The effects of COVID-19 vaccines on economic activity

Abstract

This paper empirically examines the economic effects of COVID-19 vaccine rollouts using a cross-country daily database of vaccinations and high-frequency indicators of economic activity—nitrogen dioxide (NO₂) emissions, carbon monoxide (CO) emissions, and Google mobility indices—for a sample of 46 countries over the period December 16, 2020 to June 20, 2021. Using surprises in vaccines administered, we find that an unexpected increase in vaccination per capita is associated with a significant increase in economic activity. We also find evidence for nonlinear effects of vaccines, with the marginal economic benefits being larger when vaccination rates are higher. Country-specific conditions play an important role, with lower economic gains if strict containment measures are in place or if the country is experiencing a severe outbreak. Finally, the results provide evidence of spillovers across borders, highlighting the importance of equitable access to vaccines across nations.

1 Introduction

Since the beginning of the COVID-19 pandemic, countries have been forced to put in place stringent non-pharmaceutical interventions (henceforth referred to as containment measures) in order to limit the spread of the virus. But these containment measures have come at enormous economic costs, resulting in unprecedented economic losses (Carvalho et al. 2020; Coibion et al., 2020; Deb et al., 2020b; IMF, 2020a, 2020b), despite wide-scale fiscal measures launched worldwide to mitigate some of these losses (Deb et al., 2021d). With the advent of vaccines, the focus of countries has shifted towards vaccinating their populations against the Coronavirus (SARS-nCOV-2) as quickly as possible, in an effort to raise immunity against the virus and ease containment measures, thereby helping their economies recover.

Evidence from the epidemiological literature has already established the effectiveness of COVID-19 vaccines in reducing virus transmission, curbing severe infections and hospitalizations, and lowering fatalities (Dagan et al., 2021; Polack et al., 2020, Voysey et al. 2021). However, with wide access to vaccines having only picked up since early 2021, there is thus far little empirical evidence on the effects of vaccine rollouts on economic activity in a cross-country setting (see for example, Deb et al., 2021b for an analysis based on regional data for a more limited set of countries). In this context, this paper complements our analysis of the health impact of vaccinations (Deb et al., 2021a), by providing an empirical assessment of the effects of COVID-19 vaccinations on high-frequency proxies of economic activity for a sample of 46 countries over the period December 16, 2020 to June 20, 2021. It then goes on to study potential nonlinearity in the economic effect of vaccines, including depending on the initial rate of vaccination and other country-specific conditions, such as the stringency of containment measures or the severity of the COVID-19 outbreak. Finally, the paper examines how vaccinations and new COVID-19 cases in a country’s main trading partners can affect its local economic activity through economic inter-linkages.

For this purpose, we assemble a daily database on high-frequency indicators of economic activity—nitrogen dioxide (NO₂) emissions, carbon monoxide (CO) emissions, and Google mobility indices. We use NO₂ emissions as our baseline measure for economic activity as it is most highly correlated with lower frequency measures of economic activity like industrial production and PMI indices. Economic activity data is combined with daily data on COVID-19 vaccines administered per capita (one or two doses), confirmed COVID-19 infections and containment measures. We rely on high-frequency identification to establish causality, controlling for lagged effects of economic and health variables, and accounting for expectations about the country-specific evolution of the pandemic using a set of controls that can affect future infections and economic activity—non-pharmaceutical interventions (NPIs) such as containment measures, enhanced testing, contact tracing, and public information campaigns aimed at increasing social awareness—and country-specific time trends. To further identify the causal link between COVID-19 vaccines and economic activity, we construct a novel measure of surprises in vaccines administered. The variable is computed by taking the difference between the share of population that is fully vaccinated and the predicted share that is expected to be fully vaccinated (see Data section) based on production and procurement of vaccines.

We find that surprises in vaccinations have significant economic effects: a 10% of population (unexpected) increase in vaccine is associated with an increase in daily per capita NO₂ emissions of about 0.3 standard deviation (an increase of almost 30% relative to its median). To better put this result in perspective, this is broadly equivalent to going from a full lockdown (stringency index of 1) to containment measures equivalent to a stringency level of 0.5. This comparison, however, likely underestimates the economic gains from higher vaccination rates as we find evidence for nonlinear effects of vaccines, with the marginal economic benefits being larger when vaccination rates are higher. Similar positive effects are obtained for the impact of vaccinations on mobility indicators and CO (which is only significant with a lag).

In addition, we find that country-specific conditions play an important role in determining the economic impact of vaccines. Economic gains are lower if strict containment measures are in place, as they constrain economic activity even with vaccinations picking up. Similarly, economic gains are lower if the country is experiencing a severe outbreak during the vaccine rollout as people continue to voluntarily socially distance till cases come down significantly.

Finally, we find evidence of spillovers across borders: an increase in COVID-19 cases in trading partner countries results in a slowdown in domestic economic activity due to spillovers from trade. Furthermore, an increase in COVID-19 vaccinations in the main trading partner countries has a positive and statistically significant effect on domestic economic activity. These results highlight the importance of equitable and speedy access to vaccines across nations, as higher vaccination rates in trading partners not only improve health outcomes in partners (Deb et al., 2021a), but are also likely to improve economic conditions domestically.

