Policy Research Working Paper 10918 Questioning the Climate Change Age Gap Alexandru Cojocaru Michael Lokshin Arina Nikandrova Europe and Central Asia Region & Poverty and Equity Global Practice September 2024 Policy Research Working Paper 10918 Abstract A widely held view in the media and among some research- 38 countries in Europe, Central Asia, and the Middle East ers is that younger people are more worried about climate from the 2023 round of the Life in Transition Survey. The change and more willing to support the climate agenda than findings demonstrate a positive relationship between the older generations. Such a “climate change age gap” is often respondents’ age and their concerns about climate change. explained by the longer time younger people expect to live Older people are more likely to object to higher taxes to under worsening climatic conditions. This paper develops finance public policies in general, including climate change a theoretical model that proposes an alternative explanation policies, but even this result is sensitive to the framing of for the relationship between age and attitudes toward cli- climate action questions. mate change. The empirical analysis is based on data from This paper is a product of the Office of the Chief Economist, Europe and Central Asia Region and the Poverty and Equity Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http:// www.worldbank.org/prwp. The authors may be contacted at mlokshin@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Questioning the Climate Change Age Gap. Alexandru Cojocaru, Michael Lokshin, and Arina Nikandrova1 JEL: Q54, O13, O15 Keywords: Climate change, Europe and Central Asia; government policies, perceptions. 1 Alexandru Cojocaru is Senior Economist in the Poverty and Equity Global Practice, and Michael Lokshin is Lead Economist in the Chief Economist Office of the Europe and Central Asia Region, both are at the World Bank. Arina Nikandrova is Senior Lecturer in Economics at the City University of London. This paper’s findings, interpretations, and conclusions are entirely those of the authors and do not necessarily represent the views of the World Bank, its Executive Directors, or the countries they represent. Miguel Purroy provided excellent assistance in collecting, processing, and analyzing data for this paper. We thank Thomas Farole, Ivailo Izvorski, and Richar Record for their constructive comments and suggestions. Send correspondence to mlokshin@worldbank.org. 1. Introduction In December 2023, representatives of 199 countries confirmed their commitments to addressing the global climate emergency and accelerating climate actions (UN 2024). The success of these global efforts rests on the willingness of individual countries to act upon their voluntarily-set mitigation targets, known as nationally determined contributions (NDCs). The ambitions to implement such policies vary widely across countries (Albalate et al. 2023). Among factors that influence cross−country variations in climate change actions, the “climate change age gap” − the notion that younger people care more about climate change than older people − causes concerns about the implementation of the climate change agenda. 2 Young climate activists capturing the media and public attention reinforce the perception of a widening climate change generational gap. The age structure of populations plays a significant role in climate policies. Population aging results in a growing share of older people who establish themselves as a powerful political force in decisions about the provision of public goods. The influence of older generations on public policy, including climate change, is reinforced by their greater participation in the electoral process compared with younger generations (Goerres 2007). Differences in attitudes toward climate change among generations are often attributed to the fact that younger people will live longer under deteriorating climate conditions and, thus, be more affected by its consequences. Inversely, the shorter time horizons of the aging population and the fact that the main climate change costs accrue in the future suggest that older people should be less supportive of climate change policies. In fact, the empirical evidence of the presence and the size of the climate change age gap is mixed and context-dependent (e.g., Gray et al. 2019). In this paper, we develop a theoretical model that explains why and how people of different ages perceive the increasing threat of global warming. The model guides our empirical analysis, which relies on data from 38 countries of Europe, Central Asia, the Middle East, and North Africa collected in 2022−23, during the fourth round of the Life in Transition Survey (LITS). 2 Factors affecting climate change action include: political trust and beliefs (Fairbrother et al. 2019); public opposition and political polarization (Lachapelle et al., 2012, Hornsey et al. 2016); gender, race, education, and income distribution (Ballew et al. 2020); and social norms and social capital (Bergquist et al. 2019) 2 We find no evidence of an “age gap” in perceptions of climate change. Older respondents seem to be as (or even more) concerned about climate change as younger respondents. However, we do find that older individuals are more likely to object to tax increases to finance climate adaptation and mitigation policies, which is likely to reflect the general decline in willingness to pay taxes with age rather than a specific lack of willingness to finance climate change policies. 3 Past exposure to natural disasters, a person’s physical and psychological health, the degree of risk aversion, sources of information, trust in institutions, and whether a respondent was planning to move away also affect perceptions and willingness to pay for climate change, but none of these confounding factors reverse the climate change − age gradient estimated by the most parsimonious models. The positive age gradient is also robust to a range of specifications, definitions of the dependent variable, and corrections of potential omitted variable biases. This paper contributes to the literature on the determinants of individual perceptions and engagements in climate change actions in several ways. First, it develops a theoretical model that shows that the relationship between age and concerns about climate change and the willingness to pay for climate change actions is ambiguous and depends on various factors. Second, the paper distinguishes between climate-related beliefs and actions toward climate change mitigation, an important distinction, as one can believe that climate change is real and manmade but still not be willing to bear an additional tax burden to mitigate its impacts. Third, the LiTS data allow us to account for the wide range of determinants of climate beliefs that earlier studies have identified as varying across different age groups, including ideology, attitudes toward risk and fairness, trust in government, primary sources of information, and health conditions which can exacerbate climate morbidity. The paper is organized as follows. Section 2 reviews relevant literature on generational differences in beliefs about climate change. Section 3 discusses the study’s theoretical framework and empirical methodology. Section 4 describes the data. Section 5 presents the main results of the empirical analysis. Section 6 shows the result of robustness tests. Section 7 summarizes the paper’s main findings. 3 In April 2024, the European Court of Human Rights issued a ruling in favor of a group of 2,400 Swiss women over age 64 who argued that because older women are more likely to die in heatwaves, Switzerland must take greater action to prevent the planet heating beyond the Paris agreement target of 1.5°C and demand compensation for the lost consumption from the government, rather than paying taxes to finance climate change mitigation policies (Daly 2024). 3 2. Literature review The literature finds that climate change beliefs are shaped by a wide array of socio-structural and psychological factors, including age, gender, level of education, socioeconomic status, political orientation, and value systems (Milfont et al., 2015; Echavarren, 2017; Poortinga et al., 2019). This study focuses on how beliefs about climate change and willingness to take climate change action vary by age. Many determinants of climate change beliefs may not be uniformly distributed across age groups and may thus mediate the observed gradient between age and climate change beliefs. The effect of age on perceptions of climate change has been documented across many countries. Poortinga et al. (2019) show that age is an important predictor of climate change beliefs and concerns in 22 European countries and Israel. They find that the size of the effect is generally smaller in Central and Eastern European countries than elsewhere in their sample and larger in Northern European than in Western European countries. In the EU, younger people are more likely than old people to perceive climate change as a serious problem and place the responsibility of addressing the climate change issues on business/industrial sectors and environmental groups (Skeirytė et al., 2022). Milfont et al. (2021) report a sizeable and persistent negative generational gap in the belief that climate change is real and caused by humans. They argue that the gap occurs because older cohorts started from a lower belief level but show that people of all ages increased their beliefs at a similar rate over a 10-year period. Albalate et al. (2023) find a negative association between the share of older people in a population and both the policy ambitions of climate agreements and the intensity of regulatory initiatives to fight climate change, with a 1 percent increase in the share of the elderly population associated with a 2 percent increase in the predicted NDC per capita emissions by 2030. Poortinga et al. (2023) show that levels of climate-related beliefs, risk perceptions, and emotions among the younger generations are higher than those of older generations in the UK, with a wider and more consistent gap for climate-related emotions than climate-related beliefs. Milfont et al. (2015) find both climate change skepticism and anthropogenic climate skepticism to be more prevalent among older adults than younger adults in New Zealand. Other studies do not find that older people are less likely than younger people to believe in climate change. Gray et al. (2019) report only trivial differences in concerns about climate change between 4 younger and older respondents in the United States. Their findings suggest that age is not a strong predictor of environmental concern or climate change attitudes. Instead, other factors, such as political ideology and education, play more significant roles in shaping these attitudes. Tranter and Booth (2015) find that age poorly predicts climate change skepticism in 14 advanced industrial nations in 2010−2011. Using the World Values Survey (2010–14), which covers 51 countries, Echevarria (2017) finds that age is positively associated with environmental concerns. Better-educated individuals are usually more concerned with climate change than people with less education (Poortinga et al. 2011, Milfont et al. 2015, Echevarren 2017, Poortinga et al. 2019). The level of education has increased considerably over the past century. In both advanced and developing countries, younger cohorts tend to be more educated than older cohorts, which may mediate the relationship between age and climate change beliefs. Political conservatism is associated with climate change skepticism (McCright and Dunlap 2011, Whitmarsh 2011, Milfont 2015), but conservative beliefs tend to be negatively associated with educational attainment (Pew Research Center, 2016). Poortinga et al. (2019) show that individuals on the right side of the political spectrum have lower levels of education, are more likely to prioritize self-enhancement over self-transcendence, and are more likely to be skeptical of climate change or its anthropocentric links than others. 