Equitable vaccine distribution promotes socioeconomic benets globally

Ensuring a more equitable distribution of vaccines worldwide is an effective strategy to control 17 the COVID-19 pandemic and support global economic recovery. Here, we analyze the 18 socioeconomic effects - defined as health gains, lockdown-easing benefit, and supply-chain 19 rebuilding benefit - of a set of idealized vaccine distribution scenarios, by coupling an 20 epidemiological model with a global trade-modeling framework. We find that overall a 21 perfectly equitable vaccine distribution across the world (Altruistic Age-informed Distribution 22 Strategy) would increase global economic benefits by 11.7% ($950 billion) per year, compared 23 to a strategy focusing on vaccinating the entire population within vaccine-producing countries 24 first and then distributing vaccines to non-vaccine-producing countries (Selfish Distribution 25 Strategy). With limited doses among mid- and low-income countries, prioritizing the elderly

Strategy) would increase global economic benefits by 11.7% ($950 billion) per year, compared 23 to a strategy focusing on vaccinating the entire population within vaccine-producing countries 24 first and then distributing vaccines to non-vaccine-producing countries (Selfish Distribution 25 Strategy). With limited doses among mid-and low-income countries, prioritizing the elderly 26 who are at high risk of dying, together with the key workforce who are at high risk of 27 exposure, is found to be economically beneficial. We further show that such a strategy would 28 cascade the protection to other production sectors while rebuilding the supply chains. Our 29 results point to a benefit-sharing mechanism which highlights the potential of collaboration 30 between vaccine-producing and other countries to guide an economically preferable vaccine 31 distribution worldwide. 32 The subsequent waves of COVID-19 pandemics continue to threaten public health and society across 33 the globe 1-5 . Though vaccination has proven beneficial to avert the substantial tolls, global inequity 34 in vaccine distribution and manufacturing is crucial at present 6-8 . Given the rapid evolution of SARS-35 CoV-2, it is clear that nobody wins the race until everyone wins. This motivates us to consider the 36 potential of collaboration between vaccine-producing and other countries, guiding the optimal 1 vaccine distribution across countries and allowing a faster recovery of health systems and society. 2 The dilemma is: Should vaccine-producing countries (e.g., USA, EU, UK, China, India, and Russia) 3 prioritize their own populations (a Selfish Distribution Strategy), or allocate at any time the limited 4 amount of vaccines to all countries according to their population and age structure (two Altruistic 5 Distribution Strategies)? To lay out our general insights, we, by linking epidemiological and 6 socioeconomic modeling frameworks, probe the potential gains of global vaccine allocation 7 strategies from the socioeconomic perspective. 8 It is worth noting that countries are highly connected by global supply chains. This indicates the 9 cascading effect of intervention strategies across countries. For example, evidence has shown the 10 negative impacts of the lockdown intervention to curb virus transmission in one country spread to 11 other countries along supply chains 1,5,9-11 . On the other hand, vaccination decisions in one country 12 may be beneficial to the economic recovery of other countries, which is often referred to as one type 13 of externality of vaccination [12][13][14][15] . The presence of these externalities is a major driver that makes a 14 market-oriented global vaccine distribution a socially non-optimal solution 16, 17 . Advancing our 15 understanding of the positive health and economic externalities is the key to maximize the 16 socioeconomic gains of global vaccine roll-out 12,14,16,18 . 17 Here, we quantify the socioeconomic benefits of a set of idealized COVID-19 vaccine-distribution 18 scenarios (Box 1) by linking epidemiological 3,19 and socioeconomic 1,20,21 modeling frameworks. 19 Details of our analytical approach are provided in the Methods. In brief, we base our evaluation on 20 three main outcomes: i) the health gains, i.e., the value of lives saved through vaccination. Leveraging 21 our realistic age-stratified epidemiological (RAS) model 19 , we project mortality avert under varying 22 vaccination scenarios as compared to the 'no vaccination' scenario. With the estimates, we used the 23 value of statistical life (VSL) to project the health benefit quantified in US dollars. ii) the lockdown-24 easing benefit. Assuming that the speed of vaccine rollout is equivalent to the easing of the 25 lockdown 1,19 , we multiply the lockdown reduction by sectoral value-added to obtain what we call the 26 lockdown-easing benefit. iii) the supply-chain rebuilding benefit. We developed a global trade 27 model based on the widely used ARIO approach 20,21 to assess the economic losses over 6 years, i.e., 28 2020-2025, under vaccination scenarios. We translate the estimates to the total economic benefit 29 brought by the vaccines by quantifying the difference of economic losses with vaccination versus the 30 'no vaccination' scenario. We further subtract the lockdown-easing benefit from the total economic 31 benefit to project the supply-chain rebuilding benefit. 32 We modeled three sets of idealized scenarios integrated into a tiered structure (see Box 1, 33 Supplementary Fig. 1). Basically, Tier Global set of scenarios address the issue of the cooperative 34 attitude of vaccine-producing (or vaccine-exporting) countries and vaccine-importing countries, 35 while Tier Domestic set of scenarios address the issue of how received vaccines (vaccines sent by 36 producing countries) are allocated within each destination country (Box 1). We consider different 37

Box 1 | Scenario settings, explanations, justification
