An Empirical Examination of the Effect of COVID-19 Travel Restrictions on Canadians' Cross-Border Travel and Canadian Retailers.

The coronavirus disease 2019 (COVID-19) pandemic has been devastating for many Canadian retailers. In this article, we estimate the offsetting positive effects of decreased international travel by Canadians on retail revenues. We use data from 1991 to 2021 on Canadians' travel to the United States to estimate a model of cross-border travel and establish community-level counterfactual staying rates had the pandemic not occurred. Combined with actual staying rates and elasticities of retailers' revenues with respect to staying rates, we estimate offsetting revenue gains due to the fall in cross-border travel. Our results suggest that, on average, the border closure generated a 1.49 percent offsetting gain in revenues for small Canadian retailers located within 150 kilometres of the border. We document variation across communities and sub-sectors, with estimates ranging from 0 to 125 percent. Retailers located in less-affluent communities near US shopping opportunities, and those operating in sub-sectors catering to travellers, experienced the largest gains. © Canadian Public Policy / Analyse de politiques.

Introduction

The coronavirus disease 2019 (COVID-19) global pandemic has created unprecedented economic conditions in Canada. Many of these economic implications arose as a result of public policies that encouraged or enforced mobility restrictions as a means to reduce virus transmission. In this article, we analyze the impact on the revenue of small Canadian retailers of one of these policies, the 21 March 2020 closure of the Canada–US border to non-essential travel by Canadians.

When Canadians travel across the US-Canada border to make purchases from US retailers, Canadian retailers forego those sales. Thus, the border policy prohibiting Canadians from crossing the border for non-essential travel has presumably shifted some revenue from cross-border purchases to Canadian retailers. That is, the pandemic border policies have generated offsetting revenue gains for Canadian retailers that serve to partially counter the losses those retailers may have experienced during economic shutdowns. Using geographically disaggregated data, we estimate the magnitude of such offsetting revenue gains for small retailers due to the COVID-19 border closures for Canadian communities within 150 kilometres of the border. Although we focus on COVID-19 border restrictions, our analysis contributes to an understanding of the effects on Canadian retailers of policies that limit Canadians’ cross-border expenditures more generally, such as personal exemptions and the de minimis threshold.

It is well known that Canadian retailers experienced catastrophic aggregate revenue declines in the early months of the COVID-19 pandemic. These declines were primarily the result of a combination of government-mandated or voluntary business closures and a collapse in demand as consumers avoided public spaces. As Figure 1 shows, some retailers, such as grocery stores, experienced an increase in sales, but the retail sector as a whole suffered considerably during much of 2020. In particular, in 2020 relative to the same months in 2019, retail sales in Canada were 8.8 percent lower in March, 31 percent lower in April, and 16 percent lower in May. However, the figure also demonstrates the diverse experiences of different retail sectors during the pandemic. In aggregate, retail rebounded in the latter half of 2020 and into 2021. Some sub-sectors, such as building materials, experienced large gains in the first quarter of 2021 relative to the same months in 2019. Other sub-sectors, such as gasoline stations and clothing retailers, remain well below their 2019 sales levels.

Figure 1:

Percentage Change in Sales in Overall Retail Trade and Selected Sub-Sectors, January 2020–March 2021, Relative to the Same Month in 2019

As described earlier, one potential buffer acting on Canadian retailers’ revenues is that Canadians have diverted a smaller portion of their spending to US retailers than they did pre-pandemic. Avoiding this loss of revenue from Canadians’ cross-border shopping activities may be a small bright spot in the Canadian retail landscape. This follows from the findings of Baggs, Fung, and Lapham (2018), who show that the disruption in border crossings resulting from the 11 September 2001 (9/11) terrorist attacks in the United States significantly mitigated the revenue losses that small Canadian retailers, particularly those located close to the border, could have experienced during the appreciation of the Canadian dollar from 2002 to 2007. Clearly, policies associated with the COVID-19 pandemic have created a much larger decline in border crossings, with non-essential travel at land crossings between Canada and the United States prohibited beginning in March 2020.

Source: Statistics Canada Table 20-10-0008-01.

To illustrate the magnitude of the impact of these policies on travel, Figure 2 depicts the monthly number of Canadians who traveled to the United States by automobile from January 2015 to April 2021. In addition, Table 1 presents the number of Canadian same-day and overnight travellers for each month beginning in January 2020 relative to the average over 2015–2019 for the same month. The number of travellers in January and February 2020 are close to their respective averages, but trips are significantly lower for all subsequent months. The most severe declines in cross-border travel to date occurred in April and May 2020, when the number of same-day travellers fell to about 6 percent of the previous average for those months. Overnight travellers decreased even further to around 2 percent of the monthly average. Cross-border travel recovered slightly after those months, but for the most recent data we have, April 2021, same-day travel by Canadians was still only approximately 11 percent of the monthly average, and overnight travel was about 7 percent of the monthly average.

Figure 2:

Number of Canadian Cross-Border Travellers by Automobile

Table 1:

Number of Canadian Cross-Border Travellers Relative to Five-Year Monthly Average

Month	Same Day	Overnight
January 2020	0.907	0.953
February 2020	1.050	1.075
March 2020	0.447	0.722
April 2020	0.058	0.035
May 2020	0.069	0.023
June 2020	0.087	0.031
July 2020	0.082	0.021
August 2020	0.083	0.020
September 2020	0.103	0.027
October 2020	0.108	0.035
November 2020	0.101	0.039
December 2020	0.100	0.047
January 2021	0.102	0.043
February 2021	0.111	0.045
March 2021	0.111	0.036
April 2021	0.107	0.067

Notes: The five-year monthly average is calculated over the same month from 2015 to 2019.

Source: Statistics Canada Table 24-10-0041-01.

The composition of the remaining travellers is also starkly different than before the pandemic. With non-essential travel (e.g., travel for tourism, recreation, shopping, and visiting friends or family) prohibited, the small number of remaining crossers traveled largely for essential work.[1] For example, Armenski et al. (2021) note that in 2019 and the first quarter of 2020, truck drivers accounted for 7.5 percent of arrivals, whereas from April 2020 onward, they account for approximately 50 percent of arrivals.

We begin by using regression analysis on monthly data for Canadian residents’ crossings by border post from January 1991 to April 2021. This allows us to demonstrate the quantitative importance of various factors that may affect border crossings by Canadian residents. In particular, it provides a broad picture of the relationship between border policies and Canadians’ cross-border travel over the past 30 years. It also provides some perspective on the relative magnitude of the COVID-19 border disruptions.

Next, we estimate a model of cross-border travel using pre-pandemic community-level crossing data by Canadians from January 1991 to September 2019. We use the resulting estimated parameters to estimate the number of trips that would have occurred had the pandemic not happened relative to the potential number of trips for each community. We then combine those counterfactual crossing fractions with estimated elasticities taken from Baggs et al. (2018) of a firm’s real revenue with respect to the community’s staying rate to estimate the impact on retailers’ real revenues of the COVID-19 cross-border travel restrictions imposed on Canadians.