4 Conservatism is more prevalent among older individuals (Cornelis et al. 2009; Whitmarsh 2011). In a meta-analysis of climate change beliefs across 25 polls and 171 academic studies spanning 56 countries, Hornsey et al. (2016) find ideology, world views, and political orientation to be among the strongest predictors of climate beliefs. Free-market ideology is one of the strongest predictors of climate change skepticism. In Europe and Central Asia, support for the market economy tends to be greater among younger adults, which may be related, in part, to the resistance to change of older individuals, who spent much of their lives under a socialist system (EBRD 2007). Media is an essential source of information about climate change, and more knowledge can influence concern about climate change (Lokshin et al. 2024). Malka et al. (2009) report that in the US, increased knowledge is associated with greater concern among adults who identify as 4 Schwartz (1992) defines self−enhancement as characterized by an emphasis on the pursuit of one’s own interests and dominance over others; self−transcendence emphasizes concern for the welfare and interests of others. 5 Democrats or Independents but not among Republicans. Udalov and Welfens (2021) demonstrate that Internet use is positively associated with several dimensions of environmental concern, including knowledge and pro-environmental behaviors, in high-income and low-income countries. Diehl et al. (2021) find that reliance on social media is associated with a decreased climate skepticism, based on survey data from 20 countries. At the same time, the media can also be a source of climate change disinformation (Ejaz et al., 2023). These patterns are crucial for understanding the relationship between age and climate action because younger cohorts tend to use digital communication tools more than older adults do (Seifert and Cotten 2021). 3. Theoretical framework We start from the assumption that the age-related variation in support for climate policies is determined by the length of exposure to climate change, the intensity of climate change impacts on wellbeing, and the effectiveness of mitigation and adaptation measures. The expected exposure of the elderly to climate change is shorter than that of younger people, but the risks from the effects of climate change increase with age at any given time. 5 Older demographic groups are at higher risk of being affected by natural disasters like droughts, floods, and hurricanes (Harper 2019). Comorbidities and frailty increase the adverse effects of air pollution and exposure to high temperatures by older people (Gamble et al. 2016). A simple theoretical framework illustrates how the interplay between the length of exposure to climate change impacts and the intensity of such impacts affects an individual’s attitudes toward climate change. Consider an individual who lives for periods. In every period , she derives utility ( ) from consuming and faces the cost of climate change ( ) caused by pollution stock . Function is increasing and strictly concave in consumption, that is, ′ > 0, ′′ < 0, and function is increasing and strictly convex in pollution stock, that is, ′ > 0, ′′ > 0. The assumption that is convex implies that higher pollution stock causes greater marginal damage to an individual’s wellbeing. The overall lifetime welfare of the individual is 5 Climate change is also one of the leading threats to health and wellbeing of children under five (Zhang et al. 2007). 6 =−1 ( ) − ( ) � (1) (1 + ) =0 where is the individual’s discount rate and is the perceived intensity of climate change effect on an individual’s wellbeing. 6 The size of depends on many considerations and is likely to vary with age. In the analysis, we assume that in early life, the intensity of the effect of pollution on the individual’s wellbeing is > 0, and that in the last year of life is , where > 1 to reflect that the risks from and awareness of the impact of climate change increase with age. The pollution stock grows at an exogenous rate . In each period, an individual can take costly climate change action t to slow the accumulation of the pollution stock in the next period. The pollution stock then evolves according to: Δ +1 ≡ +1 − = − , (2) where is the perceived effectiveness of climate change action . The size of depends on various factors, such as trust in the government and beliefs in its ability to deploy resources toward effective mitigation or adaptation measures. The initial stock of pollution 0 is exogenously given. As long as Δ +1 > 0, the pollution stock grows, so the negative welfare consequences of climate change intensify over time. The individual receives income in each period which she takes as given. She can borrow or save at an exogenously given rate subject to the constraint that any outstanding debt must be repaid at the end of her life. An individual’s optimal choices of consumption and climate change action thus depend on the lifetime budget constraint: =−1 =−1 + � ≤ � . (3) (1 + ) (1 + ) =0 =0 Within this framework, we can compare the optimal consumption and climate change actions that older and younger individuals choose. In the appendix, we consider the utility optimization problems of an old and a young individual. The old individual has two periods to live; the young individual has three periods to live. Both 6 Following Long (1992), for tractability, we assume that the individual’s welfare is additively separable in and . 7 individuals form lifetime plans for consumption. In period 0, they also chose climate change action, which is then applicable for the duration of their life. The individual’s optimal climate change action should be interpreted as making the desired regular contributions, such as taxes, to combat climate change. We show that the optimal climate change action of an older individual can exceed that of a younger individual if the following conditions are met. First, the impact of climate change on wellbeing intensifies significantly with age (i.e., is sufficiently high). Second, individuals discount the future sufficiently strongly – that is, is sufficiently high − so that young people are not too concerned about the higher intensity of climate change impacts in their old age. Third, the pollution stock grows sufficiently slowly − that is, is sufficiently small − so that in the future, young individuals do not expect to live in a significantly worse environment than they currently do. Linking the theoretical and empirical models The theoretical model identifies three parameters governing the impact of age on willingness to pay for climate change policies. These are the perceived effectiveness of government in mitigating the impact of climate change , the perceived intensity of climate change effect , and the individual’s discount rate . Because these parameters are not directly observable, we use proxy variables to identify their impacts on respondents’ attitudes to climate change and their willingness to pay higher taxes to support climate change actions. In the theoretical model, the effect of on concerns about climate change and willingness to finance climate change policies is ambiguous, as the analysis in the appendix shows. As increases, the climate change action becomes more effective, prompting the individual to substitute away from consumption to climate change action. At the same time, higher means that the individual can achieve higher utility with the same income, relaxing the budget constraint. This income effect pushes people to consume more and spend less on climate change action. As substitution and income effects work in opposite directions, the overall effect of is ambiguous. 7 In reality, the size of may differ depending on the specific nature of the climate change measure in question. Furthermore, it will also depend on the individuals’ trust in the government and beliefs in its ability to deploy these resources toward effective mitigation or adaptation measures (OECD 7 Similar intuition explains the backward-bending labor supply curve (e.g., Dickinson 1999). 8 2022). We use questions about trust in government and beliefs in fairness as proxies for this parameter. Several factors might affect the perceived intensity of climate change effect ( ), including age, gender, level of education, and family composition. The past exposure to adverse climate change shocks could also influence . Older individuals may perceive higher risks of future climate change impacts despite shorter future exposure, as they have lived longer and, therefore, have a higher probability of having been exposed to adverse climate shocks in the past. More nuanced relations could work through differences in the capacity to adapt to climate change. On the one hand, younger people are more mobile, can tolerate temperature fluctuations, and can more easily escape from disaster-affected areas. On the other hand, as wealth and savings tend to peak around retirement age (Modigliani 1976), older people are able to insure against the impacts of climate change and deal with the financial shocks of natural disasters. 