We designed three sets of scenarios integrated into a tiered structure. Basically, Tier Country set of scenarios address the issue of the cooperative attitude of vaccine-producing (or vaccineexporting) countries and vaccine-importing countries, while Tier Domestic set of scenarios address the issue of how received vaccines (vaccines sent by producing countries) are allocated within each destination country. Among tier domestic, subscenario A defines the allocation of the received vaccines within destination countries by age group, while sub-scenario S defines the allocation of the received vaccines within destination countries by industrial sectors. Tiered Structure of the scenario sets design 1 sub-scenarios within each scenario set (Box 1, Supplementary Fig. 1). In summary, sub-scenario C  2 represents to what extent the vaccine-exporting country is willing to share the vaccine with other 3 countries (specifically, a Selfish Distribution Strategy vs two Altruistic Distribution Strategies). Sub-1 scenario A defines the allocation of the received vaccines within destination countries by age group. 2

Scenario settings and descriptions
And sub-scenario S defines the allocation of the received vaccines within destination countries by 3 industrial sectors (Box 1). In the following analysis, when comparing the results of one dimension of 4 the scenario sets, "Altruistic Distribution Strategy", "Elderly First", and "High Risk" are used as the 5 default scenario. 6 Results 7 Figure 1 summarizes the results of a set of global vaccine distribution scenarios ( Fig. 1 only shows  8 the results under the combination of 'Elderly First' and 'High risk' scenarios, and the results under 9 other scenario combinations are documented in Supplementary Fig. 3 Altruistic Distribution Strategy; a pure per capita allocation); and panels in the right column ( Fig.  16 1c,f,i,l) show the benefits when the major vaccine-producing countries share their vaccine 17 altruistically with other countries with age profile (the Altruistic Age-informed Distribution Strategy; 18 an age-adjusted per capita allocation). Three kinds of benefits are shown in the first three rows, 19 namely health gains ( Fig. 1a,b,c), lockdown-easing benefit (Fig. 1d,e,f), and supply-chain rebuilding 20 benefit (Fig. 1g,h,i). The total benefit is then shown in the bottom row of Fig. 1 (Fig. 1j,k,l). 21 Altogether Fig. 1 shows that a more equitable distribution of vaccines across the world (i.e., 22 Altruistic Distribution Strategies) would bring more societal benefits globally than a vaccine 23 distribution that is focused on vaccine-producing countries (i.e., the Selfish Distribution 24 Strategy). If the Selfish Distribution Strategy is adopted, the total global benefit from vaccination is 25 estimated to be US$8.10 trillion (~9.6% of world GDP) per year. If the Altruistic Distribution 26 Strategy is adopted, the total global benefit from vaccination increases to $8.65 (~10.2% world GDP) 27 per year. And if the Altruistic Age-informed Distribution Strategy is adopted, the total global benefit 28 from vaccination further increases to $9.05 (~10. would not only increase health benefits by protecting more lives and direct domestic 15 production benefits by reducing the need of strict lockdowns, but would also facilitate the 16 recovery of the inter-industrial linkages and intra-/inter-regional supply chains. First, 17 compared to the Selfish Distribution Strategy, the overall health gains have increased by 1.8% under 18 the Altruistic Distribution Strategy and 7.0% under the Altruistic Age-informed Distribution Strategy 19 ( Fig. 1a-c). Under two Altruistic Distribution Strategies, the elderly with a high infection-mortality 20 rate and workforce with high exposure risk are covered more, resulting in more lives saved globally 21 (Supplementary Table 8). For example, in Mozambique (one of the least developed economy), the 22 health gains under the Selfish Distribution Strategy and Altruistic Age-informed Distribution 23 Strategy are 9.5% and 13.3% of annual GDP, respectively (Fig. 1a,c). In the other hand, vaccine-1 producing countries deliver more vaccines to other countries in two Altruistic Distribution Strategies, 2 leading to a decline in their health gains unsurprisingly. For example, in Germany, the health gains 3 under the Altruistic Age-informed Distribution Strategy would be reduced by 1.03% of annual GDP 4 compared to the Selfish Distribution Strategy (Fig. 1a,c). These results of healthy gains show that the 5 marginal health gains of vaccine (i.e., health gains created by each additional unit of vaccine) 6 in countries lacking vaccines are greater than that in countries where vaccines are relatively 7 abundant. Implicit assumptions in the above conclusion are that vaccine supply is the only 8 constraint, while demand is sufficient (e.g., no vaccine hesitancy issue), and distribution processes 9 (e.g., cold chain) are effective. increase by 1.1% compared to the Selfish Distribution Strategy (Fig. 1d-f). This is mainly because 20 the altruistic allocation strategy has been adjusted according to the age profile of each country, 21 resulting in countries with more elderly people getting more vaccines per capita. These countries 22 with more elderly people also tend to be well-developed economies representing a higher share of 23 the world GDP. Therefore, if only in terms of maximizing direct economic benefits (lockdown-24 easing benefit), the priority of vaccination should be based on GDP per capita. This is very 25 straightforward and seems to reflect the current vaccine distribution situation. But when we take 26 indirect economic effects (supply-chain rebuilding benefit) into consideration, the results will be 27 different. 28 Finally, compared to the Selfish Distribution Strategy, the overall supply-chain rebuilding benefit 29 has increased by 67.4% under the Altruistic Distribution Strategy and 74.7% under the Altruistic 30 Age-informed Distribution Strategy ( Fig. 1g-i). A better recovery within each country under two 31 Altruistic Distribution Strategies is crucial to the recovery of the global supply chains. For example, 32 Portugal has a high degree of trade openness (the total of imports and exports as a percentage of 33 GDP, 65.6%), meaning that the country has close supply-chain linkages with other countries across 34 the world. By switching from the Selfish Distribution Strategy to the Altruistic Age-informed 35 Distribution Strategy, Portugal would experience the most significant increase in supply-chain 36 rebuilding effect from 0.9% to 5.0% of annual GDP (Fig. 1g,i). Its largest trading partner, Spain, has 37 a lockdown-easing benefit of 6.3% under the Altruistic Age-informed Distribution Strategy scenario, 38 which is 46.2% higher than under the Selfish Distribution Strategy. The recovery of Spain has a 1 positive spillover effect on Portugal. 