of trade restrictions in public emergencies, but require these to be temporary, lasting only for the duration of a crisis. The reason is that use of export controls can give rise to negative spillovers, including by constraining the ability of firms to ramp up production, leading to increased prices and impeding the ability of other countries to import supplies (Atkinson et al., 2020; Evenett, 2020; Fiorini et al., 2020; Gereffi, 2020).

In most countries, government procurement of goods and services is subject to regulations that seek to ensure ‘value for money’. This is reflected in requirements and processes that enhance transparency, assure due process and accountability, and prevent corrupt practices and/or collusion among bidders. A core feature of PP processes is to mimic the market by encouraging (requiring) competition through open calls for tender.1 International agreements that cover procurement practices—such as the Treaty on the Functioning of the European Union that applies to EU member states, the WTO Agreement on Government Procurement (GPA) and recent vintage preferential trade agreements (PTAs)—not only embody generally accepted good PP practices but also require that foreign firms be treated the same as national bidders. The main thrust of such agreements is to open procurement markets to foreign competition.

As do all trade agreements, national procurement regulations and international agreements that discipline PP practices include exceptions that allow governments to respond rapidly to emergencies in ways that may be inconsistent with the rules that apply in normal times. This might take the form of direct contracting for supplies from producers without going through the processes that normally would be used (Baxter & Casady, 2020). At the time of writing, it is not yet possible to investigate the extent and effectiveness of emergency procurement measures taken by different jurisdictions.2 Instead, we examine the relationship between pre‐crisis attributes of public procurement regimes as reflected in indicators compiled by the World Bank and the trade policy behaviour of countries in the first seven months of the COVID‐19 pandemic. Most trade policy activism was observed in the initial months of the pandemic, reflecting the feasibility of applying trade policy instruments very rapidly. Similarly, trade measures can also be removed rapidly—and in principle should be to abide by WTO rules. Global excess demand for protective equipment and COVID‐19 medical supplies was met with a massive supply response, attenuating the rationale for using trade policy instruments for an extended period.

We analyse the relationship between PP regimes and trade policy activism during the first seven months of the COVID‐19 pandemic using cross‐country information on attributes of pre‐crisis national public procurement regimes (World Bank, 2020), the coverage of PP in trade agreements (Shingal & Ereshchenko, 2020) and data on changes in trade policy for medical products implemented by governments during January–July 2020 (Evenett et al., 2021). A key feature of the trade policy data is that information is available on a weekly basis, permitting analysis of when trade liberalising and restrictive policy instruments were imposed and removed.

Our results suggest that after controlling for country size, government effectiveness, economic factors and the incidence of COVID‐19 cases, restrictions on exports of medical products and import liberalisation are positively correlated with pre‐crisis attributes of PP regimes. The total number of steps, in particular, and average time taken to complete procurement processes are strongly associated with reducing import barriers and imposition of export controls. Membership of trade agreements with public procurement disciplines—both PTAs and the WTO GPA—is associated with greater openness, reflected in actions to facilitate trade in medical products.

The rest of the paper is structured as follows. Section 2 briefly discusses related literature. Section 3 provides an overview of the use of trade policy between January and July 2020. Section 4 presents the empirical methodology. Section 5 discusses the data sources for the explanatory and control variables and provides some descriptive statistics for these variables. Section 6 discusses the estimation results. Section 7 concludes.

RELATED LITERATURE

We are not aware of studies that analyse the relationship between public procurement and trade policy responses during public health emergencies. The extant studies on public procurement and COVID‐19 include a focus on strategies procuring authorities can (should) use to rapidly ramp up purchases of essential products needed by public health authorities and care providers. Procurement of health care‐related products usually involves a prolonged multi‐stage process that takes time. In times of crisis, procuring agencies may need to dispense with normal practices to meet urgent needs for critical equipment and supplies (Sanchez‐Graells, 2020). This could include responding to firms that make unsolicited proposals (Baxter & Casady, 2020) and partnership‐based approaches with the private sector or other governments (Vecchi et al., 2020). In the case of the EU, the European Commission took several measures to facilitate procurement of essential supplies, including through a (voluntary) coordinated joint procurement mechanism (Beuter, 2020; European Commission, 2020). Some governments directly contracted with large producers, bypassing standard competitive procedures stipulated in public procurement regulations. While warranted, there are risks associated with diverging from standard PP practices, including higher cost procurement, greater vulnerability to fraud, and diminished accountability and transparency in contracting (Atkinson et al., 2020).

Trade can play a pivotal role in emergencies to ensure that much needed medical supplies get to where they are needed (Gereffi, 2020; OECD, 2020b). Decisions by some governments during the early stages of the COVID‐19 outbreak—amidst shortages of medical supplies—to impose export control measures and requisition domestic supplies of essential goods may work against the goal of expanding the supply of vital equipment to healthcare workers by increasing prices, uncertainty and market volatility (Fiorini et al., 2020). The associated disruption to crisis health planning makes net importers of medical products particularly vulnerable (Evenett et al., 2021).

