Economic costs of COVID‐19 for cross‐border regions

Abstract

In Spring 2020, the first wave of the COVID‐19 pandemic hit Europe most severely. While empirical evidence regarding the economic costs of the strict lockdown measures enacted during the periods before the widespread diffusion of vaccines is now available, little is known about the economic impact of both strict lockdowns and partial closures on border regions. This is instead a relevant case study to analyze, in particular in the light of the asymmetric nature of border closures. This paper fills this gap and offers two sets of analyses: a first assessment of partial closures, enacted after the first wave of the COVID‐19 pandemic and based on the approach applied to European cross‐border regions to measure the costs of legal and administrative barriers (Camagni et al., 2019); and a second measurement based on simulating the impacts of full closures with the Macroeconometric Social Sectoral Territorial (MASST)‐4 model (Capello & Caragliu, 2021a). These analyses also allow for the pinpointing of the spatial distribution of economic losses, and to identify whether different regional typologies suffered the highest contraction.

1. SETTING THE SCENE

In Spring 2020, Europe was severely hit by the first wave of COVID‐19. Although it was not the first area to be affected, it was certainly the one with the most dramatic impact. Medical evidence suggests that Europe initially faced the hardest consequences. The virus circulated rapidly, and its tragic consequences in terms of surges in mortality rates became very clear early on.

This crisis prompted a rapid reaction by national authorities, who, first in Italy (March 9, 2020), and then in the rest of the European Union, closed inter‐ and intranational borders, prohibiting citizens' urban mobility and (with a few notable exceptions such as Sweden) public gatherings (Piccoli et al., 2020). In many cases, production plants in non‐indispensable manufacturing activities were also temporarily closed.

While early measures undertaken in Spring 2020 had a symmetric and universal nature, subsequent closures became increasingly less severe, and thus may have caused asymmetric effects in areas mostly depending on cross‐national economies and therefore directly affected by policies limiting cross‐country movements, that is, cross‐border regions (CBRs).

CBRs represent an interesting case study for analyzing the effects of COVID‐19 measures. In Europe, these areas are penalized by the presence of borders that still represent a major hurdle to full economic integration in the EU (Scott, 2011), and the recent emergence of populistic movements has made (internal) borders more important than ever (Székely & Kotosz, 2018). The consequences have already been identified, in that empirical evidence shows that borders create obstacles to the full exploitation of local assets (Camagni et al., 2019; Kerzhner et al., 2018).

The introduction of higher barriers in the period of the COVID‐19 pandemic is expected to cause drastic effects on border regions, and these effects will be asymmetric within border regions, given the country nature of such measures. A Mission Opérationnelle Transfrontalière‐commissioned study (Peyrony et al., 2021) provides a detailed account of the bilateral border closures following the “Guidelines concerning the exercise of the free movement of workers during COVID‐19 outbreak” issued by the European Commission on March 30, 2020 and the “Guidelines on EU Emergency Assistance in Cross‐Border Cooperation in Healthcare related to the COVID‐19 crisis” adopted on April 3, 2020. The report documents a detailed and precise mosaic of restrictions to the free movement of people and freight, against treaties such as the Schengen agreement, signed in 1985 with the aim of abolishing border controls in Europe.

This paper aims to provide an assessment of the economic losses caused by both total and partial closure measures undertaken to limit the diffusion of the virus, the former between March and Summer 2020, the latter after Summer 2020. In particular, costs are hereby quantified in terms of missed GDP growth due to the introduction of such measures. In particular, the paper presents two main sets of assessment regarding:

• losses in terms of missed GDP growth owing to partial measures that took place after the Spring 2020 lockdowns, and are therefore net of macroeconomic factors that accompanied the total closure period; and

• losses in terms of missed GDP growth owing to full measures, including macroeconomic factors that influenced the magnitude of losses during Spring 2020 lockdowns.1

Owing to their nature, the total and partial measures have to be treated differently, in that the macroeconomic factors characterizing the universal closure are, in the case of partial closures, not part of the story. In fact, consumption resumed in Fall 2020 and took off in early 2021, prompting an early rebound that allowed most European countries to recover 60 to 70% of GDP missed in 2020.

