key: cord-1009390-ejk3bgh7
authors: Rungcharoenkitkul, Phurichai
title: Macroeconomic effects of COVID‐19: A mid‐term review(*)
date: 2021-09-05
journal: Pacific Economic Review
DOI: 10.1111/1468-0106.12372
sha: 278405b76c2a30cbc83870566921d7e8b9d4295a
doc_id: 1009390
cord_uid: ejk3bgh7

This article provides an interim assessment of the macroeconomic consequences of the COVID‐19 pandemic. Estimates suggest a median output loss of approximately 6.5% in 2020, a gap that is expected to narrow to around 4% of the pre‐pandemic trend by the end of 2021. There is, however, a high dispersion of economic losses across economies, reflecting varying exposures to the pandemic and societies' responses. High‐frequency indicators and epidemiological models provide some insight into the interactions between the pandemic evolution and societies' strategies for combating it, including the role of vaccination. The article draws lessons from experiences thus far and discusses challenges ahead.

The economic significance of the COVID-19 pandemic cannot be overstated. The massive scale of economic losses alone makes the crisis one of the most pressing and consequential problems for many economists in their lifetimes. The unusual character of the crisis, a mixture between epidemiology and economics, makes the problem a challenging one intellectually. At the same time, in terms of policy prescription, the stakes could not be higher for economists to get it right. How costly have the lockdowns been in output terms? How much livelihood is worth sacrificing to save lives? Does vaccine arrival mean economies can reopen again? These are but some questions that economists can help shed light on. To be sure, the economics of epidemics is not a new field. However, the COVID-19 pandemic has uncertainly made it mainstream.

It is impossible to do full justice to a crisis that has spawned a whole new literature, a book collection and a new journal in a matter of months. 2 The modest objective of this article is twofold. First, the present study aims to gain a quantitative sense of the economic losses and the channels through which they have come about, in the process drawing comparisons with previous pandemics and epidemics. Second, the paper discusses the policy challenges in managing the COVID-19 pandemic. Of particular interest is the complex two-way interaction between economic activity and the spread of the virus and how one can make informed data-driven decisions to strike a balance between lives and livelihood. The remaining challenges in the next stage of the crisis are briefly discussed, with large-scale vaccination in sight but the virus also mutating continuously.

Experiences with past epidemics hold a number of lessons that resonate with the COVID-19 crisis. Losses of human capital can cause severe and long-lasting economic damage, especially in pandemics. 3 Epidemics with relatively low fatality rates can also significantly harm economic activity in affected areas. Economic losses can arise through various channels. Fears and health concerns can impact sentiment and hold back demand. Public health measures designed to contain the outbreak, somewhat paradoxically, can be detrimental to economic activity. As in any recession, financial and credit market procyclical developments can amplify and worsen output losses. The disproportionate effect of an epidemic on the service sector, which depends particularly on social interactions, is also well-recognised from prior experiences.

A key issue addressed by previous research is the magnitude of economic losses in the wake of an epidemic. Table 1 summarises selected studies on notable epidemics and estimates of output losses. The influenza pandemic of 1918 in particular has received much attention in the literature, being one of deadliest pandemics in modern history. 4 Hypothetical studies prior to the COVID-19 pandemic, indeed, often modelled pandemic scenarios around the 1918 influenza pandemic. At the same time, there have also been numerous works on more recent epidemics, such as SARS in 2003 and Ebola during 2014-2016, which serve as a basis for comparison.

It is evident from this research that severe pandemics not only resulted in many deaths but also economic costs that were an order of magnitude larger than those of typical epidemics. According to Barro, Ursua, and Weng (2020) , the 1918 influenza pandemic led to an output loss of around 6-8% of GDP, and according to Correia, Luck, and Verner (2020) , a 20% decline in manufacturing employment. Hypothetical studies suggest a present-day repeat of the 1918 pandemic would lead to an output loss of 3-6% of annual global GDP (see Global Preparedness Monitoring Board (2019) (2010)). Meanwhile, outbreaks of SARS, H5N1, and Ebola were highly disruptive for the affected countries but entailed only a negligible output loss as a percentage of global GDP (see Lee & McKibbin (2004) and Hai, Zhao, Wang, & Hou (2004) for SARS, Burns et al. (2006) for H5N1, and the World Bank (2014) for Ebola outbreaks.) While most agree on the high cost of the 1918 pandemic, the precise mechanism generating economic losses remains a subject of debate. One hypothesis is that the pandemic harmed economies by destroying human lives and the associated productive capacity. Illnesses and mortality risks could also disrupt economic activity. 5 An alternative view is that much of the output losses resulted from governments' non-pharmaceutical interventions (NPIs), such as enforced lockdowns and closures, introduced to mitigate deaths. Determining which of these two factors is more important requires a natural experiment. For example, Correia et al. (2020) exploit differences in NPI intensities across states in the United States and find that those that implemented NPIs for longer tended to experience lower mortality rates and higher manufacturing employment growth. They conclude that output loss was a direct consequence of the pandemic rather than policies designed to quell it. Lilley, Lilley, and Rinaldi (2020) dispute the finding, however, arguing that part of the observed employment expansion was driven by pre-1918 population growth. This debate, of whether the cure may be worse than the disease and whether there is a health-economic trade-off, echoes some of the present-day discussion (see next section).