2 In addition to the difference between Selfish and Altruistic Distribution Strategies, our modeling of 3 the two Altruistic Distribution Strategies, without-and with age adjustment, shows the potential 4 benefit from considering both the population and age structure of the countries when allocating 5 vaccines internationally. Until now, COVAX has not taken age or disease prevalence into account in 6 country allocation despite strongly urging age-based allocation within countries. 7 "Altruistic Distribution Strategy" scenario is the default scenario in this comparison (see 3 Supplementary Table 3-5 for the results of other scenario combinations for all countries). 4 Fig. 2 shows that economic benefits are largest when the priority of domestic vaccine 5 distribution is given to the elderly segment, i.e., 65 years old and above, of the population 6 followed by workforce with high exposure risk, such as workers in transportation, 7 accommodation, and catering industrial sectors. Fig. 2 shows that giving priority to the elderly 8 generally provides higher economic benefits, even though the difference is not as large as could have 9 been expected. For example, the total benefit from vaccination in the USA is about 7.5% 10 (7.2%~8.1%) of annual GDP when the elderly group is prioritized, whereas the total benefits 11 decrease to about 7.2% (6.9%~7.9%) of annual GDP when young people are prioritized. This 12 reduction is mainly explained by the difference in health gains from vaccination. The health benefit 13 from vaccination of the USA population is 1.1% of annual GDP when the vaccine is given first to the 14 elderly, about 4 times higher than the health benefit if priority is given to young people (0.2% of 15 USA annual GDP). The results of other countries also support this conclusion ( Fig. 2b-f, 16 Supplementary Table 3-5). The infection-mortality rate among the elderly is relatively high. 17 Prioritizing the elderly can thus save more lives (see Supplementary Fig. 6) and result in higher 18 health gains. 19 Once the elderly are fully vaccinated, moving to vaccinate workforce in sectors with high exposure 20 risk would bring higher economic benefits than an equally distribution across sectors ('Equally' 21 scenario), through both lockdown-easing benefit and supply-chain rebuilding benefit (Fig. 2). For the 22 United States, Germany, Vietnam, and Brazil, this pattern is more pronounced. For example, the 23 lockdown-easing benefit from vaccination in the USA is 4.6% of annual GDP when workforce with 24 high exposure risk is prioritized (note that 'Altruistic Distribution Strategy' and 'Elderly First' is the 25 default when we discuss results of sectoral distribution), 10.3% higher than the lockdown-easing 26 benefit if vaccinating the whole workforce equally (4.2% of annual GDP). Meanwhile, the supply-27 chain rebuilding benefit from vaccination in the USA is 2.4% of annual GDP when the vaccine is 28 given first to workforce with high exposure risk, 9.1% higher than the lockdown-easing benefit if 29 vaccinating the whole workforce equally (2.2% of annual GDP). The results of other countries also 30 support this conclusion ( Fig. 2b- with high exposure risk can greatly reduce the need of strict lockdown, which is conducive to the 32 recovery of the local economy, and further the global supply chains. 33  Table 1),  35 would maximize economic recovery in those sectors but also create strong positive spillover 36 effects to other production sectors. Generally, the entire economy would obtain the largest benefits 37 when workers in production sectors at high risk of exposure are prioritized (scenario S -High risk; 38 red dots in Fig. 3)this being due to the inter-sector spillovers 22-24 . Prioritizing high-risk groups 1 (red dots in Fig. 3) contributed 0.1%-10.9% extra spillover benefits as compared with an equal 2 distribution strategy across industrial sectors (blue dots in Fig. 3; see Supplementary Table 9). This 3 highlights the importance of considering externalities when designing domestic vaccine allocation 4 strategies. First" scenarios are the default scenario in this comparison. The x-axis represents 10 industrial sectors 10 (see Supplementary Table 10 for sector aggregation information), and y-axis represents the total economic 11 benefit of each industrial sector (expressed as a percentage of the value-added of corresponding industrial 12 sector). Color represents different vaccination scenarios: blue, any available dose will be equally allocated 13 from the outset to the working populations across all economic sectors (Equally); red, any available dose will 14 be first given to the working populations of specific sectors as ranked in terms of the exposure risk (High 15 risk); (3) orange, any available dose will be first given to the working populations in each economic sector 16 based on the proportion of critical workers (Critical, total labor requirements to meet demand for basic 17 necessities, see 25 and Supplementary Table 2). 18 The tighter the association among domestic sectors, the larger the spillover benefits from the 19 recovery of one sector to all other economic sectors. In the well-developed economies, i.e., the USA, 20 Germany, and the UK, the association among domestic sectors is relatively high. Therefore, giving 1 priority to workforce with high exposure risk not only help the rapid recovery of these departments, 2 but the recovery effect will quickly cascade to other sectors. For example, in the USA and Germany, 3 the benefits of 'Grains and Crops' sector would increase by 15.2% and 23.6%, respectively, under 4 'High risk' scenario compared to 'Equally' scenario. This is because machinery is a key input in 5 modern agriculture in well-developed countries 26,27 . The recovery of light and heavy manufacturing 6 industries is conducive to the recovery of the agricultural sector, which may otherwise suffer, e.g., 7 from missing parts for equipment. Given that in low-income countries, the association among sectors 8 along the domestic supply chains is weaker than in developed countries, the spillover benefits would 9 be lower than in high-income countries. For example, Viet Nam and in Brazil would see the spillover 10 benefit to the agriculture sector increasing by only 11.9% and 6.6%, respectively, under 'High risk' 11 scenario compared to 'Equally' scenario. 12 We also analyzed the benefits under the 'Critical' scenario (orange dots in Fig. 3) in which any 13 available dose will be firstly given to the working populations in terms of the critical worker 14 proportion in each industrial sector (see Supplementary Table 2). For example, compared to the 15 recreation sector, workers in the food manufacturing sector will be given priority. Fig. 3 shows that 16 the benefits of the 'Critical' scenario (orange dots) are generally the lowest among the three 17 vaccination strategies, which indicates that there is a trade-off between guaranteeing food/daily 18 necessities and overall economic recovery. Therefore, in the most urgent situation, we should give 19 priority to workforce in the critical sector. But after the food and necessities are met, we should give 20 priority to workforce in high-risk sectors to optimize the economic recovery. 21