A basic premise underlying the analysis in this paper is that public procurement regulation may influence incentives to use trade policy in responding to crises. The idea is that specific attributes of procurement regimes may facilitate or constrain the ability of agencies to rapidly procure needed supplies of medical products in an emergency. PP practices that are designed to control corruption, ensure accountability through due process, transparency, non‐discrimination and competitive bidding may constrain the ability to respond rapidly. Conversely, PP regimes that are efficient in the sense of allocating contracts more rapidly may be more conducive to addressing an emergency and attenuate incentives to resort to trade restrictions. ‘Buy national’ prescriptions reflecting industrial development objectives may be accompanied by restrictive import policies to support domestic production. In circumstances where domestic production capacity is too small to satisfy a crisis‐induced increase in demand, a pre‐crisis policy bias towards domestic sourcing may be associated with a temporary reduction of removal of import restrictions during a crisis. The higher the initial import barriers the greater the scope for liberalisation.

The objective of this paper is to examine the strength of such potential relationships empirically. In our analysis of PP regimes, we are limited by the availability of indicators that characterise salient attributes of PP regimes.3 One relevant feature of procurement systems on which comparable information is reported on a cross‐country basis is the average time taken to complete procurement processes. Countries where PP takes more time may be at a disadvantage in procuring supplies even if standard processes are not applied in a crisis. For example, insofar as a nation's PP ‘type’ is common knowledge, suppliers may prefer to sell products in short supply to buyers that can credibly offer rapid contracting and processing of payments. Another attribute of prevailing PP regimes is membership of the WTO GPA and PTAs that encompass government procurement. Members of PP‐liberalising trade agreements may make less use of trade restrictions in a crisis than other countries because of a presumption (commitment) to open PP markets to foreign competition. To the best of our knowledge, these are issues that have not been investigated in the extant literature on procurement and trade, which has focused on home bias in the allocation of public contracts and associated impacts on average procurement costs and firm‐level productivity.4

TRADE POLICY MEASURES DURING COVID‐19

The source of trade policy data used in the analysis is a European University Institute (EUI), Global Trade Alert (GTA) and World Bank project that tracks changes in trade policies for medical products starting on 1 January 2020. The exercise classifies measures as restrictive or liberalising and differentiates between type of instrument (tariff, quota, licensing requirement, ban, etc.). A unique feature of the project is that it includes information on the date of announcement, implementation and removal (if applicable) of each reported measure.5

As of mid‐July 2020, the dataset documented 414 trade policy measures taken by over 100 distinct jurisdictions. The measures include 209 measures liberalising imports of protective equipment, medical supplies and medicines implemented in 106 jurisdictions and 191 export controls imposed by 91 countries for the same set of products. The most common liberalising measures were reduction in import tariffs, while export bans were the most common restrictive measure across countries. Data on trade policy measures are available for 133 countries.

We focus on the first seven months of 2020 because this is the period in which most trade policy measures were imposed. Our research question is whether attributes of prevailing PP regimes are associated with the imposition of trade measures. While the supply of essential goods was largely fixed in the initial period of the crisis, over time, as supply responds to increased demand, this will attenuate the perceived need for export controls and import liberalisation, confounding inferences regarding the possible relationship between PP regulation and use of trade policy in the initial period of a public emergency which is what we are interested in. The GTA dataset shows that some governments began to roll back trade measures starting in June 2020, a pattern that strengthened in the summer and fall of 2020 (see Evenett et al., 2021).6

Figure 1 shows the weekly evolution of trade‐restrictive measures on imports and exports of medical products according to the date of implementation of the measures. The number of restrictive measures on medical products increased exponentially as of the end of March. This coincides with the beginning of the COVID‐19 pandemic and the growing demand for medical products worldwide. There were a limited number of liberalising measures implemented for medical products during the first two months of the year. However, the trend changed drastically as of the end of March, when this number started growing rapidly: the number of such measures more than doubled over a month from 77 measures at the end of March to 174 measures at the end of April. This trend is noteworthy as the period of the spike coincides with the ‘acknowledged’ outbreak of the COVID‐19 pandemic and the growing demand for protective equipment and medical products worldwide.

FIGURE 1

Trade measures for medical products (weekly, January–July 2020). Source: COVID‐19 Trade Policy database (Evenett et al., 2021), own calculations. Note: The data at the end of each week do not consider the measures that were removed (with a removal date in that week)

Close to half of all measures were implemented in March 2020. Starting in May, there is a gradual decline in the imposition of measures. Based on countries and measures where a removal date is explicitly mentioned in the database, restrictive measures were implemented for a shorter duration (58 days) on average than liberalising measures (71 days). Within these distributions, the duration of restrictive measures for medical products ranges from as brief as 2 days in the case of an export ban imposed by Slovenia to 137 days for an export ban imposed by Azerbaijan (Figure 2). Similarly, for liberalising measures the duration ranges from as brief as 11 days in the case of import tariff liberalisation by Dominican Republic to 104 days for import tariff liberalisation by South Korea.

FIGURE 2

Duration of trade measures for medical products by implementation date. Source: COVID‐19 Trade Policy database (Evenett et al., 2021), own calculations

High‐ and upper‐middle‐income countries, according to the World Bank income classification, enacted more trade policy measures targeting the medical sector than other countries (Evenett et al., 2021). As already mentioned, many of the trade‐restrictive and liberalising measures that were adopted to increase the availability of personal protective equipment and medical supplies at the beginning of the COVID‐19 pandemic outbreak were subsequently removed. We consider this dynamic in the empirical analysis.