Some previous studies have divided the transmission mechanism of an epidemic into demand versus supply channels, recognising that both can play a role. Supply effects may include a reduction in labour supply due to workers falling ill or refraining from participation in high-contact jobs due to health concerns. Demand effects include cutbacks in spending on services or other goods due to lost income. For example, Arnold, De Sa, Gronniger, Percy, & Somers (2006) examine the supply channel in a 1918-like pandemic, by combining an estimated loss of work days with an estimated productivity per worker. They conclude that in the first year, the pandemic would reduce GDP by 2.3%. The same study finds the effect through the demand side to be approximately 2% of GDP, assuming a larger effect in industries reliant on social interactions. In other words, both demand and supply channels appear to be similarly important in quantitative terms.

Another important lesson from previous epidemics is the potential long-lasting effects on the economy. On the supply side, Fan, Jamison, & Summers (2016) find that deaths and the reduction of the labour force made up the largest part of the cost associated with the 1918 influenza pandemic, due to its persistent effect. A one-time reduction in the labour force raised the capital-to-labour ratio and lowered the rate of return to capital, thus slowing the pace of capital accumulation and GDP growth for many years. 6 Aggregate demand itself could also be persistently depressed. Jordà et al. (2020) examine 12 major pandemics in Europe since the 14th century, finding that they were followed by multiple decades of low interest rates. Their interpretation is that higher precautionary saving and scarce investment opportunities kept the natural interest rates low. Unlike wars, pandemics do not destroy physical capital and typically lead to a long period of excess capital relative to labour.

While many lessons from past pandemics resonate today, the COVID-19 crisis has several distinctive characteristics that set it apart from historical benchmarks. modern-day pandemic more likely and potentially more destructive to begin with. 7 However, the most consequential and defining distinction is the large-scale lockdown policies introduced across the world. The unprecedented health policy response was designed to avert the extremely high death tolls associated with previous major pandemics. At the same time, the collateral damage of putting the economy in an induced coma has been staggering. Indeed, the proximate cause of the economic crisis this time is a conscious policy decision to save lives, rather than an unforeseen "shock" as was typically the case in past recessions. The unique character of the current crisis means the duration of economic duress is directly tied to pandemic developments. The latter are uncertain and highly nonlinear: small tweaks to policy could be the difference between infections stabilising or rising exponentially. As Table 2 shows, snapshots from March, April, and October painted very different pictures, as the situation deteriorated quickly despite strong policy responses. This section reviews some of these challenges and their macroeconomic implications.

In a prescient 2018 speech, Bill Gates warned of "a significant probability of a large and lethal, modern-day pandemic occurring in our lifetimes". He cautioned that "the health infrastructure we have for normal times breaks down very rapidly during major infectious disease outbreaks" (Gates, 2018) . The world, unfortunately, remained little prepared 2 years on when the COVID-19 pandemic struck and collectively could not prevent a full-blown pandemic. The lack of a prompt and coordinated global response illustrates the immense difficulty of assessing a rareevent risk in real time, not least when the last major pandemic was but a distant memory. Indeed, few anticipated a highly disruptive pandemic even after Italy's national lockdown and infections had started cropping up in all parts of the world. In early March, most forecasters still expected global growth in 2020 to hold up well ( Figure 1 ), with few contrarians ( Figure 2 ). Unfortunately, the lack of early decisive action made much stronger intervention later on necessary in many cases.