Our results suggest that, on average, the border closure policy prevented a further 1.49 percent annual revenue loss for small retailers located within 150 kilometres of the border resulting from the dramatic fall in Canadians’ cross-border shopping in the year after the closure. We interpret these foregone revenue losses as revenue gains that partially offset the losses Canadian retailers experienced as a result of other effects of the pandemic, such as lockdowns and decreases in consumer demand.

We document the considerable variation in these offsetting revenue gains by both geography and by retail sector. Offsetting percentage revenue gains range from approximately 0 to 50 percent across communities for overall small retailers where those differences arise from variation in the proximity of a community to the border, its income, and the availability of US shopping across the border from the community. In particular, less-affluent communities that have nearby US shopping opportunities experienced larger offsetting revenue gains as a result of Canadians’ cross-border purchases than other communities. The variation across retail sub-sectors arises from differences across those sub-sectors in the responsiveness of retailers’ revenues to fluctuations in the rates at which Canadians engage in cross-border travel. Here, the average estimated percentage offsetting revenue gains calculated across communities varies between 0.98 percent and 3.71 percent.

The magnitude of these percentage offsetting revenue gains may appear to be relatively modest given the severity of pandemic-related border disruptions. To provide perspective on our estimated revenue effects, we perform an analogous counterfactual exercise for the year after the less severe border disruptions that followed the 9/11 terrorist attacks in the United States. We estimate that the average percentage offsetting revenue gains over that period for a small retailer within 150 kilometres of the border were only 0.19 percent. We also estimate that had such severe border restrictions been imposed during the peak of cross-border travel from April 1991 to March 1992, the analogous average percentage offsetting revenue gains would have been approximately 6.30 percent. These comparisons suggest that our baseline estimate of an average offsetting gain of 1.49 percent on a retailer’s revenue reveals both the severity of pandemic-induced border controls and a general downward trend in cross-border travel by Canadians over the past 30 years. With that said, for retailers operating in communities and sectors that are most exposed to fluctuations in demand as a result of cross-border travel, the offsetting revenue gains experienced since the border closure in March 2020 were quite significant.

Beyond their relevance for the effects of pandemic policies on the Canadian economy, these results inform broader policy perspectives on domestic retail protection from cross-border shopping. Policies regarding the de minimis threshold for e-commerce imports and personal duty exemptions for travellers returning to Canada from abroad are predicated on limiting cross-border shopping by Canadians to protect domestic retailers. The COVID-19 border policies prohibiting non-essential cross-border travel have effectively also prohibited in-person cross-border shopping, creating an unexpected natural experiment from which to evaluate the revenue losses Canadian retailers might otherwise have experienced. Quantifying the heterogeneous effects across communities and retailers of policies that restrict or facilitate Canadians’ ability to make cross-border purchases contributes to optimizing future policy decisions. Moreover, understanding that there will be significant geographical and sectoral variation in the impact of universal policies is helpful in assessing the actual impact of a realized or potential policy change.

In addition to policy relevance, findings from the current article contribute to the emerging field of pandemic economics and the established literatures on retailers, border policy effects, and international mobility. Regarding pandemic economics and retailers, this article relates to Keogh-Brown et al. (2010), who predict declines of roughly 6 percent of the retail component of gross domestic product in the United Kingdom, France, Belgium, and the Netherlands in response to a theoretical severe pandemic, and Sharma et al. (2020), who document and discuss the effects of COVID-19–related uncertainty on multinational retailers. Regarding international borders and the pandemic, this article is related to Boyd et al. (2018), who examine the economic implications of border closures during a theoretical pandemic in New Zealand and find that these closures had both economic costs and benefits. Focusing on the Canada–US border, this research is related to Cardoso and Malloy (2021), who examine the impact of the COVID-19 pandemic on trade between the two countries.

This article also builds on previous research on cross-border shopping and cross-border travel. Timothy and Butler (1995), Di Matteo and Di Matteo (1996), Ferris (2000), and Chandra, Head, and Tappata (2014) provide evidence that many consumers cross the border to take advantage of international price differences, and their findings suggest that cross-border trips from Canada to the United States are sensitive to the value of the Canadian dollar. We also contribute to a literature focused on the impact of cross-border shopping on retailers. Campbell and Lapham (2004), Asplund, Friberg, and Wilander (2007), and Baggs et al. (2016, 2018) find that changes in international relative prices affect retailer performance with effects that diminish with distance of the retailer from the border.

This article is organized as follows. In the following sections, we describe the data used in the empirical analysis and summarize findings from a preliminary empirical analysis of aggregate cross-border travel patterns. Next, we present the results of our counterfactual community-level border crossings analysis and provide estimates of the impact of COVID-19–related travel restrictions on Canadian retailers’ revenues. Then we consider some policy implications of our research, discuss the robustness of our results, and provide some historical context. The final section concludes.

Data

For the majority of our empirical analysis, we begin with data at the Canadian census division level (hereinafter, CD). We then disaggregate CDs into our primary unit of analysis, a community. If a CD contains a single border post, then the CD itself is a community. If a CD contains multiple border posts, it is divided into multiple communities in which a community is an area within the CD such that firms in that area share the same closest land border post. Baggs et al. (2018) provide further details regarding the use of firm-level data to facilitate the division of CDs into communities.[2] In total, there are 108 border posts, 280 CDs, and 490 communities in our data set. We use monthly data from January 1991 to April 2021.[3]

Distance Measures

A community’s border post distance is measured by the median driving distance across firms in a community to the closest border post. Median driving distances are calculated using data from Statistics Canada’s T2-LEAP database, as in Baggs et al. (2018).[4] Because travellers’ cross-border travel decisions are affected by the travel distance from their community to their border post and from their border post to the closest shopping destination in the United States, we also include a measure of distance from border posts to US shopping destinations. A community’s shopping distance is the driving distance from the relevant border post to the closest US shopping city or shopping outlet.[5] The overall distance measure for a community that we use, effective distance, is the sum of a community’s border post distance and its shopping distance.

Traveller Data

Analyzing border-crossing patterns at the community level requires detailed traveller data. We use data from Statistics Canada on the number of Canadian residents who cross and return through each border post by automobile, separated into three categories according to the length of their stay in the United States. As in Chandra et al. (2014) and Baggs et al. (2018), we also use International Travel Survey data from Statistics Canada. This database has information from a sub-sample of Canadian travellers, including their CD of residence and the length of their stay in the United States. We consider three categories of trip length: same day, overnight, and combined. On a same-day trip, a traveller crosses the border and returns within a 24-hour period; overnight trips include all trips that last longer than a same-day trip. Combined trips includes all trips with any length of stay. Over our sample, on average, 72 percent of trips by Canadian residents are same-day trips.

Following the methodology used in Baggs et al. (2018), for each category of trip length, we combine the two travellers’ data sets described earlier to estimate the monthly number of Canadian travellers emanating from each community. The estimation method accounts for the fact that some CDs have multiple border posts and some border posts serve multiple CDs. We first use firm shares to apportion surveyed travellers from each CD to communities within the CD. We then estimate the fraction of travellers through a border post that should be allocated to each community. This fraction is then used to portion actual total travellers through each border post to each relevant community, giving us estimates of the number of cross-border trips by community. Finally, we calculate community-level crossing fractions by dividing the estimated community-level number of cross-border trips by the number of potential trips (30 days multiplied by the community population). A detailed explanation of the estimation method is contained in Appendix A.

