John Lewis
Covid travel restrictions limited movement of people but also made cross-border goods trade more difficult. Did this contribute to the fall in global goods trade during the pandemic, and if so by how much? In a recent paper using a structural gravity model on global trade flows with domestic trade, I show that a full closure reduced trade for a typical country pair by around 19%, implying a peak hit to global trade of about 23% in 2020 Q2. Hits were larger for nearby partners, and were concentrated in road and air freight, with seaborne trade unaffected. These differences explain why some countries could close borders with smaller trade hits than others. Trade rebounded as restrictions eased, suggesting no lasting scarring.
In the first quarter of 2020, governments tightened borders at unprecedented speed, introducing testing, quarantine and, in some cases, full closures. Global goods trade also fell sharply. Was that fall purely down to contracting global activity, or did border frictions rise making it more costly to trade internationally? In a recent paper, I answer this using a gravity framework that includes domestic trade flows and time-varying exporter and importer controls, allowing me to isolate the role of the ‘extra’ cost of selling abroad versus at home: the so-called ‘border friction’ which controls for the effect of reduced supply capacity in the exporting country and reduced demand in the importing country. I find that the rise in border frictions was substantial, and implies a significant hit to global trade, over and above that which the generalised contraction in economic activity would have implied. To my knowledge this is the first attempt to explore this issue on cross-country data over the full pandemic period (and beyond) using a multi-country gravity model.
Why might travel restrictions affect goods trade?
Travel restrictions target people, but goods trade depends on people crossing borders too. Testing, quarantine and entry bans add paperwork, cause delays and increase uncertainty, all of which might raise the cost of trading across borders relative to selling domestically and thus reduce cross-border trade. Effects also differ by transport mode: road freight is exposed to queues and checks at crossings air freight lost capacity when passenger flights were cancelled (reducing ‘belly cargo’); and shipping faced stricter port and crew protocols, though containerised cargo could often keep moving with limited contact.
To quantify the trade impact, I adopt a key innovation from the recent gravity literature – including domestic trade: ie goods produced at home which are consumed domestically, (proxied by GDP minus exports) alongside international trade – to uncover (changes in) frictions to moving goods across borders (see, Yotov (2012); Yotov et al (2016)). By comparing how a country’s cross-border trade moved relative to its domestic trade, I can isolate changes in the extra costs of trading across borders.
Econometrically, the model is estimated with the standard PPML estimator and a rich set of fixed effects. Exporter-by-quarter and importer-by-quarter fixed effects absorb country-specific shocks to supply and demand (including domestic lockdown effects). Country-pair fixed effects capture time-invariant bilateral factors (distance, common language and so on). Finally, seasonal ‘border-by-quarter-of-year’ dummies remove regular seasonality in cross-border relative to domestic trade. The resulting border coefficients can be read as changes in border frictions relative to 2019.
How did border frictions evolve during the pandemic?
To estimate how border frictions moved through the pandemic, I allow the ‘border effect’, the gap between trading domestically and trading across an international border, to vary quarter by quarter by interacting a cross-border indicator with time dummies. These time-specific border coefficients are plotted below.
Chart 1: Border coefficients over time
Before Covid, estimated border frictions were broadly stable. When the pandemic hit, the model identifies a sharp, temporary increase in the ‘border cost’ for selling abroad rather than domestically. At its trough in 2020 Q2, the estimated border effect implies around a 27% decline in international trade over and above what would be predicted by the collapse in economic activity. The border friction then fell back as restrictions were relaxed, and the estimates turn temporarily positive in late 2021, implying an ‘overshoot’, as firms caught up on delayed shipments and rebuilt inventories.
How big was the trade impact of travel restrictions?
I then relate this time variation in border frictions to international travel restrictions, as captured by the Oxford Covid-19 Government Response Tracker, which ranges from no travel restrictions, up to full border closure. Including this as an explanatory variable in the gravity equation shows that even after controlling for the broader pandemic shock, tighter travel restrictions are associated with lower international trade.
