The Representative Agent Must Die: Using Demographics to Inform Social Discount Rates January 30, 2016 Eli P. Fenichel, Matthew J. Kotchen, and Ethan Addicott Abstract Benefit-cost analyses of long-lived public projects—such as those related to environmental protection, infrastructure, education, and health—require a social discounting process. The Ramsey Rule provides the backbone for modern discounting. A critical parameter in the Ramsey Rule is the utility discount rate, UDR, (or rate of pure time preference). There is a split in the literature as to whether the UDR is a purely normative decision or should be backed out from market behavior. We offer a third approach, based life expectancy. We exploit the fact that multiple generations are extant at any point in time, and the UDR can reflect an aggregation over how these generations care about their own future utility. We use the approach to derive estimates of the UDR for nearly all countries of the world, and find UDRs range between 1.5 and 4.0 percent, with a global estimate of 2-3 percent. We also place our empirical results in the context of deriving the overall social discount rate, and find results that are strikingly close to those often used in practice. For example, we derive social discount rates between 2 and 6 percent for the United States, which compares well to OMBs 3-7 percent range. Introduction Few topics in economics are as fundamental and generate as much controversy as discounting. Benefit-cost analyses of long-lived public projects—such as those related to environmental protection and investments in infrastructure, education, and health—require a social discounting process. While most regulatory agencies have established procedures for discounting future benefits and costs, economists continue to debate what constitutes an appropriate discount rate and fundamentally how discounting should be applied—and the stakes are high. Small changes to discount rates and procedures can significantly affect the present 1 value calculations for long-lived projects, so questions about discounting often play a critical role in policy evaluation. Economic analysis of climate change has brought many of the important issues to the fore. Policies to address climate change are generally associated with near term costs of reducing emissions in order to avoid damages that would otherwise occur far into the future. Hence the extent to which more or less aggressive climate policy passes a benefit-cost test will depend critically on the social discount rate. This is the main insight that comes from the highly influential and contrasting research of Stern (2007) and Nordhaus (2007).1 Both employ the Ramsey-Koopmans-Cass framework for discounting, yet they differ primarily in their underlying assumptions about the utility discount rate (UDR), which reflects the time discount rate between generations irrespective of differences in consumption.2 While Stern uses a very low rate that supports more aggressive climate policy relative to Nordhaus, the differences in how they justify their assumptions are of greater interest here. Stern follows classical economists and argues that the choice should be based on ethical considerations, whereas Nordhaus argues that discounting should be backed out of behavior reflected in actual market interest rates. The Stern-Nordhaus exchange rekindled a long-standing debate about “descriptive” versus “prescriptive” approaches to discounting (Arrow et al. 1996), and subsequent papers have sought to further clarify the role of positive and normative assumptions implicit in the economics of climate change and discounting more generally.3 While this literature furthers the understanding of conceptual issues surrounding the choice of social discount rates, there remains relatively little empirical guidance on how to choose the underlying parameters, especially with regard to the UDR. Researchers and policy makers are left to choose based on various ethical criteria, some of which can push in different directions, or to back out values after determining other parameters, including an a priori assumption about what the social discount rate should be. In this paper, we develop a demographic approach for estimating the UDR that serves as a useful benchmark. Rather than make a judgment about how a social planner is to compare utilities across generations, we exploit the fact that multiple generations are extant at any point in 1 Both Nicholas Stern and William Nordhaus have written several subsequent papers directly on the topic, but here we reference the original two in which the distinction first arose. 2 This parameter of the Ramsey equation is sometimes referred to as the rate of societal pure time preference, in addition to other variants. 3 Examples include Weitzman (2007), Dasgupta (2008), Goulder and Williams (2012), Arrow et al. (2012), and Gollier (2012). 2 time, and the UDR can reflect an aggregation over how these generations care about their own future utility. We begin with an (admittedly extreme) assumption that individuals care about the future only to the extent that they live to enjoy it; that is, individuals discount future utility according to mortality risk. We then derive estimates of the UDR based on the age structure of a population and life expectancy at each age. Given life expectancies, the estimates inform periods considered long-term for many policies (50 to 80 years), and they can apply even further into the future to the extent that population structures are stable, which we investigate. In effect, our approach is “descriptive” at the individual level, and “prescriptive” in the aggregation, for which we estimate and discuss various alternatives. One advantage is that normative judgments are anchored in tradeoffs among the current population rather than abstract notions about future generations not yet born. We use the approach to derive estimates of the UDR for nearly all countries of the world using detailed