COVID-19 underscores the urgency of just transition alongside green recovery

12 13 Green recovery has been highly advocated as a promising strategy to balance climate actions and 14 economic reset after COVID-19. However, the potential inequality risk associated with the green 15 recovery hasn’t been fully assessed. Here, enabled by an extended adaptive regional input-output 16 (E-ARIO) model, we quantify the short-term impacts of COVID and various recovery packages on 17 labor demand and income equality. The findings reveal that in the pandemic, low- and medium- 18 income labor suffered more income decrease (by 36%) than those with high-level income (by 24%), 19 leading to a 24% increase of income inequality at the global level (measured by the Oshima 20 coefficient). The high-income labor benefits more from a low-carbon pathway to economic recovery, 21 which further exacerbates the income inequality across the world by 3~5% compared to those in a 22 traditional, carbon-intensive recovery scenario. The findings reveal the tradeoffs between income 23 equality and green development and underscore the urgency of just transition alongside green 24 recovery. 25


28
COVID-19 adds unprecedented health and economic challenges to the existing poverty and 29 inequality in the world. The pandemic is aggravating economic divisions, which, in turn, worsens 30 the negative effect of the crisis. On the one hand, the poor and the vulnerable are more likely to 31 suffer income loss from distancing measures and economic recessions. This is because these 32 people's work is usually labor-dependent (such as planting and construction) or require face-to-face 33 contact with others (e.g., accommodation and restaurant service), which makes it less likely to work 34 remotely from home 1 . According to the World Bank's report 2 , COVID-19 put 71 million people into 35 extreme poverty in the baseline scenario, and the number reaches up to 100 million in a downward 36 scenario. On the other hand, economic inequality weakens the societies' resilience to pandemics 37 since it acts as a multiplier on the virus' spread speed and mortality rate 3 . People with lower 38 socioeconomic status have to continue working in an environment with a higher level of exposure 39 to the virus and have less access to preventive protection 4 . If, unfortunately, infected by the virus, 40 they have higher rates of death due to unaffordable health care costs and the accompanying chronic 41 diseases associated with poverty 5,6 . The self-reinforce feedback loop reveals the necessity and 42 urgency of protecting the poor and the vulnerable after the COVID-19 pandemic 7 . 43 44 Meanwhile, the pandemic knocked climate change down the agenda. COVID-19 has striking 45 similarities with climate change because both are irreversible, spreading across country borders, 46 exerting uneven impacts among people, and less costly to prevent than to cure 8,9 . However, huge 47 differences exist as well: the pandemic occurs anytime with rapid expansion and direct cause-effect 48 relationships, while climate change is a slow process with ambiguous and controversial attributes 8 . 49 Such difference might lead to a viewpoint that the current world should prioritize battling COVID, 50 improving health, restoring jobs, and stabilizing the economy over climate change mitigation 10,11 . 51 However, others argue that the urgent need for economy reboot doesn't mean a delay in climate 52 change mitigation but underscores the necessity to accelerate the process 12 . How governments spend 53 billions of fiscal recovery money in recent years will determine the trend of climate change in the 54 next few decades. The committed emissions of carbon-intensive investments in post-COVID-19 55 economic recovery might jeopardize the Paris Agreement goals because of the carbon "lock-in" 56 effect of infrastructure 13 . Consequently, it is vital to make the right decisions to tackle the compound 57 climate risks in the pandemic. 58 59 Green recoveries are called for as a solution to balance climate actions with economic recovery. 60 Researchers have pointed out that green investment not only benefits the environment but also 61 flattens the economic curves and creates job opportunities 14-16 . The multiplier effects of green 62 recovery packages on economic reboots and job creation can be competitive, or even superior to, 63 traditional carbon-intensive stimulus pathways 17 . The advocacy of green recovery and the focus on 64 the possibilities of the co-benefits dominates current discussions on economic reset, leaving the 65 potential risks overshadowed. The asymmetric information description and delivery might cause 66 biased perception and improper decision making. 67 68 One of the potential risks associated with green recovery is its impacts on social equity 18 . It has been 69 widely acknowledged that low-carbon transition will bring about structural changes in labor demand 70 and possible risks of 'structural unemployment' 19,20 . The transition needs a painful period where 71 low-skilled labor and people whose livelihoods depend on fossil fuel energy suffer wage reductions 72 and unemployment, exacerbating social inequality at the time. Later, with the improvement of 73 production efficiency and the continuous absorption of unemployed labor by other sectors, social 74 inequity will be alleviated. Although social justice considerations are not novel, the pandemic 75 fundamentally changes its nature, and the scale of the equity challenge remains unclear. The 76 pandemic has reduced society's tolerance for the duration and extent of the challenging period. Any 77 further deterioration can become the last straw that breaks the camel's back. Therefore, we need to 78 rethink and comprehensively assess how green recovery packages affect social equity after the 79 pandemic. Policymakers need to know who is most affected by the pandemic, to what extent these 80 groups benefit from recovery policies, and how to avoid stark inequality while rebooting the 81 economy.