Discussion 22
Our modeling of vaccines distribution demonstrates the potentially significant differences in the 23 socioeconomic benefits brought by different global and domestic vaccines distribution modes. 24 Insights extracted from our scenario analysis suggesting that economic benefits will be maximized 25 by giving priority to vaccinating high-risk populations all over the world, including the elderly with 26 high infection-mortality rate and workers with a high risk of exposure. Indeed, recent results from 27 related research seem to support this same conclusion 28,29 . Our modeling of the benefits of 28 vaccination in terms of health gains, lockdown-easing benefit, and supply-chain rebuilding benefit, 29 quantifies the spillover effects of vaccination providing further support, beyond the arguments in 30 terms of infection-mortality rate and probability of exposure, to the choice of an altruistic vaccine 31 distribution strategy across the world. 32 Our study has some limitations. We do not build a feedback mechanism between the epidemiological 33 and the economic model, i.e., the interaction between the intensity of economic activity and the 34 spread of the virus. We acknowledge that a feedback mechanism is theoretically feasible. The current 35 practical knowledge in this area, however, is still very limited, which means that the introduction of a 36 feedback mechanism will bring about very large uncertainties. Our model is also limited by taking no 37 consideration of technological changes and adjustment of behaviors and by assuming that production 1 and consumption patterns remain the same as pre-crisis. Our model has a focus on the short-term 2 scenarios, and therefore the above two assumptions are rather unlikely to have a significant impact 3 on the results 1 . Our model is further constrained by the trade relationship at the sectoral level among 4 countries, and has no ability to capture the complexity of supply-chain networks at the firm level and 5 may therefore underestimate the benefits 30 . In addition, this study only focuses on economic benefits. 6 We acknowledge that maximize the aggregate benefit is not the only criteria that need to be 7 accounted for, but also fairness and feasibility. But these are beyond the scope of this study. 8 While our scenarios are idealized cases, the current vaccine distribution is closer to the Selfish 9 distribution one 6,7 . As a final analysis, we explore opportunities for Pareto improvement with 10 an improved distribution of vaccines. In order to simplify the analysis, 1) we group all countries 11 into three categories: (i) the major vaccine-producing countries; (ii) the non-vaccine-producing 12 countries with high-income levels (>US$4046 per year; World Bank high-income and upper-middle-13 income countries); and (iii) the non-vaccine-producing countries with low-income levels (<US$4046 14 per year; World Bank low-middle-income and low-income countries); 2) we assume that, in our 15 scenarios, vaccines are traded at their production costs so that vaccine-producing countries do not 16 benefit from vaccine sales (mimicking the current situation of vaccine-producing countries donating 17 vaccines to low-income countries). 18 Figure 4 shows that globally the Altruistic Distribution Strategy would create 6.8% additional 19 benefits as compared to the Selfish Distribution Strategy, implying that there is room for 20 Pareto improvement from the Selfish Distribution Strategy (i.e., a change in which the 21 reallocation of vaccines can make at least one group better off without making any of them 22 worse off). Flows of money and vaccines without any benefit-sharing mechanism are shown in Fig.  23 4a. In this case, the distribution of vaccines is quite unequal due to differences in consumption 24 capacity of different groups of countries. If the vaccine-producing countries choose the Selfish 25 Distribution Strategy, the world's annual average benefit from vaccination is estimated at US$8.10 26 trillion (about two times the annual US government spending; Fig. 4b). The benefit to the vaccine-27 producing countries is US$5.31 trillion, while the benefit to non-vaccine-producing countries with 28 high-income and low-income levels are US$2.23 trillion and US$0.55 trillion, respectively. When 29 vaccine-producing countries adopt the Altruistic Distribution Strategy, the world's annual average 30 benefit from vaccination is US$8.65 trillion (Fig. 4b). In this case, the benefit to the vaccine-31 producing country is US$4.58 trillion, whereas the benefit of non-producing countries with high-32 income levels and low-income levels are US$3.23 trillion and US$0.84 trillion, respectively (Fig.  33 4b). Based on these results, it appears clear why the current global vaccine distribution tends to 34 be a Selfish Distribution Strategy rather than an Altruistic Distribution Strategy even if only 35 economic benefits are considered. That is, without any benefit-sharing mechanism, vaccine-36 producing countries are more willing to choose the Selfish Distribution Strategy that is most 37 beneficial to themselves, while other countries have no option but to accept an unequal distribution 38 of vaccines. Note that political pressure faced by governments of vaccine-producing countries to 39 prioritize their population before exporting, and even the rise of the so-called "vaccine nationalism" 31 1 during the COVID-19 pandemic can also be major reasons for the current unequal distribution 2 situation, which is out of the scope of this study. 3 This dilemma has reproduced the current unequal situation. That is, vaccine-producing countries 4 prefer to give priority to vaccinating their residents, high-income non-producing countries buy large 5 amounts of vaccines for their domestic use, while, middle-and low-income non-producing countries 6 can only obtain very few vaccines due to their insufficient consumption capacity. The reason for this 7 situation is that the benefits of health (health gains) are straightforward and easy to be taken into 8 consideration, but the benefits of the economic recovery (lockdown-easing benefit and supply-chain 9 rebuilding benefit) are often not well understood and considered. A key significance of the 10 quantitative analysis presented in this article is that it provides countries with a comprehensive 11 understanding of their potential payoffs in the global vaccine distribution game, which is a 12 prerequisite for players to make the right decision. 13 14