Our paper contributes to two main strands in the literature. The first is that which studies the economic effects of COVID-19 vaccines and remains somewhat limited. Sandmann et al. (2021) examine the potential health and economic value of COVID-19 vaccinations in the UK and find that introducing vaccinations leads to a reduction in community transmission and incremental monetary gains from a health-care perspective. Deb et al. (2021b) employ a regional database of 17 countries (326 states) to study the impact of COVID-19 vaccinations on economic activity proxies—night-time lights, aerosol optical depth (AOD) emissions, and mobility. They find that vaccine deployment has persistent positive effects on the level of economic activity. Agarwal and Gopinath (2021) propose a cost–benefit analysis for an expedited rollout of vaccines in an equitable manner across all countries, and find that while vaccinating 40 percent of the world’s population by 2021 could cost around $50 billion, its engendered benefits could reach about $9 trillion in economic gains. This paper contributes to this literature by: (1) extending on Deb et al. (2021b) to examine the effects of surprises in vaccines administered on economic activity for 43 countries; (2) studying the role of country-specific conditions in amplifying/dampening the effects of vaccine surprises; and (3) examining the impact of COVID-19 cases and vaccines in main trading partners on a country’s own economic activity levels.

This paper also contributes to the literature which uses high-frequency indicators to proxy economic activity. Fernández-Villaverde and Jones (2020) and Sampi et al. (2020) establish that Google mobility data is an adequate proxy for economic activity by finding a high correlation between GDP and mobility data. Deb et al., (2020a, 2020b) use Google mobility indicators to capture the economic impact of the COVID-19 pandemic. Lin and McElroy (2011) show that variation in NO₂ emissions in China resembles its GDP growth during and after the Global Financial Crisis. Deb et al. (2020b) quantify the economic costs of the COVID-19 pandemic using nitrogen dioxide (NO₂) emissions and estimate losses in NO₂ emissions 30 days after the implementation of containment measures to be equivalent to about a 15% loss in industrial production. This paper contributes to this strand of the literature by putting together a novel database of daily high-frequency indicators of NO₂ emissions, CO emissions, and Google mobility indicators for 43 countries, to examine how they are affected by surprises in vaccines administered.

The paper is structured as follows. Section 2 describes the data and Sect. 3 the methodological approach. Section 4 discusses the results on effects of vaccinations on economic outcomes, the role of country-specific factors, and the effects of COVID-19 cases and vaccines in main trading partners on a country’s own economic activity. The last section concludes.

2 Data

Our empirical analysis relies on a comprehensive country-level database of daily COVID-19 cases and vaccinations, high-frequency proxies of economic activity (NO₂, CO, and mobility), and government responses to the pandemic in the form of different non-pharmaceutical interventions. Appendix Table 7 provides a summary of the data used.

2.1 COVID-19 related variables

COVID-19 vaccines data is sourced from the Our World in Data COVID-19 repository.¹ Vaccines data is disaggregated by first and second shots, with data covering up to 202 countries starting in December 2020.

COVID-19 cases: Daily data on COVID-19 cases is collected from the COVID-19 Data Repository by the Center for Systems Science and Engineering (CSSE) at Johns Hopkins University.² Coverage begins from January 22, 2020 for 208 countries.

Expected vaccinations: Data on expected vaccination rollout is taken from Airfinity, a science information and analytics company.³ Airfinity uses a supply-driven model to construct country-level daily time series for the number of people expected to be fully vaccinated. Their model tracks vaccine production facilities and links supply from each facility to actual and expected deliveries to each country, including through international sharing arrangement like the COVAX facility. Based on actual/expected deliveries, Airfinity produces a time series for people expected to be fully vaccinated, allowing for different speeds of vaccine rollout depending on the income level of the country. In addition, while their model predicts the expected number of people fully vaccinated, it takes into account country specific policies, such as greater emphasis on first doses in some countries like Canada, Finland, or the UK.

Vaccine surprise: We construct a novel measure of vaccine surprises by taking the difference between actual vaccination rates (percent of population fully vaccinated) in the data and expected vaccination rates. The vaccine surprise variable has two key advantages over simply using vaccination rates in our empirical analysis. Economic activity is more likely to increases following surprises in vaccination rates rather than to actual vaccination rates in the population (after controlling for number of cases) as economic agents will likely internalize expected vaccine rollouts. In other words, if future vaccine rollouts are anticipated, using actual vaccination rates would lead to underestimating the economic effect of vaccines (for a similar argument, related to fiscal policy actions see Ramey, 2011). In addition, surprises in vaccination are less likely to be endogenous to economic developments and COVID-19 trends, as well as other shocks affecting vaccine supply, allowing for better causal identification.

To check that indeed our vaccine surprise variable can be deemed as exogenous, we analyze the relationship between the vaccine surprise and other variables affecting high-frequency indicators of economic activity.⁴ The results reported in Appendix Table 8 show that the vaccine surprise variable is uncorrelated with daily contemporaneous developments related to the pandemic (new cases and stringency of containment measures), the procurement of vaccines, as well as lags of our high-frequency measures of economic activity (NO₂, CO, mobility discussed below).