As the situation deteriorated rapidly over the course of March, many authorities placed their countries under lockdowns, including France and the UK. These stringent measures prompted a major reassessment of the macroeconomic impact. 8 As shown in Table 3 , impact assessments underwent a regime shift around mid-March. The usual depiction of economies being subject to occasional disturbances gave way to sudden stop economics. The rough-and-ready calculations of the lockdown implications replaced sophisticated model-based estimates and suggested a much larger decline in activity than in previous assessments. The April consensus forecasts revealed a sharp worsening in the 2020 outlook, with output loss estimated to be approximately 6% for the year (Figure 1 ). This order of magnitude is similar to the 1918 pandemic loss, previously seen by many to be a conservative worst-case scenario (e.g., see Barro et al., 2020) .

Pushing back the first infection wave would end up taking an even heavier toll than first estimated. Lockdowns put a swathe of economic activities into a frozen state overnight and subjected businesses and households to an unprecedented period of lost incomes. Despite massive policy support, insolvency risks and unemployment loomed large, especially for the most affected economic sectors. This led to further downward growth revisions between spring and summer. By June, the expected output loss was 8% for the median economy ( Figure 1 ). The subsequent easing of restrictions in the summer led to only small improvements in the economic outlook, as the median output loss for 2020 was revised to around 6.5%. The economic outlook F I G U R E 2 Disagreement on output projections. Panel shows GDP forecast disagreement at each survey, measured as the cross-sectional standard deviations over different forecasters. Lines are cross-country medians, and shaded areas are cross-country interquartile ranges [Color figure can be viewed at wileyonlinelibrary.com] F I G U R E 1 Output losses implied by forecasts. The output losses implied by Consensus GDP growth forecasts are shown over successive survey vintages. For each survey, output losses are calculated as the difference between annual GDP growth forecasts and the corresponding forecasts made in January 2020 before the pandemic. Similarly, the difference between cumulative GDP growth forecasts over 2020-2021 between prepandemic and post-pandemic surveys produces the expected output loss over the 2020-2021 period [Color figure can be viewed at wileyonlinelibrary.com] started to improve more noticeably in late 2020, buoyed in part by the sooner-than-expected arrival of vaccines. Even so, the median output level was still projected to remain 4% below the pre-pandemic trend at the end of 2021.

While the COVID-19 pandemic is truly a global crisis, the economic impact has also been highly differentiated across countries. As Figures 1 and 2 show, this became particularly clear in the second half of 2020, with the range of expected output losses widening markedly across countries, just as disagreements about individual estimates narrowed. The latest estimates of 2020 output losses vary from around 3% in Norway and Korea to more than 10% in the UK, Spain, India, Peru and the Philippines (Figure 3 ). This heterogeneity partly reflects varying effectiveness in dealing with the pandemic: for example, Korea and China managed to contain the virus spread in 2020 without resorting to prolonged and widespread lockdowns. It is also related to strengths of policy responses and weights societies attach to health and economic objectives. Countries, moreover, differ in their inherent vulnerabilities to the pandemic (e.g., dependence on tourism and services industries, population density, and compliance with government orders). For countries hardest hit by the pandemic, economic damage has, thus, been extremely significant, in some cases surpassing even those of previous financial crises. 

With so much at stake and the situation being highly fluid, tracking economic activity in real time has become indispensable during the pandemic. Traditional macroeconomic statistics were not timely enough for monitoring events that could take a sharp turn in a matter of days. Various high-frequency indicators have thus been adapted for macroeconomic monitoring and to guide policy decisions. Since the first lockdown in Hubei, indicators such as electricity consumption, daily underground passengers, and traffic congestion indices have been used to gauge the economic fallout from the large-scale containment. Similarly, as the containment measures were relaxed, these indicators helped illuminate how quickly economic activity resumed. One simple and useful set of high-frequency indicators is based on the location history of mobile devices, made available by Apple and Google in the wake of the COVID-19 crisis. 9 They provide a timely pulse of economic activity, proxied by the extent to which people frequent recreational spots and their workplaces. Figures 4 and 5 show the Google mobility indices for a range of countries, illustrating how economic activity came to a sudden stop in real time. Mobility in most countries dipped sharply in late March and April as lockdowns came into effect, before gradually recovering in subsequent months. Mobility declined again toward the end of 2020 as countries faced the second wave of infection, and recovered in early 2021 as the F I G U R E 4 High-frequency indicators. Panel shows the cross-country interquartile ranges of Google mobility indices (shaded area), alongside three other high-frequency indicators. Each Google mobility index is the average of retail and recreation, transit, and workplace activities subindices. "GAF" is the cumulative global activity factor from Leiva-Leon et al. (2020), a proxy for global economic activity based on a dynamic factor model. "Export shipment" is the deadweight tonnage of exports relative to 2017-2019 from automatic identification system (AIS) transponder data, computed by Cerdeiro et al. (2020) . "Tracker" is the cross-country interquartile ranges of OECD Weekly Tracker, which uses Google Trends as predictors of GDP growth (see Woloszko, 2020) [Color figure can be viewed at wileyonlinelibrary.com] F I G U R E 5 Mobility indices. Cross-country mobility ranges together with mobility indices of selected countries [Color figure can be viewed at wileyonlinelibrary.com] vaccine rollout accelerated. Mobility continued to increase through the second quarter of 2021, although remained more than 15% below pre-pandemic levels as of June 2021 in most countries.