EMPIRICAL METHODOLOGY

We assess the relationship between different attributes of public procurement regulation and trade policy measures imposed by countries on imports and exports separately by estimating the following equations:7 where is the number of import (‘M’) policy measures imposed by type (‘T’ = liberalising, restrictive) in implementing jurisdiction j; is the number of export (‘X’) policy measures imposed by type (‘T’ = liberalising, restrictive) in implementing jurisdiction j; Proc is a vector of PP regulation variables for country j; z is a vector of country‐ and country‐sector‐specific control variables; is the constant term and ε is the error term. Equations (1) and (2) are estimated separately for liberalising and restrictive measures imposed on medical products and personal protective equipment.

The procurement vector comprises variables reflecting the timeliness (Total_time), administrative procedures (Total_steps), efforts to lower transactions costs (Eproc) and commitments to open government procurement regimes to foreign competition. The first two variables denote the pre‐crisis average total time and number of steps to complete procurement processes in each country or jurisdiction. Use of e‐procurement is measured as the share of e‐procurement in total procurement, based on the range categorisation reported in the World Bank Doing Business Contracting with the Government indicator: less than 25%, 25%–50%, 50%–75% and 100%.8 Openness of procurement regimes is proxied by a binary variable indicating GPA membership (GPA) and by the number of deep procurement agreements (DPAs) signed by each country with trading partners (Num_DPA). Membership of the GPA and the number of DPAs signed by a country implies more open PP regimes, which may be associated with a lower likelihood of imposing trade‐restrictive measures.

The control vector includes country size, the log of population (POP); a measure of geographic distance to global markets, the log of market penetration (MP), computed as a distance (d) weighted measure of other countries’ GDP (GDP), that is MP = Σ (GDP/d); and a measure of government effectiveness (GE). Both equations also include (a) the share of imports of medical goods in country j's total imports (Msh); (b) country j's standardised revealed comparative advantage index (RCA) for medical goods;9 and (c) the (log of) simple average applied tariff rate [ln(1 + Tar)] in country j on medical goods. We also control for the number of COVID‐19 cases (Covid_cases) and the number of related deaths (Covid_deaths) as of 22 July 2020. We do this because the number of cases and deaths may affect the likelihood of removing (retaining) trade policy instruments, with greater case numbers potentially associated with more export restrictions, keeping them on longer and/or deeper liberalisation of imports of essential supplies. We do not use number of cases/deaths for the early months of the pandemic because in most countries case numbers and deaths were low in the initial period—February–March—when most of trade measures were put in place by those countries that decided to do so.

The logic that informs the choice of potential explanatory variables and controls is straightforward, reflecting two types of factors that may influence trade policy responses to an unexpected public health crisis. The first relates to the ‘quality’ of PP regimes that apply to public agencies charged with procuring essential products, captured by indicators of efficiency (speed, ‘red tape’) of domestic PP regimes and commitments to maintain transparent, non‐discriminatory, open procurement markets (captured by membership in trade agreements). The second pertains to standard economic and political economy factors that may influence trade policy responses, including market size, comparative advantage, initial levels of trade protection and government effectiveness. Large, populous countries may have market power and greater supply capacity. A more effective government is more likely to be able to adapt procurement processes to source needed medical supplies. Import share, RCA and import tariff variables proxy for political economy forces that can be expected to have shaped pre‐crisis trade policy and that may influence the direction and duration of crisis trade measures.10 For example, higher import dependence may reflect a pro‐liberalisation domestic political economy, while high levels of pre‐crisis import restrictions may reflect industrial development goals. Countries that have higher import restrictions pre‐crisis may have both greater incentives to relax import barriers temporarily to improve access to essential commodities and may also be expected to re‐impose import barriers more—and more rapidly—than countries that had low or no import tariffs pre‐crisis. Countries with an RCA >1 for exports of medical products will have supply‐side capacity, and thus, the governments concerned may remove export controls more rapidly if such measures are imposed in response to the pandemic spike in demand.

The dependent variable in all equations is characterised by over‐dispersion, which biases log linear OLS estimation. Given the scale dependence of the negative binomial pseudo‐maximum likelihood estimator, we estimate the equations using the Poisson pseudo‐maximum likelihood estimator (PPML) (Santos Silva & Tenreyro, 2006). Given that our interest is in trade policy measures imposed in response to COVID‐19 in the first half of 2020 and all the explanatory variables pertain to 2017–2018, endogeneity emanating from reverse causality is unlikely to be a concern in estimating Equations (1) and (2). There could however be omitted variable biases, especially given that we cannot include any fixed effects. We therefore refrain from drawing any inferences about causality in the presentation of our results.

DATA SOURCES AND DESCRIPTIVE STATISTICS

The procurement variables used as explanatory variables in the empirical analysis are sourced from the World Bank Doing Business Contracting with the Government indicator and pertain to the year 2018. Data are available for all 133 jurisdictions reported to have used trade measures in the first half of 2020 in the GTA trade policy dataset (Table A1).11