Finally, we looked at the time series of vaccine surprises as well as their distribution across countries. Figure 1 shows the box plot for the vaccine surprise variable for advanced economies as well as emerging and developing economies. The median vaccine surprise is close to zero for both groups, although the distribution is skewed with larger negative surprises indicating, consistent with anecdotal evidence, slower than expected vaccine rollout. Figure 2 shows the time series for the vaccine surprise variable for a few specific countries. The USA has had relatively small surprises, with vaccinations lagging model predictions till April, but a subsequent pickup in rollout resulting in actual vaccination rates catching up to model predictions. By contrast, vaccine rollout has consistently underperformed model predictions in India and overperformed model predictions in Israel.

Fig. 1

Source: Airfinity, Our World in Data, IMF staff calculations

Vaccine surprises, people fully vaccinated (percent of population).

Fig. 2

Source: Airfinity, Our World in Data, IMF staff calculations

Vaccine surprises, people fully vaccinated (percent of population).

2.2 Economic activity variables

Emissions Data: Following Deb et al. (2020b), we use emissions as a proxy for economic activity. We gather nitrogen dioxide (NO₂) and carbon monoxide (CO) emissions data at a daily frequency from The World Air Quality Index project, a non-profit project whose mission is to provide a unified worldwide air quality information.⁵ Our sample covers 64 countries for NO₂ and 59 countries for CO starting on January 1st, 2021. The data is reported using EPA standards and is based on the median emissions reported by city-specific stations which are updated three times a day.

Mobility: We collect retail and recreation mobility data from Google Mobility Community Reports, which have been shown to be a good proxy for economic activity (IMF, 2020a).⁶ The reports provide country-level daily data by country and highlight the percent change in visits to places related to retail and recreation activity (e.g. restaurants, cafes, shopping centers, movie theaters, museums, and libraries). The data is reported as the change relative to a pre-pandemic baseline value for that corresponding day of the week, said baseline is calculated as the median value for that corresponding day of the week, during the five-week period between January 3rd and February 6th, 2020. Daily data is available for 135 countries in our dataset, with coverage beginning from February 15th, 2020.

While the daily indicators used for the analysis do not capture every aspect of economic activity, they are strongly correlated with more traditional, monthly measures of economic activity such as industrial production (IP), Purchasing Managers Index (PMI), and OECDs composite leading indicator (CLI). Table 1 shows a statistically significant relationship using a monthly database of indicators from January 2019 to June 2021. We find that an increase in all 3 indicators is associated with a corresponding pickup in the average level of NO₂ emissions during the estimation period. The results for CO emissions and mobility also go in the same direction but are statistically weaker, suggesting that NO₂ is a more robust proxy of economic activity.

Table 1 High-frequency indicators and monthly measures of economic activity (2019–2021)

We further establish that NO₂ emissions are also strongly associated with the level of economic activity over a longer horizon. Using data available from the OECD database for total man-made emissions of nitrogen oxides from 1990 to 2018, we test the sensitivity of such emissions to conventional measures of economic activity such as GDP growth, growth in manufacturing value added and growth in measures of industrial production. Table 2 shows a robust relationship between these economic variables and NO₂ emissions.

Table 2 NO₂ emissions and economic activity—historical data (1990–2018)

Summarizing, the results in Tables 1 and 2 validate our choice of NO₂ emissions as the main proxy of interest for the empirical work in this paper.

2.3 Government responses

Containment measures: We use data from Oxford’s COVID-19 Government Response Tracker (OxCGRT).⁷ OxCGRT collects information on government policy responses across eight dimensions, namely: (1) school closures; (2) workplace closures; (3) public event cancellations; (4) gathering restrictions; (5) public transportation closures; (6) stay-at-home orders; (7) restrictions on internal movement; and (8) international travel bans. The database scores the stringency of each measure ordinally, for example, depending on whether the measure is a recommendation or a requirement and whether it is targeted or nation-wide. We normalize each measure to range between 0 and 1 to make them comparable. In addition, we compute and aggregate a Stringency Index as the average of the sub-indices, again normalized to range between 0 and 1. The data starts on January 1, 2020 and covers 151 countries/regions.

3 Methodology

We conduct three distinct exercises to study: (1) the impact of vaccines on economic outcomes; (2) the heterogeneity in the impact of vaccines depending on country conditions; and (3) the effects from increased COVID-19 infections in main trading partners on economic activity.

3.1 Effect of vaccinations on economic outcomes

For the analysis of the economic impact of vaccinations, we use our country-time panel dataset at the daily frequency that allows for high-frequency identification of the impact of vaccinations on economic outcomes. Establishing causality is difficult because vaccine rollout may depend on current or expected economic conditions, either directly or through the evolution of the pandemic which in turn impacts economic activity. We try to mitigate reverse causality by controlling for lagged values of number of COVID-19 cases as well as lagged values of our high-frequency economic indicators. We also control for country fixed effects which effectively control for vaccine procurement, structural factors (such as health capacity) affecting the speed of vaccine rollout, and for differences of the structure economic activity across countries—such as the share services or tourism. To further account for expectations about country-specific evolution of vaccine rollout and economic activity, we also control for a set of variables which may affect future infections and economic activity such as non-pharmaceutical interventions (NPIs)—including containment measures—and country-specific time trends.⁸ We include time fixed effects to account for global factors affecting the evolution of the virus (such as new variants), vaccination (supply disruptions), and economic activity (global shifts in confidence).