8 We use health self-assessment and information about past exposure to climatic shocks to proxy the effect of this parameter. In the theoretical model, strong discounting of the future (higher ) increases the asymmetry between old and young individuals and may lead to higher optimal investment in climate−change action for old individuals. Older individuals experience the intensification of the impacts of climate change sooner and so do not discount them as much as young individuals do. For a sufficiently high , young people are not too concerned about the higher intensity of climate change impacts in their old age. We approximate the future discussion rate with responses to a question that asks respondents to select between less money now and more money in a month. Willingness to support climate change actions can be constrained by either a lack of awareness of climate change impacts, which may recede with age, or ideological beliefs that may be correlated with age. Political conservatism, which has been associated with climate change skepticism, is more prevalent among older cohorts. But older adults may have greater concerns about future generations and express stronger pro-social generational orientation compared to younger adults (Maxfield et al. 2014). 8 In the proposed theoretical framework, individuals can freely borrow against their future income; in reality, uncollateralized borrowing against future income is limited, so savings matter for determining capacity to deal with climate change shocks. 9 The theoretical framework motivates our estimation strategy. Our empirical model estimates the relationship between responses to climate change questions and age of the respondent controlling for respondent’s characteristics: = + + + , (4) where, is a set of individual characteristics such as the respondent’s age, gender, highest level of education, marital status, religious affiliation, indicators of mental health, risk aversion, trust in government, main sources of information, household composition, and a proxy for household income. is a vector of regional characteristics such as type of location (urban or rural). is time-invariant country c fixed effect, and is an innovation term. 4. The data We use the latest 2023 round of the Life in Transition Survey (LiTS), implemented by the European Bank of Reconstruction and Development (EBRD) and the World Bank (EBRD 2023). It covers the transition economies of Europe and Central Asia and several comparator countries in Western Europe, the Middle East, and North Africa, including a total of 38 countries. The survey included a nationally representative sample of around 1,000 households per country. The analysis focuses on climate beliefs and climate actions. Beliefs about climate change are captured by the following questions: “How convinced are you personally that climate change is real?” and “How convinced are you personally that climate change is manmade?” Responses are recorded on a 5-step Likert scale ranging from “entirely unconvinced” to “entirely convinced.” We distinguish between impacts on one’s self and impacts affecting the next generation. The former is based on responses to the yes/no question “Do you think climate change seriously affects or will seriously affect you during your lifetime?”, while the latter is based on responses to the question “Do you think climate change seriously affects or will seriously affect the children of today during their lifetime?”. We assess past impacts of climate change using the question, “Have you personally experienced disruption or damage due to flooding, drought, or other natural disasters? (Yes/No).” We capture climate action through questions concerning support for government and personal action in the climate space. Support for government action is based on responses to the question 10 “Here are two statements people sometimes make when discussing the environment and economic growth. Which of them comes closer to your own point of view? (a) “Protecting the environment should be given priority, even if it causes slower economic growth and some loss of jobs, (b) Economic growth and creating jobs should be the top priority, even if the environment suffers to some extent,” or (c) “Neither.” We create a dummy variable that evaluates to 1 if respondents choose option (a). Support for personal action is based on responses to questions related to tax-financed investments in climate change actions. The first is: “Would you be willing to give up part of your income or pay more taxes if you were sure that the extra money was spent on combatting climate change?” We create a dummy with a value of 1 if the respondent answers yes and zero if he or she answers no or does not know. The second question is: “To what extent do you agree or disagree with each of the following statements. I would be willing to pay more in taxes if the extra money were used to… (a) Reduce/prevent pollution such as improve the quality of air or water, or deal with waste/sewage, (b) fight global warming or the greenhouse effect; or (c) prevent the loss of plant or animal species or biodiversity. In all three cases, the responses are on a 5-step scale ranging from “strongly disagree” to “strongly agree.” For each domain, we construct a binary indicator, that takes a value of 1 for agree or strongly disagree and 0 otherwise. An individual’s risk preferences are captured by the following question: “Please rate your willingness to take risks in general, on a scale from 1 to 10, where 1 means that you are not willing to take risks at all and 10 means that you are very much willing to take risks.” 9 We approximate individual future discount rate by constructing a dummy variable based on the following question: “Would you prefer to receive [around 55% of median household daily income in the country] guaranteed today, or [around 85% of median household daily income in the country] guaranteed in 1 month?” The dummy variable takes value of 1 if a respondent selects the first option, corresponding to a high future discount rate. Beliefs about fairness are captured through the question “In your opinion, which of the following factors is the most important to succeed in life in [COUNTRY] now?” We construct a binary variable that evaluates to 1 in the case of meritocratic responses (intelligence, skills, effort, and 9 To measure risk aversion, we reverse the categories so that higher values imply higher risk aversion and lower values imply lower risk aversion. 11 hard work) and to 0 if success is attributed to political connections, breaking the law, or other factors. Free-market ideology has been found to be one of the strongest predictors of climate change skepticism. We capture it based on the question “Which of the following statements do you agree with most? (a) market economy is preferable to any other form of economic system (b) under some circumstances, a planned economy may be preferable to a market economy, (c) for people like me, it does not matter whether the economic system is organized as a market economy or a planned economy.” We construct a binary variable that takes a value of 1 if respondents prefer a market economy to other economic system alternatives and 0 otherwise. Willingness to pay more taxes to combat climate change may depend on one’s beliefs about the ability of the government to allocate tax revenue toward the issues that respondents care about. We proxy these beliefs with answers to the question “To what extent do you trust the government / cabinet of ministers?” The responses are recorded on a 5-step scale ranging from “complete distrust” to “complete trust.” We create a dummy variable that evaluates to 1 for “some trust” or “complete trust.” To capture the impact of social media on attitudes toward climate change, we rely on responses in the “social media” category of the question “People use different sources to learn what is going on in their country and the world. For each of the following sources, please indicate how often you use it” with responses ranging from never to daily on a seven-step scale. To capture individuals’ health status, we rely on the question “How would you assess your health?” with responses on a 5-step scale ranging from “very good” to “very bad.” We assess mental health through responses to the question “How often, if at all, do the following apply to you?”, where the following refers to “you feel depressed” and/or “you feel very anxious, nervous or worried,” with responses ranging from never, to a few times a year, to monthly, weekly, and daily. Table 1 presents the summary statistics for the variables used in our analysis. 12 5. Empirical results Figure 1 shows the changes in the perceptions of climate change by age and gender normalized by the levels of 20-year-olds. 10 The values shown are the percentage differences between the average responses for 20-year-olds and the responses of respondents in different age groups. It demonstrates that perceptions of whether climate change is real or not by the age and gender of the respondents vary only slightly by age and gender. For example, about 66 percent of 20-year- olds consider climate change to be real versus 62 percent of 75-year-olds, representing about a 6 percent difference. Men appear to be more skeptical about the reality of climate change than women, but even at older ages, the male-female differences in perceptions of climate change do not exceed 6 percent. The perceptions of female respondents of all ages do not differ from those of the youngest group of women. The proportion of respondents who think climate change is manmade increases slightly with age (Figure 1, panel b.). Women 60 to 80 are about 5 percent more likely than the reference group to believe in the anthropogenic nature of climate change. The correlation of such beliefs with age is weaker among men. About 60 percent of men 20 to 50 believe that human activities cause climate change, and that proportion increases with age. The answers to the question of whether respondents believe that climate change will affect them during their lifetime show different tendencies (Figure 1, panel c.). Among men, almost 70 percent of the youngest respondents, but just 52 percent of men 75 and older believe they will be affected by climate change. Among women, perceptions are similar for women under 60; after 60, the proportion of positive answers drops about 10 percent. The share of respondents who think climate change will directly affect them is greater among women than men at all ages. More than three-quarters of 20-year-olds think climate change will affect the lives of current children. The share of respondents who agree with this statement rises with age for both men and women, reaching almost 90 percent for the oldest female respondents. As in all other questions, women are more pessimistic than men about the impact of climate change. 