TABLE A1

Jurisdictions included in the sample

Albania, Algeria, Angola, Anguilla, Antigua & Barbuda, Argentina, Armenia, Australia, Azerbaijan, Bahamas, Bahrain, Bangladesh, Belarus, Belgium, Belize, Bermuda, Bhutan, Bolivia, Botswana, Brazil, Brunei, Darussalam, Bulgaria, Burkina Faso, Cambodia, Cameroon, Canada, Chad, Chile, China, Colombia, Costa Rica, Cyprus, Czech Republic, Côte d’Ivoire, DR Congo, Dominican Republic, Ecuador, Egypt, El Salvador, Estonia, Eurasian Economic Union, European Union, Fiji, France, Gambia, Georgia, Germany, Greece, Guatemala, Guinea, Guyana, Honduras, Hungary, Iceland, India, Indonesia, Iran, Israel, Italy, Jordan, Kazakhstan, Kenya, Kuwait, Kyrgyz Republic, Lao PDR, Latvia, Lebanon, Libya, Malawi, Malaysia, Maldives, Mali, Mauritania, Mauritius, Mexico, Moldova, Montserrat, Morocco, Mozambique, Myanmar, Namibia, Nepal, Netherlands, New Caledonia, New Zealand, Niger, Nigeria, North Macedonia, Norway, Oman, Pakistan, Panama, Paraguay, Peru, Philippines, Poland, Qatar, South Korea, Sudan, Romania, Russian Federation, Samoa, Saudi Arabia, Senegal, Serbia, Seychelles, Singapore, Slovakia, Slovenia, South Africa, Southern African Customs Union, Sri Lanka, St. Kitts and Nevis, St. Vincent and the Grenadines, Suriname, Switzerland, Syria, Taiwan, Tajikistan, Thailand, Togo, Turkey, Turks & Caicos Islands, Uganda, Ukraine, United Kingdom, United States, Uruguay, Uzbekistan, Venezuela, Vietnam, Zambia and Zimbabwe.

PP processes in richer countries tend to be more efficient on average, that is require both a smaller number of steps and less time for completion (Figure 3, bottom panels). The total number of steps range from a low of 12 (Singapore) to a high of 21 (Honduras, Hungary, Iran, Laos, Myanmar and Oman). The average time taken to complete a procurement process ranges from 270 (South Korea) to 2062 (Venezuela) days. The sample mean is 18 steps and 717 days, respectively. Richer countries also tend to use e‐procurement more on average (Figure 3, top right panel). The share of e‐procurement in total procurement ranges from <25% for most African countries to more than 75% for EU and ASEAN Member States.

FIGURE 3

Scatterplots of procurement variables against per capita income. Source: Shingal and Ereshchenko (2020), World Bank World Development Indicators; World Bank, Doing Business database

GPA is constructed using information on membership of the WTO’s Agreement on Government Procurement as of July 2020; and Num_DPA is constructed using data from Shingal and Ereshchenko (2020), which cover all PTAs in effect until March 2017.12 On average, richer countries tend to be members of more DPAs (Figure 3, top left panel). The number of non‐zero DPAs ranges from 1 for West Asian countries to 26 for the EU. There are 32 WTO GPA members in the sample, mostly comprising high‐income countries.

The control variables are sourced as follows: population and GDP data are from the World Bank World Development Indicators; market penetration (MP) is computed using bilateral distance data from CEPII (Head et al., 2010); and government effectiveness (GE) is sourced from the Worldwide Governance Indicators (Kaufmann et al., 2011). Trade data to construct the import share and RCA variables are from UN Comtrade. Import tariffs are from UNCTAD TRAINS/WITS. Data on COVID‐19 cases are from the WHO (https://covid19.who.int/table). Table A2 reports summary statistics for all variables used in the analysis. Apart from COVID‐19 cases and deaths that pertain to the third week of July 2020, all control variables are the average for the years 2017 and 2018.

TABLE A2

Summary statistics

Variable name	Variable description	Obs	Mean	Std. dev.	Min	Max
Dependent variable
lib_m	Count of import liberalising measures on medical products	133	1.91	2.68	0	22
res_m	Count of import restrictive measures on medical products	133	0.24	1.12	0	10
lib_x	Count of export liberalising measures on medical products	133	0.16	0.37	0	1
res_x	Count of export restrictive measures on medical products	133	1.49	2.98	0	30
Control variables
pop	Population (mln)	128	57	179	0.04	1390
Ge	Government effectiveness	127	0.06	0.88	−1.81	2.23
mp	Market penetration (USD mln)	130	168	481	0	5200
Msh	Share of medical imports in total imports	129	0.06	0.03	0.01	0.17
RCA	Standard RCA for medical products	129	−0.47	0.44	−0.99	0.58
Tar	Simple average applied tariff rate on medical products	119	5.93	4.57	0	27.59
COVID_cases	Cumulative count of COVID‐19 cases	130	107,070	400,949.8	3	3,805,524
COVID_deaths	Cumulative count of deaths due to COVID‐19	130	4294	15,539.5	0	140,437
Procurement variables
tot_steps	Total steps required to complete procurement process	122	17.72	1.76	12	21
tot_time	Total time to complete procurement process (# of days)	122	716.61	245.10	270	2062
eproc	Share of e‐procurement in total procurement	124	0.69	0.32	0.25	1
gpa	Membership of WTO's GPA	133	0.24	0.43	0	1
num_dpa	Number of deep procurement agreements	133	4.36	8.71	0	26

RESULTS

Table 1 reports the results from PPML estimation of Equations (1) and (2) separately for liberalising and restrictive measures imposed on medical goods, with standard errors clustered by country in each case. Results reported in Table 2 replicate the analysis in Table 1 but distinguish between measures that were imposed and subsequently removed within the sample period and those that remained in effect as of mid‐July 2020. Table 3 reports the results from the PPML estimation of Equations (1) and (2) applied to specific types of trade measures, that is restriction or liberalisation of exports or imports, respectively. Finally, based on the dates of implementation and removal, Table 4 reports the results from estimating Equations (1) and (2) using as dependent variable the duration of the respective measure and not the total number of measures.