Fig. 4 | Vaccination-related benefits of three groups under different global vaccine-distribution 15
strategies and the potential incentives (benefit-sharing mechanism) to promote a more 16 equitable distribution. a) shows flows of money and vaccines without any benefit-sharing 17 mechanism; b) shows the benefits of the three groups of countries without any benefit-sharing 18 mechanism; c) shows flows of money and vaccines with the proposed benefit-sharing mechanism; d) 19 shows the new situation (payoffs of the three groups of countries under different distribution 1 strategies) with the benefit-sharing mechanism. The yellow money symbol in Fig.4a and c indicates 2 money used to buy vaccines and the bule vaccine symbol in Fig. 4a and c indicates vaccines  3 purchased. The red money symbol with a hand below in Fig. 4c indicates money donations and red 4 vaccine symbol with a hand below in Fig.4c indicates vaccine donations. The "Selfish" and 5 "Altruistic" in Fig.4b and d represent the "Selfish Distribution Strategy" scenario and "Altruistic 6 Distribution Strategy" scenario. The number on the horizontal bars in Fig.4b and d indicates the 7 benefit (expressed in trillion US dollars). The light red horizontal bars in Fig.4d represent the aids 8 from high-income countries to promote a more equitable distribution of vaccines around the world. 9 The 'Elderly First' and 'High risk' scenario is the default scenario in this comparison. 10 Fig. 4b, we propose a benefit-sharing 11 mechanism, shown in Fig. 4c, that can incentivize vaccine-producing countries to share 12 vaccines early and promote global vaccine distribution towards a "win-win" equilibrium. As 13 shown in Fig. 4c, through the proposed benefit-sharing mechanism based on an international 14 platform, high-income non-producing countries donated vaccine aid to the platform to seek a 15 globally equitable distribution of vaccines. Vaccine-producing countries deliver vaccines to the 16 platform and obtain corresponding benefits. Middle-and low-income countries actively cooperate 17 with the platform in completing vaccine delivery and capacity building. The basis for this proposal to 18 become economically rational is the positive externality of vaccination implicit in the global supply 19 chain recovery. Note that the benefit-sharing mechanism proposed here only provides a potential 20 economically rational way of international cooperation on the basis of the modeling of the externality 21 of vaccination. It does not mean that the current situation is not an economically rational equilibrium, 22

Based on the distribution of benefits shown in
while it highlights that the current situation has room for Pareto improvement through cooperation. 23 Figure 4d shows that the three parties are willing to implement this mechanism when the 24 donation required is within a certain amount. High-income non-producing countries can share 25 part of the additional benefits gained as a result of the Altruistic Distribution Strategy (US$0.87 26 trillion) with vaccine-producing countries in order to persuade vaccine-producing countries to choose 27 the Altruistic Distribution Strategy. If the extra cost is less than US$1.00 trillion, high-income non-28 producing countries are willing to do so because their benefit will be still greater than the benefit in 29 the Selfish Distribution Strategy after they have paid the cost. Meanwhile, if the transfer is greater 30 than US$0.73 trillion, vaccine-producing countries will be willing to choose the Altruistic 31 Distribution Strategy because their benefit will exceed the benefit of choosing a Selfish Distribution 32 Strategy (US$5.31 trillion). And undoubtedly, middle-and low-income countries are willing to 33 receive vaccines or build local production capacity as these are beneficial to them. With this 34 incentive mechanism, the benefits of the three groups of countries will be improved simultaneously. 35

Our quantification of benefits shows that only when the incentive reaches a certain level can all 36
groups achieve the "win-win" situation. Current proposal made by G7 countries to provide US$10 37 billion to COVAX 32 is, however, insufficient to motivate vaccine-producing countries to largely 38 distribute the vaccines to mid-and low-income countries. To ease off the large divide of vaccine 1 distribution, global governance is needed. On the other hand, the benefit-sharing mechanism 2 proposed here is based on the perspective of externalities and market failures. The specific measures 3 are not necessarily in monetary form. High-income countries would need to provide necessary 4 capacity building to key personnel in establishing production facilities in mid-and low-income 5 countries, where the local government would need to provide necessary space and tax waiving 6 mechanisms for fast and scale productions in order to minimize the cost of vaccinations (including 7 the manufacturing, transportation and logistics, and implementing). All these actions that can 8 increase global vaccine production capacity and reduce distribution costs are part or complementary 9 of the benefit-sharing mechanism proposed here. 10 In preparing for the next pandemics, a global benefit-sharing instrument should be developed so as to 11 remove some of the disincentives for early equitable vaccines distribution globally. Such an 12 instrument would provide enormous global health and economic benefits in an economically 13 sustainable manner. 14 15

Vaccine-distribution scenario sets 17