10 Figure 1 shows graphs estimated by running mean smoother normalized by the level of the corresponding variable for 20-year-olds. 13 While the differences in perceptions about climate change are indicative of attitudes of different age groups towards climate change and related policies, it could be more informative to analyze the responses to questions about “climate actions.” Figure 2 shows the relation between respondents’ willingness to pay through higher taxes for policies that combat climate change and global warming. Both panels demonstrate responses to similar questions asked in a slightly different form. The trends show a strong age gradient in willingness to pay for climate change policies by age with little or no differences by gender. Among young people, 40 percent of men and 46 percent of women express willingness to pay higher taxes to finance climate change actions. That proportion declines by 15 percent among people 50 years old and further by more than 35 percent (panel a.) or even by 45 percent (panel b.) among older respondents. Factors affecting attitudes toward climate change and willingness to finance climate change actions Table 2 shows the estimations of the model (4) on four questions on beliefs about the impact of climate change. The results indicate that education strongly predicts whether a respondent is concerned by climate change. 11 For example, the share of the population with only primary education who thinks climate change is real (Specification 1) is 16.0 percentage points lower, and the share of people with upper-secondary education is 11.6 percentage points lower than the post- graduate education baseline. Women, people living in larger households, and those residing in urban areas are also more convinced that climate change is real. In contrast, respondents who place themselves higher on the societal income ladder are more skeptical about climate and its risks for themselves and children. Higher levels of education are associated with a greater awareness of the anthropogenic nature of climate change and a higher degree of concern about climate risks; the conditional correlations of other socio-demographic variables are less consistent across the different domains of climate beliefs. Age appears to be correlated with greater concerns about climate change in three of four estimations, holding constant other personal and household characteristics. For example, a 10-year increase in the age of respondents raises the probability of thinking that climate change is real or manmade by 0.7 percentage points, and the probability of believing that climate change will affect children by 1.4 percentage points. Older people are thus less skeptical about climate change and 11 This finding is consistent with the recent literature that studied similar questions (e.g., Dechezlepretre et al. 2023). 14 more concerned about its impacts on children, but the age gradient is not very pronounced, conditional on other socio-demographic covariates. Age does not significantly affect the perception that climate change will affect respondents during their lifetime. How and to what extent are the environmental concerns discussed above related to environmental behavior? Table 3 presents estimations of the relationship between respondents’ characteristics and respondents’ beliefs that mitigating climate change should be a government priority (Column 1), and the willingness to pay for climate change through higher taxes (Column 2). 12 Older respondents are more inclined to consider climate change policies as a priority but less likely to support climate change policies financed by higher taxes. Better-educated respondents and those higher on the societal income ladder are more likely to support such policies (consistent with the “affluence hypothesis” by Franzen (2003)). Is the negative age gradient in willingness to pay higher taxes specific for climate change or is it common for any tax-finance policy? A number of studies demonstrate a negative association between the willingness to finance public policies by higher taxes and age (e.g., Jacques 2023, Cattaneo and Wolter 2009). Table A1 in the appendix shows the probit marginal effects of individual characteristics on the willingness to pay for improvements in education and health systems and reduction in income inequality. All three policy dimensions have a strong, significant, and negative age gradient. The marginal effects of age are much stronger (−0.210, −0.095, and −0.191, respectively) than the marginal effect of age on willingness to pay for climate change (−0.065) shown in Table 3. The negative effect of age on willingness to pay to combat climate change may simply reflect the general decline in willingness to pay more in taxes as people age. 13 Explaining the climate change age gap We propose and test several hypotheses to explain these relationships. The idea of omitted variable bias (OVB) provides a framework for structuring our effort. The hypotheses suggest specific characteristics that might influence respondents’ attitudes towards and willingness to pay for 12 We obtain qualitatively similar results using monthly per capita household income instead of the income ladder variable. However, the sample size dropped by about a third when using monthly income; therefore, we used the income ladder in our main specifications. 13 Such a decline in willingness to support policies through taxes is especially pronounced in post−communist countries where older generations lived through the shock of transition that undermined their trust in government and institutions (Pop−Eleches and Tucker 2014). 15 climate change actions. Our list of potential confounders is not exhaustive, but it is sufficient for us to be confident about the answer to the main question posed in this paper. The OVB framework allows us to test our hypotheses by estimating whether the correlation between respondents’ age and the variables reflecting their climate concerns is affected after adding regressors to our empirical model. Table 4 shows the marginal effects of confounding variables (top panel) and respondent’s age (bottom panel) on the dependent variables used in Tables 2 and 3 based on regressions that add one confounding variable at a time to the specification and, in the last row, add the full vector of confounding variables together into the regressions. The perceptions of climate change-specific risks could influence individual attitudes and actions, which may be heightened by past exposure to natural disasters. To test this effect, we add a dummy variable to our model to show whether a respondent experienced such an event. The first row of the top panel in Table 4 shows that exposure to a natural disaster has a strong positive effect on climate attitudes and beliefs about the risks of climate change. For example, past exposure to climate impacts increases a respondent’s probability of believing that she will be affected by climate change by almost 18 percentage points and increases the belief that children will be affected by 14 percentage points. Including this variable mitigates the effect of the respondent’s age for most of the questions; differences with the baseline effects are not statistically significant. For a given level of exposure to climate change, vulnerability depends on health conditions. Individuals with poor health are more sensitive to temperature fluctuations and air pollution and less capable of avoiding climate change risks. We try to capture this pathway by adding to the regression three variables that capture self-reported health assessment as well as experience with anxiety and depression. Health self-assessments exhibit no statistically significant correlation with climate beliefs; adding this variable to our baseline specifications results in no significant changes in the effect of age. Experiencing anxiety during the week before the interview heightened concerns about climate change (by about 6 percentage points for the questions about whether climate change will affect respondents directly). Respondents who experienced depression seem more skeptical about whether climate change is real or manmade. Depression does not change perceptions of climate change risks, either to one’s self or children. Neither of these variables affects the size or statistical significance of the age effect. 16 Accounting for the respondent’s primary source of information results in a significant increase in the age effect for three out of six questions. The age effect almost doubles for the question of whether climate change is real and whether it is manmade and increases the share of respondents who believe that it will affect their children by about 36 percent. These large and statistically significant changes do not alter the relationships uncovered by our baseline estimations, however; they do not eliminate the positive age gradient in climate change awareness and risk perceptions. Respondents who plan to move away from their current place of residence, either within their country or abroad, do not exhibit significantly different levels of skepticism regarding the reality of climate change or its anthropogenic character, but they are more likely to think that climate change will affect both them and children. These correlations do not affect the age gradient observed in our baseline specification. Risk aversion, beliefs that success is meritocratic, trust in government, and a preference for the market economy are all associated with climate awareness. Beliefs in meritocratic success and preference for the market economy are associated with greater perceived risks of climate impacts. Respondents with high future discount rates appear less convinced that climate change is real, and that variable has no significant effect on other climate beliefs. Including these confounding factors does not affect the age gradient. If all confounding variables described above are added to the regression, the age-climate gradient becomes similar in magnitude to the regression that controls for the use of social media, which appears to account for the bulk of the overall increase in the marginal effect of age on climate beliefs. For all countries in our sample except nine countries in the Middle East, North Africa, and Western Europe, we have sub-national administrative data on a number of environmental risks as well as critical infrastructure that can contribute to climate resilience. We capture the following dimensions of environmental risks: (i) flood frequency – the number of flood incidents between 1985 and 2022 from the Dartmouth Flood Observatory; (ii) drought risk – frequency of severe drought when more than 30 percent of cropland is affected over 1984−2022 from the Food and Agriculture Organization; (iii) heat risk – extreme heat hazard based on daily maximum 100 year return period, from the Global Facility for Disaster Reduction and Recovery; (iv) landslide risk – 17 average annual frequency of significant landslides during the period from 1980 until 2018 from the World Bank; and (v) pollution – annual average levels of PM 2.5 in 2019 from NASA. For infrastructure, we rely on data from the Harmonized Global Critical Infrastructure