Readings from mobility indices and other high-frequency indicators align reasonably closely, at least in the initial phase of the pandemic. As shown in Figure 4 , alternative measures of real-time activity, such as the global activity factor (GAF) based on a dynamic factor model of (Leiva-Leon et al., 2020), export shipment from transponder data due to Cerdeiro et al. (2020) and the Weekly Tracker based on Google Trends of Woloszko (2020) , all pointed to a sharp output contraction in the spring of 2020 and a gradual recovery in the summer. Mobility indices and the Tracker are arguably the most timely indicators, having reacted most promptly during both contraction and recovery phases. At the same time, mobility appears less correlated with other indicators over the more recent sample: mobility plunged during the second wave while both GAF and Weekly Tracker held up well. One interpretation is that societies may have adapted to the pandemic over time, making output less sensitive to mobility restrictions. That said, structural breaks between high-frequency indicators and aggregate output are potential problems for most of these measures.

In addition to timeliness, other advantages of the mobility indices are their standardised format and availability for a large group of economies, which allow cross-country comparisons. Figure 5 shows selected countries at both tails of the distribution: those whose mobility declined the most, such as India and the UK, versus the other end of spectrum, such as South Korea and the Nordic countries. This cross-country heterogeneity provides a basis for policy evaluation and comparative analysis. For example, what are the implications of different approaches to lockdown in terms of health and macroeconomic outcomes?

Differences in average mobility, indeed, explain almost 40% of cross-country variations in growth outcomes in 2020, suggesting that the mobility index has performed relatively well as a nowcasting tool, at least in the early stage of the pandemic. Figure 6 shows average mobility on the horizontal axis and the corresponding growth forecast revisions for 2020 on the vertical axis. There is a positive relationship, as one might expect, with 10 percentage points of mobility being equivalent to approximately 2.4 percentage points of GDP growth on average. The estimate is statistically significant, despite leaving out many relevant factors. Countries such as Spain, Italy, Greece and Thailand experience larger falls in expected GDP growth than suggested by their average mobility alone, which could be explained by their heavy reliance on tourism.

The mobility index is sometimes interpreted as a policy variable, a measure of government restriction stringency. Mobility is, indeed, negatively correlated with the "stringency index", as shown in Figure 7 : countries whose governments impose more stringent measures than others tend to have lower mobility. That said, mobility and policy stringency are strictly distinct concepts. Public health policies could be ineffective in reducing mobility if public compliance is low. By the same token, relaxing restrictions does not guarantee free-flowing economic activity if the general public remains concerned about the virus. In fact, the mobility index reflects both government policies and preferences of the general public, and it is better thought of as the society's policy choice rather than that of the government's alone.

Most scientists, and the WHO, considered strict containment measures, including lockdowns, as necessary on medical grounds to counter rapidly rising new infections. 10 An influential study Ferguson et al. (2020) shows that in the absence of a "suppression" strategy, surges of new patients could overrun hospital capacity and lead to more deaths: Bill Gates' doom scenario. The conclusion follows from simple epidemiological arithmetic: if each infected person continues passing on the virus to more than one person, infections will continue rising at an exponential rate. With the inherent infectiousness of SARS-CoV-2, the explosive reproductive dynamic will continue until much of the population has either developed immunity or died. Lockdowns are designed to drastically reduce the infection rate and avoid exposing a large F I G U R E 7 Mobility as a policy stance measure. "Mobility" on the horizontal axis is computed as in Figure 4 . "Stringency index" is the sample average of the aggregate policy stringency index from Hale et al. (2021) F I G U R E 6 Mobility as a nowcasting tool. "Mobility" on the horizontal axis is the Google mobility index (computed as in Figure 4) , averaged over the first 60 days since the first confirmed case. "Growth revision" is the difference between December and January 2020 Consensus forecasts of GDP growth, in percentage points portion of the population to the virus. Some scientists argued against strict lockdowns in response to the first wave on the grounds that a significant part of the population will likely acquire immunity naturally (see Boissay, Rees, & Rungcharoenkitkul (2020) for a review of the debate). However, subsequent infection waves have, by and large, dispelled any perception that "herd immunity" was close to being achieved early on in the pandemic.