To evaluate the economic benefits of allocating the vaccines across the globe, we propose three 18 scenario sets which are designated in a tiered structure. Basically, Tier Global scenario sets address 19 the issue of the cooperative attitude of vaccine exporting countries and importing countries, while 20 Tier Domestic scenario sets address the issue of allocating the received vaccines within the 21 destination countries. We treat each scenario set as an individual parameter in the model, such that 22 we will have three parameters (i.e., C, A, S). We vary the value of each parameter by considering 23 different sub-scenarios within each scenario set (see Box 1, Supplementary 错误!未找到引用源。). 24 In summary, parameter C indicates whether the vaccine exporting country is more willing to share 25 the vaccine with other countries. Considering labors in different sectors, critical workers first or mass distribution. The acronym S is 22 'industrial sector'. 23 • High risk: Any available dose will be firstly given to the working populations in terms of the 24 exposure risk rank of economic sectors. 25 • Equally: Any available dose will be firstly given to the working populations equally distributed 26 to all economic sectors. 27 • Critical: Any available dose will be firstly given to the working populations in terms of the 28 critical worker proportion in each economic sector. 29 Combination of each variation of above 3 parameters gives a distribution strategy. In this analysis, 30 we will have 3 x 3 x 3 = 27 scenarios. When comparing the results of a scenario set, "Altruistic 31 Distribution", "Elderly first", and "High risk" are used as the default scenario. 32

Estimation of vaccine production capacity 33
We consider seven major vaccine-manufacturing counties, including China, US, Germany, India, 34 UK, Netherlands, and Russia 33 . We collected the current vaccine production capacity of these 35 countries, and based on this, we predicted the future vaccine production capacity. 36 • The overall capacity of all manufacturing countries in the "Approved in use" development 1 stage and in the future (i.e., 2022-2023) is collected from the United Nations International 2 Children's Emergency Fund (UNICEF) 33 . (see Supplementary Fig. 6-7, Supplementary  3 Table 11) 4 • With the data, we project the annual capacity of all manufacturing countries by using the 5 logarithmic function to fit the growth trend of capacity. (Supplementary Fig. 6) 6 • Assuming an invariant relative capacity across countries over time, we further partition the 7 annual capacity to each of the manufacturing country according to their capacity 8 documented on March 3, 2021. (Supplementary Table 11, Supplementary Fig. 7) 9 The epidemiological model 10 Model structure. Built upon our age-structured SIR model 19 , we project the fraction of incidence and 11 mortality over age groups by using chains of differential equations: 12 where , are the number of susceptible individuals and primary infections in age group ; 18 accordingly, is the number of non-primary infections. The recovered individuals ( ) may lose 19 immunity and return to susceptibility ( ) after an average duration of immunity of ⁄ . The 20 force-of-infection on susceptible in age-class is designated as = ∑ ( + )⁄ , where 21 = is the baseline rate of transmission and is the contact rate between age group and . 22 ⁄ to be the average duration of infection which is taken to be 7 days 34 . For simplicity, we assume 23 a uniform 1-year duration rate of aging ( ) across ages, i.e., = for all . Assuming 0 = 3.5, 24 we parameterize the model with country-specific population pyramid 35 and social mixing 36   , respectively 37,38 ; is the vaccine efficacy. We do not explicitly model the timing of the 1 vaccination campaign; instead, we assume that vaccines are uniformly distributed over the year. 2 With the simulation we estimate the age-specific fraction of infection and further infer the fraction of 3 deaths by multiplying the age-infections with infection fatality ratio (IFR) 39 . Considering the current 4 spread and variation of COVID-19, we used the current infection and death data to scale up the 5 simulation results. By comparing the results of scenarios with or without vaccines, we can obtain the 6 benefits of different vaccine distribution strategies. 7 Model assumptions. To appropriately lay out our insights, we make several assumptions. First, we 8 assume the homogeneous susceptibility to infection, clinical fraction and infection vs case-fatality 9 ratio as well as the immunity-dependent infectiousness of reinfection across age classes. 10 Additionally, we assume a one-year duration of immunity, given the brief immunity from natural 11 infection of seasonal coronavirus 40 . Moreover, we assume a uniform distribution of vaccine roll-out 12 over the year. Relaxing the assumptions by explicitly consider age-specific heterogeneities, differing 13 durations of immunity and the timing of vaccination are easy extensions giving the general nature of 14 our model framework. 15 The estimation of VSL 16 The value of statistical life (VSL) is widely used throughout the world to monetize fatality risks in 17 benefit-cost analyses 41 . The VSL represents the individual's local money-mortality risk tradeoff 18 value 42 , which is the value of small changes in risk not the value attached to identified lives. The 19 country based VSL estimation used in this research is adopted from the COVID-19 global health 20 risks pricing study by Viscusi (Table 6) 43 . The estimation is based on the estimated VSL in the U.S. 21 (11 million in 2019 US dollor). Based on this, we do two VSL estimation. We first use an income 22 elasticity (=1.0) to adjust the VSL to other countries using the fixed effects specification 44 . 23 Supplementary Fig. 10, shows the spatial distribution of estimated VSL for 175 countries used in this 24 approach. And then, to keep in line with the idea that every life is equal, we value each life equally 25 with a global average a uniform global VSL (= 11 USD million times average global GDP per 26 capita/US GDP per capita; 2.94 USD million). 27

Estimation of required strictness of control measures 28