TABLE 1

Number of export and import measures targeting medical products

Variables	(1)	(2)	(3)	(4)
	lib_m	res_m	lib_x	res_x
Ln(Total_stepsj)	2.20**	6.28**	2.89**	1.35
	(1.08)	(3.02)	(1.47)	(1.22)
Ln(Total_timej)	0.39	1.34	−0.60	1.72**
	(0.31)	(0.91)	(0.76)	(0.71)
Eprocj	0.03	−0.84*	0.47	0.30*
	(0.11)	(0.46)	(0.34)	(0.18)
Ln(Num_DPAj)	−0.31**	−1.97***		−0.04
	(0.15)	(0.63)		(0.16)
GPAj	0.72**	0.56	2.68***	0.67*
	(0.33)	(0.79)	(0.85)	(0.38)
Ln(POPj)	0.09	0.16	0.08	0.40***
	(0.07)	(0.17)	(0.17)	(0.13)
Ln(MPj)	−0.08	0.38	−0.24**	0.23**
	(0.07)	(0.27)	(0.11)	(0.12)
GEj	0.34*	2.06**	−0.62	0.75**
	(0.18)	(0.87)	(0.38)	(0.35)
Mshj	1.76	3.91	7.32	−4.03
	(3.34)	(11.62)	(10.88)	(5.97)
RCAj	0.18	1.50***	1.72**	−0.07
	(0.23)	(0.56)	(0.82)	(0.53)
Ln(1 + Tarj)	0.44**	0.23	0.44	−0.46
	(0.22)	(0.48)	(0.46)	(0.32)
Ln(Covid_casesj)	0.05	0.37*	−0.49	−0.09
	(0.10)	(0.21)	(0.32)	(0.17)
Ln(Covid_deathsj)	0.12	0.06	0.32	0.06
	(0.10)	(0.23)	(0.29)	(0.14)
Constant	−10.15**	−32.18***	−6.67	−16.60***
	(3.98)	(7.22)	(9.12)	(6.25)
Observations	108	108	108	108
R‐squared	.54	.84	.65	.45

Robust standard errors, clustered by country, in parentheses. Levels of significance: *10%, **5%, ***1%.

Abbreviations: DPA, deep procurement agreement; GE, government effectiveness; GPA, WTO Government Procurement Agreement; lib, liberalising; m, import policy; MP, measure of geographic distance to global markets; Msh, import share; POP, population; RCA, revealed comparative advantage; res, restrictive; Tar, import tariff; x, export policy.

TABLE 2

Number of trade measures by status of implementation

Variables	Measures still in effect as of mid‐July 2020				Measures no longer in effect as of mid‐July 2020
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	lib_m	res_m	lib_x	res_x	lib_m	res_m	lib_x	res_x
Ln(Total_stepsj)	3.86***	3.72	20.61***	2.75*	−0.12	274.81***	2.06	−1.52
	(1.31)	(4.45)	(6.56)	(1.60)	(3.11)	(3.72)	(2.33)	(1.88)
Ln(Total_timej)	0.31	1.71*	−13.33***	2.04***	1.57	−47.89***	0.73	1.10
	(0.32)	(1.04)	(3.24)	(0.79)	(2.05)	(1.37)	(1.56)	(0.75)
Eprocj	−0.06	−1.02**	−0.18	0.31	0.19	−0.61***	1.06***	0.08
	(0.10)	(0.49)	(0.25)	(0.23)	(0.26)	(0.10)	(0.30)	(0.17)
Ln(Num_DPAj)	−0.51***	−2.50***	4.35***	−0.06	−0.16			0.11
	(0.15)	(0.71)	(1.35)	(0.20)	(0.25)			(0.21)
GPAj	1.14***	1.16		0.71	0.09			0.75*
	(0.32)	(1.10)		(0.53)	(0.63)			(0.44)
Ln(POPj)	0.13*	0.22	0.26	0.40**	−0.19	−0.20	−0.29	0.16
	(0.07)	(0.24)	(0.19)	(0.16)	(0.16)	(0.16)	(0.25)	(0.15)
Ln(MPj)	−0.13*	0.25	−0.11	0.31**	0.21	11.23***	−0.21	0.16
	(0.07)	(0.33)	(0.39)	(0.13)	(0.24)	(0.18)	(0.27)	(0.11)
GEj	0.35**	3.20***	−3.03***	1.08***	0.63	8.37***	−0.31	0.21
	(0.16)	(0.52)	(0.97)	(0.38)	(0.54)	(0.67)	(0.83)	(0.43)
Mshj	4.72	−5.45			−0.77	−299.18***
	(3.37)	(13.71)			(10.44)	(8.57)
RCAj			−0.75	−0.26			1.01*	−0.28
			(0.87)	(0.63)			(0.61)	(0.44)
Ln(1 + Tarj)	0.31	0.03	2.25**	−0.22	1.10**	29.81***	−1.61**	−0.42
	(0.21)	(0.41)	(0.93)	(0.29)	(0.52)	(0.54)	(0.67)	(0.40)
Ln(Covid_casesj)	−0.03	0.15	−0.34	0.01	0.10	8.20***	−1.48***	0.08
	(0.12)	(0.23)	(0.24)	(0.17)	(0.31)	(0.29)	(0.46)	(0.26)
Ln(Covid_deathsj)	0.20	0.47	−0.52	0.01	0.04	−7.06***	1.20***	−0.15
	(0.13)	(0.32)	(0.35)	(0.15)	(0.26)	(0.23)	(0.34)	(0.20)
Constant	−13.92***	−26.98***	20.79*	−24.84***	−14.54	−642.77***	−5.16	−3.85
	(4.09)	(9.28)	(11.41)	(7.86)	(9.93)	(8.66)	(13.02)	(6.68)
Observations	97	97	75	97	50	33	43	50
R‐squared	.66	.86	.35	.45	.30	1.00	.63	.48

Robust standard errors, clustered by country, included in parentheses. Levels of significance: *10%, **5%, ***1%.