To deal with these two assessment exercises, we employ two well‐established methodologies. The first method, originally developed in Capello et al. (2018a) for measuring generic border effects and adapted in Camagni et al. (2019) to the assessment of legal and administrative barriers, is used to assess the costs of partial measures. The second method is based on the Macroeconometric Social Sectoral Territorial (MASST) model (Capello, 2007; Capello & Caragliu, 2021a), and exploits the simulations at NUTS2 level of the costs of full closures discussed in Capello and Caragliu (2021b) to translate them at NUTS3 level (baseline geography for European CBRs) through the use of the MASST at NUTS3 (MAN) model (Camagni & Capello, 2011).

This paper aims to highlight whether costs are different within border regions because not all border regions are equal. In particular, we examine whether urban areas suffer more than rural ones; whether rich border regions (located in EU14 Countries) are characterized by higher losses than those affecting less advanced countries (Central and Eastern European Countries [CEECs]); and whether their land or maritime position make them suffer more or less severely.

The rest of the paper is organized as follows. Section 2 presents the literature framework of two relatively disconnected branches that this paper merges, one dealing with the identification of border effects (economic losses owing to the disruption caused by transnational border closures), and the other dealing with the economic costs of the COVID‐19 closures. Section 3 describes the methodologies used to simulate losses due to both partial and full closure measures. Section 4 illustrates the results of both these simulations. In Section 5 we conclude and discuss possible policy implications stemming from our overall findings.

2. BORDER EFFECTS AND THE ECONOMIC COSTS OF MEASURES TO PREVENT THE DIFFUSION OF COVID‐19

2.1. Border effects

European economic integration went through three major steps, increasingly abating the relevance of economic and institutional barriers: the creation of a common market in 1957, the evolution toward the single market in the 1985–1992 period, and the birth of the Economic and Monetary Union (EMU) in the 2000s. The first step led to the abolition of international economic barriers, causing a major increase in (mostly intraindustry [Balassa, 1975; Sapir, 1992]) international trade. In the second step, non‐tariffs and frontier effects were tackled, with a progressive harmonization of technical standards, quality certifications of products rules, and most importantly, the free movement of people, goods, and services (Badinger, 2005). The last step witnessed the creation of a common currency (Micco et al., 2003).

Within the integration framework, border regions represent a particularly relevant case study, in that in these areas the rupture caused by political borders, with the burden of administrative, legal, language, cultural, and institutional barriers they cause, is particularly evident. For this reason, two parallel strands of literature have developed over the past three decades focusing on the identification of border effects. In the international trade literature, studies following in the footsteps of McCallum (1995) provided convincing empirical evidence on the role of international borders in hampering the free flow of freight (the so‐called home bias in trade puzzle; Wolf, 2000). In the political science and geography traditions, borders have instead been studied mostly through the lens of geopolitics, and hence seen as partially exogeneous elements worth of being analyzed because of their impact on citizens' self‐perception and identity.

While in international trade a shortcut is typically made by assuming that more intense trade automatically leads to growth effects, in the second approach qualitative insights are not often matched with sound quantitative assessments of the economic implications of border effects.

For the sake of our work, it is also worth stressing that border regions represent particularly relevant case studies in the light of the ongoing COVID‐19 pandemic. The diffusion of the virus has, from the very outset of the pandemic, caused an increase in barriers among European countries, and a natural consequence of such closures is to wonder whether closure effects were symmetric or whether their impact is different for border regions with respect to areas closer to a country's center. Owing to their peculiar nature, CBRs are object of specific policy actions, in particular of the European Cross‐Border Cooperation Program (CBCPs; European Commission, 2015), created to enhance the effective and coordinated intervention on specific common transnational challenges.

While the two strands of literature discussed above remain rather disconnected, an attempt to merge them to obtain a methodology amenable to the empirical assessment of growth effects in border regions has first been proposed in Capello et al. (2018a), and will be discussed in Section 3.1.