demographic data from the World Health Organization. A striking feature of our results—which range between 1.5 and 4.0 percent—is that they fall within the range many economists consider reasonable.4 Overall, the approach yields a global estimate of the UDR between 2 and 3 percent depending on whether aggregation is based on the median or mean individual, respectively. We emphasize how the results are dependent on the aggregation rule, which reflects normative judgments. While there is heterogeneity in the estimates across countries, differences tend to be attenuated because of two demographic features that push in opposite directions. When one’s concern about the future depends entirely on whether s/he will live to enjoy it, we would expect countries with a younger aged population to discount the future less. However, the demographics, which are influenced by levels of economic development, are such that countries with younger populations also tend to have lower life expectancies at age, and this implies greater discounting of the future all else equal. The result, as we discuss, is that two countries can have very similar estimates of the UDR, but for very different reasons. Our primary contribution is the development of a demographically based approach for estimating the UDR, along with its application to all countries individually and collectively, but we also place our empirical results in the context of deriving the overall social discount rate (also referred to as the consumption discount rate). Specifically, we combine our estimates of the UDR 4 For example, in various studies, Nordhaus uses rates that range from 1.5 to 3 percent (see Nordhaus and Yang (1996) and Nordhaus (2007),(2014)). Moreover, a recent survey of expert opinion among economists finds a mean estimate of the UDR of 1.1 percent, with a range between 0 and 8 percent (Drupp et al. 2015) 3 with other parameters of the Ramsey equation using the International Monetary Fund’s 20-year forecasts of economic growth for each country, along with expert opinion about the elasticity of the marginal utility of consumption (Drupp et al. 2015).5 We again find results that are strikingly close to those often used in practice. For example, we derive social discount rates between 3 and 6 percent for the United States when aggregation is based on the median or mean individual, respectively. These compare very closely with the Office of Management and Budget’s range between 3 and 7 percent. Across all countries we find that social discount rates are not driven primarily by estimates of the UDR or forecasts about economic growth, but rather a roughly equal weighting of both. Our approach thus provides a new methodology for deriving regionallyspecific discount rates for use in integrated assessment models of climate change, in addition to more general applications of cost-benefit analysis. 2. Conceptual Framework To motivate our conceptual framework, we begin with the standard discounted utility model for a representative individual, who has an intertemporal utility function of the form ,…, where 1/ 1 , is the discount factor and is the individual’s pure rate of time preference. It has long been recognized that uncertainty about survival affects the way individuals discount the future and therefore make intertemporal choices. Fredrick, Loewenstein, and O'Donoghue (2002) provide a nice quotation from Rae (1834) that goes back almost two centuries: When engaged in safe occupations, and living in healthy countries, men are much more apt to be frugal, than in unhealthy, or hazardous occupations, and in climates pernicious to human life. Sailors and soldiers are prodigals. In the West Indies, New Orleans, the East Indies, the expenditure of the inhabitants is profuse. The same people, coming to reside in the healthy parts of Europe, and not getting into the vortex of extravagant fashion, live economically. War and pestilence have always waste and luxury, among the other evils that follow in their train. (Rae 1834 p. 57) There are many more recent references that make the same point with specific reference to life expectancy. Moreover, it is mortality risk in a catastrophe that has been used to motivate 5 We discuss other possibilities as well, which can easily be incorporated as alternative assumptions. 4 discounting in an intergenerational context (Weitzman 2009; Dasgupta 2007; Goulder and Williams 2012). We can explicitly account for life expectancy in the discounted utility model.6 We begin simply by assuming a constant probability of death, , between periods. The discount factor can account for this as follows: 1 , 1 where represents the pure rate of time preference assuming certain survival. When 1, the future is discounted entirely, because of certain death. In contrast, immortality leads to 0, in which case the discount factor entirely depends on the pure rate of time preference. Two observations are helpful before generalizing the framework further. First, the original formulation of the discount factor can account for the probability of death, whereby 1 1 (1) 1 Second, it is easy to see that the probability of survival 1 assuming is an upper bound for , 0. In other words, the individual discounts future utility by at least the probability of not being around to enjoy it. A clear limitation of this formulation is that is constant through time, whereas life expectancies are such that the probability of death generally increases with age. To account for this reality, assume the individual is age 0, and that the probability of death is known at at . The model can then be written as each age, denoted as ,…, ,where 1 1 . Assuming that the probability of death in any year is increasing with age, this model implies that is decreasing over time, or equivalently, the discount rate is getting larger, reflecting less concern about the future when one is less likely to be around to enjoy. A further implication is that for any , a younger initial age implies a lower discount rate in period because of greater chances of survival. These results, as we will show, play an important role as offsetting effects in our empirical estimates over populations. 