83
This research addresses the social equity concern and reveals the severe structural weakness of green 84 recoveries belied in the win-win potentials of economic growth and green development. Enabled by 85 an extended adaptive regional input-output (E-ARIO) model, we quantify the short-term impacts of 86 COVID and various recovery packages on social equity through the changes in income and labor 87 demand. The findings demonstrate that the pandemic has an uneven impact on the labor market, 88 with more negative impacts on lower-skilled and lower-income groups but less on high-skilled and 89 higher-income groups. The less affected population, however, receives more assistance in green 90 recovery plans compared to those in traditional recovery plans, leading to an increase in global 91 income inequality. The findings highlight the importance of just transition alongside green recovery 92 and provide new insights for developing green recovery strategies.

94
Who suffers the most from the pandemic recession?

96
Although most people's life and work have been negatively affected by the pandemic, low-and 97 medium-skilled labors are more affected than those higher-skilled ones (Fig.1). Globally, more than 98 86% of the reduced labor demands are low-and medium-skilled workers, who account for 83% of 99 the global labor market (Fig.1a). Due to the decrease in labor demand, the average income of low-100 and medium-skilled workers decrease by more than 32%, 6% higher than the decrease rate of the 101 average income of high-skilled workers (Fig.1b). Assuming that the unemployment risk is 102 proportional to the reduction of labor demand, the unemployment risks faced by low-and medium-103 skilled workers in the pandemic are about 1.2 times that of high-skilled workers. 104 105 Fig.1 The impacts of COVID-19 on labor demand and average income. Graph (a) shows the 106 structure of the labor force in each region in the initial situation (left bars) and the structure of the 107 labor force affected by the COVID-19 lockdown (right bars). Graph (b) describes the income change 108 of different skill groups (displayed as points) and the average level (displayed as bars) in each region 109 in the lockdown period.

111
At the national scale, the uneven impacts of the pandemic on the labor market are also evident, albeit 112 with a different extent across countries. In China, the average income of low-and medium-skilled 113 workers, who account for 95% of the labor demand reduction, decrease by more than 41%. In 114 contrast, the average income of high-skilled workers only decreases by about 29%. The 115 unemployment risks that low-and medium-skilled workers faced is 1.3 times those of high-skilled 116 ones. In the United States (USA), 71% of the reduced labor demand is low-and medium-skilled, 117 who account for 6% and 58% of the labor market, respectively. The average income of the low-and 118 medium-skilled workers decrease by about 26% in the pandemic while the figure for high-skilled 119 workers is less than 20%. As for the EU, low-and medium-skilled workers account for 64% of the 120 reduced labor demand, of whom the average income loss is about 40%, 16% higher than high-skilled 121 workers.

123
The most affected industries at the global level are low-and medium-income ones, whose 124 employees have limited ability to resist the impacts (Fig.2). Before the COVID, 26% and 38% of 125 the global industries are low-and medium-income industries, and 36% are high-income (see more 126 details of the sector classification by income level). Among the industries with a substantial decline 127 in average income (the decline rate is more than the sectorial average), 36% are low-income 128 industries, 46% are middle-income industries, and 18% are high-income industries. The average 129 wage of the low-and medium-income industries decreased by 36%. In particular, low-income 130 agriculture industries, including fruit and vegetable planting, cereal grains planting and farming, 131 suffer particularly heavy losses due to the shutdown of the transportation industry and downstream 132 processing industry. At the national level, the situations distinct across counties (Fig.2). In China, 61% of the low-income 142 industries suffer substantial income decrease (higher than the average level of income decrease) 143 while the proportion is only 34% for high-income industries. In the EU, the average wage of low-144 and medium-income industries decrease by 40% while that of high-income industries decrease by 145 34%. The situation is slightly different in the USA, where high-income industries suffer as many 146 negative impacts as low-and medium-income industries. About 77% of the low-income industries, 147 54% of the middle-income industries, and 68% of the high-income industries in the USA went 148 through substantial income decrease.