Demographic Data and Nominal Exchange Rates

We use annual estimates of Canadian population by CD from Statistics Canada and annual population estimates for our US destination shopping cities from the US Census Bureau to estimate the monthly population series by applying the linear trend between the unit’s (Canadian CD or US city) annual data points. Monthly Canadian community-level population series are estimated by multiplying monthly CD population estimates by the share of firms in a community.

We also have median income for each Canadian CD from Statistics Canada for census years 1991, 1996, 2001, 2006, 2011, and 2016. We use these to estimate monthly median income by CD by applying the linear trend between the unit’s census year observations.[6]

We use monthly averages of the bilateral nominal exchange rate between Canada and the United States from the Pacific Exchange Rate Service. We include this variable in our detailed analysis of travel patterns because previous studies have clearly demonstrated that cross-border travel is correlated with bilateral nominal exchange rates. In particular, appreciations of the Canadian dollar tend to be positively correlated with cross-border trips by Canadians, particularly same-day trips.[7]

Summary Statistics

Because cross-border trips are highly sensitive to distance, following Baggs et al. (2018), our detailed data analysis in the “Counterfactual Analysis” section restricts attention to the sub-sample of 238 communities that are located within 150 kilometres of the southern Canadian border with the United States. Moreover, as explained in that section, we undertake estimation using data from January 1991 to September 2019 so as to capture pre-pandemic travel patterns.

Table 2 reports summary statistics for key variables for the 238 communities over the 345 months between January 1991 and September 2019. We note that our measures of crossing fractions appear to be low because, as described earlier, they reflect the ratio of actual crossings to a relatively high number of potential crossings.

Table 2:

Summary Statistics

Variable	No. of Observations	Mean (SD)
Same-day crossing fraction	82,110	0.013 (0.043)
Overnight crossing fraction	82,110	0.003 (0.010)
Combined crossing fraction	82,110	0.016 (0.047)
Border post distance (km)	82,110	72.414 (43.372)
Shopping distance (km)	82,110	93.378 (77.084)
Effective distance (km)	82,110	165.792 (87.332)
Population (persons)	82,110	172,756 (380,771)
US city population (persons)	82,110	200,604 (462,243)
Median income (dollars)	82,110	24,623 (6,436)
Nominal exchange rate (US$/C$)	82,110	0.805 (0.113)

Notes: Summary statistics are based on monthly data for the sub-sample of the 238 communities within 150 km of the border from January 1991 to September 2019.

Source: Authors’ calculations.

Preliminary Analysis of Cross-Border Trips

Before presenting our detailed empirical analysis of the impact of COVID-19–related travel restrictions on Canadians’ cross-border travel and retailer revenues, we present some exploratory estimates related to cross-border travel. We briefly examine empirical regularities in the monthly number of border crossings by Canadian residents for the two categories of trip length, same day and overnight, using monthly data from January 1991 through April 2021.[8] This analysis provides a broad picture of the impact on travel of changes in border policies that have occurred over the past 30 years. It also provides some perspective on the relative magnitude of the COVID-19 border disruptions.

The regression specification in this section is motivated by previous articles such as Baggs et al. (2016, 2018), Chandra et al. (2014), Di Matteo and Di Matteo (1996), and Ferris (2000, 2010). We include bilateral nominal exchange rates and indicator variables to reflect three important changes in border policies that occurred during the period we analyze. The first change was tighter border controls after the 9/11 terrorist attacks, the second was the introduction of a passport requirement for crossers in June 2009, and the most recent was the travel restrictions imposed in March 2020 in response to the COVID-19 outbreak.

We estimate the following equation by ordinary least squares: where b denotes border post and t denotes time (year–month). The dependent variable, N, is the number of Canadian residents who travel across the border and return to Canada through border post b at time t. The regressors include the nominal exchange rate expressed as the number of US dollars per Canadian dollar, S;[9] an indicator variable that equals 1 after September 2001 (inclusive), 911; an indicator variable that equals 1 after June 2009 (inclusive), Pass; and an indicator variable that equals 1 after March 2020 (inclusive), COVID. On the basis of the approach in Baggs et al. (2018), we interact each border policy indicator with the nominal exchange rate. We include border post, month, year, province–year, and state–year fixed effects, denoted by θ, θ, θ, θ, and θ, respectively. The province–year and state–year fixed effects are included to account for changes in local economic conditions. The error term is denoted ε.

The results of estimating Equation (1) using 108 border posts are reported in Table 3. Columns 1–2 summarize results for same-day trips, and Columns 3–4 present results for overnight trips. Consistent with previous studies, we find that, before the pandemic, an appreciation of the Canadian dollar is associated with a rise in Canadian residents’ cross-border trips of all lengths. The results suggest that the border disruptions after the 9/11 terrorist attacks significantly weakened this relationship for some period of time. The positive coefficient estimate on the nominal exchange rate–passport interaction term is likely reflecting that by 2009, travellers had adjusted to some degree to the 2001 border shock rather than reflecting the impact of that particular policy change on the sensitivity of travellers to exchange rate movements. Finally, because of the severity of the pandemic-induced border controls, standard economic reasoning behind the usual expected relationship between exchange rates and number of crossers does not readily apply after March 2020. In particular, the large negative coefficient estimate on the nominal exchange rate–COVID interaction term likely reflects opposing trends between the exchange rate and the number of crossers during the Canadian dollar appreciation from May 2020 to April 2021 rather than an economically meaningful correlation between these two variables.

Table 3:

OLS Regressions of Number of Cross-Border Trips by Border Post

Variable	Same-Day Trips		Overnight Trips
	(1)	(2)	(3)	(4)
ρ0: constant	6.476***	6.674***	1.835***	2.052***
	(0.092)	(0.092)	(0.118)	(0.121)
ρ1: ln(Nominal Exchange Rate)	0.779***	2.314***	0.690***	2.370***
	(0.080)	(0.147)	(0.089)	(0.227)
ρ2: ln(Nominal Exchange Rate) × Post-9/11		−1.566***		−1.589***
Indicator (St × 911t)		(0.199)		(0.262)
ρ3: ln(Nominal Exchange Rate) × Passport		0.333*		0.169
Requirement Indicator (St × Passt)		(0.194)		(0.208)
ρ4: ln(Nominal Exchange Rate) × COVID-19		−9.148***		−12.052***
Indicator (St × COVIDt)		(0.727)		(0.708)
ρ5: Post-9/11 Indicator (911t)	−0.190***	−0.883***	−0.089*	−0.793***
	(0.072)	(0.140)	(0.053)	(0.152)
ρ6: Passport Requirement Indicator (Passt)	−0.072***	−0.062**	−0.083**	−0.105**
	(0.027)	(0.028)	(0.035)	(0.042)
ρ7: COVID-19 Indicator (COVIDt)	−2.887***	−5.603***	−3.228***	−6.806***
	(0.105)	(0.292)	(0.116)	(0.276)
Year FE	Yes	Yes	Yes	Yes
Month FE	Yes	Yes	Yes	Yes
Border post FE	Yes	Yes	Yes	Yes
Province-year FE	Yes	Yes	Yes	Yes
State-year FE	Yes	Yes	Yes	Yes
Adjusted R2	0.932	0.933	0.919	0.920
No. of observations	39,312	39,312	39,312	39,312

Notes: The data are based on 108 border posts and are monthly from January 1991 to April 2021. Robust standard errors, adjusted for clustering at the border post level, are in parentheses. OLS = ordinary least squares; FE = fixed effects.

p = 0.1;

p = 0.05;

p = 0.01.