COVID-19 has resulted in varying degrees of social lockdown in countries all around the world. The 29 lockdown strictness by each country is measured by the percentage by which labour availability and 30 transportation capacity are reduced relative to pre-pandemic levels. The Google Community 31 Mobility Reports (COVID-19 Community Mobility Reports (google.com)), which aim to provide 32 insights into changes in response to policies aimed at combating COVID-19, are used to measure the 33 strictness specifically. The reports chart movement trends over time by geography, across different 34 categories of places such as retail and recreation, groceries and pharmacies, parks, transit stations, 35 workplaces, and residential. We averaged the changes in the five types of visitors as a parameter to 36 measure the extent of a country's lockdown strictness. The data are monthly, starting in February 37 2020 and the latest up to April 2021. For countries where Google data are not available, we 1 supplement them with data from the nearest country based on geographical location. 2 The recursive dynamic disaster impact assessment model (estimation of lockdown-easing effect 3 and supply-chain rebuilding benefit) 4 In addition to the life-saving benefits calculated by infectious disease models, vaccine distribution 5 also generates lockdown-easing effect and supply-chain rebuilding benefits through the global 6 supply chains. The global economic benefits will be calculated using the following recursive 7 dynamic disaster impact assessment model 1,20,21,45 . The simulation code and examples can be found 8 in GitHub (https://github.com/DaopingW/economic-impact-model). 9 We do two simulations and compared the results to obtain the benefits of vaccination. The first 10 simulation is the counter factual scenario, i.e., a world with no vaccines at all. The results of the first 11 simulation represent global economic loss (includes direct lockdown losses and supply-chain 12 propagations damages) if there is no vaccine. The second simulation is used to calculate the global 13 economic loss under a specific vaccine distribution scenario. The amount by which the loss of the 14 second simulation is less than the loss of the first simulation is defined as the economic benefit of 15 vaccination. 16 In detail, our disaster impact assessment model is an extension of the adaptive regional input-output 17 (ARIO) model 20,21 , which was widely used in the literature to simulate the propagation of negative 18 shocks throughout the economy 1,5,30,46 . Our model improves the ARIO model in two ways. The first 19 improvement is related to the substitutability of products from the same sector sourced from different 20 regions. Second, in our model, clients will choose their suppliers across regions based on their 21 capacity. These two improvements contribute to a more realistic representation of bottlenecks along 22 global supply chains. 23 Our disaster footprint model mainly includes 4 modules, i.e., production module, allocation module, 24 demand module and simulation module. The production module is mainly designed for 25 characterizing the firm's production activities. The allocation module is mainly used to describe how 26 firms allocate output to their clients, including downstream firms (intermediate demand) and 27 households (final demand). The demand module is mainly used to describe how clients place orders 28 to their suppliers. And the simulation module is mainly designed for executing the whole simulation 29 procedure. 30 Production module. The production module is used to characterize production processes. Firms rent 31 capital and employ labour to process natural resources and intermediate inputs produced by other 32 firms into a specific product (see Supplementary Fig. 11). The production process for firm can 33 expressed as follows, services to satisfy demand from their clients. After a disaster, output will decline. From a production 17 perspective, there are mainly the following constraints: 18 Labour supply constraints. Labour constraints after a disaster may impose severe knock-on effects 19 on the rest of the economy 20,48 . This makes labour constraints a key factor to consider in disaster 20 impact analysis. For example, in the case of a pandemic, these constraints can arise from employees' 21 inability to work as a result of illness or death, or from the inability to go to work and the 22 requirement to work at home (if possible). In this model, the proportion of surviving productive 23 capacity from the constrained labour productive capacity ( ) after a shock is defined as: 24 Where ( ) is the proportion of labour that is unavailable at each time step during containment. 26 (1 − ( )) contains the available proportion of employment at time . 27 ( ) = ( ̅ − ( ))/ ̅ 28 The proportion of the available productive capacity of labour is thus a function of the losses from the 1 sectoral labour forces and its pre-disaster employment level. Following the assumption of the fixed 2 proportion of production functions, the productive capacity of labour in each region after a disaster 3 ( ) will represent a linear proportion of the available labour capacity at each time step. Take  4 COVID-19 as an example, during an outbreak of an infectious disease, authorities often adopt social 5 distancing and other measures to reduce the risk of infection. This imposes an exogenous negative 6 shock on the economic network. 7 Constraints on productive capital. Similar to labour constraints, the productive capacity of 8 industrial capital in each region during the aftermath of a disaster ( ) will be constrained by the 9 surviving capacity of the industrial capital 46,49,50 . The share of damage to each sector is directly 10 considered as the proportion of the monetized damage to capital assets in relation to the total value of 11 industrial capital for each sector, which is disclosed in the event account vector (EAV) for each 12 region ( ), following 50 . This assumption is embodied in the essence of the IO model, which is 13 hard-coded through the Leontief-type production function and its restricted substitution. That is, as 14 capital and labour are considered perfectly complementary as well as the main production factors, 15 and the full employment of those factors in the economy is also assumed, we assume that damage in 16 capital assets is directly related with production level and therefore, value added level. Then, the 17 remaining productive capacity of the industrial capital at each time step is defined as: 18 Where, ̅ is the capital stock of firm in the pre-disaster situation, and ( ) is the surviving capital 20 stock of firm at time during the recovery process. The actual production of firm , ( ), depends on both its maximum supply capacity and the total 1 orders the firm received from its clients (see the Demand Module), 2 The inventory held by firm will be consumed during the production process, 4 , ( ) = * ( ) 5 Allocation module. The allocation module mainly describes how suppliers allocate products to their 6 clients. When some firms in the economic system suffer a negative shock, their production will be 7 constrained by a shortage to primary inputs such as a shortage of labour supply in the outbreak of 8 COVID-19. In this case, a firm's output will not be able to fill all orders of its clients. A rationing 9 scheme that reflects a mechanism based on which a firm allocates an insufficient amount of products 10 to its clients is needed. For this case study, we applied a proportional rationing scheme according to 11 which a firm allocates its output in proportion to its orders. Under the proportional rationing scheme, 12 the amounts of products of firm allocated to firm and household ℎ is as follows, 13