6 Our approach here is similar to that of Eckstein (1961) and Kula (1997). 5 Thus far we have considered only a single individual, but our aim of deriving a social UDR requires consideration of an entire population. We assume the population consists of individuals denoted same and 1, … , , and they differ only by age; that is, all individuals share the 0 in order to focus on the extent of heterogeneity . We assume further that in the pure rate of time preference owing to life expectancy alone. This implies, using equation (1) subscripting for time and superscripting for the individual, that each individual ’s 0 is demographically determined discount rate at 1 This makes immediately clear that as discounting of the future. Moreover, as . → 0 it follows that → 0 and → 1 it follows that → ∞ and 0 → 1, implying no 0 → 0, implying complete discounting of the future. 0 is given by The age profile of the population in year ,…, . We make the further simplifying assumption that the age distribution is stable through time, which means that the growth rate is identical for all ages in the population. While this assumption is common in demography (Casewell 2001) and used to study human capital (Arrow, Dasgupta, and Mumford 2014, UNU-IHDP and UNEP 2014), we test and discuss this assumption in the results section. The assumption greatly simplifies our analysis here because current population structures are sufficient for providing estimates, eliminating the need for demographic projections, although the approach could be extended to account for demographic shifts. Although the effective discount rates derived for individuals are based entirely on the demographics associated with life expectancy, aggregating them into a social UDR requires normative judgments. Any aggregation requires some sort of weighting across individuals of different ages, and there is a literature on aggregating individuals’ time preferences into a planner’s objective function.7 In this paper, we are agnostic about how such weighting should occur, but employ two aggregation rules because of their simplicity and frequent application. Our first approach is to estimate a social UDR based on the mean discount rate in the population. Of course, one justification for using the mean is that the social UDR represents the discount rate of a representative individual. This is simply 7 See the recent paper by Millner and Heal (2014) and the references cited therein. 6 ̅ 1 , where the assumption of a stable population structure means that we need only consider life expectancies at time 0. The intuition for why we need only consider one point in time is that while individuals’ discount rates change as they age, the fraction of the population at each age remains constant. Thus, the population’s social UDR does not change over time. Our second approach is to estimate the social UDR using the discount rate of the median individual such that Median ,…, . While this approach provides a useful point of comparison, it can also be justified as consistent with demographic principles of majority rule. In what follows, we estimate the different social UDRs for nearly all countries of the world, in addition to the world as a whole. 3. Estimation of Social Utility Discount Rates We obtain demographic data from the World Health Organization’s (WHO) 2012 life tables. 8 The tables report vital statistics binned in ages 0, 1-4, and then in 5-year increments. The important variables for our purposes are the number of individuals and the life expectancy in each bin. We treat the number of individuals in each bin identically. We assume that individuals know their age specific life expectancy at age (Smith, Taylor, and Sloan 2001), but are uncertain about their own specific survival. Based on the available data, let | denote the expected number of years left to live for someone at age , which we now treat as the binned age classes and drop subscripts for time. Then for individuals in each age class, we assume survival from year to year is independently and identically distributed such that survival is a series of Bernoulli . It follows that an individual’s probably of trials, with mortality probability at a given age survival as time passes is geometrically distributed. This means that the geometric distribution describes probabilistically the number of periods passing until the individual’s death. We thus have the relationship whereby mean of the geometric distribution is the inverse of the expect number of periods until death—that is, | . Using this relationship and equation (2), 8 Global Health Observatory Data Repository, Life Tables by Country 1990-2013 (webpage on the Internet). Geneva: World Health Organization; 2013. http://apps.who.int/gho/data/node.main.692?lang=en> [cited November 2015] 7 we are able to estimate a demographically based discount rate for all individuals in each age class. We then calculate social UDRs for all countries using the mean and median aggregation rules described in equations (3) and (4). We also calculate corresponding UDRs for the world as a whole using data on the population structure of all countries combined. 