Source: Authors’ calculations.

In addition to their moderating effects on the relationship between the exchange rate and crossers, the three restrictive border policy changes that we consider are directly associated with a decrease in the number of Canadian travellers. The severe COVID-19–related travel restrictions, unsurprisingly, have by far the largest negative effects. Using the estimated coefficients, we estimate that COVID-19 border restrictions are associated with a fall of 94 percent to 97 percent for same-day crossers and a 96 percent to 98 percent for overnight crossers, depending on the specification (calculated as for Columns 1 and 3 and for Columns 2 and 4, where ln S is evaluated at its mean).

Counterfactual Analysis of Cross-Border Travel and Retailer Revenues

Quantifying the impact of COVID-19 on Canadians’ cross-border travel requires a comparison between estimated counterfactual crossing fractions had the COVID-19 outbreak not occurred in North America and actual crossing fractions. To measure counterfactual crossing fractions, we estimate a crossing fraction equation using monthly pre-pandemic data that reflect crossing fraction patterns during normal times. We define the pre-pandemic period as January 1991–September 2019 because Asian countries may have started to be affected by COVID-19 in the fourth quarter of 2019.

Crossing Fraction Regressions

A detailed analysis of Canadian residents’ cross-border travel for use in our counterfactual predictions involves estimating the theoretical model of travel for geographically disaggregated units developed in Chandra et al. (2014) and Baggs et al. (2018). Following the latter article, we specify the following fractional probit model for estimation of crossing fractions by Canadian communities: where c denotes community and t denotes the time period.[10] The dependent variable, x, is the estimated fraction of Canadian residents’ potential trips from community c to the United States with a return to Canada at time t (see Appendix A). The additional regressors relative to Equation (1) include median income in the census division in which the community resides, I; the ratio of community population to the population of the nearest US shopping city, Rpop; and the distance between the community and the nearest US shopping location, D. We also include province, month, and year fixed effects, denoted by φ, and φ, respectively.

The results of estimating Equation (2) for trips of various lengths are reported in Table 4.[11] Columns 1 and 2 report results for same-day trips, Columns 3 and 4 reflect longer trips, and Columns 5 and 6 display results for all types of trips (combined trips). Many of our findings based on monthly community-level data are consistent with the results of Baggs et al. (2018) based on annual data, as well as with our preliminary results based on monthly number of crossings by border post contained in Table 3. In particular, crossing fractions are positively associated with Canadian dollar appreciations and negatively associated with more restrictive border policies. Moreover, the strength of the relationship between exchange rates and crossing fractions weakens in 2001 but, in line with our findings using number of crossers in the “Preliminary Analysis of Cross-Border Trips” section, we find that the relationship between crossing fractions and exchange rates regains some of its strength by 2009.

Table 4:

Fractional Probit Estimation of Community-Level Crossing Fractions

Variable	Same-Day Crossing Fractions		Overnight Crossing Fractions		Combined Crossing Fractions
	(1)	(2)	(3)	(4)	(5)	(6)
β0: Constant	9.767***	9.811***	−2.79	−2.736	6.935***	6.982***
	(2.588)	(2.586)	(2.227)	(2.226)	(2.220)	(2.219)
β1: ln(Nominal Exchange Rate) (St)	0.449***	0.800***	0.355***	0.768***	0.465***	0.836***
	(0.054)	(0.112)	(0.033)	(0.102)	(0.048)	(0.101)
β2: ln(Nominal Exchange Rate) × 9/11 Indicator (St × 911t)		−0.461***		−0.508***		0.486***
		(0.089)		(0.105)		(0.086)
β3: ln(Nominal Exchange Rate) × Passport Indicator (St × Passt)		0.073		0.081***		0.104*
		(0.072)		(0.031)		(0.062)
β4: ln(Median Income) (Ict)	−1.010***	−1.010***	0.06	0.06	−0.726***	−0.726***
	(0.296)	(0.296)	(0.239)	(0.239)	(0.252)	(0.252)
β5: ln(Relative Population) (Rpopct)	−0.171***	−0.171***	−0.061**	−0.061**	−0.153***	−0.153***
	(0.044)	(0.044)	(0.029)	(0.029)	(0.040)	(0.040)
β6: ln(Effective Distance) (Dct)	−0.245*	−0.245*	−0.086	−0.086	−0.231**	−0.231**
	(0.129)	(0.129)	(0.054)	(0.054)	(0.112)	(0.112)
β7: 9/11 Indicator (911t)	−0.043***	−0.246***	−0.032***	−0.255***	−0.041***	−0.255***
	(0.015)	(0.038)	(0.009)	(0.050)	(0.013)	(0.036)
β8: Passport Indicator (Passt)	−0.040***	−0.024***	−0.042***	−0.026***	−0.037***	−0.018***
	(0.006)	(0.006)	(0.007)	(0.007)	(0.006)	(0.007)
Provincial FE	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
Month FE	Yes	Yes	Yes	Yes	Yes	Yes
Log-likelihood	−4,872.09	−4,872.02	−1,600.68	−1,600.66	−5,933.95	−5,933.86
No. of observations	82,110	82,110	82,110	82,110	82,110	82,110

Notes: Results are for the subsample of communities within 150km from the border from January 1991 to September 2019. Robust standard errors adjusted for clustering at the census division level are in the parentheses. FE = fixed effects.

p = 0.1;

p = 0.05;

p = 0.01.

Source: Authors’ calculations.

Regarding the role of community characteristics, there is a negative relationship between community income and crossing fractions for same-day travel and a negative association between relative population and crossing fractions and between community distance and crossing fractions for trips of all lengths. The results for combined travel are mainly driven by same-day crossing fractions because same-day trips account for a large share of total cross-border travel. Our findings regarding cross-border travel patterns are all quite intuitive, and we refer the reader to Baggs et al. (2018) for a more detailed explanation of the forces behind those relationships.

Counterfactual Staying Rate Results

To examine the impact of COVID-19–related travel restrictions on cross-border travel by Canadians and the subsequent offsetting gains in revenue for Canadian small retailers, we first compare estimated counterfactual staying rates with actual staying rates. The staying rate is the fraction of potential cross-border trips that did not occur from a community and is equal to one minus the crossing fraction. Focusing on staying rates allows us to directly use estimates from Baggs et al. (2018) based on firm-level data to quantify the offsetting revenue gains for Canadian retailers resulting from the policy-induced fall in cross-border purchases by Canadians beginning in March 2020.