As a first step, we use an econometric specification as follows:

$$\Delta Y_{i,t} = \mu_{i} + \gamma_{t} + \beta V_{i,t - l} + \theta X_{i,t - l} + \varepsilon_{i,t}$$

(1)

where ${Y}_{i,t}$ alternatively denotes: the level of NO₂ emissions as a share of the population, the level of CO emissions as a share of the population, and Google’s retail mobility indicator, of country i at time t. ${V}_{i,t-l}$ denotes the share of the individuals in the population which have been vaccinated. The coefficient $\beta$ gives us the impact of higher vaccination rates on various economic variables. We include country and time fixed effects (${\mu }_{i}$ and ${\gamma }_{t}$) to control for country-specific characteristics and global trends that can affect the evolution of the pandemic. We also include a vector of control variables,${X}_{i,t-l}$, which comprises of the lagged level of COVID-19 cases, the lagged levels of NO₂ emissions, CO emissions, mobility, and stringency of containment measures, as well as country-specific time trends. We opt for a one-day lags as a baseline to reduce the risk of reverse causality but examine various lags as a robustness check.

We also explore nonlinear effects of vaccines by adding the square of ${V}_{i,t-l}$ in some specifications.

Despite the extensive set of controls used in Eq. (1), residual concerns about endogeneity may remain. To further address this issue, we construct a novel measure of vaccine surprises that accounts for expected rollout given procurement and use this as the independent variable instead of actual vaccination rates. We test the impact of unexpected vaccinations by modifying the econometric specification as follows:

$$\Delta Y_{i,t} = \mu_{i} + \gamma_{t} + \beta SV_{i,t - l} + \theta X_{i,t - l} + \varepsilon_{i,t}$$

(2)

where ${SV}_{i,t-l}$ is a measure of vaccine surprises constructed by taking the difference between the share of people that are fully vaccinated and the expected share as predicted by Airfinity’s supply-driven vaccine rollout model (see Sect. 2 for details). All other variables are the same as in Eq. (1).

3.2 Role of country-specific conditions on the effect of vaccines on economic activity

We also test the role of country-specific conditions in shaping the effects of vaccinations on economic activity. Namely, we examine whether the impact of vaccines on economic outcomes varies depending on the stringency of containment measures, or the severity of the outbreak itself. For this, we use a semi-parametric approach in which we interact vaccination surprises with quartiles (“bins”) of country-specific conditions. This approach does not impose the strong parametric restriction of the effectiveness of vaccines changing linearly with country conditions. Rather, it allows us to flexibly explore variation in vaccine effectiveness across the distribution of country conditions.⁹ We augment Eq. (2) with the following:

$$\begin{aligned} & \Delta Y_{i,t} \,=\, \mu_{i} + \gamma_{t} + \beta_{1} Q_{1} *SV_{i,t - l} + \beta_{2} Q_{2} *SV_{i,t - l} + \beta_{3} Q_{3} *SV_{i,t - l} + \beta_{4} Q_{4} *SV_{i,t - l} \\ & \quad + \mathop \sum \limits_{j = 1}^{4} \delta_{j} Q_{j} + \theta X_{i,t - l} + \varepsilon_{i,t} \\ \end{aligned}$$

(3)

where ${Q}_{1}$, ${Q}_{2}$, ${Q}_{3},$ and ${Q}_{4}$ are dummy variables that denote alternatively quartiles of the stringency of containment measures, or the level of new COVID-19 cases in a country. Quartiles are interacted with our vaccine surprises variable. Interaction terms are also lagged 1 day, consistent with the vaccine surprise variable. If the coefficients on the interaction terms of higher quartiles differ from those at lower quartiles, it signifies that the effectiveness of vaccines depends on country-specific conditions. In the robustness checks, we also test for alternative nonlinear specifications such as those based on linear interactions and smooth transition functions.

3.3 Effect of COVID-19 cases and vaccines in trading partners on economic outcomes

We further test whether a pandemic outbreak in a countries’ close trading partners can affect economic activity locally. Similarly, we also explore whether increased vaccines administered in a country’s main trading partners can help boost economic activity locally. To investigate whether this may be the case, we create the following:

$$Trading\, Partner_{i,t} = \mathop \sum \limits_{j = 1}^{N} w_{i,j} *Outcomes_{j,t}$$

(4)

where ${Trading\, Partner}_{i,t}$ is a term which alternatively denotes the COVID-19 cases or COVID-19 vaccinations in country i’s main trading partners. ${Outcomes}_{j,t}$ refer to either country j’s COVID-19 cases or vaccinations as a share of population at time t. These outcomes are combined with ${w}_{i,j},$ trade weights constructed based on bilateral trade flows (exports and imports from the 2019 Directional of Trade Statistics) between country i and country j that are scaled by total exports and imports such that $\sum_{j=1}^{n-i}{w}_{i,j}$=1. The weights thus capture each country’s relative trade exposure to its different trading partners, and the spillover term ${Trading Partner}_{i,t}$ captures COVID-19 cases/vaccines in a country’s trading partners, assigning higher weights to countries with strong trade linkages under the assumption that countries with closer trading relationships will have a larger impact on domestic economic activity.¹⁰ This term is introduced to Eq. (2) as following:

$$\Delta Y_{i,t} =\, \alpha + \mu_{i} + \gamma_{t} + \beta SV_{i,t - l} + \gamma Trading\, Partner_{j,t - m} + \theta X_{i,t - l} + \varepsilon_{i,t}$$

(5)

Equations (1) through (5) are estimated using OLS, with standard errors clustered at the country level.