Controversies surrounding lockdowns also arise from the potential conflict between health and economic objectives. One view is that there is a health-economic tradeoff, between "lives and livelihood". Lockdowns are detrimental to the economy but help save lives. The question is whether severe lockdowns go too far along the latter dimension from a welfare perspective. One approach to answering this question is to quantify lives saved in economic terms using the value of a statistical life (VSL), which measures how much a person values his or her own life. The VSL is one metric of social welfare, on the grounds that a priori a pandemic could affect just anyone. Greenstone and Nigram (2020) apply this approach in the case of the United States, and find that saving 530,000 lives is worth over one-third of US annual GDP, achievable through moderate social distancing. Cutler and Summers (2020) use the VSL to calculate losses from long-term and mental health problems, finding that these are as large as the economic costs of mortality (see for further discussion of the VSL approach and its limits). A different approach to quantifying the tradeoff is to examine the impact of livelihood on lives. Recessions could also have harmful effects on health and mortality, especially in economies with weak social safety nets and limited fiscal resources. These considerations tend to nudge the balance towards the economic objective (see Hausmann & Schetter (2020) and Doerr & Hofmann (2020) ).

Another view on the debate is that there may not be a tradeoff at all. The argument is that normal economic activity cannot resume as long as a pandemic is still raging. Households conscious of health risks cannot freely spend, while businesses will put off investment until the coast is clear. In this environment where a more severe health crisis only leads to a deeper economic fallout, the solution is to do whatever it takes to beat the pandemic (see Portes (2020) and Chang & Velasco (2020) ). Indeed, the data seem to support this view, as the supposed tradeoff between economic activity and health outcomes is elusive. Figure 8 presents a cross-country scatter plot of deaths per million on average mobility during the early phase of the pandemic. 11 The negative slope suggests that on average those countries that suffer more economically are also the ones with worse health outcomes. There is seemingly no tradeoff. 12 One way to reconcile these viewpoints is to distinguish two independent forces. The first is the epidemiological link between social interactions and the pandemic severity: more interactions lead to higher infections and deaths. This is summarised schematically by thes upwardsloping "pandemic curve" in Figure 9 , where greater activity implies more deaths. The second force is the society's preferences over health-economic outcomes, represented by the "society's response" schedule. It is downward-sloping because the society exercises greater precaution in response to a worsening pandemic, by curtailing economic activity. 13 The relative positions of these two schedules then dictate differences in cross-country outcomes. An explanation for the negative slope in Figure 8 is that shifts in the pandemic curve dominate those of society's response. Large epidemiological variations across countries thus trace out the average societies' response curve, revealing the estimated downward-sloping relationship. In truth, the society as a whole still faces an epidemiological constraint, as defined by the pandemic curve, as the pandemic inescapably narrows the set of feasible health-economic outcomes available to the society. 14 Having an estimate of the pandemic curve would provide useful information about the set of feasible health-economic outcomes for policymakers, whatever their preferences may be. Estimating it requires a structural epidemiological model that internalises how the infection rate interacts with the society's policy choices. 15 Figure 10 shows the estimated pandemic curves for selected countries, based on the epidemiological--economic model of Rungcharoenkitkul (2021) , which augments a "susceptibl-exposed-infectious-removed (SEIR)" model with an optimising unit representing the societies' preferences over lives and livelihood. The model is fit to daily health data (cases and deaths) and mobility, using a fast filtering method (see https://github.com/phurichai/covid19macro for the open-source code). The estimated model is used to construct counterfactual scenarios by simulating over alternative social welfare parameters. As shown in Figure 10 , the feasible set is always upward sloping at any given moment: saving more lives comes at the expense of lower mobility. F I G U R E 9 Tale of two forces. A schematic of how society's policy choices and the pandemic evolution jointly determine health-economic outcomes F I G U R E 8 Reduced-form empirical relationship. The horizontal axis is the average mobility over the first 60 days after the first confirmed case. The vertical axis shows deaths per million, as of end of October 2020