2.2. Economic costs of measures to prevent the diffusion of COVID‐19

The unexpected and dramatically pervasive nature of the COVID‐19 pandemic has drawn almost universal attention from research institutions ever since March 2020. While this literature is burgeoning, here we critically summarize studies with a specific regional focus.

An excellent synthesis of this literature is presented in a special issue of Regional Science Policy and Practice, published in December 2020. Among these works, Bonet‐Morón et al. (2020), who, focusing on Colombia and on the basis of input–output methodologies, forecast a contraction of average regional GDP levels ranging between 0.5% and 6.1% of pre‐COVID levels, depending on the scenario considered. In the same issue, Porsse et al. (2020) exploit computable general equilibrium (CGE) simulations to show that the Brazilian economy could have shrunk by between 5% and 10% of pre‐COVID levels, depending on scenario assumptions and the length of lockdowns. Modrego et al. (2020), in turn, warn that job losses in Chilean regions may range from 1.5% to 13.6% of the total workforce depending on a region's sensitivity to lockdown measures.

In the European case, owing to the substantial lag with which statistics at the subnational level are made available by EUROSTAT, regional economic losses engendered by full and partial closure measures enacted to minimize the diffusion of COVID‐19 have been calculated mostly on the basis of model simulations. For instance, Conte et al. (2020) use the spatial CGE model RHOMOLO, the main regional economic modelling instrument of the EU's Joint Research Centre, to portray substantial heterogeneity in the regional responses to the enactment of measures aiming to contain COVID‐19; their baseline scenario forecasts a drop of about 13% of regional GDP in European regions. In another special issue,2 Capello and Caragliu (2021b) further document spatial polarization of full closure impacts, with relevant country effects and substantially higher GDP losses for regions located in countries hit the hardest by the initial wave of the pandemic (including France, Italy, Spain, and Belgium).

Across this literature, no specific attention is paid to a typology of regions – those located within EU internal borders – that should theoretically be expected to incur the highest losses, precisely because of their border location. Section 2.3 discusses the few studies tackling this aspect.

2.3. Economic costs of measures to prevent the diffusion of COVID‐19 in border regions

A final branch of the literature important to summarize here is related to emerging studies discussing the impacts of COVID‐19 and related closure measures with a specific focus on border regions.

Given the paramount relevance of this global pandemic, it comes as no surprise that, despite the short amount of time that has passed from the first wave of the pandemic, studies have been rather abundant. Empirical work focused on two main issues: (i) the role of border location, settlement structure, hard and soft institutions, and other regional and urban characteristics in determining the spatial distribution of COVID infections (Chilla et al., 2022; Gerritse, 2020) and mortality rates (Del Bo, 2021); and (ii) the economic impacts of full or partial closure measures on border regions. This subsection focuses on the second branch.

This is in turn made up of a mosaic of studies that cut across the scientific, the gray, and the policy literature. The latter subbranch is made up of studies elicited by European (e.g., the excellent collection of studies on border impacts in Scott, 2021) and local (Wøien Meijer & Giacometti, 2021) institutions for providing evidence‐based policy suggestions on how to deal with the pandemic in these particular areas.

Academic research has instead focused mainly on case studies, both qualitative and quantitative, on single border regions. For instance, Wallin Aagesen et al. (2022) quantify the mobility reduction due to COVID restrictions in Nordic Countries (Denmark, Finland, Iceland, Norway, and Sweden). On the basis of a 4‐year panel of Twitter data, they show that cross‐border mobility decreased by 68% during the first wave of the pandemic. Along the same lines, quali‐quantitative methods are used in Böhm (2021) to highlight the multifaceted impacts of border closures in the Czech–Polish CBR. A subregional breakdown of estimated GDP contraction is presented in Paül et al. (2022), who show a spatially heterogeneous set of economic impacts across municipalities in the Spanish–Portuguese CBR of Lower Minho.

