What determines the values of environmental benets? Evidence from a worldwide survey

13 One of the key obstacles to building international cooperation for environmental problems is the fact 14 that environmental benefits are valued differently in different countries. But where does the disparity 15 come from? This study gives an answer to this question by analysing large-scale survey data collected 16 across G20 countries. Combining lifecycle impact assessment and economic valuation techniques, we 17 found that people's perceptions of environmental benefits are in fact diverse, but are highly correlated 18 with a few social indicators such as life expectancy, the Gini index, and subjective well-being. Our 19 findings suggest that improving these social indicators in otherwise ill-equipped countries will 20 facilitate convergence of people's perceptions and will thereby establish a common ground for 21 tackling global environmental issues.

The relationship between economic growth and environmental conservation has been a longstanding 27 concern for environmental and resource economists and policymakers. In 2015, the global agenda 28 shifted from the United Nations' Millennium Development Goals to its 17 Sustainable Development 29 Goals (SDGs), which comprise comprehensive targets, such as poverty eradication and sustainable 30 use of natural resources. Since the common global future vision is now sustainable development, 31 understanding the trade-off and balance between economic activity and environmental conservation 32 has become increasingly relevant for policymakers to achieve consensus among different countries 33 with diverse values and conditions [1]. 34 To achieve this future vision, researchers should physically assess the following impacts. For 35 example: To what degree has air pollution and its probability of occurrence affected society? What 36 types of effects do the loss of ecosystems in certain areas (through land use) have on various 37 functions, such as purification capacity, provisioning of biological resources, and climate regulating 38 functions supported by the distribution of surrounding species? However, answering these questions is 39 insufficient for internalizing the environmental impact into the socioeconomic system automatically 40 [2,3]. We must also identify the degree of urgency and importance perceived by the society regarding 41 such environmental effects-social weighting. Bateman and Mace [4] discuss the importance of 42 incorporating such (especially monetary) social weightings into decision-making practice and 43 estimate the benefit-to-cost ratios for investments in natural assets in the UK. 44 However, we know that such social weighting is not universal because the achievement degree for 45 each SDG target varies from country to country [5]. They likely differ depending on the current 46 situation and the type of goods/services affected [6]. In this study, interdisciplinary approaches (biodiversity and primary production, respectively) are trending upward. This indicates that people in 150 lower-income countries focus more on health damage, while those in higher-income countries focus 151 more on natural resource damage, on a national average. 152 Figure 2b shows a scatter plot of WF1s focusing on human health and biodiversity, both of which 153 are qualitative endpoints. These revealed particularly large differences in each category (i.e. human 154 society and ecosystem). People living in the countries located in the upper left area, place more 155 weight on biodiversity than human health. By contrast, people living in the countries located in the 156 lower right area, place more weight on human health than biodiversity. People located around the 45-157 degree line in the figure place equal weight on the factors. Figures 2a and b show that, when facing a 158 trade-off between mitigating damages to society (e.g. economic activity) and the ecosystem (e.g. 159 environmental conservation), the priority of public policymaking depends on the country.  which had less than a middle-income level; for example, the difference among national average 176 weighting factors may reflect the situation that environmental goods, such as good health and clean 177 air, are normal goods that are easily available without inequality in high-income countries, while these 178 goods are still a luxury because of the basic needs' insufficiency or uneven distribution in other 179 countries with less than middle income levels. These findings suggest that the benefits estimated in a 180 high-income country (especially more than 40,000 USD (PPP)) may be transferable to other high-181 income countries with small errors in a cost-benefit analysis of each national project. By contrast, in 182 lower-income countries, the decision-making may not reflect the actual preferences (i.e. the local 183 trade-off perception) if using equivalent values estimated in other countries (even if they are estimated 184 in a similar-income-level country) as the local benefit. Considering factors other than the national 185 income level is thus critical to estimating the local benefit of a global public policy, particularly in 186 lower-income countries (especially less than 16,000 USD (PPP)). 187 [ Figure 2] 188 189

Four types in preference 190
To explore the determinants of heterogeneity, we analysed the same response data simultaneously by

Factors predicting heterogeneity 209
An important feature of the latent class approach is that the membership parameters show which type 210 of people tend to belong to each group, with such distinct preferences as shown in Table 1. 211 People living in countries with longer (shorter) life expectancy, smaller (larger) income inequality, 212 and lower (higher) urban population density tend to belong Class A1 (A2) with larger weight on 213 biodiversity (human health). Although Class A4 has similar characteristics to A2, a large proportion of 214 forest area is correlated to the tendency of belonging to this class, with equal weight on biodiversity 215 and human health. People living in countries with lower (although higher than A2 and A4) life 216 expectancy, smaller income inequality, lower urban population density, and smaller proportion of 217 forest areas tend to belong to Class A3, with equal importance on all four environmental goods. 218 Regarding Figure 3b for the nine countries with less than middle-income levels, the share of the 219 sample is similar between Class M2 with a larger weight on biodiversity, and M3 with larger weight 220 on human health. People living in countries with a higher (lower) urban population density and 9 shorter (longer) life expectancy tend to belong to M3 (M2). Class M4, which places a larger weight on 222 qualitative subjects, is a segment to which people living in countries with a significantly higher Gini 223 index (larger income inequality), such as Brazil and South Africa, tend to belong. Class M1 is similar 224 in terms of equal weight on human health and biodiversity, but quite different from M4, as it does not 225 focus on qualitative subjects alone. In addition, considering individual attributes, respondents with 226 relatively higher income than their neighbours tended to belong to Class M2. Interpreting this result in 227 conjunction with the fact that the national averages of these nine countries are located in the lower 228 right area (as shown in Figure 2b), we found that some respondents with higher relative income within 229 each survey site are more likely to place a higher weight on the ecosystem. 230 Despite potential biases from the 'subjective' index measured by the self-report rating scales [33,34]  Similar social weightings imply that a common ground are ready to establish a cooperative 277 relationship between groups. Note that life expectancy, domestic income inequality, and population 278 density are significantly associated with such classification; for example, longer life expectancy is 279 correlated to the higher weight on biodiversity. Larger domestic income inequality (Gini index) is 280 correlated to the higher weight on human health. Hence, such national statistics are significant 281 indicators for predicting the regional preference of environmental goods. Moreover, improving these 282 social indicators in ill-equipped countries will facilitate convergence of people's perceptions and will 283 thereby contribute to establish a common ground for tackling global environmental issues.