Our method for estimating COVID-19 counterfactual monthly staying rates for the 238 communities within 150 kilometres of the border from October 2019 to April 2021 is as follows. We generate counterfactual crossing fractions for those months using the point estimates of the coefficients from Columns 2, 4, and 6 of Table 4 with right-hand side variables set to following levels. We use the linear projections for community median income, community population, and the relevant US shopping city population as described in the “Data” section. We use the actual measure of a community’s effective distance to the nearest US shopping area. We set the 9/11 and passport indicators equal to one. We set the month and province fixed effects at their relevant estimated levels, and we set the year fixed effects for all observations equal to the estimated fixed effect for 2019. The only remaining variable is the nominal exchange rate. According to Ca’Zorzi and Rubaszek (2018), a random-walk forecast for the Canada–US bilateral nominal exchange rate outperforms other relatively standard forecasting methods at the forecasting horizons we use in this article (between one and 15 months ahead).[12] Hence, using the random-walk forecast, we set the nominal exchange rate at its September 2019 level of 0.755 for each subsequent month in our baseline counterfactual analysis.

After generating counterfactual estimates of crossing fractions, we measure counterfactual staying rates as one minus those counterfactual crossing fractions. We denote our monthly counterfactual estimates of the staying rate for community c at time t for October 2019 to April 2021 as . We generate predicted staying rates from April 2014 to September 2019 using in-sample predicted crossing fractions based on point estimates from Table 4. We also denote these in-sample predicted staying rates as for community c for April 2014 to September 2019. Henceforth, we refer to this variable as the predicted–counterfactual staying rate.

We also calculate the actual staying rate series for each community and denote this as v, where v, where x is the observed crossing fraction for community c at time t. In the final steps, we first calculate percentage differences between actual staying rates and predicted–counterfactual staying rates for each community from April 2014 to April 2021: Finally, to account for systematic prediction errors, we construct adjusted counterfactual percentage differences in staying rates for each community for October 2019 to April 2021 as follows: Here is the median percentage difference between predicted and actual staying rates over 2015 to 2019 for the same month as in period t if the month is between January and September and the median over 2014 to 2018 if the month is between October and December. We refer to δ as the counterfactual percentage difference in staying rates based on point estimates. These are our primary estimates of the percentage changes in community-level staying rates from March 2020 to April 2021 as a result of the COVID-19 pandemic. We also report parallel measures using 95 percent confidence interval bounds in place of point estimates for the coefficients from the crossing fraction regressions to allow for sampling uncertainty.[13]

To further clarify our method for estimating counterfactual percentage differences in staying rates, Figure 3 depicts actual and predicted–counterfactual staying rates for combined trips for one example community in our sample, along with 95 percent confidence interval bounds for the predicted–counterfactual rates. The example community is the combined census division of Fraser Valley, British Columbia, and border post Aldergrove, British Columbia. Referring to the figure, consider the observations for March 2020. For our main measure based on point estimates, we take the percentage difference between the solid line and the heavy dashed line for that month and subtract the median distance between those two lines for the March observation across the previous five years. This gives our measure of δ based on point estimates for that community for March 2020. Appendix C presents analogous graphs for an example community from each of the remaining six provinces. Those figures illustrate the heterogeneity across communities with respect to levels of staying rates and the magnitude of prediction errors.

Figure 3:

Example Community Staying Rates in Fraser Valley–Aldergrove, British Columbia: March 2014 to April 2021

Source: Statistics Canada International Travel Survey and authors’ calculations.

In what follows, we focus on two time periods in reporting our results regarding the impact of COVID-19 travel restrictions: the third quarter of 2020 when travel rates are historically at their peak during each year and the full year from April 2020 to March 2021 when COVID-19 border restrictions were in place. We use our monthly actual, predicted, and counterfactual staying rates to construct monthly staying levels, aggregate those to quarterly and annual frequencies, and then use those to construct quarterly and annual actual, predicted, and counterfactual staying rates.[14] We then use those and the methods described earlier to construct quarterly and annual estimated counterfactual percentage differences in staying rates for each community in our sample.

Table 5 presents summary statistics for our estimated counterfactual percentage differences for staying rates for the third quarter of 2020 across the communities in our sample. Figure 4 provides a visual presentation of the results based on point estimates, allowing us to present variation within and across provinces in our estimated impact on staying rates of COVID-19-related travel restrictions for that time period.

Table 5:

Counterfactual Percentage Differences in Staying Rates: Third Quarter 2020

Based on	Mean (SD)	Median	Minimum, Maximum
Point estimates (%)
Same day	0.775 (2.924)	0.062	−0.038, 32.792
Overnight	0.467 (1.588)	0.073	−0.000, 18.446
Combined	1.253 (4.098)	0.209	−0.022, 46.745
95% CI lower bounds (%)
Same day	0.810 (3.088)	0.020	−0.319, 35.089
Overnight	0.480 (1.603)	0.090	−0.001, 18.521
Combined	1.290 (4.292)	0.165	−0.106, 49.633
95% CI upper bounds (%)
Same day	0.769 (2.862)	0.072	−0.005, 31.653
Overnight	0.463 (1.581)	0.071	−0.000, 18.412
Combined	1.243 (4.002)	0.222	−0.008, 45.122

Notes: Statistics are calculated across the subsample of communities within 150 km of the border. CI = confidence interval.

Source: Authors’ calculations.

Figure 4:

Counterfactual Staying Rate Percentage Differences, Third Quarter 2020

We first note that there is significant variation in estimated staying rate counterfactual percentage differences across communities, and a handful of communities have negative differences. The figure suggests that cross-border travel by residents of Alberta and Saskatchewan is relatively unaffected by the travel restrictions. In contrast, British Columbia, Ontario, and Quebec house communities in which there were very large predicted effects. As expected from the results of Table 4, the estimated impact on staying rates of the border closure tends to be larger for relatively less-affluent communities that reside near more favorable US shopping opportunities. In particular, the cross-community Spearman rank correlation coefficients between estimated changes in staying rates and community characteristics are −0.305 for effective distance, −0.254 for median income, and −0.153 for relative population. All of these correlation coefficients are significant at the 2 percent level or below.[15] We also note the considerable variation in our estimates across the three categories of trip length. As expected, we estimate larger effects on same-day trips that are reflected in combined trips. For the remainder of this article, we focus on combined trips.

To provide further information on geographical variation in combined staying rates counterfactual percentage differences, we aggregate our results to the census division level for mapping purposes. Figure 5 depicts the geographic variation in the size of those differences for the third quarter of 2020 across the 126 Canadian census divisions within 150 kilometres from the border. According to our estimates, the darkest census divisions on that map will experience the largest increases in their staying rates during this quarter, with increases of more than 3 percent. These are communities for which we typically estimate counterfactual crossing rates of more than 3 percent, which is a very high fraction considering that crossing fractions are measured with a base of total monthly potential same-day trips equal to 90 times the population of the census division. For our example census division, Fraser Valley, British Columbia, the estimated rise in its staying rate represents a fall of approximately 324,600 cross-border travellers during the third quarter of 2020 as a result of COVID-19–related travel restrictions. We estimate that the census division with the largest fall in travellers over this period, Greater Vancouver, British Columbia, experienced a decrease of approximately 1.6 million travellers. The gains in retail sales for firms located in these types of census divisions as a result of those would-be Canadian travellers who instead stay and make purchases in their local community can be significant. In the next section, we provide estimates of those offsetting revenue gains.