Evidence has also been collected outside the geographical scope of European CBRs. For instance, Silva‐Sobrinho et al. (2021) exploit a self‐collected micro dataset containing information from 2,510 individual interviews to uncover the economic impact of COVID‐19‐induced border closures in the Brazilian state of Paraná, located on the border with Paraguay. Their result, while providing evidence of substantial negative income effects of border closures, also document relevant support toward the latter among the interviewed sample.

Across this whole literature, there seems to be no quantitative assessment of the economic effects of full and/or partial closure measures across all European CBRs to date. A notable exception in this respect is Rosik et al. (2022), who focus on the impacts of border closures on overall regional accessibility. This paper fills this gap, in particular addressing the following two research questions:

• RQ. 1What is the economic impact of COVID‐19‐related partial and full closure measures in European cross‐border regions?
• RQ. 2Within border regions, which type of regions suffered the heaviest economic losses owing to partial and full closure measures?

The methods employed to answer these two research questions are described in the next section.

3. MEASURING THE IMPACTS OF MEASURES PREVENTING THE DIFFUSION OF THE VIRUS

3.1. Methodology for measuring the impacts of partial closure measures preventing the diffusion of the virus

The method used to quantify the impacts of partial closures has first been presented in Capello et al. (2018a) and subsequently exploited in several applications (Capello et al., 2018b, where the overall border effect is broken down into its building blocks; Capello et al., 2018c, where the method is amended so as to also encompass border effects acting on external regional assets; and Capello et al., 2018d, where the approach is employed to simulate the regional costs of Brexit before it actually took place). More recently, the method has been specifically applied to the measurement of the costs stemming from administrative barriers persisting among EU regions (Camagni et al., 2019; Caragliu, 2022). This last approach is applied for the measurement of the costs of partial barriers discussed in this paper.

The method entails three steps (Figure 1):

1. we first econometrically estimate the efficiency of border regions in a pre‐COVID‐19 period, leading to an increase in GDP between 2008 and 2019, and obtaining GDP in 2019 as a projection of GDP in 2008;

2. we then simulate what GDP would have been if partial closures were enacted, thus causing GDP to shrink;

3. lastly, we calculate the difference between the estimated GDP and the simulated one. This difference represents the loss of GDP owing to the increase in borders as a result of partial measures.

More specifically, in stage 1 we estimate the following specification:

$$\mathrm{\Delta }Y=\alpha +\mathrm{\Sigma }{\beta }_k*{\mathit{\text{control}}}_k+\gamma *\mathit{\text{border}}+{\delta }_i*{\mathit{\text{assets}}}_i+{\vartheta }_i*\mathit{\text{border}}*{\mathit{\text{assets}}}_i+{\mu }_{\mathit{ij}}*\mathit{\text{border}}*{\mathit{\text{assets}}}_i*\mathit{\text{barrier}}+\mathrm{\Sigma }{\rho }_c*\mathit{\text{country}},+{\epsilon }_{i,j},c=1\dots n$$ (1)

where ΔY is 2008–2019 regional GDP growth rate, i indicates individual assets, barrier whether the region suffers from the three types of barriers discussed below more than the EU average, and c is countries. δ _i measures the impact of each regional growth asset i on regional growth, ϑ _i captures the impact of asset i on the growth of border regions with respect to all other regions, and μ _ij measures the impact of assets i on the growth of border regions characterized by barrier j, with respect to all other regions. In Equation (1), when the μ _ij coefficient is associated with a negative and significant parameter estimate, regions characterized by barriers have a lower growth impact from a specific asset with respect to all other EU regions, which means that when barriers affect the region, that asset is not exploited as efficiently as in other regions.3 ^, 4

FIGURE 1.

Structure of the simulation procedure for partial‐closure measures Source: Camagni et al. (2019), author creation.