150
Such results imply that the pandemic has an uneven impact on the labor market, with more negative 151 impacts on low-and medium-income groups. The finding implies that the pandemic may exacerbate 152 income inequality. After calculating the Oshima coefficients (a measurement of income equality) in 153 countries, we find that this implication is supported at the global scale, but the situations vary across 154 countries. For example, the Oshima coefficients increase by 24% at the global level, increase by 16% 155 and 29% in China and the EU, but decreases by 4% in the USA. The decrease in the USA, which 156 implies slight elimination in income equality, is more or less out of expectation 4 . The contradictory 157 result might be explained that our estimation only captures the impacts of COVID on income 158 equality through lockdown measures on labor supply and consumer demand. Other influencing 159 channels on inequality, including unaffordable economic burden brought by the access to healthcare, 160 healthcare spending and overcrowded housing conditions 4 , are not included in our estimation, which 161 might underestimate the income inequality in the pandemic. Who will benefit more from a green recovery?

168
We designed four scenarios to simulate the impacts of economic recovery policy packages on 169 economic growth and labor demand (see more details in the Methods). The four scenarios are the 170 business as usual scenario (BAU), traditional scenario (TES), low-carbon scenario (LCS), and low-171 carbon and digital scenario (LDS). The results reveal some common implications across the world.

173
First, green recovery plans show comparable or even better multiplying effects on economic growth 174 and labor demand compared to the traditional scenarios (Fig.4). A stimulus package equal to 10% 175 of national GDP drives a 10%~14% increase of GDP under the three stimulating scenarios. The 176 differences in economic stimulus between traditional recovery and green recovery are less than 0.2%. 177 Regarding the impacts on employment, differences are minor too. represents business-as-usual scenario. TES represents traditional recovery plan. LCS represents 189 low-carbon recovery plan. LDS represents low-carbon and digital recovery plan.

191
The second similarity shared by most of the countries is that high-skilled workers benefit more from 192 green recovery compared to traditional recovery plans. On the global scale, high-skilled workers 193 account for 23% of the additional job creation in the LDS scenario, which is 11% higher than that 194 in the TES scenario. At the national level, the proportion of high-skilled jobs in total job demand 195 increase in the LDS scenario is 4%, 14%, and 18% higher than that in the TES scenario in China,196 the USA, and the EU, respectively. The benefits of green recovery on high-skilled workers are also 197 apparent from the perspective of income change. In the LDS scenario, the income of high-skilled 198 works is 5% higher than that in the TES scenario, while the difference between these two scenarios 199 for low-and medium-skilled workers are imperceptible.

201
At the sectoral level, the TES scenario favors three sectors whose job demands are most affected by 202 the pandemic: the construction industry, mining of copper ores and concentrates industry, and the 203 land transport industry. In this scenario, the average revenues in these three sectors decrease by 37%, 204 46%, and 26%, respectively. The LCS and LDS scenarios favor the sector of telecommunication 205 and education, which are affected less in the pandemic. These two sectors account for an increase 206 of 14~16 million new jobs in the green recovery scenario. However, the most affected sectors, 207 including the industries of fruit and vegetable planting and hotels and catering, only create 5~8 208 million new jobs. As the nature of job changes, about 120 million people worldwide (4% of the 209 initial state employment) may need a career transition.

211
In general, economic recovery offset some adverse effects of the pandemic on income inequality. 212 However, compared with the TES scenarios, the LCS and LDS scenarios generally increase income 213 equality. On the global level, the Oshima coefficients in LCS and LDS scenarios increase by 3~5% 214 compared to the TES scenario. This is consistent with the observation that LCS and LDS scenarios 215 provide more benefits for high-skilled workers and high-income sectors compared to TES scenarios. 216 At the country level, this finding is also valid, albeit with some exceptions. For example, in the EU 217 and the USA, the TES brings more inequality than LCS and LDS, which can be explained by the 218 limited pulling effect of TES scenarios on job creation. First, it is essential to reassess the synergies and tradeoffs between various Sustainable Development 234 Goals (SDGs) after COVID-19 and select an optimal economic recovery pathway that reboots the 235 economy with the least harm to other sustainable goals. The pandemic might alter the priority of the 236 SDG achievement and the tradeoffs among SDGs. Our study provides a template for the assessment, 237 which considers not only economic growth and job creation but also the impacts on income 238 inequality. The primary purpose of the assessment is to answer two questions: 1) who is most 239 affected by the pandemic? and 2) could those who suffer the most in the pandemic receive timely 240 and effective assistance during the recovery process? For more comprehensive pathway selection, 241 future research can include other dimensions in the analysis to best balance the tradeoffs among the 242