index (Nirandjan et al., 2022), capturing access to (i) energy infrastructure, including cable, line, minor line, power tower, power pole, plant, and substation; (ii) water infrastructure, including water towers, water wells, reservoirs, and water works; (iii) waste management facilities for solid and water waste; (iv) transportation infrastructure, including roads, railways, and airports; (v) education infrastructure, including kindergartens, schools, colleges, universities, and libraries; and (vi) health infrastructure, including ambulatory and non-ambulatory medical facilities. The last line in the top panel of Table 4 reports the χ2-test of joint significance of the administrative risk and resilience variables, based on the regression in which the 11 risk and resilience variables are added to the baseline specification; the last line in the bottom panel reports the marginal effect of age in specifications that include the administrative data on environmental risk and resilience. The risk and resilience variables are jointly significantly correlated with climate beliefs, but they have no significant effect on the magnitude of the marginal effect of age on climate beliefs. In Table 5, we undertake a similar exercise for the two climate action variables related to government priorities and willingness to pay higher taxes to support the climate agenda. One key difference is that when examining the determinants of climate action, we also include the two dependent variables from Table 4 related to perceived risks of climate impacts as explanatory variables. Their inclusion is motivated by the notion that perceived risks of climate change could partially determine climate actions. The estimates bear out this intuition: respondents who believe climate change will affect them personally or children are considerably more likely to prioritize environmental protection over economic growth and be willing to pay more taxes to combat climate change. Most of the confounding variables affect beliefs about the importance of government actions and willingness to pay for climate change action in similar ways – past exposure to climate impacts, perceived risks of future impacts, preferences for the market economy, trust in government, and beliefs that success is meritocratic positively correlate with climate action, both personal and by the government. Greater risk aversion, reliance on social media, and plans to move away in the near future also correlate positively with attitudes about climate action. 18 In terms of our variable of interest (age), older individuals are more likely to believe that combating climate change should be prioritized over economic growth, but they are also less willing to pay higher taxes or give up a part of their income to achieve that objective. The positive correlation between age and beliefs that climate action should be prioritized at the expense of economic growth is robust to the inclusion of other determinants of climate beliefs. In fact, when all the additional confounding variables are included in the model, the marginal effect of age increases relative to the baseline specification. In the case of willingness to pay for climate action, the negative association with age is robust to the inclusion of most confounding variables but becomes insignificant once we account for differences in risk aversion in the specification that includes all the confounding variables. 6. Sensitivity analysis This section examines the sensitivity of our results to the potential bias introduced by the endogeneity of climate beliefs in our climate action regressions, as well as the sensitivity of the age gradients to alternative framing of the climate action questions. Addressing the endogeneity of climate beliefs One concern with the results presented in Table 5 has to do with the possibility that perceived risks of climate change for one’s self or children could be correlated with unobserved characteristics of respondents that affect their willingness to pay for climate action or their beliefs about the tradeoff between climate action and economic growth. For example, political conservatism, which we do not capture with LiTS data, could be correlated with less pessimistic views about the risk of future climate impacts and, at the same time, a lower willingness to support climate action, either politically or financially. To account for such unobservable factors, we instrument the perceived risks to one’s self or children with past exposure to climate shocks based on the question “Have you personally experienced disruption or damage due to flooding, drought, or other natural disasters?” Regressing the perceived risk for one’s self (children) on the instrumental variable, conditional on other variables in the model, results in t statistics associated with the coefficient of the instrumental variable of 16.4 and 17.4, respectively, confirming that our instrument is correlated with the endogenous variable, conditional on other model covariates. The exclusion restriction for this 19 instrument is based on an argument that past exposure to climate-related disasters is correlated with beliefs about climate action by affecting people’s perceptions of the risks of experiencing similar disasters in the future. Having been affected by disasters in the past may make an individual more likely to support policies aimed at limiting the impacts of such disasters, but this increased willingness to pay operates through changed perceived risks of such events happening in the future. In tables 6 and 7 we report marginal effects of age in regressions without the perceived risks variable (Column 1), with the “perceived risks for self” variable added (Column 2) and “perceived risks for children” added in column 4, and 2SLS estimates in which the perceived risks for self are instrumented with past exposure to climate disasters (Column 3) and with perceived risks for children instrumented with past exposure in Column 5. In all specifications, the tests reject the null of exogeneity and pass the weak instrument tests. 14 Subject to the validity of the assumptions underlying the instrumentation strategy, the age gradient in beliefs that climate change action should be a government priority remains positive when the model includes perceived future risks to one’s self. The gradient becomes smaller but remains statistically significant when the model includes perceived risks for children. IV estimates do not change the significance of the age gradient in willingness to pay additional taxes or give a part of one’s income to combat climate change when risks to one’s self are instrumented with past exposure to climate disasters. However, the correlation between age and willingness to pay becomes negative when perceived risks to children are instrumented with past exposure. Alternative framing of climate action questions So far, the willingness to pay for climate action results were based on a question that implies that the money would go to “combating climate change” without specifying what combating climate change means. The LiTS survey includes three questions on willingness to pay for specific climate- related outcomes: reducing or preventing pollution by improving the quality of air or water or dealing with waste/sewage, fighting global warming or the greenhouse effect, and preventing the loss of plant or animal species or biodiversity. These questions allow us to test the sensitivity of the age gradient to the framing of the question related to willingness to pay for climate change action. 14 We cannot perform overidentification tests as we only have one instrument per endogenous variable. 20 Table 8 shows no statistically significant correlation between age and willingness to pay for climate action across the three alternative dimensions of climate change actions at baseline specification. Older respondents are just as willing as younger respondents to pay for actions to combat pollution, global warming, or biodiversity. The factors that affect support for climate action, such as past exposure and perceived risks of future exposure, risk aversion, beliefs about fairness, trust in government, preference for the market economy, high future discount rates, or use of social media, exhibit similar correlations with specific dimensions of climate change action as in the main climate change action regressions (Table 5). Adding most of these variables in the regressions does not affect the age gradient. The exceptions are reliance on social media and risk aversion, inclusion of which leads to the age coefficient to become positive and statistically significant. Controlling jointly for all the confounding variables results in a positive correlation between age and willingness to pay for pollution reduction and a marginally positive correlation with willingness to pay to combat global warming and prevent the loss of biodiversity. For every 10-year increase in age, the predicted willingness to pay for the reduction in pollution increases by 1 percentage point – 70-year-old respondents are thus 5 percentage points more likely to be willing to pay for the pollution reduction than 20-year-olds. 7. Conclusions Younger people are much more active and visible in the climate change debate than older generations. This led some observers and researchers to believe in a climate change age gap. This paper questions this premise and demonstrates that older people are at least as concerned as younger people about climate change. Older people are also more likely to believe that governments should prioritize combating climate change, even if economic growth suffers. These results are robust to several alternative definitions of the dependent variable, empirical specifications, and corrections for unobserved omitted variable biases. We do find that older individuals are less inclined to agree with raising taxes to combat climate change, but this effect most likely reflects the fact that older people are less inclined to pay for any expenditures in the public sector (Cattaneo and Wolter 2009). Our theoretical model offers several empirically testable hypotheses that are suggestive of potentially confounding variables that might explain such an age effect. Accounting for 21 confounding variables, primarily for differences in individual risk aversion, eliminates the negative age gradient in tax-financed climate action. Replacing a generic combating climate change formulation with more specific questions about willingness to pay additional taxes to combat pollution, global warming, or loss of biodiversity, we find a positive correlation between age and these alternative, dependent variables. Older people represent a rapidly growing share of the consuming and voting public; they could help shape policies to reduce environmental risks for future generations. How to translate the elevated environmental concerns of older generations into actionable support and financing of climate change policies is an important direction for future research. 