Barriers allow us to model the effects of partial closures put in place by different types of measures among countries. Legal and administrative barriers were drastically increased to ban the free movement of people and goods, to limit contagion, and to keep track of the number of infections. Peyrony et al. (2021) have also vastly documented the spatial distribution of the increases in legal and administrative barriers being erected between country couples in the wake of the pandemic. Border closures have been associated with a rise in the burden of paperwork associated with moving people and freight across borders. While a comprehensive account of all such measures goes beyond the scope of this paper, it is important here to point out the guidelines for safe travelling in COVID‐19 by the European Commission, first issued on October, 13 2020, and subsequently updated on February 1, 2021, June 14, 2021, and January 25, 2022. Until a few months before the writing of this paper, for instance, most EU countries adopted the mandatory request of a green certificate, often with the additional request of the negative result of a recent antigen COVID‐19 test, to allow international travel within Europe.

Another type of barrier that was worsened by the partial measures, and by the presence of the virus, was trust among countries, in many cases for the different attitudes and partial restriction measures adopted. The worsening of this barrier complicated the way people living across international borders within CBRs could travel between countries. Early evidence suggests that the erection of partial measures to contrast the diffusion of COVID‐19 within CBRs has caused a significant decrease in the stock of bilateral trust (Haist & Novotný, 2022) in all its forms, in particular increasing trade costs, hampering the diffusion of knowledge, and limiting firm cooperation.

Lastly, partial restrictions for fighting the diffusion of COVID‐19 also simply meant that the Schengen treaty has been repeatedly suspended by individual countries in a patchy way across the whole continent.

Our simulation is based on the increase in these three barriers, namely:

• Legal and administrative barriers;

• Bilateral trust;

• Schengen barriers.

The simulation proceeds as follows. For legal and administrative barriers and bilateral trust, measured in a continuous way, we raise each region's level of barrier to the maximum level observed in all EU NUTS3 regions, which represent the baseline geography of European CBRs. For the Schengen barrier, we instead switch a dichotomous variable off, implying that we simulate a universal dropping of the Schengen treaty.

3.2. Methodology for measuring the impacts of full closure measures preventing the diffusion of the virus

Spring 2020 witnessed a nearly unprecedented example of a symmetric shock to European countries. The negative effect on EU economies has been fully exogenous; its diffusion caused a nearly universal reaction translating in plant closures, bans on public gatherings and events, and border controls that caused a severe macroeconomic downturn in all EU countries with no exception, hitting almost all major components of aggregate income (i.e., consumption, investment, exports, and imports). The only exception was represented by a major surge in public expenditure almost everywhere in Europe (Carraro et al., 2022), with the aim to react to the medical emergency and support the increased need for healthcare financing with extra public expenditure on the one hand, and to counterbalance negative labor market outcomes due to forced plant closures on the other hand.

Owing to the symmetric nature of the Spring 2020 closures and in the absence of official EUROSTAT statistics at regional level, Capello and Caragliu (2021b) simulated the costs of COVID‐related closures in Spring 2020 with a multi‐equation model (described below), taking both the border increase and the macroeconomic measures into account. In fact, within this rather dramatic framework, border closures represent only part of the whole story, the core of which is the diffusion of contagion‐preventing measures to the macroeconomic sphere.

In this paper we treat the costs of partial measures simulated at NUTS2 level as a first step to eventually obtain an assessment of the costs of these measures at NUTS3 level. In a second step, we estimate NUTS3 GDP levels from the simulated NUTS2 ones, as summarized in Figure 2.

FIGURE 2.

Structure of the simulation procedure for full‐fledged stay‐at‐home measures Source: Author creation.

As for the first stage, simulations were obtained by applying a forecasting regional macroeconometric growth model, MASST (presented in its fourth version in Capello & Caragliu, 2021a), built to simulate regional growth scenarios in the medium‐ and long‐run (typically, over 15–20 years' time horizon; see Capello & Caragliu, 2021a, and Technical Appendix A.2.1 for additional details on the MASST4 model). On the basis of estimates of structural relations among economic variables at both the national and regional levels, the model projects these relations to future periods on the basis of assumptions on the macroeconomic and territorial economic conditions captured through variables that remain exogenous to the model and represent levers in the hands of the modeler.