4 Results

4.1 Baseline results

We begin by assessing the impact of vaccinations on high-frequency proxies of economic activity—the level of nitrogen dioxide (NO₂) emissions, the level of carbon monoxide (CO) emissions, and the decline in retail and recreation mobility. Table 3 column 1 shows estimates for Eq. (1) with change in NO₂ as the dependent variable. The coefficient on vaccinations is positive and significant, indicating that higher vaccination rates are associated with an increase in economic activity. Next, we introduce the second vaccine dose as an additional variable and find that second vaccine doses also have a significant effect on NO₂ emissions (Table 3, column 2).

Table 3 Effect of vaccines on economic activity—alternative high-frequency economic indicators

The impact of vaccination on economic activity based on Eq. (1) might, however, be biased due to residual concerns regarding endogeneity. As discussed before, to address this, we focus on surprises in vaccines administered (per capita) measured as the difference between actual vaccinations and the expected vaccination rollout, which we show is more likely to be exogenous. We find that surprises in vaccinations are strongly associated with higher NO₂ emissions (Table 3, column 3).¹¹ The magnitude of the coefficient of vaccine surprise is typically larger than those of first and second doses suggesting that expectations play a role. Quantitatively, a surprise increase in vaccinations by 10% of population is associated with an increase in daily per capita NO₂ emissions of around 0.112, which is about a 0.28 standard deviation increase in NO₂, or an increase of almost 30 percent relative to median.¹² As expected, stronger containment measures are also associated with lower economic activity. Taking the results at face value they imply that a 10-percentage points vaccine surprise has about the same impact on economic activity as going from a full lockdown (stringency index of 1) to containment measures equivalent to a stringency level of 0.5.¹³

Columns 4 through 8 of Table 3 repeat the analysis with changes in CO as the dependent variable while columns 9 through 12 use mobility as the dependent variables. The results for CO are statistically weaker and generally not significant at one lag. The insignificant impact on CO at short horizons may reflect the weaker correlation between CO and other measures of economic activity documented in Table 1. As discussed below, the impact on CO becomes significant at longer lags. For mobility measures, vaccine surprises have a significant positive impact, with a 10-percentage points vaccine surprise associated with an increase in mobility of 5 points (Table 3, column 11), which is equivalent to the average difference in mobility in the USA in March 2021 when a bulk of containment measures remained in place versus mobility in May 2021 when access to vaccines improved and states started easing restrictions gradually.

4.2 Lag structure

Turning to the lag structure, Figs. 3, 4, and 5 show that the impact of vaccination increases with greater lags, as economic agents gain greater protection from the virus and increasingly resume economic activities. In particular, the results for CO, which are not statically significant immediately, become significant over time.

Fig. 3

Effect of vaccinations on NO₂ emissions per capita, at different lags. Coefficient $\beta$ is reported for each lag ℓ (1–40), and based on $\Delta {Y}_{i,t}=\alpha + {\mu }_{i}+{\gamma }_{t}+{\beta SV}_{i,t-l}+ {\theta X}_{i,t-l}+ {\varepsilon }_{i,t}$ for a sample of 44 countries using daily data from December 20, 2020–June 16, 2021. ${Y}_{i,t}$ denotes NO₂ emissions per capita and ${SV}_{i,t-l}$ is the surprises in vaccines administered (per capita).${\mu }_{i}$ and ${\gamma }_{t}$ are the country and time fixed effects. X is a vector of control variables which includes the level of new cases, NO₂ and CO emissions per capita, the stringency of containment measures index, and mobility indices at t-1. Lightly shaded bars denote 90 percent confidence bands, and dark-shared bars denote 95 percent confidence bands

Fig. 4

Effect of vaccinations on CO emissions per capita, at different lags. Coefficient $\beta$ is reported for each lag ℓ (1–40), and based on $\Delta {Y}_{i,t}=\alpha + {\mu }_{i}+{\gamma }_{t}+{\beta SV}_{i,t-l}+ {\theta X}_{i,t-l}+ {\varepsilon }_{i,t}$ for a sample of 44 countries using daily data from December 20, 2020–June 16, 2021. ${Y}_{i,t}$ denotes CO emissions per capita and ${SV}_{i,t-l}$ is the surprises in vaccines administered (per capita).${\mu }_{i}$ and ${\gamma }_{t}$ are the country and time fixed effects. X is a vector of control variables which includes the level of new cases, NO₂ and CO emissions per capita, the stringency of containment measures index, and mobility indices at t-1. Lightly shaded bars denote 90 percent confidence bands, and dark-shared bars denote 95 percent confidence bands