The terms of health-economic tradeoff can vary across economies depending on the context and initial conditions. As illustrated in Figure 10 , reducing mobility by 5% can save more than 200 lives per million in India (almost 300,000 lives), but less than 100 lives per million in the United States. Country-specific factors such as population density and effectiveness of policy may partly explain these differences. Also important is the size of the susceptible population. Those that have already experienced previous large infection waves and/or have made substantial progress in vaccination would be less exposed to additional deaths. In these cases, restricting mobility would entail lower marginal benefits.

The terms of tradeoff can also vary over time in a given country as the situation evolves. The arrival of vaccination, by granting immunity to individuals without exposing them to the virus, eases the terms of tradeoff and allows economies to remove restrictions with fewer fatalities. Improving treatment standards, by suppressing the case-fatality rate, can, similarly, deliver better outcomes. Conversely, any setback to the vaccination prog,ramme could justify a reenactment of restrictions. The tradeoff could also deteriorate if new virus strains emerge that overcome existing immunity in the population. The epidemiological-economic model can again provide some quantitative implications of these contingencies. Figure 11 depicts the tradeoff schedules under the baseline and two downside scenarios, using India as an illustration. If vaccination progress stalls or virus mutations compromise immunity, mobility may need to be further reduced by 5-10% to keep new deaths below 300 per million.

At the time of writing (June 2021), many countries were still grappling with rising new cases and most countries still had some form of restrictions in place. As foreshadowed by the 1918 flu F I G U R E 1 0 Tradeoffs across countries. Simulation from the epidemiological-economic model of (Rungcharoenkitkul, 2021) , estimated as of the end of June 2021. Each line represents the feasible set of health and economic outcomes (respectively, new fatalities and average mobility projected in the second half of 2021). The panel shows the feasible sets for four selected countries [Color figure can be viewed at wileyonlinelibrary.com] pandemic, subsequent infection waves have proved to be more deadly. At the same time, containment measures have generally become less stringent compared to those during the first wave of infection. One reason was the greater perceived effectiveness of even moderate and targeted lockdowns, given better sanitation practices (e.g., public mask wearing), improved treatment protocols, and more rigorous trace-and-quarantine practices. Another reason was a concern that another strict lockdown could be much more debilitating economically than in the early stages, especially due to higher risk of bankruptcies. 16 Lockdown fatigue and dwindling political support in some cases may have also played a role.

A quick end of the pandemic is far from assured. The early arrival of vaccines promised to be a game changer, but scaling vaccination globally has been an immense challenge. Under pressure to contain infections at home, many countries rushed to secure vaccine doses amid limited global supply. Relatively few managed to obtain more than adequate supply for their population, while others had to contend with infection surges with little vaccine protection. The lack of global coordination has likely made it easier for new virus variants to emerge and take hold. The much more contagious Delta variant has already become rampant in many countries, and is capable of causing breakthrough infections among those vaccinated. As the race between vaccination and virus mutations gets protracted, eradication appears further out of reach while an endemic COVID-19 in the longer term becomes an increasingly likely prospect.

Looking beyond the pandemic, it is likely that the global economy would not return to what it once was. For one, the large and uneven impact on the economy is providing strong impetus for resource allocations. Those industries reliant on face-to-face interactions, hardest hit by the pandemic, may encounter a more protracted recovery and falling output share (see . Meanwhile, winning sectors from the pandemic could benefit from permanent preference shifts, for example towards conducting virtual meetings and working from home. The pandemic also disproportionately affected smaller firms and low-income groups, which could have longterm ramifications on employment prospects and income distribution (see Yonzan, Lakner, Mahler, Aguilar, & Wu, 2020). F I G U R E 1 1 Tradeoffs across scenarios. Each line represents the feasible set for India (computed as in Figure  10 ), under baseline and two additional scenarios. Under "slow vaccination", the vaccine supply/take-up slows to a third of the baseline, while under "virus mutation", each vaccinated/recovered person loses immunity after 60 days [Color figure can be viewed at wileyonlinelibrary.com] The pandemic could also leave lasting impressions on people's beliefs and influence their behaviour long after it is gone. Kozlowski, Veldkamp, and Venkateswaran (2020) argue that because a pandemic is so rare, once it happens its perceived likelihood jumps and persists. In addition, because a pandemic precipitates resource allocations and capital obsolescence, concerns about future pandemics can have significant effects on today's risk premium and investment. They estimate that this "belief scarring effect" could generate a long-term cost 10 times larger than the short-term GDP loss.