In the case of full‐fledged closures, 2018 data are projected into 2020 for the period of total closures (March–June 2020). The assumptions regard both macroeconomic variables (e.g., interest rate, inflation rate) as well as regional ones (e.g., level of trust, innovation rates, industrial growth), among which are also those related to increased barriers due to total closures. In particular, two major barriers are assumed to drastically increase for this simulation; namely the decrease of trust among countries and the weakening of input–output relationships due to total closures. These assumptions (the details of which is provided in Technical Appendix A.2.2) have been cross‐checked for consistency with respect to the fine‐grained evidence collected in Peyrony et al. (2021).

The second step in Figure 2 entails moving from the NUTS2 to the NUTS3 level. To do this, we exploit a methodology first presented in Camagni and Capello (2011). This is obtained by explaining the difference between NUTS2 GDP growth and NUTS3 GDP growth (the regional shift s) as a function of local context conditions. The shift s is added to the growth of GDP at NUTS2 level to obtain NUTS3 GDP growth as follows (Equation 2):

$$\Delta {\mathit{GDP}}_{\mathit{\text{NUTS}}\mathbf{3}}=\Delta {\mathit{GDP}}_{\mathit{\text{NUTS}}\mathbf{2}}+\mathit{s}$$ (2)

Through the simulation of NUTS3 GDP growth rates, we obtain an estimate of the loss of GDP for CBRs due to full closures.

4. IMPACTS OF PARTIAL AND FULL CLOSURES ON BORDER REGIONS

This section presents results of our empirical assessment exercise. Results are discussed starting with the assessment of partial closures, and then of full ones. The rationale for this choice5 is owing to the different methodologies adopted for testing the two impacts. As anticipated in Section 3, a single equation model is used to assess the impact of partial restrictions. While this is compatible with the need to incorporate multiple barriers in a growth equation, we then need a multi‐equation model to take account of the macroeconomic factors playing the lion's share in determining the effects of full lockdowns.6

4.1. Impacts of partial closures on border regions

Aggregate losses caused by partial COVID‐19 restrictions for all of Europe are shown in Table 1.

TABLE 1.

GDP losses in Europe due to partial closures

Source of loss	Absolute loss	% EU27	% border regions
GDP	−12,419,969,936.17 €	−1.08	−2.44
	[−13,268,969,936.17 €, −11,570,969,936 €]	[−1.15, −1.01]	[−2.61, −2.27]

Note: 95% confidence intervals reported.

Source: Author creation.

Table 1 highlights a rather relevant role played by partial COVID‐19 closures through the increase in barriers in CBRs in the aftermath of the first (European) wave of the pandemic. All in all, our method suggests a GDP loss of 12 billion euros in CBRs. This corresponds to roughly 1% of the EU27's overall GDP, and tops at almost 2.5% of CBRs’ GDP. In fact, the method identifies losses for NUTS3 regions located in CBRs, which are only a subset of all NUTS3 regions. For each assessed loss, we also present a confidence interval, which suggests the boundaries of an ideal area defining the intensity of the loss with a given (chosen) probability. In our case, confidence intervals are based on a 95% confidence level. The precision of the estimates is reflected in relatively narrow intervals, ranging from 11 to 13 billion Euros in aggregate terms; from 1.15% to 1% in EU27 percentage terms; and from 2.6% to 2.3% in CBR percentage terms.

Even though aggregate results provide a general and informative picture of the extent of the losses, it is now worth focusing on their spatial dispersion. Here we focus on the typologies of regions most affected by partial closures; for a full breakdown of NUTS3‐specific losses, see Technical Appendix A.3.1.

The first typology seeks to identify whether losses have been most severe in urban areas or in rural ones. To this aim, we employ the regional classification of NUTS3 areas as agglomerated, urban, or rural, first presented in the ESPON 1.1.1 Project Final Report and then in Capello et al. (2015). The classification is reported in Technical Appendix A.4, but the main idea is to allocate a class to regions depending on a combination of population and density indicators.