TABLE 3

Types of trade measures used (number)

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	m_lib_t	m_lib_tx	m_lib_oth	m_res_t	x_res_b	x_res_lic	x_res_oth
Ln(Total_stepsj)	−0.02	0.24	1.49*	2.36***	1.63**	2.38**	−0.39
	(0.43)	(0.62)	(0.81)	(0.86)	(0.79)	(0.99)	(0.91)
Ln(Total_timej)	1.23	3.38**	2.29	7.86***	−1.45	5.37***	−0.48
	(1.35)	(1.62)	(3.40)	(2.66)	(1.57)	(1.66)	(2.19)
Eprocj	0.03	0.01	0.01	−0.30	0.39**	0.44	−0.59
	(0.15)	(0.17)	(0.22)	(0.28)	(0.17)	(0.36)	(0.38)
Ln(Num_DPAj)	−0.42**	0.17	−0.50	−1.69***	−0.04	−0.11	0.67**
	(0.18)	(0.26)	(0.34)	(0.42)	(0.19)	(0.25)	(0.32)
GPAj	0.72*	0.86	0.36	0.30	0.66*	0.89	0.95
	(0.41)	(0.54)	(0.63)	(0.72)	(0.35)	(0.69)	(0.99)
Ln(POPj)	0.05	0.00	0.18	0.41*	0.29*	0.55***	0.33
	(0.10)	(0.12)	(0.18)	(0.23)	(0.16)	(0.17)	(0.27)
Ln(MPj)	−0.09	−0.08	0.07	0.38	0.26**	0.25*	−0.23*
	(0.09)	(0.08)	(0.20)	(0.24)	(0.12)	(0.15)	(0.13)
GEj	0.25	0.03	0.65	0.84	0.33	1.29***	0.46
	(0.23)	(0.28)	(0.48)	(0.68)	(0.40)	(0.46)	(0.64)
Mshj	4.21	−2.14	1.33	9.77	−4.49	−3.82	4.35
	(3.90)	(5.69)	(8.99)	(11.11)	(6.88)	(10.05)	(8.17)
RCAj	−0.05	0.35	0.73	1.42***	0.12	−0.07	0.29
	(0.28)	(0.38)	(0.51)	(0.49)	(0.56)	(0.61)	(0.89)
Ln(1 + Tarj)	0.37	0.86**	0.43	−1.44***	−0.27	−0.82	1.25**
	(0.30)	(0.42)	(0.39)	(0.40)	(0.34)	(0.54)	(0.63)
Ln(Covid_casesj)	0.14	−0.09	−0.03	0.37	−0.12	−0.07	0.59*
	(0.13)	(0.18)	(0.29)	(0.30)	(0.20)	(0.22)	(0.35)
Ln(Covid_deathsj)	0.12	0.21	−0.01	0.28	0.13	−0.00	−0.55**
	(0.13)	(0.16)	(0.22)	(0.27)	(0.16)	(0.19)	(0.24)
Constant	−6.14	−13.91**	−18.48*	−45.21***	−9.11	−33.88***	−2.61
	(4.84)	(6.98)	(10.91)	(7.92)	(7.49)	(10.93)	(9.31)
Observations	108	108	108	108	108	108	108
R‐squared	.51	.29	.17	.91	.28	.56	.29

Robust standard errors, clustered by country, included in parentheses. Levels of significance: *10%, **5%, ***1%.

Abbreviations: b, ban; lib, liberalising; lic, licensing requirement; m, import policy; oth, other; res, restrictive; t, tariff; tx, tax; x, export policy.

TABLE 4

Correlates with the duration of aggregate measures

Variables	(1)	(2)	(3)
	lib_m	res_m	res_x
Ln(Total_stepsj)	2.98	385.15***	−1.33
	(2.80)	(4.08)	(1.84)
Ln(Total_timej)	0.50	−49.43***	−0.47
	(1.34)	(2.29)	(0.69)
Eprocj	−0.33	−1.77***	−0.29
	(0.35)	(0.13)	(0.20)
Ln(Num_DPAj)	0.11		−0.11
	(0.33)		(0.18)
GPAj	1.27		1.37***
	(0.85)		(0.51)
Ln(POPj)	0.14	3.77***	−0.20
	(0.22)	(0.20)	(0.13)
Ln(MPj)	0.07	15.76***	−0.09
	(0.23)	(0.27)	(0.10)
GEj	1.11**	10.06***	−0.29
	(0.50)	(0.89)	(0.36)
Mshj	−11.95	−238.30***	−18.44**
	(14.68)	(10.82)	(9.18)
RCAj	−2.43**	0.32	0.04
	(1.01)	(0.47)	(0.55)
Ln(1 + Tarj)	1.04**	38.76***	−0.13
	(0.49)	(0.69)	(0.33)
Ln(Covid_casesj)	−0.11		0.42
	(0.40)		(0.27)
Ln(Covid_deathsj)	0.14	−2.06***	−0.22
	(0.31)	(0.16)	(0.25)
Constant	−10.91	−955.74***	10.36**
	(7.52)	(9.43)	(4.28)
Observations	50	33	50
R‐squared	.20	1.00	.31

Robust standard errors, clustered by country, in parentheses. Levels of significance: *10%, **5%, ***1%.