Fig. 5

Effect of vaccinations on mobility, at different lags. Coefficient $\beta$ is reported for each lag ℓ (1–40), and based on $\Delta {Y}_{i,t}=\alpha + {\mu }_{i}+{\gamma }_{t}+{\beta SV}_{i,t-l}+ {\theta X}_{i,t-l}+ {\varepsilon }_{i,t}$ for a sample of 44 countries using daily data from December 20, 2020–June 16, 2021. ${Y}_{i,t}$ denotes changes in retail and recreation mobility and $S{V}_{i,t-l}$ is the surprises in vaccines administered (per capita).${\mu }_{i}$ and ${\gamma }_{t}$ are the country and time fixed effects. X is a vector of control variables which includes the level of new cases, NO₂ and CO emissions per capita, the stringency of containment measures index, and mobility indices at t-1. Lightly shaded bars denote 90 percent confidence bands, and dark-shared bars denote 95 percent confidence bands

4.3 Other robustness

The results are robust to different subsamples. Appendix Table 10 summarizes the robustness results for NO₂: (1) the results hold when the data is winsorized to ensure that the results are not driven by outliers; (2) our results go through if we drop countries that started vaccinating late such as Colombia and Vietnam—started their vaccination campaigns after March 1; (3) the results are also robust to dropping countries that started vaccinations very early such as USA and UK—already reached 5% of the population by February 1; and (4) the results hold if we drop one region at a time or one country at a time, suggesting that they are not driven by a particular region or country with high levels of vaccinations (such as the USA, the UK, or Israel).

4.4 Nonlinear effects

Next, we test for nonlinearities in the impact of vaccines on economic activity. Economic activity may not respond significantly to vaccinations in the initial phase of the vaccine rollout which often targets specific groups (health-care workers or the elderly). As the rollout continues and vaccinations spread to a wider set of people, confidence is more likely to improve, in part because individuals are more willing to reduce voluntary social distancing as the risk of infections go down.

Column 1 of Table 4 reports results for change in NO₂ as the dependent variable, where we include the share of population that has received one vaccine dose as well as the square of this variable as independent variables. We find evidence for nonlinear effects, with the square term being positive and significant, indicating that the economic benefits of vaccinations are larger when vaccination rates increase. Similar nonlinear effects are seen for the second vaccine dose (Table 4, column 3).

Table 4 Nonlinear effects of vaccines on economic activity

Columns 4–6 of Table 4 test for nonlinear effects of our more exogenous vaccine surprise variable and finds similar results. Column 4 allows for the simple interaction between the vaccine surprise variable and the share of population that is fully vaccinated. The interaction term is positive and significant, indicating that vaccine surprises have larger economic effects when the level of vaccinations is higher. In column 5, we present the results obtained by interacting the vaccine surprise variable with different quartiles of the share of people fully vaccinated, while in column 6 those by interacting vaccination rates with the surprise variable using a smooth transition function. Results are similar across specifications.

Appendix Tables 11 and 12 repeat the nonlinear regressions with changes in CO and mobility as the dependent variable, respectively. Results for these other indicators are generally less precisely estimated compared to NO₂.¹⁴

4.5 Role of containment measures and severity of outbreak

This section examines the extent to which the impact of vaccines on economic outcomes depends on other factors such as the stringency of containment measures and the severity of the outbreak.

4.6 Stringency of containment measures

In addition to the impact of vaccines, economic activity is also dependent on the severity of non-pharmaceutical interventions in the form of containment measures. A quick vaccine rollout may not lead to an immediate improvement in economic outcomes if strong containment measures need to be maintained at the same time. Column 1 of Table 5 adds an interaction term between the vaccine surprise term and the stringency of containment measures categorized into quartiles (Eq. 3). The interaction terms are negative for the higher quartiles and significantly different from zero for the 4th quartile. This indicates that an increase in vaccines leads to a smaller positive impact on NO₂ emissions when accompanied with stringent containment measures, potentially because movement restrictions prevent individuals from ramping up economic activity in response to higher vaccination rates. Column 3 of Table 5 repeats the analysis using retail mobility as the dependent variable. The results are similar to the NO₂ regressions.

Table 5 Effect of vaccines on economic activity—role of containment new cases

4.7 Severity of the outbreak

The impact of vaccines on economic activity is also likely to depend on the stage of the outbreak. If a country is in the middle of a large outbreak, an increase in vaccine rollout may have only a muted impact on activity as people continue to voluntarily socially distance till cases come down significantly. To test this hypothesis, column 2 of Table 5 adds an interaction term between the vaccine surprise variable and the number of new cases (moving average over seven days) in the country categorized into quartiles (Eq. 2). The interaction terms are negative and significant for the higher quartiles, supporting the hypothesis that voluntary social distancing may limit the beneficial impact of vaccines on economic activity when cases are high. Results are qualitatively similar when using mobility as the dependent variable, though statistically weaker.¹⁵

4.8 Spillovers from foreign COVID-19 cases and vaccines to economic activity

The pace of COVID-19 vaccinations across countries has been uneven, producing divergent economic and health outcomes across nations. In this section, we explore whether global health and vaccination outcomes can have an indirect effect on a country’s own economic activity levels through economic linkages such as trade. We also investigate whether the rapid vaccination pace in a systemically important economy such as the USA would have any spillover effects to the rest of the world.