22 References Albalate D., Bel, G., and J. Teixido. (2023). “The influence of population aging on global climate policy.” Population and Environment, 45:13, https://doi.org/10.1007/s11111−023−00425−4. Ballew, M. T., Pearson, A. R., Goldberg, M. H., Rosenthal, S. A., and A. Leiserowitz. (2020). “Does socioeconomic status moderate the political divide on climate change? The roles of education, income, and individualism.” Global Environmental Change, 60, 102024. Bergquist, M., Nilssona, A., and W. Schultz. (2019). “A meta−analysis of field−experiments using social norms to promote proenvironmental behaviors.” Global Environmental Change, 59, 10194. Cattaneo, A., and S. Wolter. (2009). “Are the elderly a threat to educational expenditures?” European Journal of Political Economy, 25(2): 225−236, Cornelis, I., Van Hiel, A., Roets, A., and M. Kossowska. (2009). “Age differences in conservatism: Evidence on the mediating effects of personality and cognitive style,” Journal of Personality, 77(1): 51−88. Daly, A. (2024). “Older Swiss women just set a global legal precedent for challenging their nation’s climate change policy,” The Conversation, April 11, 2024. https://theconversation.com/older-swiss-women-just-set-a-global-legal-precedent-for- challenging-their-nations-climate-change-policy-227629. Dechezlepretre, A., Fabre, A. Kruse, T., Planterose, B., Chico, A., and S. Stantcheva (2023). “Fighting Climate Change: International Attitudes Toward Climate Policies,” NBER Working Paper 30265. Dickinson, D. (1999). “An examination of labor supply and work intensities,” Journal of Labor Economics, 17: 638−670. Diehl, T., Huber, B., Gil de Zúñiga, H., and J. Liu (2021). “Social Media and Beliefs about Climate Change: A Cross−National Analysis of News Use, Political Ideology, and Trust in Science,” International Journal of Public Opinion Research, 33(2): 197–213. EBRD (European Bank for Reconstruction and Development) (2007). Transition report 2007: People in transition. London. EBRD (2023). Life in Transition Survey. London. Echavarren, J., (2017). “From objective environmental problems to subjective environmental concern: a multilevel analysis using 30 indicators of environmental quality,” Society and Natural Resources 30 (2): 145–159. Ejaz, W., Mukherjee, M., and R. Fletcher. (2023). “Climate change news audiences: Analysis of news use and attitudes in eight countries,” Reuters Institute for the Study of Journalism, University of Oxford. Fairbrother, M., Sevä, I., and J. Kul. (2019). “Political trust and the relationship between climate change beliefs and support for fossil fuel taxes: Evidence from a survey of 23 European countries.” Global Environmental Change, 59, 102003. Franzen, A. (2003). “Environmental Attitudes in International Comparison: An Analysis of the ISSP Surveys 1993 and 2000.” Social Science Quarterly, 84: 297–308. 23 Gamble, J., et al. (2016). “Ch. 9: Populations of concern.” In: The impacts of climate change on human health in the United States: A scientific assessment. US Global Change Research Program, Washington, DC, p. 257. Goerres, A. (2007). “Why are Older People More Likely to Vote? The Impact of Ageing on Electoral Turnout in Europe.” The British Journal of Politics and International Relations, 9(1): 90−121. Gray, S., Raimi, K., Wilson, R., and J. Árvai. (2019). “Will Millennials save the world? The effect of age and generational differences on environmental concern,” Journal of Environmental Management, 242: 394−402. Harper, S. (2019). “The Convergence of Population Aging with Climate Change.” Journal of Population Aging, 12: 401–403. Hornsey, M., Harris, E., Bain, P., Fielding, B., and K. Fielding. (2016). Meta−analyses of the determinants and outcomes of belief in climate change. Nature Climate Change, 6: 622– 626. Jacques, O. (2023). “Explaining willingness to pay taxes: The role of income, education, ideology.” Journal of European Social Policy, 33(3): 267−284. Lachapelle, E., Borick, C., and B. Rabe. (2012). “Public attitudes toward climate science and climate policy in federal systems: Canada and the United States compared.” Review of Policy Research, 29: 334–357. Long, N. (1992). “Pollution control: A differential game approach,” Annals of Operations Research, 37(1): 283–296. Malka, A., Krosnick, J., and G. Langer. (2009). “The Association of Knowledge with Concern About Global Warming: Trusted Information Sources Shape Public Thinking.” Risk Analysis, 29: 633−647. Maxfield, M., Greenberg, J., and T. Pyszczynski. (2014). “Increases in Generative Concern among Older Adults following Reminders of Mortality.” International Journal of Aging and Human Development, 79(1):1−21. McCright, A. and R. Dunlap. (2011). “The politization of climate change and polarization in the American public’s views of global warming, 2001–2010.” Sociological Quarterly 52 (2): 155–194. Milfont, T., Milojev, P., Greaves, L., and C. Sibley. (2015). “Socio−structural and psychological foundations of climate change beliefs,” New Zealand Journal of Psychology, 44 (1): 17– 30. Milfont, T., Zubielevitch, E., Milojev, P. and C. Sibley. (2021). “Ten−year panel data confirm generation gap, but climate beliefs increase at similar rates across ages.” Nature Communications, 12, 4038. Modigliani, F. (1976). “Life Cycle, Individual Thrift, and the Wealth of Nations,” The American Economic Review, 76(3): 297–313. Nirandjan, S., Koks, E., Ward, P., and J. Aerts. (2022). “A spatially-explicit harmonized global dataset of critical infrastructure.” Scientific Data 9, 150. OECD. (2022). Building Trust to Reinforce Democracy: Main Findings from the 2021 OECD Survey on Drivers of Trust in Public Institutions, OECD Publishing, Paris. 24 Pew Research Center (2016). “A wider ideological gap between more and less educated adults: Political polarization update,” Washington, DC. Poortinga, W., Spence, A., Whitmarsh, L., Capstick, S., and N. Pidgeon. (2011). “Uncertain climate: an investigation into public skepticism about anthropogenic climate change,” Global Environmental Change, 21 (3, SI): 1015–1024. Poortinga, W., Demski, C. and K., Steentjes. (2023). “Generational differences in climate−related beliefs, risk perceptions and emotions in the UK.” Communication, Earth and Environment, 4, 229. Poortinga, W., Whitmarsh, L., Steg, L., Böhm, G., and S. Fisher. (2019). “Climate change perceptions and their individual−level determinants: A cross−European analysis,” Global Environmental Change, 55: 25−35. Pop−Eleches, G. and J. Tucker. (2014). “Communist socialization and post−communist economic and political attitudes.” Electoral Studies, 33: 77−89. Schwartz, S., (1992). “Universals in the content and structure of values: theoretical advances and empirical tests in 20 countries,” Advances in Experimental Social Psychology, 25, 1–65. Seifert A., and S. Cotten. (2021). “Use of information and communication technologies among older adults: usage differences, health-related impacts, and future needs” in The Oxford handbook of sociology and digital media. Eds. Rohlinger D., and S. Sobieraj. Oxford: Oxford University Press, 1–27. Skeirytė, A., Krikštolaitis, R., and G. Liobikienė. (2022). “The differences of climate change perception, responsibility and climate-friendly behavior among generations and the main determinants of youth’s climate-friendly actions in the EU,” Journal of Environmental Management, 323. Tranter, B., and K. Booth, (2015). “Scepticism in a changing climate: a cross−national study.” Global Environmental Change, 33, 154–164. Udalov, V. and P. Welfens, (2021). “Digital and competing information sources: Impact on environmental concern and prospects for international policy cooperation,” International Economics and Economic Policy, 18: 631−660. United Nations (2024). COP 28, https://unfccc.int/cop28/5−key−takeaways. Whitmarsh, L. (2011). “Skepticism and uncertainty about climate change: Dimensions, determinants and change over time,” Global Environmental Change, Vol. 21(2): 690−700. Zhang, Y., Bi, P., and J. Hiller. (2007). “Climate change and disability−adjusted life years.” Journal of Environmental Health. 70:32–36. 25 Figure 1: Perception of climate change by gender and age. Note: Graphs estimated by running mean smoother normalized by the level of the corresponding variable for 20-year-olds. The values shown are the percentage difference between the average responses for 20- year-olds and the average responses of respondents in different age groups. 26 Figure 2: Willingness to pay higher taxes to fight climate change by gender and age. Note: Graphs estimated by running mean smoother normalized by the level of the corresponding variable for 20-year-olds. The values shown are the percentage difference between the average responses for 20- year-olds and the average responses of respondents in different age groups. 27 Table 1: Descriptive statistics for the dependent and explanatory variables Variables and controls Mean Std.Dev. Minimum Maximum N Dependent variables Convinced climate change is real 3.733 1.209 1 5 34,663 Convinced climate change is manmade 3.578 1.294 1 5 34,383 Climate change will affect your lifetime 0.663 0.473 0 1 33,699 Climate change will affect your children 0.804 0.397 0 1 33,441 Agree to more tax to combat climate change 2.936 1.267 1 5 35,572 Gov. should increase taxes for climate change 0.296 0.457 0 1 32,339 Controls Female 0.591 0.492 0 1 35,572 Age 0.479 0.173 0.180 0.950 35,572 Education No degree / No education 0.028 0.166 0 1 35,572 Primary education 0.055 0.228 0 1 35,572 Lower secondary education 0.150 0.357 0 1 35,572 (Upper) secondary education 0.423 0.494 0 1 35,572 Post−secondary non−tertiary education 0.084 0.278 0 1 35,572 Tertiary education (not a university diploma) 0.051 0.220 0 1 35,572 Bachelor’s degree or more 0.159 0.366 0 1 35,572 Master’s degree or PhD 0.049 0.216 0 1 35,572 Marital status Single (never married) 0.213 0.410 0 1 35,435 Married 0.543 0.498 0 1 35,435 Widowed 0.136 0.342 0 1 35,435 Divorced 0.092 0.288 0 1 35,435 Separated 0.016 0.127 0 1 35,435 Religion Atheistic/agnostic/none 0.089 0.285 0 1 34,164 Buddhist 0.018 0.135 0 1 34,164 Jewish 0.001 0.025 0 1 34,164 Christian 0.495 0.500 0 1 34,164 Muslim 0.384 0.486 0 1 34,164 Other 0.013 0.112 0 1 34,164 Household size 2.930 1.749 1 10 35,572 Urban 0.407 0.491 0 1 35,572 Income ladder 4.199 1.723 1 10 25,566 Confounding variables Was affected by a natural disaster 0.181 0.385 0 1 35,572 Self−assessed poor health 0.094 0.292 0 1 35,572 Felt anxious last week 0.320 0.467 0 1 35,572 Was depressed last week 0.164 0.370 0 1 35,572 Main information source: social media 0.462 0.499 0 1 35,572 Risk aversion 4.784 2.955 1 10 