The average number of pre‐crisis steps required to complete procurement processes is found to be positively correlated with the number of trade liberalising measures, on both the import and export sides. This result reflects measures that were still in force at the end of the sample period (see Table 2, Columns (1) and (3)). In contrast, this variable is positively correlated with the duration of import restrictive measures (Table 4, Column (2)).13 There is also a relatively strong correlation between the total time taken for procurement and the number of trade restrictions, on both the import and export sides. This result reflects measures that were still in force at the end of the sample period (see Table 2, Columns (2) and (4)). On the import side, this result is driven by import tariff levels, while on the export side it reflects licensing requirements (Table 3, Columns (4) and (6)). In contrast, the duration of import restrictions is found to be inversely related to pre‐crisis average time to complete procurement processes (Table 4, Column (2)). Thus, both attributes of pre‐crisis PP regimes are strongly correlated with trade policy activism during the initial months of the pandemic.

This is not the case for e‐procurement. Greater use of e‐procurement is only weakly associated with the number of export restrictions (Table 1), although the association is stronger for use of export bans (Table 3). In contrast, e‐procurement is found to be inversely related to import restrictions that were imposed but subsequently removed (Table 2, Column (6)) as well as the duration of such measures (Table 4, Column (2)).

Turning to the international procurement policy variables, membership of the GPA is positively associated with the number of liberalising measures for imports of medical products as well as removal of export restrictions (Table 1). The former result is also found when attention is restricted to measures that are still in force at the end of the sample period (Table 2). These findings provide some support for a presumption that members of the GPA are more inclined (committed) to maintaining open markets. At the more disaggregated trade instrument level, we find a weakly positive association between GPA membership and the use of export bans, a finding that attains strong statistical significance when focussing on the duration of export restrictions: GPA membership is positively correlated with maintaining export controls during the whole period under analysis if a member decides to use this instrument (Table 4).

More DPAs are associated with fewer import restrictive measures on medical products—the coefficient on Num_DPA is negative and statistically significant at conventional levels. This result seems to be driven by measures that were still in effect at the end of our sample period (see Table 2, Columns (1) and (2)). The analysis focussing on disaggregated measures reveals this is driven mainly by import tariffs (Table 3, Columns (1) and (4)). These results suggest that countries with more DPAs are more open, liberalise imports and are less quick to (re‐)impose import barriers.

Among the control variables, country size is positively correlated, significant at the 1% level, with the number of export restrictions on medical products (Table 1). The same result is obtained for the measure of distance to markets—greater distance from markets is associated with greater use of export restrictions. These findings seem to be driven by measures that were still in effect at the end of the sample period (i.e., that were not removed within the seven‐month period under consideration; Table 2).

These findings across the four main categories of trade policy actions are ‘unpacked’ in Table 3, which reports results for seven different types of trade policy instruments most frequently observed in the GTA dataset. This more disaggregated focus reveals large countries focused more on import tariffs as well as export bans and export licensing requirements.14 Distance to markets is positively associated with restrictive export measures and inversely correlated with liberalising export control measures; the former finding driven by measures still in force (see Table 2, Column (4)). Measures that were removed during the period are strongly correlated with re‐imposition of import restrictions, with a coefficient estimate that is statistically significant at the 1% level (Table 2, Column (6)). These findings are consistent with the patterns of trade activism discussed by Evenett et al. (2021), who note the large number of countries that operate on both the import and the export trade policy margin.

Government effectiveness is positively correlated with most categories of trade policy measures, especially export licensing requirements (Table 3). Higher initial tariffs are positively correlated with liberalisation of imports (Table 1), which may reflect the fact that the higher the pre‐crisis tariffs the greater the potential for attenuating price rises for essential products. Focussing on measures that are removed during the sample period, pre‐crisis tariffs are strongly associated with re‐imposition of import barriers—which may reflect a return to the initial level of protection for the products concerned (Table 2, Column (6)), consistent with the large positive correlation between pre‐crisis tariff levels and the duration of import policy measures imposed on medical products (Table 4).

Supply capacity (proxied by an RCA >1 for pre‐crisis exports of medical products) is positively correlated with import restrictive and export liberalising measures (Table 1). Conversely, greater reliance on imports of medical products is not correlated with the number of import or export measures imposed on medical products (Table 1), but import reliance is strongly negatively associated with re‐imposition of import barriers (Table 2).15 This variable is also found to be negatively correlated with the duration of restrictive measures imposed (Table 4, Columns (2) and (3)).