Finally, the pandemic could cast a long shadow on the policy landscape for years if not decades to come. The provision of fiscal backstops during lockdowns totalled US$12 trillion globally, equivalent to 12% of global GDP (International Monetary Fund, 2020a) . Most of the stimulus is accounted for by advanced economies, whose public debt is now at over 120% of GDP, the highest level since World War II. Emerging market economies have been more constrained fiscally but have themselves seen public debt rising to historical highs, at over 60%. Any major bankruptcy events, a non-negligible risk, could make further fiscal commitments necessary. High public debt levels coupled with low growth could present challenges for central banks when the time comes to rebuild policy space (Borio, 2020). Monetary policy normalisation, even if justified on macroeconomic grounds, could be subordinate to fiscal sustainability concerns. As a result, the low-for-long interest rate environment could be further entrenched by the pandemic, placing even greater burden on macroprudential policies to deliver on financial stability . It is not clear if they can achieve this by themselves.

The pandemic is by no means over and there may be more surprises in store. Nonetheless, the last 18 months already hold some useful lessons for the future.

First, it is now clear that future highly disruptive pandemics remain possible, and underestimating the risk can come at great cost. This urgently calls for enhancing global preparedness to quickly identify and thwart future threats, as well as a concrete well-orchestrated contingency plan of action once an outbreak has occurred. Second, severe lockdowns of the type seen in the spring of 2020 can be introduced only so many times over the course of a pandemic. Economic damages are extremely high, and entail large fiscal resources to alleviate the impact. Any large-scale restrictions must, therefore, be accompanied by other less disruptive measures as part of the overall strategy. Third, maximising the effectiveness of these other measures is key for limiting the total pandemic cost. Draconian but targeted policies are far less expensive than indiscriminate measures, but there may be only a small window of opportunity where they are effective. Enforcing strict quarantine rules, tightening border controls, and conducting extensive contact tracing very early on in an outbreak could be the key difference.

1 As of June 2021, there have been 4 million confirmed deaths from COVID-19. An uncontrolled pandemic could have claimed many more: for example, applying age-specific infection fatality rates to the global demographic structure implies 55 million total fatalities. Surges in demand for ICU beds would likely lead to additional deaths. 2 See Brodeur, Gray, Islam, and Bhuiyan (2021) and Bloom, Kuhn, and Prettner (2021) for literature reviews on the economic impact and comparisons with previous epidemics, echoing some of the same themes in this article. 3 This section draws in part from . 4 The world has seen deadlier epidemics in the more distant past. The bubonic plague of the 14th century claimed up to 200 million lives, while the European viruses in Mexico effectively decimated an entire civilisation. However, much poorer sanitary standards during these times arguably make these episodes less useful as a benchmark for the current situation. 5 Barro et al. (2020) note that the illness of US President Woodrow Wilson may have contributed to the harsh terms of the Versailles Treaty in 1919, with tragic ramifications in the following decades. 6 Jordà, Singh, and Taylor (2020) also find that real wages remained elevated over more than three decades after major pandemics. The concurrent World War I presented an identification challenge in quantifying the pandemic's contribution to mortality, however (see Barro et al., 2020). 7 In 2018, 4 billion people boarded flights, an average of one in two global citizens. In 2013, when SARS hit, flight passengers totalled just 1.7 billion. Today's close economic integration permeates the production network as well as cross-border credit and financial market linkages. Increased connectedness amplifies damage of a pandemic and strengthens the international spillover effects, through both supply and demand. 8 For a detailed chronicle of the pandemic and lockdowns, see Bank for International Settlements (2020) and