4.9 Effect of foreign COVID-19 cases on local economic activity

Deb et al. (2021a) find that neighboring COVID-19 cases can have a significant effect on a country’s own pandemic, amplifying its own caseload despite vaccinations or containment measures. Similarly, we find significant spillover effects on country i’s economic activity, proxied by NO₂ emissions, through an increase in new COVID-19 cases in its main trading partners (Table 6, column 1). Namely, a one standard deviation increase in foreign COVID-19 cases would lead to a 0.10 standard deviation decrease in domestic NO₂ emissions, effectively reducing domestic economic activity through traditional economic linkages. This effect is lagged (Fig. 6), with the impact becoming statistically significant after around 21 days. This is plausible, given that the negative effects of an outbreak on a country’s economic activity are likely to take time to materialize, which implies that negative spillover effects would show after some lag.

Table 6 Effects of foreign new COVID-19 cases and vaccines on economic activity

Fig. 6

Effect of foreign new COVID-19 cases on economic activity. Coefficient $\gamma$ is reported for each lag ℓ (1–40), and based on $\Delta {Y}_{i,t}=\alpha + {\mu }_{i}+{\gamma }_{t}+{\beta SV}_{i,t-l}+\gamma {Trading Partner}_{j,t-m}+ {\theta X}_{i,t-l}+ {\varepsilon }_{i,t}$ for a sample of 43 countries using daily data from December 20, 2020–June 16, 2021. where ${Y}_{i,t}$ denotes: the number of new COVID-19 cases. ${Trading Partner}_{i,t}$ is a spillover term for COVID-19 cases in main trading partner countries. ${SV}_{i,t-l}$ denotes the share of the individuals in the population which have received at least one vaccine shot.${\mu }_{i}$ and ${\gamma }_{t}$ are the country and time fixed effects. X is a vector of control variables which includes the level of new cases, NO₂ and Co emissions per capita, the stringency of containment measures index, and mobility indices at t-1. ℓ denotes the lags in the response of new COVID-19 cases. Lightly shaded bars denote 90 percent confidence bands, and dark-shared bars denote 95 percent confidence bands

4.10 Effect of foreign COVID-19 vaccines on local economic activity

While foreign COVID-19 cases can have a dampening effect on local economic activity, the opposite seemingly holds for foreign COVID-19 vaccinations. First, we look at the spillovers from US vaccinations, given the relatively high rate of vaccinations in the USA and the important global economic linkages. Table 6, column 2 adds a US spillover term to the regression which is calculated by multiplying US vaccination rates with each country’s bilateral trade exposure to the USA. We find that there are positive spillover effects from increased vaccinations in the USA.¹⁶ Figure 7 shows the effect of US vaccinations on NO₂ emissions at different lags, with economic gains materializing with a 20-day lag.

Fig. 7

Effect of US COVID-19 vaccinations on economic activity. Coefficient $\gamma$ is reported for each lag ℓ (1–40), and based on $\Delta {Y}_{i,t}=\alpha + {\mu }_{i}+{\gamma }_{t}+{\beta SV}_{i,t-l}+\gamma {Trading Partner}_{j,t-m}+ {\theta X}_{i,t-l}+ {\varepsilon }_{i,t}$ for a sample of 43 countries using daily data from December 20, 2020–June 16, 2021. where ${Y}_{i,t}$ denotes: the number of new COVID-19 cases. ${Trading Partner}_{i,t}$ is a spillover term for COVID-19 vaccinations in the USA. ${SV}_{i,t-l}$ denotes the share of the individuals in the population which have received at least one vaccine shot. ${\mu }_{i}$ and ${\gamma }_{t}$ are the country and time fixed effects. X is a vector of control variables which includes the level of new cases, NO₂ and Co emissions per capita, the stringency of containment measures index, and mobility indices at t-1. ℓ denotes the lags in the response of new COVID-19 cases. Lightly shaded bars denote 90 percent confidence bands, and dark-shared bars denote 95 percent confidence bands

Next, we look at spillovers from vaccinations more broadly by adding the average vaccination rate of trading partners as described in Eq. 4 and 5. Results in Table 6 column 3 show that foreign COVID-19 vaccines have a positive and statistically significant effect on economic activity, with one standard deviation increase in foreign COVID-19 vaccines leading to a 0.13 standard deviation increase in NO₂ emissions. The result is persistent, and also materializes with a lag: Fig. 8 shows the impact of trading partner vaccination at different lags, with a persistently higher trend again materializing around the 20-day mark. The results provide additional evidence that a higher vaccination pace worldwide can also boost domestic economic activity.