The second classification relate to types of borders. In fact, this paper focuses on losses due to restrictive measures enacted in all border regions, irrespective of whether borders are land or maritime ones. For a subset of NUTS3 areas, the two may actually coexist, in that some areas can join multiple cooperation programs on both land and sea.7

Lastly, we deal with a classification of aggregate losses into those accruing to CBRs located in countries joining the EU prior to 2004 versus those joining the EU in the last three waves of enlargement. The rationale of this last classification is to verify whether regions belonging to countries with a long history of centralized planning suffer more or less from the losses induced by partial restrictions.

Table 2 shows results for GDP losses in border regions by regional typologies.

TABLE 2.

CBRs' GDP losses due to partial closure by regional typologies

Typology	Loss of GDP (€)	GDP loss as a share of CBRs' GDP (%)
Agglomerated	−5,871,258,515.28	−1.53
Non‐agglomerated	−6,548,711,420.89	−5.47
Land border	−8,919,796,590.52	−2.19
Maritime border	−3,725,990,980.85	−2.70
Region located in EU14	−11,177,972,942.55	−2.36
Region in CEECs	−1,241,996,993.62	−2.39

Table 2 highlights the following main messages:

• in absolute terms, we find evidence of a relatively evenly spread loss of GDP in agglomerated regions and non‐agglomerated ones. However, given that the latter represent a substantial share of the EU's CBRs, losses in percentage terms turn out to be substantially larger for rural areas and regions hosting second and third tier cities;

• taking the uneven distribution of GDP production and looking at percentage losses for all remaining classifications, we find no substantial difference between land border regions versus maritime ones nor EU14 regions versus CEECs regions. Across all these classes, losses oscillate around two and a half a percentage points of regional GDP.

While a full‐fledged cost–benefit analysis of the net effect of fully removing barriers into account goes beyond the scope of this paper, it is here worth relating our findings to prior estimates of the costs of eliminating barriers.8 However, there is no direct assessment of such costs, and this literature focuses mainly on the gains that would stem from the removal of such barriers.

From an economic theory standpoint, gains are usually estimated to be quite large and to cover a number of positive economic outcomes, including international trade and mobility (McKenzie, 2007); faster economic growth (Spolaore & Wacziarg, 2005); and increased competition (Coşar et al., 2015). Consequently, we could indirectly infer that costs of removing barriers would be at least as high as the gains stemming from their removal, unless some form of market failure prevents the action of removing barriers. While a direct assessment of the extent of these costs seems, to the best of our knowledge, not to have been attempted before, some studies provide indirect evidence, for instance documenting the relevant persistence of costs once they are removed (see for example, Nitsch & Wolf, 2013 who discuss the hysteretic behavior of the home trade bias between East and West Germany despite more than two decades having passed since their reunification).

4.2. Impacts of full closures on border regions

This section presents the findings of our simulations based on the methodology introduced in Section 3.2. Findings represent the loss of GDP due to both the increase in barriers and the deep worsening of the macroeconomic conditions that total closures engendered. Aggregate results are presented in Table 3, along with the classical confidence intervals at the 95% level.

TABLE 3.

GDP loss in Europe (March–June 2020)

Absolute loss	% EU27	% border regions
−51,117,350,000 €	−4.44	−10.04
[−89,461,270,000 €, −9,068,400,000 €]	[−7.78, −0.79]	[−17.58, −1.78]

Losses are roughly five times as large as those due to partial closures owing to the symmetric and universal nature of closures in border regions in the wake of the first lockdown period. All in all, border regions lost about 10% of their potential GDP in 2020, and about 4.5% of EU27 GDP (Table 3).9

Moving on to the analyses of results by regional typologies, losses are presented in Table 4.

TABLE 4.