1
Households issue orders to their suppliers based on their demand and the supply capacity of their 2 suppliers. In this study, the demand of household ℎ to final products , ℎ ( ), is given 3 exogenously at each time step. Then, the order issued by household ℎ to its supplier is 4 The total order received by firm is 6 Simulation module. At each time step, the actions of firms and households are as follows: 8 1. Firms plan and execute their production based on three factors: a) inventories of intermediate 9 products they have, b) supply of primary inputs, and c) orders from their clients. Firms will 10 maximize their output under these constraints. 11 2. Product allocation. Firms allocate outputs to clients based on their orders. In equilibrium, the 12 output of firms just meets all orders. When production is constrained by exogenous negative 13 shocks, outputs may not cover all orders. In this case, we use a proportional rationing scheme 14 proposed in the literature 20,21 (see Allocation Module) to allocate products of firms. 15 3. Firms and household issue orders to their suppliers for the next time step. Firms place orders 16 with their suppliers based on the gaps in their inventories (target inventory level minus existing 17 inventory level). Households place orders with their suppliers based on their demand. When a 18 product comes from multiple suppliers, the allocation of orders is adjusted according to the 19 production capacity of each supplier. 20 This discrete-time dynamic procedure can reproduce the equilibrium of the economic system, and 21 can simulate the propagation of exogenous shocks, both from firm and household side, or 22 transportation disruptions, in the economic network. From the firm side, if the supply of a firm's 23 primary inputs is constrained, it will have two effects. On the one hand, the decline in output in this 24 firm means that its clients' orders cannot be fulfilled. This will result in a decrease in inventory of 25 these clients, which will constrain their production. This is the so-called forward or downstream 26 effect. On the other hand, less output in this firm also means less use of intermediate products from 27 its suppliers. This will reduce the number of orders it places on its suppliers, which will further 28 reduce the production level of its suppliers. This is the so-called backward or upstream effect. 29 Similarly, these two effects can also occur if the transport of a firm to its clients or suppliers is 30 restricted. For instance, during the outbreak of COVID-19 in China, the authorities adopted strict 31 isolation measures. These measures have placed constraints on the supply of labour and the 1 transportation of products. This led to a decline in China's output and also triggered the forward and 2 backward effect, which make the shock to propagate to the global economic network. From the 3 household side, the fluctuation of household demand caused by exogenous shocks will also trigger 4 the aforementioned backward effect. Take tourism as an example, during the outbreak of COVID-19 5 in China, the demand for Chinese tourism from households all over the world will decline 6 significantly. This influence will further propagate to the accommodation and catering industry 7 through supplier-client links. 8 Economic footprint. We define the value-added decrease of all firms in a network caused by an 9 exogenous negative shock as the disaster footprint of the shock. For the firm directly affected by 10 exogenous negative shocks, its loss includes two parts: a) the value-added decrease caused by 11 exogenous constraints, and b) the value-added decrease caused by propagation. The former is the 12 direct loss, while the latter is the indirect loss. A negative shock's total economic footprint ( , ), which contains 65 production sectors. If we treat each sector as a firm (producer), and assume that 24 each region has a representative household, we can obtain the following information in the MRIO 25 table: a) suppliers and clients of each firm; b) suppliers for each household, and c) the flow of each 26 supplier-client connection under the equilibrium state. This provides a benchmark for our model. 27 When applying such a realistic and aggregated network in the disaster footprint model, we need to 28 consider the substitutability of intermediate products supplied by suppliers from the same sector in 29 different regions. The substitution between some intermediate products is fairly straightforward. For 30 example, for a firm that extracts spices from bananas it does not make much of a difference if the 31 bananas are sourced from the Philippines or Thailand. However, for a car manufacturing firm in 32 Japan, which use screw from Chinese auto parts suppliers and engines from German auto parts 33 suppliers to assemble cars, the products of the suppliers in these two regions are non-substitutable. If 1 we assume that all goods are non-substitutable as in the traditional IO model, then we will 2 overestimate the loss of producers such as fragrance extraction firm. If we assume that products from 3 suppliers in the same sector can be completely substitutable, then we will significantly underestimate 4 the losses of producers such as Japanese car manufacturing firm. In order to alleviate the 5 shortcomings of the evaluation deviation under the two assumptions, we set the possibility of 6 substitution for each firm based on the region and sector of supplier supply (see Allocation Module 7 of the model). 8 9