Fig. 8

Effect of foreign COVID-19 vaccines on economic activity. Coefficient $\gamma$ is reported for each lag ℓ (1–40), and based on $\Delta {Y}_{i,t}=\alpha + {\mu }_{i}+{\gamma }_{t}+{\beta SV}_{i,t-l}+\gamma {Trading Partner}_{j,t-m}+ {\theta X}_{i,t-l}+ {\varepsilon }_{i,t}$ for a sample of 43 countries using daily data from December 20, 2020–June 16, 2021. where ${Y}_{i,t}$ denotes: the number of new COVID-19 cases. ${Trading Partner}_{i,t}$ is a spillover term for COVID-19 vaccinations in main trading partner countries. $S{V}_{i,t-l}$ denotes the share of the individuals in the population which have received at least one vaccine shot. ${\mu }_{i}$ and ${\gamma }_{t}$ are the country and time fixed effects. X is a vector of control variables which includes the level of new cases, NO₂ and Co emissions per capita, the stringency of containment measures index, and mobility indices at t-1. ℓ denotes the lags in the response of new COVID-19 cases. Lightly shaded bars denote 90 percent confidence bands, and dark-shared bars denote 95 percent confidence bands

Agarwal and Gopinath (2021) stress the importance of vaccinating a large share of the global population as quickly as possible, noting that “the pandemic is not over anywhere unless it is over everywhere.” The economic spillover results in this paper and the health spillover results in Deb et al. (2021a), Deb et al. (2021b), Deb, Ganslmeier, et al. (2021)) provide strong empirical evidence in support of such a policy, showing that a virus outbreak in trading partners is likely to have severe negative health and economic consequences domestically. Thus, ensuring an even distribution of vaccines, especially by sharing any vaccine oversupply in advanced economies, can help bring the pandemic to a speedier end, minimizing the global loss of lives and promoting a robust economic recovery.

5 Conclusion

Vaccines against the coronavirus disease are key to exiting the health and economic crises that COVID-19 has brought about. In this paper, we provide an empirical assessment of the effects of COVID-19 vaccines on economic activity. We put together a novel daily database of high-frequency economic indicators—NO₂ emissions, CO emissions, and Google mobility indices, combined with data on new COVID-19 cases, vaccinations, and surprises in vaccines administered, as well as data on non-pharmaceutical intervention. To the best of our knowledge, this is the first empirical assessment of the economic effects of COVID-19 vaccine surprises on a large-scale sample (46 countries), and of the role of country-specific conditions and the impact of COVID-19 vaccines and cases in main trading partners on a country’s local economic activity.

The results suggest that COVID-19 vaccines have a large and statistically significant effect on economic activity. A surprise increase in vaccinations by 10% of population is associated with an increase in daily per capita NO₂ emissions of about 0.3 standard deviations (an increase of almost 30 percent relative to its median). This is equivalent to going from a full lockdown (stringency index of 1) to containment measures equivalent to a stringency level of 0.5. We get similar significant results for mobility, with a 10-percentage points vaccine surprise associated with an increase in mobility of 5 percentage points. CO emissions also increase in response to increased vaccine surprises, but with a lag. These results are robust to alternative specifications and, as our result show, the magnitude of the effect is likely to increase with higher vacation rates.

We also find that the effect of COVID-19 vaccines on economic activity varies depending on the level of stringency measures imposed in a country, as well as on the severity of the pandemic outbreak in a country. Namely, the results suggest that the impact of vaccines rollouts may not lead to an immediate improvement in economic outcomes if strong containment measures need to be maintained at the same time. In addition, we find that the effect of vaccines on economic activity is more muted when a country is experiencing a severe outbreak, or when COVID-19 cases are high.

Finally, the results provide evidence on the importance of an even access to vaccines across nations, as we find that countries can be affected by pandemic spillovers via trade linkages with their main trading partners. Namely, we find that while COVID-19 cases in main trading partner countries can dampen local economic activity in a country, rollouts of vaccines have the opposing effect, boosting local economic activity and thus speeding up the global recovery. This highlights the potential gains from vaccine sharing, and the importance of vaccinating early and broadly not only a country’s own population, but all populations, in order to bring a swifter end to the global economic crisis the pandemic had brought about.¹⁷

The findings in this paper, combined with results from Deb et al., (2021a, 2021b) on the beneficial effects of vaccines on health outcomes, highlight the importance of vaccines to address the crisis instigated by the COVID-19 pandemic (see also IMF, 2021). In addition to the direct health and economic benefits of vaccines, this paper finds evidence for the dampening effect of containment measures and local outbreaks on economic activity, and the importance of sharing excess vaccine doses to boost global economic activity. We hope the empirical analysis provides evidence to policymakers on the importance of vaccinating swiftly and efficiently, both locally and globally, in order to exit the COVID-19 crisis.

Appendix

See Tables 7 , 8 , 9 , 10, 11, 12, and 13.

Table 7 Summary statistics

Table 8 Bivariate regressions of vaccine surprises with various variables

Table 9 Robustness checks—instrumenting using vaccine surprises

Table 10 Robustness checks—baseline using NO₂ emissions

Table 11 Nonlinear effects of vaccines on economic activity using CO

Table 12 Nonlinear effects of vaccines on economic activity using mobility

Table 13 Effect of vaccines—role of containment and new cases using CO and Mobility