35,256 Believes in fairness 0.708 0.455 0 1 34,528 Trust in government 0.313 0.464 0 1 35,562 Preference for market economy 0.301 0.459 0 1 35,352 High future discount rate 0.461 0.498 0 1 33,656 28 Table 2: Responses to climate change related statements by respondent’s characteristics. Climate change is real is manmade will affect self will affect children ME Std.Err ME Std.Err ME Std.Err ME Std.Err Age 0.069*** 0.026 0.071*** 0.026 −0.035 0.027 0.145*** 0.022 Female 0.016** 0.006 0.006 0.006 0.052*** 0.006 0.032*** 0.005 Education level Reference category: Master’s degree or PhD No degree / No education −0.094*** 0.025 −0.128*** 0.026 −0.082*** 0.025 −0.077*** 0.022 Primary education −0.160*** 0.019 −0.154*** 0.021 −0.095*** 0.020 −0.073*** 0.017 Lower secondary education −0.126*** 0.016 −0.129*** 0.017 −0.098*** 0.016 −0.059*** 0.014 (Upper) secondary education −0.116*** 0.013 −0.095*** 0.014 −0.067*** 0.014 −0.054*** 0.011 Bachelor’s degree or more −0.031** 0.014 −0.010 0.015 −0.033** 0.015 −0.014 0.012 Marital status Reference category: Single (never married) Married −0.025*** 0.009 −0.024** 0.010 −0.021** 0.009 −0.017** 0.008 Widowed −0.017 0.012 −0.037*** 0.013 −0.029** 0.013 −0.013 0.010 Divorced −0.048*** 0.012 −0.035*** 0.013 −0.051*** 0.013 −0.026** 0.010 Separated −0.023 0.023 −0.033 0.026 −0.006 0.023 −0.034* 0.020 Religion Reference category: Atheistic/agnostic/none Jewish −0.182 0.130 −0.081 0.144 −0.181 0.142 −0.120 0.130 Christian 0.003 0.015 −0.013 0.016 0.013 0.014 0.016 0.013 Muslim −0.043* 0.024 −0.086*** 0.026 −0.018 0.026 −0.025 0.024 Other −0.070** 0.033 −0.128*** 0.030 −0.047 0.034 −0.019 0.026 Household size 0.011*** 0.002 0.003 0.003 0.011*** 0.003 0.011*** 0.002 Share of children −0.009 0.017 0.015 0.018 −0.008 0.017 −0.001 0.014 Urban 0.030*** 0.011 0.020* 0.012 −0.002 0.012 −0.001 0.010 Income ladder −0.005** 0.002 −0.002 0.002 −0.010*** 0.002 −0.011*** 0.002 Number of observations 32,889 32,620 31,979 31,740 Note: Table shows marginal effects after probit regressions. Results account for the square term of age. Marginal effects on 38 county dummies are omitted. Standard errors are clustered at the PSU level. *** significant at the 1% level, ** significant at the 5% level, * significant at the 10% level. 29 Table 3: Willingness to pay or finance through higher taxes climate change mitigation policies by respondent characteristics. Willingness to pay more tax for Climate change is a priority climate change ME Std. Error ME Std. Error Age 0.101*** 0.025 −0.065** 0.025 Female 0.025*** 0.006 −0.003 0.006 Education level Reference category: Master’s degree or PhD No degree / no education −0.130*** 0.024 −0.063** 0.025 Primary education −0.123*** 0.020 −0.083*** 0.021 Lower secondary education −0.115*** 0.017 −0.113*** 0.018 (Upper) secondary education −0.100*** 0.015 −0.100*** 0.016 Bachelor’s degree or more −0.042** 0.016 −0.065*** 0.017 Marital status Reference category: Single (never married) Married −0.020** 0.009 0.002 0.009 Widowed −0.056*** 0.012 −0.028** 0.012 Divorced −0.023* 0.013 −0.018 0.013 Separated −0.002 0.024 0.071*** 0.026 Religion Reference category: Atheistic/agnostic/none Jewish −0.173* 0.098 0.423*** 0.096 Christian −0.035** 0.016 −0.012 0.015 Muslim −0.051** 0.025 0.012 0.024 Other −0.078** 0.030 −0.048 0.031 Household size 0.004* 0.003 0.006** 0.002 Share of children 0.016 0.017 −0.019 0.016 Urban 0.007 0.011 −0.016 0.011 Income ladder 0.005** 0.002 0.016*** 0.002 Number of observations 33,669 30,671 Note: Table shows marginal effects after probit regressions. Results account for the square term of age. Marginal effects on 38 county dummies are omitted. Standard errors are clustered at the PSU level. *** significant at the 1% level, ** significant at the 5% level, * significant at the 10% level. 30 Table 4: Marginal effect of age (accounting for the squared term) on the probability of giving a positive answer to climate perceptions questions. Climate change Confounding variables will affect will affect is real is manmade one’s self children Marginal effect of confounding variable (pp) Affected by natural disaster 8.929*** 8.887*** 17.804*** 13.811*** Self−reported bad health −0.973 0.226 2.331* 0.814 Anxiety 1.037 1.978* 5.851*** 2.491*** Depression −3.633*** −2.627** 1.648 −2.083* Social media 6.672*** 7.869*** 5.973*** 6.260*** Plan to move 0.434 2.044* 3.466** 3.352** Risk aversion −0.662*** −0.486*** −0.662*** −0.124 Fairness 3.772*** −1.639* 3.569*** 2.682*** Trust in government 3.180*** −0.013 0.574 −0.224 Preference for market economy 5.527*** 4.736*** 4.386*** 4.006*** High future discount rate -1.809* -1.333 -0.330 0.168 Climate exposure and resilience infrastructure 24.15** 24.25** 20.78** 39.04*** (χ2(11) test of joint significance) Marginal effect of age in specification with confounding variable (pp) Baseline (no confounding variables included) 0.069** 0.071** −0.035 0.145*** Affected by natural disaster 0.058* 0.060* −0.056* 0.129*** Self−reported bad health 0.072** 0.071** −0.042 0.142*** Anxiety 0.068** 0.070** −0.039 0.143*** Depression 0.070** 0.072** −0.036 0.146*** Social media 0.129*** 0.143*** 0.017 0.198*** Plan to move 0.070** 0.076** −0.028 0.151*** Risk aversion 0.093*** 0.088*** −0.012 0.151*** Fairness 0.062* 0.078** −0.037 0.139*** Trust in government 0.064* 0.072** −0.036 0.145*** Preference for market economy 0.074** 0.076** −0.031 0.149*** High future discount rate 0.069*** 0.080** -0.050* 0.143*** All confounding variables together 0.131*** 0.152*** 0.012 0.181*** All together + PSU admin data 0.133*** 0.127*** −0.005 0.198*** Note: Baseline marginal effects are calculated based on the estimations of the models shown in Table 2. Marginal effects for other specifications are calculated based on the models where the baseline specification was extended with corresponding variable(s). The χ2-test of joint significance of climate exposure and resilience infrastructure is based on a regression in which the baseline models are augmented with the PSU-level environmental risk variables. Results account for the square term of age. *** significant at the 1% level, ** significant at 5% level, * significant at 10% level. 31 Table 5: Marginal effect of age on the probability of giving a positive answer to climate action questions. Willing to pay higher Climate change is a Confounding variables taxes to combat government priority climate change Marginal effect of confounding variable (pp) Climate change will affect self during lifetime 17.875*** 14.074*** Climate change will affect children during their lifetime 20.499*** 13.057*** Affected by natural disaster 5.286*** 5.913*** Self−reported bad health −1.781 −4.067*** Anxiety −0.959 2.745*** Depression −3.743*** 0.044 Social media 3.297*** 1.448* Plan to move 4.851*** 4.419*** Risk aversion −0.664*** −1.425*** Fairness 4.947*** 5.033*** Trust in government 2.105** 6.163*** Preference for market economy 11.244*** 1.432* High future discount rate -0.097 -2.260*** Climate exposure and resilience infrastructure 28.39*** 11.04 (χ2(11)-test of joint significance) Marginal effect of age in specification with confounding variable (pp) Baseline (no confounding variables are included) 0.101*** −0.065* Climate change will affect self during lifetime 0.109*** −0.062* Climate change will affect children during their lifetime 0.075** −0.082** Affected by natural disaster 0.094*** −0.073** Self−reported bad health 0.107*** −0.052* Anxiety 0.102*** −0.067** Depression 0.102*** −0.065* Social media 0.131*** −0.052* Plan to move 0.112*** −0.055* Risk aversion 0.128*** −0.012 Fairness 0.109*** −0.060* Trust in government 0.098*** −0.076** Preference for market economy 0.112*** −0.064* High future discount rate 0.102*** -0.062* All confounding variables together 0.145*** −0.011 All together + PSU admin data 0.159*** −0.015 Note: Baseline marginal effects are calculated based on the models’ estimations in Table 2. Marginal effects for other specifications are calculated based on the models where the baseline specification was extended with corresponding variable(s). The χ2-test of joint significance of climate exposure and resilience infrastructure is based on a regression in which the baseline models are augmented with the PSU-level environmental risk variables. Results account for the square term of age. *** significant at the 1% level, ** significant at the 5% level, * significant at the 10% level. 32 Table 6: Instrumental Variable (IV) estimates of the effect of age on beliefs that mitigation of climate change is more important government priority than economic growth. Baseline Baseline + self risk IV (self risk) Baseline + child IV (child risk) (OLS) (OLS) risk (OLS) Age 0.164*** 0.158*** 0.153*** 0.126*** 0.072** (0.026) (0.026) (0.026) (0.026) (0.032) Wh−Hausman endogeneity test 7.731 10.041 P>F 0.005 0.002 Weak instrument F test 302.525 261.686 P>F 0.000 0.000 Notes: Results account for possibility of endogeneity of perceived future climate risks. Marginal effects expressed in percentage points after OLS (columns 1,2 and 4) and 2SLS (columns 3 and 5) regressions. Included in the regressions but omitted from the output are demographic controls from table 2, self−reported bad health, anxiety, depression, use of social media, plan to move, risk aversion, attitudes towards fairness, trust in government, preference for a market economy, and a set of country dummies. *** significant at the 1% level, ** significant at the 5% level, * significant at the 10% level. Table 7: Instrumental Variable (IV) estimates of the effect of age on willingness to pay for actions to mitigate climate change. Baseline Baseline + self risk IV (self risk) Baseline + child IV (child risk) (OLS) (OLS) risk (OLS) Age 0.014 0.006 −0.001 0.008 −0.086** (0.026) (0.026) (0.027) (0.026) (0.032) Wh−Hausman endogeneity test 17.013 23.284 P>F 0.000 0.000 Weak instrument F test 287.792 242.723 P>F 0.000 0.000 Notes: Results account for possibility of endogeneity of perceived future climate risks. Marginal effects expressed in percentage points after OLS (columns 1,2 and 4) and 2SLS (columns 3 and 5) regressions. Included in the regressions but omitted from the output are demographic controls from table 2, self−reported bad health, anxiety, depression, use of social media, plan to move, risk aversion, attitudes towards fairness, trust in government, preference for a market economy, and a set of country dummies. *** significant at the 1% level, ** significant at the 5% level, * significant at the 10% level. 