Analysis by level of economic development

Most attributes of PP regimes are strongly correlated with the level of economic development (Figure 3). While our country sample is dominated by high‐ (30%) and upper‐middle‐income countries (40%) the analysis in Evenett et al. (2021) reveals considerable heterogeneity in the imposition of trade policy measures across all country groups. Upper‐ and lower‐middle‐income countries in the sample display similar patterns of trade policy activism in terms of liberalising and restricting trade in medical products. High‐income countries, in contrast, imposed more restrictive measures, while low‐income countries, which tend to be net importers, mostly implemented measures to liberalise and facilitate trade.

We examine this heterogeneity by introducing interaction terms in Equations (1) and (2) to distinguish between OECD and non‐OECD member economies. Results, reported in Table 5, suggest that country size, membership of the GPA and government effectiveness are correlated more with observed trade policy responses of non‐OECD countries than attributes of PP regimes. Most developing countries have not joined the GPA or signed bilateral or regional DPAs, reflecting perceptions that there is little prospect for national firms to contest foreign procurement markets even if discriminatory PP provisions are removed.16 Our analysis of the initial trade policy responses to the COVID‐19 pandemic suggests an export‐oriented metric to evaluate the prospective net benefits of GPA participation may be too narrow.

TABLE 5

Analysis of aggregate trade policy measures for OECD vs non‐OECD countries

Variables	(1)	(2)	(3)
	lib_m	lib_x	res_x
Ln(No. of DPAj)	−0.27		−0.12
	(0.17)		(0.18)
GPAj	0.68*	1.75***	0.92***
	(0.39)	(0.59)	(0.35)
Ln(Total_stepsj)	2.05	19.35***	2.09
	(1.58)	(4.90)	(1.50)
Ln(Total_timej)	0.71*	−7.91***	0.76
	(0.41)	(1.57)	(0.51)
Eprocj	0.07	−0.01	0.14
	(0.12)	(0.18)	(0.11)
Ln(POPj)	0.15*	−0.27*	0.33***
	(0.08)	(0.14)	(0.10)
Ln(MPj)	−0.13	0.06	−0.04
	(0.10)	(0.21)	(0.09)
GEj	0.55**	−1.23**	0.48**
	(0.25)	(0.61)	(0.24)
Mshj	2.47	29.00*	4.42
	(3.79)	(15.64)	(4.39)
RCAj	−0.06	2.20**	0.58**
	(0.30)	(0.87)	(0.29)
Ln(1 + Tarj)	0.35	0.89**	−0.06
	(0.23)	(0.43)	(0.27)
Ln(Covid_casesj)	0.00	−0.22	0.14
	(0.11)	(0.30)	(0.13)
OECD*Ln(No. of DPAj)	−0.06	24.31***	2.93***
	(0.43)	(2.98)	(0.71)
OECD*GPAj	1.05	1.14	−3.19**
	(1.15)	(2.36)	(1.52)
OECD*Ln(Total_stepsj)	−2.18	−17.75***	−6.48***
	(1.84)	(4.91)	(1.93)
OECD*Ln(Total_timej)	−1.32*	6.20***	−0.67
	(0.69)	(1.58)	(0.87)
OECD*Eprocj	−0.32	0.99***	−2.61***
	(0.36)	(0.26)	(0.76)
OECD*Ln(POPj)	−0.43	−1.51***	−0.03
	(0.28)	(0.28)	(0.27)
OECD*Ln(MPj)	0.00	−0.36*	0.28**
	(0.13)	(0.22)	(0.11)
OECD*GEj	−1.20**	−1.41*	1.29
	(0.59)	(0.73)	(0.80)
OECD*Mshj	−11.90	−48.74***	−42.06**
	(9.16)	(16.00)	(17.46)
OECD*RCAj	1.02	−5.25***	−0.72
	(0.86)	(0.96)	(0.85)
OECD*Ln(1 + Tarj)	−0.21	−9.10***	−2.29***
	(0.52)	(1.30)	(0.70)
OECD*Ln(Covid_casesj)	0.13	0.58*	−0.57**
	(0.26)	(0.30)	(0.23)
OECDj	16.76**	−56.91***	33.38***
	(6.63)	(12.52)	(9.14)
Constant	−11.56***	−10.34	−13.03***
	(4.18)	(11.26)	(3.82)
Observations	108	108	108
R‐squared	.62	.90	.90

Robust standard errors, clustered by country, in parentheses. Levels of significance: *10%, **5%, ***1%. OECD is a binary dummy variable that takes the value 1 if the reporting jurisdiction is an OECD country and zero otherwise.

CONCLUSION

Controlling for country size, government effectiveness and economic factors that may influence trade policy, we find evidence that the use trade measures targeting medical products in the first seven months of the global COVID‐19 pandemic is positively correlated with pre‐crisis attributes of national public procurement regimes. Jurisdictions with PP systems that are characterised by more steps or stages and processes that take longer on average to complete made greater use of trade policy to increase domestic availability of protective equipment and medical supplies. At the same time, we find that GPA membership and participation in DPAs is associated with maintaining more open markets for medical products. Jurisdictions that have signed agreements that require non‐discrimination between foreign and domestic firms do more to reduce import barriers and are slower to re‐impose import restrictions.

Our results point to the need for more in‐depth analysis that includes information on what was done by different jurisdictions to procure medical products and protective equipment on an emergency basis, and the global supply response by business producing the relevant products. The main conclusion we draw from our analysis is that future research assessing policy responses to the COVID‐19 pandemic should consider both the efficacy and efficiency of public procurement processes and the recourse made to trade policy in efforts by governments to address the sharp rise in domestic demand for personal protective equipment and medical supplies.