Figure 5:

Combined Staying Rate Estimated Counterfactual Differences (%): Third Quarter 2020

Counterfactual Percentage Offsetting Retailer Revenue Gains

The direct effect of the COVID-19 pandemic on most Canadian retailers was negative. In this section, our primary objective is to estimate partial offsetting revenue gains that some of those retailers experienced as a result of the dramatic decrease in cross-border travel by Canadians facing COVID-19 border closures. In particular, for each community within 150 kilometres of the border, we seek to estimate firm-level percentage offsetting revenue gains: actual revenues minus counterfactual estimates of what revenues would have been had the COVID-19 border restrictions not been imposed, expressed as a percentage of the latter. We now explain our approach to estimating those offsetting revenue gains.

We first aggregate our staying rate counterfactual percentage differences to annual percentage differences for the year from April 2020 to March 2021 because border restrictions were put in place in the middle of March 2020. We focus on annual rates because our method uses the estimated firm-level elasticities for real revenue with respect to combined staying rates as reported in Baggs et al. (2018), based on annual data. Baggs et al. provide estimates of revenue elasticities for small retailers for a retail aggregate and for a set of retail sub-industries.[16] We multiply those estimated elasticities by the annual counterfactual percentage staying rate differences for each community to generate estimates of percentage offsetting revenue gains to Canadian retailers from restricting cross-border travel by Canadians from April 2020 to March 2021

Table 6 provides summary statistics for those estimated percentage offsetting gains in retailer revenues across the 238 communities in our sample within 150 kilometres of the border for overall small retailers and for a selection of retail sub-industries. For all small retailers, the estimated percentage offsetting revenue gains based on point estimates of crossing fractions vary from −0.440 percent to 50.085 percent, and the average is 1.487 percent.[17] The table also reports analogous measures that use 95% confidence interval bounds for crossing fractions to measure estimated staying rate differences.

Table 6:

Estimated Percentage Offsetting Gains in Canadian Retail Revenues from COVID-19 Border Closures

Based on	Mean (SD)	Median	Minimum, Maximum
Point estimates (%)
All small retailers	1.487 (4.639)	0.208	−0.440, 50.085
Gasoline service stations	3.707 (11.561)	0.519	−1.096, 124.828
Accommodations	2.959 (9.229)	0.415	−0.875, 99.646
Apparel and general retail	1.854 (5.782)	0.260	−0.548, 62.429
Furnishings and appliances	1.737 (5.416)	0.243	−0.514, 58.479
Sporting and hobby goods, books	0.975 (3.042)	0.137	−0.288, 32.835
95% CI lower bounds (%)
All small retailers	1.500 (4.883)	0.151	−1.122, 53.227
Gasoline service stations	3.737 (12.170)	0.377	−2.796, 132.659
Accommodations	2.983 (9.715)	0.301	−2.232, 105.897
Apparel and general retail	1.869 (6.087)	0.189	−1.398, 66.346
Furnishings and appliances	1.751 (5.702)	0.177	−1.310, 62.148
Sporting and hobby goods, books	0.983 (3.201)	0.100	−0.736, 34.895
95% CI upper bounds (%)
All small retailers	1.489 (4.533)	0.219	−0.140, 48.356
Gasoline service stations	3.712 (11.298)	0.545	−0.348, 120.518
Accommodations	2.963 (9.018)	0.435	−0.278, 96.206
Apparel and general retail	1.856 (5.650)	0.272	−0.174, 60.274
Furnishings and appliances	1.739 (5.293)	0.255	−0.163, 56.460
Sporting and hobby goods, books	0.976 (2.972)	0.143	−0.092, 31.701

Notes: Estimated percentage offsetting revenue gains reflect actual revenues minus counterfactual estimates of what revenues would have been had the COVID-19 border restrictions not been imposed, expressed as a percentage of the latter. Calculations are for the year April 2020 to March 2021. Counterfactual crossing fractions are based on crossing regression results using data from January 1991 to September 2019. Statistics are calculated across the sample of communities within 150 km of the border. COVID-19 = coronavirus disease 2019; CI = confidence interval.

Source: Authors’ calculations.

Figure 6 depicts the geographic dispersion of the estimates for all small retailers across census divisions. The figure suggests that retailers in communities closer to the border experienced larger offsetting revenue gains, particularly in British Columbia, New Brunswick, and Ontario. More generally, more affluent communities located relatively far from advantageous US shopping centers tend to have smaller offsetting revenue gains. In particular, the cross-community Spearman rank correlation coefficients between estimated annual offsetting revenue gains and community characteristics are −0.327 for effective distance, −0.316 for median income, and −0.134 for relative population. The correlation coefficients for effective distance and median income are significant at the 1 percent level, whereas the relative population coefficient is significant at the 4 percent level.[18]

Figure 6:

Estimated Percentage Offsetting Gains in Retailers’ Revenues, April 2020–March 2021

Table 6 also demonstrates considerable variation in estimated offsetting revenue gains across sub-industries. The impact is larger for sub-industries that are known to be particularly exposed to fluctuations in demand as a result of cross-border travel and tourism. In particular, the offsetting gains are largest for gasoline stations and accommodations, which are two industries for which we expect a significant level of expenditures by traveling Canadians on US service providers. Moreover, in contrast to the other sub-industries in the table, there are no enforceable Canadian restrictions on the amount of those services that Canadians are able purchase while outside of Canada. Nonetheless, our estimates also suggest significant offsetting gains in the more goods-oriented sub-industries in our sample. We estimate offsetting percentage revenue gains of similar magnitude for apparel and general retail and furnishings and appliances and smaller gains for sporting and hobby goods and books.

Overall, our estimates can be thought of as representing the relatively modest buffer that the closure of the Canada–US border to non-essential travel offers Canadian retailers during COVID-19–related economic disruptions. By limiting Canadians’ ability to engage in cross-border shopping and tourism, this closure is especially relevant to retailers located in less-affluent communities close to good US shopping opportunities and those in sectors particularly exposed to fluctuations in demand because of cross-border travel.

Policy Implications

A growing body of work suggests that low-income households and communities have been more affected by many COVID-19–related policies than their more affluent counterparts. Lemieux et al. (2020) and Gallacher and Hossain (2020), for example, note that poorer workers were more likely to experience employment loss during the early months of the pandemic. Our work indicates that retailers in less affluent communities should have experienced larger offsetting revenue gains as a result of the border restrictions, which is a positive effect. However, if this effect arises because lower-income consumers are more likely to participate in cross-border shopping, these retailer gains may have come at the expense of lower-income consumers foregoing cost savings on goods purchased from US retailers. This suggests that less affluent consumers may be more adversely affected by the border policies associated with COVID-19.