33 Table 8: Marginal effect of age on willingness to pay higher taxes to reduce pollution, combat global warming, and prevent loss of biodiversity. Willingness to pay higher taxes to Confounding variables reduce combat global prevent loss of pollution warming biodiversity Marginal effect of confounding variable (pp) Climate change will affect self during lifetime 18.978*** 21.281*** 19.767*** Climate change will affect children during their lifetime 22.190*** 23.375*** 21.389*** Affected by natural disaster 9.242*** 7.588*** 9.080*** Self−reported bad health −3.362** −1.808* −2.326* Anxiety 3.030*** 3.446*** 3.470*** Depression −0.632 0.998 −0.057 Social media 5.891*** 4.909*** 5.902*** Plan to move 6.819*** 4.668*** 5.465*** Risk aversion −.356*** −1.196*** −1.439*** Fairness 5.699*** 5.732*** 5.227*** Trust in government 7.240*** 6.440*** 6.113*** Preference for market economy 2.164* 2.483** 1.958* High future discount rate -2.724*** -2.868*** -3.277*** Marginal effect of age in specification with confounding variable (pp) Baseline (no confounding variables are included) 0.029 0.011 −0.003 Climate change will affect self during lifetime 0.031 0.010 −0.002 Climate change will affect children during their lifetime 0.004 −0.017 −0.026 Affected by natural disaster 0.017 0.001 −0.015 Self−reported bad health 0.040 0.017 0.005 Anxiety 0.026 0.008 −0.005 Depression 0.029 0.011 −0.003 Social media 0.082** 0.055* 0.050* Plan to move 0.043* 0.022 0.009 Risk aversion 0.079** 0.055* 0.050* Fairness 0.027 0.008 −0.004 Trust in government 0.016 −0.000 −0.014 Preference for market economy 0.031 0.013 −0.001 High future discount rate 0.027 0.011 -0.009 All confounding variables together 0.111*** 0.072* 0.071* Note: Baseline marginal effects are calculated based on the model estimations in Table 2. Marginal effects for other specifications are calculated based on the models in which the baseline specification was extended with corresponding variable(s). Results account for the square term of age. *** significant at the 1% level, ** significant at the 5% level, * significant at the 10% level. 34 8. Appendix: Analysis of the theoretical model This appendix examines the optimization problem of two individuals: one old and one young. The old individual has two periods to live; the young individual has three periods to live. Both individuals form lifetime plans for consumption. In period 0, they choose climate change action, which is then applicable for the duration of their lives. Old individual’s problem As the marginal utility of consumption is always positive, the individual’s lifetime budget constraint holds as equality. Hence, an old individual solves the following utility maximization problem: (1 ) − (1 ) max (0 ) − (0 ) + (A.1) 0 ,1 , 1 + subject to: 1 + 1 0 + + = 0 + (A.2) 1 + 1 + where 1 is defined as 1 = (1 + )0 − . (A.3) The Lagrangian for this optimization problem is: (1 ) − (1 ) 1 1 + ℒ = (0 ) − (0 ) + + �0 + − 0 − − �. (A.4) 1 + 1 + 1 + The first-order conditions are ∂ℒ = ′ (0 ) − λ = 0 (A.5) ∂0 ∂ℒ 1 1 = ′ (1 ) − λ =0 (A.6) ∂1 1 + 1 + ∂ℒ ηθκ′ (1 ) 1 (A.7) = − λ �1 + �=0 ∂ 1 + 1 + ∂ℒ 1 1 + (A.8) = 0 + − 0 − − =0 ∂λ 1 + 1 + 35 From first-order conditions (A.5) and (A.6), ′ (0 ) = ′ (1 ), which implies that 0 = 1 = as utility function is strictly concave. Then, by (A.8), 1 + 1 = �0 + � − (A.9) 2 + �� � ��� 1+� �� � PV of lifetime income ≡ where is pinned down using (A.5) and (A.7): ηθκ′ (1 ) ′ ( ) = . (A.10) 2 + Substituting the expression for 1 into (A.10) and using expression for from (A.9) yields ηθκ ′ 1 + ′ ( ) = �(1 + γ)0 − η � � − ��, (A.11) 2 + 2 + � is the present value of the old individual’s lifetime income. where Young individual’s problem The young individual solves the following utility maximization problem: (1 ) − (1 ) (2 ) − (2 ) (A.12) max (0 ) − (0 ) + + 0 ,1 ,2 , 1 + (1 + )2 subject to budget constraints: 1 + 2 + 1 2 0 + + + = 0 + + (A.13) 1 + (1 + ) 2 1 + (1 + )2 where 1 is defined in (A.3) and 2 = (1 + )1 − = (1 + )2 0 − (2 + ) . (A.14) The Lagrangian for a young individual’s optimization problem is: (1 ) − (1 ) (2 ) − (2 ) ℒ = (0 ) − (0 ) + + 1 + (1 + )2 (A.14) 1 2 1 + 2 + + �0 + + − 0 − − − �. 1 + (1 + )2 1 + (1 + )2 The first-order conditions are as follows: ℒ = ′ (0 ) − = 0 (A.15) 0 36 ℒ 1 1 = ′ (1 ) − =0 (A.16) 1 1 + 1 + ℒ 1 1 = ′ (2 ) − =0 (A.17) 2 (1 + ) 2 (1 + )2 ℒ ′ (1 ) (2 + )′ (2 ) 3 + 3 + 2 = + − =0 (A.18) 1 + (1 + )2 (1 + )2 ℒ 1 2 1 + 2 + = 0 + + − 0 − − − =0 (A.19) 1 + (1 + )2 1 + (1 + )2 From first-order conditions (A.15)−(A.17), ′ (0 ) = ′ (1 ) = ′ (2 ), which implies that 0 = 1 = 2 = , as utility function is strictly concave. Then, by (A.19), (1 + )2 1 2 = � + + � − , 3 + 3 + 2 ��� 0 1 ���� + ����� (1 + ����� ) 2 (A.20) � PV of lifetime income ≡ where is pinned down using (A.15) and (A.18): ′ ( ) = ((1 + )′(1 ) + (2 + ) ′(2 )). (A.21) 3 + 3 + 2 Substituting expressions for 1 and 2 into (A.21) and using the expression for from (A.20) yields: (1 + )2 ′ ( ) = � (1 + )′ � (1 + ) 0 − � � − �� 3 + 3 + 2 3 + 3 + 2 (A.22) (1 + )2 + (2 + ) ′ �(1 + )2 0 (2 + ) � � − ���, 3 + 3 + 2 � is the present value of the young individual’s lifetime income. where Comparison of results on old and young individuals To compare the willingness to pay of old and young individuals, suppose both individuals receive income in every period. Then, equation (A.11), which implicitly defines the optimal consumption of an old individual, simplifies to: ηθκ ′ ′ ( ) = �(1 + γ)0 − η( − )�, (A.23) 2 + 37 and (A.22), the similar equation for the young individual, becomes: ′ ( ) = �(1 + )′�(1 + )0 − ( − )� 3 + 3 + 2 (A.24) )2 + (2 + ) ′�(1 + 0 − (2 + ) ( − ��. ) Consider the following function: θ 1 + ( ): = � − � ′ �(1 + )0 − ( − )� 2 + 3 + 3 + 2 (A.25) (2 + ) ′ − 2 �(1 + )2 0 − (2 + ) ( − )�. 3 + 3 + It is the difference between an old person’s marginal utility for consumption and a young person’s marginal utility for consumption at the optimum. If at = , this function is positive, then ′( ) > ′( ) at the optimum, which implies that < (since is concave) and so > . Thus, to show that old individuals may have a higher willingness to pay for climate change action, we need to show that for some parameter values, function is positive. Taking the derivative ∂ ′ = �(1 + γ)0 − η( − )� ∂θ ��������������������� 2 + >0 (A.26) ′ (2 + γ)(2 + ) �(1 + γ)2 0 − (2 + γ)η( − )� × �1 − � 3 + 3 + 2 ′ �(1 + γ)0 − η( − )� ��������������������������������������� >0 if γ→0 and →1 (2+γ)(2+ ) when γ is sufficiently small and is sufficiently large, < 1 and (1 + γ)2 0 − 3+3+ 2 (2 + γ)η( − ) < (1 + γ)0 − η( − ). In this case, the second line in the expression above is positive and is increasing in . Consequently, for large enough, function is positive. To sum up, the sufficient conditions for > are: • The impact of climate change on wellbeing intensifies significantly with age − that is, is sufficiently high. • Discount rate is sufficiently high so that individuals discount the future sufficiently strongly. • The pollution stock grows sufficiently slowly (i.e., is sufficiently small) − so that young individuals do not expect to live in a significantly worse environment than they currently do. 38 Intuitively, when > 0, both individuals expect the impacts of climate change to intensify in the last year of their life. But a high has a stronger effect on the old individual’s optimal choice of climate change action, because old individuals experience the intensification of the climate change impacts sooner and do not discount them as much as young individuals do. Strong discounting (higher ) increases the asymmetry between old and young individuals and may lead to higher optimal climate−change action for old individuals. The growth of the pollution stock may undo some of the effects of discounting, but the last condition ensures that this attenuating effect is not too strong. In the framework, parameter captures the perceived effectiveness of climate change action. The effect of an increase in on the optimal climate change action is ambiguous and depends on the curvature of cost function . To see this formally, first consider an old individual. Differentiating implicitly (A.23) with respect to yields: ∂ θκ�′ (1 ) − η ′′ (1 )� = . (A.27) ∂η (2 + ) ′′ ( ) − η2 θκ′′ (1 ) The denominator is negative because is concave, while is convex. The sign of thus depends on the sign of the numerator. If is linear, then ′′ = 0, and the numerator is positive. In this case, an increase in results in lower consumption (i.e., individuals optimally divert more of their budget to climate change action, which is now more effective). However, if is sufficiently convex − that is, ′′ is sufficiently high − individuals may respond by lowering their optimal climate change action in response to an increase in . Now consider a young individual. Differentiating implicitly (A.24) with respect to yields: ∂ (A.28) =κ , ∂ where: ≔ (1 + )�′ (1 ) − η ′′ (1 )� + (2 + γ)θ(′ (2 ) − η(2 + γ) ′′(2 )) (A.29) and 39 ≔ (3 + 3 + 2 ) ′′ ( ) − (1 + )2 ′′(1 ) − (2 + )2 2 ′′(2 ). (A.30) The denominator is negative as is concave and is convex; the sign of thus depends on the sign of the numerator . As in the case of an old individual, if is linear, then ′′ = 0, and the numerator is positive, in which case an increase in results in higher climate change action . When is strictly convex, the sign of is ambiguous. 40 Table A1: Willingness to pay higher taxes for improvements in education and health systems and to reduce income inequality as a function of individual characteristics. Probit marginal effects (ME). LITS 2023. Willingness to pay more tax for education health reduce inequality ME Std. Error ME Std. Error ME Std. Error Age −0.210*** 0.028 −0.095*** 0.028 −0.191*** 0.028 Female −0.001 0.007 0.003 0.006 −0.016** 0.006 Education level Reference category: Master’s degree or PhD No degree / No education −0.204*** 0.027 −0.155*** 0.027 −0.082*** 0.028 Primary education −0.211*** 0.022 −0.139*** 0.022 −0.063*** 0.023 Lower secondary education −0.224*** 0.018 −0.149*** 0.017 −0.078*** 0.019 (Upper) secondary education −0.173*** 0.015 −0.117*** 0.015 −0.061*** 0.017 Bachelor’s degree or more −0.095*** 0.017 −0.054*** 0.016 −0.041** 0.018 Marital status Reference category: Single (never married) Married 0.017* 0.010 0.008 0.010 0.009 0.010 Widowed 0.001 0.013 −0.039*** 0.014 0.005 0.014 Divorced −0.013 0.014 −0.022 0.014 0.009 0.014 Separated 0.039 0.026 0.022 0.026 0.075*** 0.026 Religion Reference category: Atheistic/agnostic/none Jewish 0.328*** 0.063 0.288*** 0.076 0.426*** 0.056 Christian −0.039** 0.018 0.024 0.017 0.028 0.017 Muslim −0.015 0.029 0.031 0.029 0.082*** 0.030 Other −0.066 0.040 −0.006 0.039 −0.010 0.039 Household size 0.012*** 0.003 0.010*** 0.003 0.014*** 0.003 Share of children 0.054*** 0.019 −0.006 0.018 −0.026 0.019 Urban −0.025* 0.013 −0.011 0.013 −0.016 0.014 Income ladder 0.027*** 0.003 0.026*** 0.003 0.016*** 0.003 N 31,818 32,073 31,435 Note: Marginal effects after probit regressions. Marginal effects on 38 county dummies are omitted. Standard errors are clustered at the PSU level. *** indicates that the coefficient is significant at the 1% level, ** at the 5% level, * at the 10% level. 41