CBRs' GDP loss owing to full closure by regional typologies (March–June 2020)

Typology	Loss of GDP (€)	GDP loss as a share of CBRs' GDP (%)
Agglomerated	−24,607,330,000	−9.46
Non‐agglomerated	−26,510,020,000	−10.20
Land border	−38,427,860,000	−9.61
Maritime border	−15,010,660,000	−10.72
Region located in EU14	−44,923,550,000	−9.98
Region in CEECs	−6,193,794,000	−9.37

The main result emerging from reading Table 4 is related to the relatively spatially even diffusion of losses across all typologies. While trade and cross‐border mobility become more complicated as a result of border closures, forcing border areas to reroute toward other regions located in the same country, in 2020 macroeconomic factors played the lion's share in causing aggregate GDP contraction. As a result, losses appear slightly higher for:

• rural areas with respect to urban ones, which could be interpreted as the result of cities being able to cope more efficiently owing to the nature of large markets (Fujita, 2012);

• maritime regions with respect to land ones, probably owing to the fact that land borders are by definition more permeable, and more complicated to control and seal compared with sea ones;

• EU14 regions with respect to CEECs regions, arguably owing to the more severe impact of the COVID‐19 pandemic in the Western part of the continent during the first wave in 2020.10

5. CONCLUSIONS AND POLICY IMPLICATIONS

This paper documented a rather relevant role played even by partial restrictions in causing losses in terms of GDP in European CBRs. All in all, our simulations suggest that partial measures may cause a decrease of around 1% of the EU's GDP. In the long run, this may engender substantial losses in terms of potential levels of welfare for EU citizens. These costs are roughly twice as large for EU CBRs, which suffer most directly because of their geographical location.

The paper also complemented these analyses by showing results of measuring losses due to the Spring 2020 lockdowns, obtained with simulations through the MASST4 model at NUTS2 level, and allocated to NUTS3 regions in our sample by means of the MAN model. These analyses suggest that losses due to COVID‐19 restrictions, which encompass both more severe restrictions in terms of border closures, as well as macroeconomic factors, are an order of magnitude larger than the losses induced by partial closures. All in all, border regions lost about 10% of their potential GDP in 2020. Despite a certain degree of uncertainty about the estimates, our estimates appear in line with recent EUROSTAT figures now available at the country level.

As in other related studies, this is not exempt from limitations, which call for important caveats in the interpretation of our findings. In particular, the inclusion of one single growth asset in each single growth regression testing Equation (1), owing to a degrees of freedom issue, suggests that the losses estimated here could probably represent a maximum threshold, and their extent, especially in case barriers modeled in Equation (1), would be mutually correlated and could be lower in absolute terms.

Despite the lack of regional statistics for both 2020 and 2021, and even taking the random component of all estimates into account, our results appear in line with the mounting evidence that is becoming available at the country level. For future reference, results could be compared against the backdrop of real NUTS2‐ and NUTS3‐level statistics, both for checking the methodology's statistical degree of approximation as well as to better help evidence‐based policy design.

In terms of policy suggestions, while this paper does not deal with the cost of removing barriers and/or avoiding partial restrictions, it does hint at a rather relevant cost caused by the latter, and highlights substantial gains stemming from policies favoring full labor and freight mobility, as well as the complete removal of remaining barriers in terms of trust, the Schengen treaty, and administrative differences.

While policies aiming to remove barriers could be rather complicated to enact, both supply‐side and demand‐side actions could work in the direction of barrier removal.

On the demand side, the very existence of the European Union is (also) meant to foster further integration by coordinating demand across political borders. The removal of border effects in trade may proceed through incentives to demand of goods produced across borders. Signaling (which has already been fostered by EU institutions through the enforcement of a compulsory 2‐year minimum guarantee on all products sold on the EU markets) is one way to achieve a better perception of foreign goods. Along the same lines, a strengthening of the Schengen treaty, and an attempt not to drop it partially or fully in case of local shocks, would enhance citizen mobility and also allow citizens of cross‐border regions to increase their mutual trust.

Supply‐side policies would instead require additional investment at both the national and supranational levels to increase the supply of growth‐enhancing factors, especially in cross‐border areas. One instance of this would be tax credits for research and development (R&D) investment by also co‐financing international cooperation among R&D partners located across political borders.

Supporting information

Appendix S1. Technical Appendix

Capello, R. , Caragliu, A. , & Panzera, E. (2022). Economic costs of COVID‐19 for cross‐border regions. Regional Science Policy & Practice, 1–14. 10.1111/rsp3.12590