In addition to contributing to our understanding of the effects of pandemic-related border policies on retailers, our analysis also informs discussion around two other specific policies: the personal duty-free exemption for Canadian residents returning from foreign travel and, more indirectly, the de minimis threshold for postal and courier imports. Both of these policies act to insulate domestic retailers from cross-border shopping. Our results look specifically at retailers and do not consider the direct effect of cross-border shopping restrictions on consumers or the broader economy. Without these perspectives, it is difficult to suggest any particular change to either policy. However, our work does inform policy-makers regarding how universal border policies will affect regions and retailers differently depending on their characteristics. Our methodology also provides insight into how policy-makers might assess regional variation in the impacts associated with potential changes to policies associated with cross-border shopping.

We first consider the personal duty-free exemption policies. For Canadian travellers, these exemptions were increased in 2012, allowing returning residents to bring up to $200 of goods into Canada without paying taxes or duties if they were away for 24–48 hours and up to $800 if their absence exceeded 48 hours. There is no exemption for absences of less than 24 hours. The COVID- 19 border policies analyzed in this article effectively created a situation in which there has been only negligible in-person cross-border shopping. Hence, we can consider Canadian retailers offsetting revenue gains in this situation as analogous to what might occur if duties and taxes applied at the border, and the enforcement thereof, were sufficiently high so as to almost completely deter in- person cross-border shopping.[19] Our results suggest that the offsetting revenue gains are relatively modest on average, but much larger for retailers closer to the border and those in sub-sectors particularly prone to cross-border shopping. They are also larger for retailers in lower-income communities.

Understanding this heterogeneity in how border communities and retailers are affected by the personal exemption threshold, and enforcement of that policy, will contribute to optimizing future policy decisions that restrict or facilitate Canadians’ ability to engage in cross-border shopping. Differentiating border policies such as the duty-free exemption by retail sector or community of residence would be impractical at best. However, understanding that there will be significant variation in the impact of universal policies is helpful in assessing the actual impact of a realized or potential policy change. For example, our results highlight the larger revenue gains accrued by retailers in lower-income communities from the pandemic-related border closures. This suggests equity considerations in policies such as the duty-free exemption that would be obscured by considering only aggregate-level effects.

A second policy focused on cross-border shopping is the de minimis threshold. From a consumer perspective, this threshold determines the dollar value of e-commerce imports (although technically also mail-in and phone orders, etc.) that can enter Canada without application of duties or taxes. For Canadian retailers, the de minimis acts as domestic protection from foreign competition, but it also raises the cost of purchasing intermediate goods in some cases. Historically, Canadian brick-and-mortar retailers have generally opposed increases in the de minimis, and consumer groups and foreign and domestic e-tailers have been in favour of higher thresholds.[20] In May 2020, the Canada–United States–Mexico Agreement increased the de minimis threshold for courier imports from the United States and Mexico from $20 to $40 for taxation and to $150 for customs duties. However, the de minimis threshold for postal imports remained at $20 for both taxes and duties.

Latipov, McDaniel, and Schropp (2017) provide evidence that Canada’s historically low de minimis threshold has been inefficient and costly for the Canadian economy, including retailers in aggregate. The analysis in this article suggests that these aggregate effects may obfuscate subsets of retailers that experience particularly large consequences of increased e-commerce imports. At first glance, this seems particularly important because the pandemic disruptions experienced by retailers have occurred concurrently with the policy changes affecting the de minimis thresholds for courier imports. Although our results speak directly to in-person cross-border shopping, they reflect only indirectly on policies relating to e-commerce imports. The potential heterogeneity in the effect of changes to the de minimis is an interesting area for future study.

Robustness, Historical Comparisons, and Caveats

In our first robustness exercise, we predict counterfactual crossing fractions based on crossing-rate regressions using data that begin in 2010 instead of 1991. In particular, we use fractional probit regression on a variation of Equation (2) that does not include indicator variables for September 2001 or June 2009. Our estimation results are reported in Table D.1 in online Appendix D, and the resulting estimated offsetting retailer revenue gains for April 2020 to March 2021 are reported in Table D.2. The results for offsetting revenue gains are very similar to those from our primary specification.

In a second robustness exercise, we estimate offsetting retailer revenue gains for April 2020 to March 2021 under alternate exchange rate paths. In our baseline specification, we set the nominal exchange rate to its September 2019 value for each month after that date on the basis of best practices for forecasting the Canada–US nominal exchange rate. Historically, however, the Canada–US nominal exchange rate has varied significantly in some years. For example, during the past decade, the largest annual Canadian dollar appreciation vis-à-vis the US dollar was 23 percent in 2010, and the largest depreciation was 17 percent in 2015. Hence, here we estimate offsetting revenue gains between April 2020 and March 2021 for continual monthly changes consistent with a 10 percent annual rate of nominal exchange rate appreciation and a 10 percent annual rate of depreciation. The estimated percentage offsetting gains in retailers’ revenues under those scenarios are reported in Table 7. Comparing those results with our baseline specification for all small retailers, we estimate percentage offsetting gains in revenues, which are 12.64 percent larger if the Canadian dollar had appreciated at an annual rate of 10 percent over that time period and which are 12.44 percent smaller if it had experienced a 10 percent annual rate of depreciation.

Table 7:

Estimated Percentage Offsetting Gains in Canadian Retail Revenues from COVID-19 Border Closures with Alternate ER Paths

Point Estimates (%) Based on	Mean (SD)	Median	Minimum, Maximum
10% annual nominal ER appreciation
All small retailers	1.675 (4.735)	0.336	−0.205, 50.918
Gasoline service stations	4.174 (11.801)	0.838	−0.511, 126.902
Accommodations	3.332 (9.420)	0.669	−0.408, 101.302
Apparel and general retail	2.087 (5.902)	0.419	−0.255, 63.467
Furnishings and appliances	1.955 (5.529)	0.393	−0.239, 59.451
Sporting and hobby goods, books	1.098 (3.104)	0.220	−0.134, 33.380
10% annual nominal ER depreciation
All small retailers	1.302 (4.553)	0.121	−0.947, 49.248
Gasoline service stations	3.244 (11.347)	0.300	−2.361, 122.739
Accommodations	2.589 (9.058)	0.240	−1.885, 97.979
Apparel and general retail	1.622 (5.675)	0.150	−1.181, 61.385
Furnishings and appliances	1.520 (5.316)	0.141	−1.106, 57.501
Sporting and hobby goods, books	0.853 (2.985)	0.790	−0.621, 32.286

Notes: Estimated percentage offsetting revenue gains reflect actual revenues minus counterfactual estimates of what revenues would have been had the COVID-19 border restrictions not been imposed, expressed as a percentage of the latter. Calculations are for the year April 2020-March 2021. Counterfactual crossing fractions are based on crossing regression results using data from January 1991 to September 2019. Statistics are calculated across the sample of communities within 150 kilometres of the border. COVID-19 = coronavirus disease 2019; ER = exchange rate.

Source: Authors’ calculations.

To provide historical perspective, we undertake two additional counterfactual exercises, with methodological details presented in online Appendix D. First, we undertake an exercise aimed at estimating the counterfactual impacts that would have resulted had a similar level of border restrictions been imposed during the peak period of cross-border travel during the 1990s. For reference, Figure D.1 in online Appendix D depicts actual and predicted staying rates for the example community, Fraser Valley–Aldergrove, British Columbia, for the span of our time series, January 1991 to April 2021, and illustrates the relatively low staying rates in the 1990s.