SDG targets according to local situations. 243
The second implication is that just transition should be addressed as much as green recovery. Just 244 transition can be designed from both short-run and long-run perspectives. In the short-run term, it 245 has been widely acknowledged that determining a detailed plan of decarbonization at the national 246 and sectoral levels is the premise of just transition policy design. For example, a detailed schedule 247 of the early decommissioning pathway of the coal-fired power industry informs policymaking when 248 and in which regions workers will be affected. Based on such information, policymakers could 249 establish precise transitional assistance mechanisms for the affected. Transitional assistance in the 250 short term includes three sections: financial assistance, social protection, and employment training. 251 The first and the most direct way is to provide financial assistance to the low-skilled and low-income 252 workers directly affected by the green recovery. Forms of financial assistance include compensation 253 fees, relocation cost, wage subsidies, etc., and should adapt to the actual development needs of 254 specific areas with local characteristics. Funding sources can be fiscal support for economic 255 recovery or can be a sound green financial system with a payment transfer mechanism. The second 256 way of just transition is to strengthen social protection networks and labor market policy. A just 257 transition requires improvement of social welfare systems, including minimum living standards, 258 unemployment insurance, and early retirement benefits. It is also essential to promote labor 259 migration by reducing relocation costs and breaking down the policy barriers for cross-regional 260 labor mobility. Moreover, training and skill development is another essential measure to assist the 261 unemployed with career transfer. Based on identified skill needs, restarting the apprenticeship 262 program, fostering entrepreneurship, and promoting the cross-sector flow of human resources are 263 vital steps to improve the overall adaptive ability or workers. Apart from short-term aid measures, 264 just transition also needs a long-term plan to enhance the flexibility of the human resource market 265 and economic resilience. Energy transition puts forward a higher demand for cross-disciplinary 266 talents. In the long term, cultivating innovative talents and preparing innovation curricula are 267 fundamental ways to solve the structural contradiction between labor supply and demand. In For example, although green recovery might cause structural unemployment and aggravate income 277 inequality, the co-benefits of air quality improvement brought by climate change mitigation might 278 alleviate the unequal harms to the poor. This is because the low-income and the vulnerable have 279 been identified as exposed more to severe air pollution, and they may gain the most from the 280 reduction of air pollution in the green recovery 21,22 . Moreover, research could explore the impact on 281 the job quality of disadvantaged groups, such as ethnic minorities and women, as they usually 282 benefit less from job creation 23 . Thus, an integrated assessment with more factors is essential to 283 provide more comprehensive social support for achieving just transition when implementating green 284 recovery policies. 285 In sum, our analysis quantitatively reveals that the low-and medium-income groups are the primary 286 victim in the COVID-19, while the high-income is the main beneficiary of green recovery. Such 287 mismatch alerts that COVID-19 stresses the tradeoffs among SDGs (between climate change 288 mitigation and income equality) and highlights the necessity of performing just transition alongside 289 green recovery. We recommend that policymakers pay attention to the immediate needs of the poor 290 and the vulnerable during and after the pandemic and take transition assistance measures to facilitate 291 a smooth transition in the green recovery. 292 293

295
Modelling of short-term economic impact. We adopt and develop an improved Adaptive Regional high-income (20%) group. As shown in Fig.6, the low-and medium-skilled workers account for 328 more than 97% of low-income group, while high-skilled workers dominate the high-income group 329 (about 40%). Mobility Report 29 also reports transportation to other destinations (retail store, grocery and 356 pharmacy, parks, transportation hubs, and residential areas), which is used in this research to 357 calibrate the demand data during the pandemic. Since Google data excludes China, we calculate 358 Chinese situations as the strictest of all countries during the same period of pandemic. 359 three scenario sets 10% of GDP economic stimulus for each region, which is put to the markets 373 before the end of the year. Different scenarios allocate economic stimulus to different sectors based 374 on the initial final demand. Due to the diversity of industrial situations among countries/regions, we 375 adjust the scenario setting for each country/region to fit the scenario description. Besides recovery 376 packages, the recovery rate of labor supply for each region is set at 4%, and economic stimulus is 377 set to start at 4 weeks after the controlling measures are stopped. The stimulating sectors in each 378 scenario are listed in Table 2 and the detailed setting are provided in the SI. 379