Measures of environmental damages 296
We assessed the current level of global damages as the four endpoints presented below. The 297 explanations provided in the questionnaire are presented in Figures S1a-d. We assessed the damage to social assets from consumption of fossil fuels and mineral resources as 336 quantitative damages to society, in addition to qualitative damage (human health). We adopted the 337 user cost approach [45], which exhaustively evaluates non-biological resources. In this study, the 338 amount of damage is expressed in USD applying a 5% discount rate [10]. User cost is broadly used as 339 a measure of loss of income production capacity due to the depletion of natural resources (i.e. in green 340 GDP). It is defined as the amount of savings required to ensure that the product sales incomes from 341 resource extraction in the current and future generations are balanced, based on the idea of weak 342 sustainability (c.f. replaceable assumption). While economic loss (i.e. decrease in income from 343 agriculture, forestry, and fisheries) due to climate change can be considered in the LIME framework, 344 these damages are excluded from the scope of this study at the time of our cross-national social survey 345 because of insufficient environmental science knowledge; for example, large regional differences are 346 expected between a region with revenue gain and one with revenue loss. Such distributions comprise 347 important information to express the conditions of these regions, but it is generally lost during the 348 process of aggregating the information to the global scale. This may lead to an underestimation of the 349 amount of damage. We applied the most appropriate method for each survey site after consulting with a local research 382 company. Table S1 in the supplementary information shows the national statistics, including GDP 383 (economic scale), population, life expectancy, forest area, GNI per capita, Gini index, region category, 384 and income group, as defined by the World Bank database. Table S2 shows survey information and 385 regional statistics of the survey site. We adopted face-to-face interviews for emerging countries,

Random parameter logit model for national average preference 398
The choice experiment data were analysed statistically using a random parameter logit model by each 399 country. We estimated the marginal utilities by applying the following function: Respondents were asked to respond subjectively to the question 'How would you describe the current 462 state of your health?' They answered by choosing a number from 0 to 10, with 10 being 'very 463 healthy', 5 being 'neither healthy nor unhealthy', and 0 being 'very unhealthy'. SRH is closely related 464 to perceived happiness and life satisfaction; For example, the influence of GDP per capita on SWB 465 tends to reduce when considering additional explanatory variables, such as the health condition [60]. 466 Thus, it is reasonable that an increase in wealth leads to a higher SWB, partially because of better 467 health conditions. 468 469

Income class (relative income level per region) 470
Respondents were asked to specify their total monthly/annual household income based on 5 income For sampling efficiency, we selected an urban area with the largest economic scale and a high 486 population density as the survey site for each country and implemented random sampling. This 487 possibly resulted in higher monetary social weightings (i.e. WF2s defined as WTP) and higher SWB 488 than the national average, reflecting less scarcity of wealth because of their higher income level. The 489 income gap between urban and other areas tends to be larger for middle-income countries as shown in 490 Figure S3 in SI. As our estimates are representative of these cities (survey sites), further investigation 491 is needed to clarify whether this can be extrapolated to the national level. Nevertheless, this concern 492 may be alleviated because the factors mainly associated with heterogeneity (shown in the table below 493 The LIME model used in this study provides the framework incorporating multiple impacts into 495 social decision-making using an interdisciplinary LCIA approach. The public importance of 496 mitigating the multiple environmental damages is evaluated using the current damages physically 497 assessed as a reference status. Since potential damages cannot be fully captured in the assessment 498 because of insufficient knowledge of environmental science, the estimated physical damages of the 499 four endpoints (i.e. DALY, social assets, EINES, and NPP), posed to the respondents as the current 500 damage levels, may be underestimated. This could result in smaller estimated marginal utilities of 501 environmental goods. We can expand the same framework with up-to-date knowledge. 502 Despite the complications and difficulties of incorporating biodiversity mainly due to its 503 multifunctional features, its appraisals need to incorporate socio-economical urgency/importance 504 regarding its damages into public decision-making. Bateman and Mace [4] suggest a comprehensive 505 framework of ecosystem service assessment. This study's social weighting of the ecosystem (i.e. 506 biodiversity and primary production) is one example, wherein the weighting factor of biodiversity is 507 defined as the WTP (or the relative size of marginal utilities) to avoid the extinction of one species. 508 This does not consider differences that are associated with the organism involved. Increased spending 509 is prevalent to avoid extinction of animals [66], suggesting that it may be desirable to estimate the 510 social weighting per species for each organism. In this study, another social weighting of the 511 ecosystem (i.e. primary production) is defined as a factor to capture the importance of the essential 20 foundation of energy flow in the ecosystem. Further improvement and expansion of ecosystem service 513 assessment is needed, which can promote interdisciplinary collaboration and help achieve progress in 514 sustainable development.  For damage assessment, we covered eight impact categories including climate change, air pollution, 677 photochemical oxidants, water use, land use, fossil fuel use, mineral resource, and forest resource use. 678 significantly correlated to higher social weighting on biodiversity (M2) among the middle-income 734 countries, despite having no significant impact on segmentation for the high-income countries. 735