With that period in mind, we consider a counterfactual in which we allow community-level crossing rates for each month from April 1991 to March 1992 to fall to the levels observed for the relevant month from April 2020 to March 2021 and use analogous methods to those preceding to estimate the resulting counterfactual offsetting revenue gains for Canadian retailers. We estimate that the average offsetting revenue gains for a small retailer would have been 6.308 percent, with maximum offsetting gains for a gasoline station retailer of 15.722 percent.[21] These can be compared to our earlier estimates for April 2020 to March 2021 of 1.487 percent for all small retailers and 3.707 percent for gasoline stations. This suggests that the relatively modest offsetting revenue gains we estimate due to pandemic-related border restrictions partially reflect the general downward trend in cross-border travel by Canadians over the past 30 years as a result of a variety of forces, including improvements in shopping opportunities in Canada and fluctuations in the value of the Canadian dollar.

In a second exercise aimed at a historical comparison, we perform a directly analogous counterfactual exercise as in our baseline but for a scenario in which the fall in border crossings that resulted after the 9/11 terrorist attacks did not occur. We focus on the 12-month period from October 2001 to September 2002. Our estimates indicate that over that period, the average offsetting revenue gains for a small retailer were 0.193 percent, with maximal gains on the order of 0.481 percent for gasoline retailers. As in the previous historical exercise, these offsetting gains should be compared with our estimates for the year April 2020 to March 2021 of 1.487 and 3.707, respectively. Hence, according to our estimates, the offsetting percentage revenue gains for the year after the border restrictions imposed in March 2020 were approximately seven times as large as those experienced during the year after the 9/11 terrorist attacks.

Finally, we note some potential limitations of our analysis. First, we note that it is possible that Canadians’ physical cross-border shopping might have been partially replaced by cross-border e-commerce purchases. However, as shown in Figure 7, e-commerce imports remained below 2 percent of total retail sales in Canada in 2020 and below their 2019 levels in most individual months. For example, e-commerce imports were roughly $760 million, or 1.32 percent of Canadian total retail sales, in May 2019 and $667 million, or 1.44 percent of sales, in May 2020.[22] The shifts in potential physical cross-border shoppers are much larger: in May 2019 there were 1.9 million same-day trips by Canadian residents to the United States, whereas in May 2020 there were only 0.1 million same-day trips—a 93 percent decline.[23]

Figure 7:

Domestic versus Imported Canadian E-Commerce

However, perhaps not surprisingly, Canadian domestic e-commerce purchases did increase significantly during 2020. Aston et al. (2020) indicate that domestic retail e-commerce sales in Canada increased by 99.3 percent from February to May 2020, reaching 6.6 percent of total retail sales in the first half of 2020 compared with 4 percent in 2019. This trend may affect our estimates of the geographical dispersion of the impact on retailers’ revenues of the COVID-19 travel restrictions. It may also affect our estimates if domestic e-commerce is concentrated in larger firms. In principle, a switch to either domestic in-store purchases or domestic e-commerce implies Canadian consumers diverting a smaller portion of their spending to US retailers and offsetting revenue gains for Canadian retailers. However, if Canadians switched from cross-border shopping to domestic e-commerce at larger firms rather than to domestic purchases (either in store or online) at the small retailers we are considering, our estimated revenue gains for small retailers may be overstated.

A second limitation is that we are only considering the effect of Canadian travellers; because of data limitations, we are unable to include US travellers. Of course, US travellers have also been prevented from crossing the border into Canada for non-essential purposes, which has presumably led to revenue decreases for Canadian retailers who cater to those visitors. While acknowledging this, we note that the number of US travellers who cross the border is significantly lower than the number of such Canadian travellers. Figure 8 demonstrates this phenomenon.

Figure 8:

Canadian and US Cross-Border Trips

We also note that US visitors play a much more limited role as shoppers in the Canadian retail landscape than do Canadians shopping in the United States. Specifically, following Li (2012) and Government of Canada (2011), we note that retail prices are consistently significantly lower in the United States than in Canada, and the diversity and variety of products are higher. In addition, unlike Canada, the US population is not concentrated along the border, suggesting longer travel times and higher travel costs for potential shoppers. Moreover, we expect that the smaller number of US travellers are generally motivated by travel purposes other than shopping, such as tourism. This suggests that revenue losses that Canadian retailers may have experienced emanating from limitations on Americans’ travel are likely to be most prominent in the hospitality sectors.

Source: Statistics Canada Table: 20-10-0072-01.

Conclusions

For most retailers, the economic circumstances created by the COVID-19 pandemic have been extremely difficult. In this article, we note one policy that has provided some indirect, but very modest, relief for Canadian retailers. By restricting cross-border travel, the closure of the Canada–US border to non-essential travel has acted to divert at least some retail spending that would have gone to US retailers if the border was open to Canadian retailers.

We use detailed travel data on Canadians to estimate counterfactual crossing rates from March 2020 to April 2021 in the absence of the pandemic, and we pair these data with revenue elasticities for Canadian retailers to estimate the magnitude of this effect. Our results suggest that the border closure does act to partially insulate small retailers, particularly those close to the border, from the other negative shocks associated with COVID-19. Our findings also indicate that the degree of insulation varies significantly by industry and by local conditions. The effect is larger for communities that were more exposed to fluctuations in demand as a result of cross-border travel before the pandemic such as communities with lower income levels and those with nearby superior shopping opportunities across the border.

Overall, our results suggest that the diversion to domestic spending by Canadians arising from travel restrictions offers a relatively modest buffer from the economic disruptions caused by COVID-19 for Canadian small retailers in most communities. However, margins are typically very low in retail firms, especially among the small retailers in this study, so an offsetting increase in revenues of 1.5 percent is not trivial, especially during a time of overall falling revenues as a result of other effects of the pandemic. Moreover, we estimate offsetting revenue gains above 5 percent for several retail industries because of their industry and location.

Future research on the economic impact of COVID-19–related travel restrictions would include analysis of incoming travellers into Canada, particularly from the United States. The International Travel Survey data also include some information on purpose of trip and expenditures. These components of the data could be exploited in future work. Using data on the number of retailers operating in industries and firm-level employment, revenue, and profit for retailers of all sizes for more recent years, especially 2020, will be important when those data become available.

The pace of change in economic conditions experienced during 2020 and 2021 is unprecedented. This rapid transformation creates considerable uncertainty for any prediction or counterfactual analysis, including our own. Ex post analysis of the COVID-19 period is sure to offer refinement and nuance that we are currently unable to model.

Click here for additional data file.

Table B.1:

Number of Canadian Travellers by Border Post

Variable	No. of Observations	Mean (SD)
Same-day trips (persons)	39,312	21,891.23 (49,710.51)
Overnight trips (persons)	39,312	7,762.41 (19,129.80)

Notes: Variable consists of Canadian residents returning from the United States, by automobile, on the same day, after one night, and after two or more nights. “Overnight” is the sum of one night and two or more nights. The data are based on 108 border posts and are monthly from January 1991 to April 2021.

Source: Statistics Canada Table 24-10-0041-01