I then interact travel restrictions with bilateral distance. This tests whether restrictions change trade costs mainly through a distance-invariant ‘border’ component (paperwork, checks and uncertainty at the border) rather than the per‑kilometre cost of moving goods. If so, we would expect larger percentage trade losses for nearby partners, which is exactly what the estimates show. The central estimate implies that moving to a full closure for an entire quarter reduced trade between a typical country pair (ie the trade weighted average distance between partners) by around 19%.
Importantly, the effect varies strongly across distance. The trade hit is larger for geographically closer trading relationships. That pattern fits a simple intuition that border frictions are ‘distance-invariant’, while transport costs rise with kilometres travelled. When two countries are close, distance-related costs are small, so any increase in border friction is a large percentage increase in total trade costs, and trade falls by more. In the estimates, a full closure reduces trade by roughly 27% at the 10th percentile of trading distances (around 450km), but by around 11% at the 90th percentile (around 11,500km).
Chart 2: Effect of border closures by distance
How did transport mode shape the trade hit?
Distance is only part of the story: how goods travel also matters. To explore this, I draw on UNCTAD data on the value of trade carried by sea, air, road, rail and other modes. Because the transport data are annual and do not cover domestic trade, I calculate a pre-pandemic ‘exposure’ measure: for each country pair, how intensively their 2019 trade relied on each mode.
The results are striking. Once I allow the effect of restrictions to vary with transport exposure, the trade impacts are concentrated in road and air (and the small ‘other’ category). In contrast, there is no evidence that seaborne trade was significantly reduced by travel restrictions, and rail effects are also insignificant. This helps reconcile seemingly different national experiences during Covid: for an island economy where most trade arrives by ship, even strict border measures need not translate into a large hit to goods trade, while land-transport based economies heavily reliant on trucking can face much larger disruption.
Putting distance and transport exposure together generates large cross-country differences in the implied trade cost of closing borders. The paper calculates the hit at country level. The blue dots below show the hit to air/road/other flows given by applying the coefficients on travel restrictions, and the interaction between travel restrictions and distance. By definition, the only source of heterogeneity here is differences in average distance travelled. The red dots then show the hit to total flows: this is the hit in blue dots times exposure to air/road/other flows which allows differences in transport model to play a role. This shows that implied hit to total trade from a full closure ranges from low single digits for some sea-reliant economies with distant partners (such as Australia and New Zealand) to close to 30% for the most exposed countries (such as Slovakia or Bosnia. This demonstrates how some countries were able to close their borders at a much lower cost than others.
Chart 3: Estimated hit from border closures by country
Did restrictions leave lasting scars on trade?
Did temporary border disruptions could permanently reshape trade relationships, for example by causing firms to switch suppliers or breaking logistics links? I test this by including ‘backlog’ variables that capture earlier restrictions. The evidence points away from long-run scarring. Instead, once restrictions begin to ease, trade tends to rebound strongly and temporarily ‘overshoot’, consistent with catch-up trade that makes up for earlier shortfalls. Aggregating the estimates across country pairs implies a peak hit to global goods trade of around 23% in 2020 Q2.
Chart 4: Dynamic effects
What are the broader conclusions?
Three broad lessons stand out. First, even when goods are formally exempt, restricting cross-border movement of people can raise the relative cost of selling abroad. This can happen in ways that look like a classic border friction. Second, incidence is uneven. The same policy can have very different trade consequences depending on geography and logistics: restrictions matter more for nearby trading partners, consistent with a distance-invariant ‘border’ cost making up a larger share of total trade costs at short distances, and road- and air-reliant supply chains are particularly exposed. Third, temporary disruption need not mean permanent damage. trade recovered strongly once restrictions eased, with evidence of catch-up rather than scarring.
John Lewis works in the Bank’s Centre for Central Banking Studies.
If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.
Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.