International Monetary Fund (2020b). 9 See https://covid19.apple.com/mobility and https://www.google.com/covid19/mobility/ 10 The health emergency proviso is important. Some countries have managed to avoid severe lockdowns through a combination of strategies, including earlier responses, extensive use of masks, and aggressive testing and quarantining. 11 I allow for some time lag between mobility and eventual health outcomes because of epidemiological inertia, but other lag choices deliver similar results. This supposedly "perverse" relationship has also been widely documented by others, including in the press. 12 See also Oliu-Barton et al. (2021) , who use the Weekly Tracker of Woloszko (2020) instead of GDP and arrive at the same conclusion. 13 In keeping with the previous section, I refer to the society to make clear that such a decision is jointly made by the general public and the government. More precisely, the society's response curve is defined by the more stringent of the two. If government regulations are more lax than the public prefers, the public response curve will prevail, and vice versa assuming perfect compliance. 14 From this perspective, the argument of Portes (2020) may be more about the relevance rather than the existence of a tradeoff. If the society cannot tolerate a very high level of fatality (its response curve has a ceiling), then the upper portion of the pandemic curve is never binding. 15 A number of papers have explored this interaction in a stylised setting. See Alvarez, Argente, and Lippi (2021), Eichenbaum, Rebelo, and Trabandt (2021) , Glover, Heathcote, Krueger, and Rios-Rull (2020) , Jones, Phillipon, and Venkateswaran (forthcoming) , and Krueger, Uhlig, & Xie (2020), among others. See for a review of this new literature. 16 Firms and households drew on much of their liquidity and equity cushions during the first lockdown, although bankruptcies have so far been fewer than suggested by output (see and Banerjee, Kharroubi, & Lewrick (2020) ). Extraordinary government support has clearly played an important role as a shock absorber, but scope to provide a blanket backstop has significantly narrowed.

A simple planning problem for COVID-19 lockdown

A potential influenza pandemic: Possible macroeconomic effects and policy issues Report to the

The financial vulnerabilities driving firms to the exit

Bankruptcies, unemployment and reallocation from Covid-19. BIS Bulletin, 31. Bank for International Settlements

The coronavirus and the great influenza pandemic: Lessons from the 'spanish flu' for the coronavirus' potential effects on mortality and economic activity (NBER Working Papers No. 26866)

Modern infectious diseases: Macroeconomic impacts and policy responses (CEPR Discussion Papers No. 15997)

Dealing with covid-19: Understanding the policy choices

Macroeconomic effects of covid-19: An early review

The economic effects of the 1918 influenza epidemic. London: CEPR Discussion Papers 3791

A literature review of the economics of covid-19

Evaluating the economic consequences of avian influenza

World seaborne trade in real time: A proof of concept for building AIS-based nowcasts from scratch (IMF Working Papers No. 20/57)

Economic policy incentives to preserve lives and livelihoods

Pandemics depress the economy

The covid-19 pandemic and the $16 trillion virus

Recessions and mortality: A global perspective

The macroeconomics of epidemics. The Review of Financial Studies

The inclusive cost of pandemic influenza risk (NBER Working Papers No. 22137)

Back-of-the-envelope estimates of next quarter's unemployment rate, On the economy blog

Impact of non-pharmaceutical interventions (NPIs) to reduce COVID-19 mortality and healthcare demand

A world at risk: Annual report on global preparedness for health emergencies

Health versus wealth: On the distributional effects of controlling a pandemic

Does social distancing matter?

Coronavirus: Impact on stock prices and growth expectations, Working paper

The short-term impact of SARS on the chinese economy

A global panel database of pandemic policies (Oxford Covid-19 Government Response Tracker)

International Monetary Fund. (2020a). Fiscal monitor, october 2020 -Policies for the recovery

Optimal mitigation policies in a pandemic

Longer-run economic consequences of pandemics

The macroeconomic impact of pandemic influenza: Estimates from models of the United Kingdom

Scarring body and mind: The long-term belief-scarring effects of COVID-19

Macroeconomic dynamics and reallocation in an epidemic: Evaluating the "swedish solution

Learning from SARS: Preparing for the next outbreak

Real-time weakness of the global economy: A first assessment of the coronavirus crisis

Public health interventions and economic growth: Revisiting the spanish flu evidence

The global macroeconomic impacts of Covid-19: seven scenarios. Canberra

Evaluating the initial impact of Covid containment measures on activity

SARS-CoV-2 elimination, not mitigation, creates best outcomes for health, the economy, and civil liberties

Don't believe the myth that we must sacrifice lives to save the economy. The Guardian, Opinion

An integrated macroprudential framework in the post-pandemic world. How Do Monetary, Micro-and Macroprudential Policies Interact?

Macroeconomic consequences of pandexit (BIS Working Papers No. 932)

Global trade impact of the coronavirus (Covid-19) epidemic

Tracking activity in real time with google trends (OECD Economics Department Working Papers No. 1634)

The impact of COVID-19 on global poverty under worsening growth and inequality