3.1 Country-Specific Results We find substantial country-specific heterogeneity in social UDRs. We report the results for every country in Appendix Table 1 and summarize the general pattern in Figure 1. When using the mean aggregation (top panel), country-specific social UDRs mostly vary between 2 and 4 percent. When using the median aggregation, the social UDRs are generally lower and range between 1.5 and 3 percent. The distribution of social UDRs is left skewed, and developed countries generally have higher social UDRs than developing countries. The variation in social UDRs is driven by two offsetting and yet related effects. The first is countries with longer life expectancies will have lower social UDRs, all else equal. These tend to be countries with greater incomes and better access to medical care. These same countries, however, also tend to have older populations, and this is the second effect that pushes towards a higher social UDR, all else equal. It follows that two countries may have the same social UDR but for very different reasons. One may be a developed country with long life expectancy and an older population, while the other may be a developing country with shorter life expectancy and a younger population.9 These offsetting effects illustrate the importance of age distribution when using life expectancy to estimate a social UDR. We further illustrate the role of demographic structure by showing how country-specific UDRs differ based on the fraction of a country’s population over the age of 65 and life expectancy at birth. For all countries, Figure 2 plots life expectancy against the fraction of the population over the age of 65, and the points are colored according to the size of the country’s social UDR based on the mean aggregation. The figure shows how countries with lower life expectancies at birth—e.g., those less than 65 years of age—have little variation in the social UDR, and the magnitudes are relatively low (below the world estimate discussed below). This is because these same countries tend to have very young populations, in large part because they have shorter life expectancies. Focusing instead on the countries with life expectancies at birth 9 NOTE: We should pull out a few specific examples to illustrate this point. 8 greater than 65 years of age indicates the importance of the age distribution as captures by the fraction over the age of 65. Among these countries, increases in the fraction from 2 to 10 percent in of the population over 65 results yields substantial differences in the social UDR. These results provide one slice through the data that underscore the importance of demographic structure on the social UDR. 3.2. Global Results We follow the same procedures described above to estimate social UDRs for the world. The estimates are based on demographic information for the world constructed from the WHO’s country-specific life tables. This approach effectively treats all people of the world as a single population as described above. Using the mean aggregation, we estimate a global, social UDR of 2.9 percent, which is very close that that used in previous research (e.g., Nordhaus and Yang (1996)). Using the median aggregation lowers the estimate to 2.0 percent. While both rates are substantially greater than zero, it is useful to emphasize that the aggregation rule plays an important role. We focus on the mean and median, but alternative normative judgments will produce different results. For example, if only the utility of newborns are taken into account, the social UDR would be close to zero. In contrast, accounting for only the elderly would produce a social UDR that easily exceeds 100 percent. This range emphasizes the importance of normative judgments implicit in the aggregation rule across a population, be it a country or the world as a whole. Indeed, the estimates that we report in this subsection treat the people of the world as a single population, but alternative estimates are possible whereby social UDRs are estimated for each country and then aggregated based on country-weights. Alternatives to consider include the mean and median social UDR across countries, which differ from our approach here, which essentially represent population-weighted mean and median estimates across countries. 4. Relation to Social (Consumption) Discount Rates The primary focus of our analysis is on deriving demographically based estimate of the social UDR. But this is only one component of deriving the overall social (consumption) discount rate that is used for intertemporal economic evaluation of long-term or intergenerational projects. In this context, the overall social discount rate most commonly employed is based on the Ramsey rule: 9 , where is the UDR that we seek to inform with our analysis. The other term is the product of the elasticity of the marginal utility of consumption , and the rate of growth in consumption . While there exists a literature on whether the social discount rate should be constant or declining (see Arrow et al. 2014), we assume a constant rate here for the illustrative purpose of placing our results in the larger context of deriving the overall social discount rate. We use the following procedures to obtain estimates of the other parameters in the Ramsey equation. Our estimate of the elasticity of the marginal utility of consumption is based on the expert opinion survey conducted by Drupp et al. (2015). This implies a single value of 1.35 for our purposes. For estimates of the growth rate in consumption, we use the International Monetary Fund’s mean 20-year projection for real per capita GPD growth in each country.10This provides different estimates of for each country. We then combine these estimates with our estimates of the social UDR to derive overall social discount rates for all countries. We use the mean aggregation for the UDR to be consistent with our use of GDP per capita. We estimate an overall social discount rate for the world of 6.8 percent. As a point of reference, the U.S. Office of Management and Budget recommends using discount rates that range between 3 and 7 percent for social regulatory impact analysis (OMB 2003). Using our procedure to estimate a social discount rate for the United States in particular, we find a rate of 6.0 percent.11 We report the results for all countries in the last column of Appendix Table 1. There is a large degree of heterogeneity across countries owing to variation in our estimates of the social UDR and differences in the forecasted rate of grown in countries. To determine the extent to which heterogeneity is social discount rate is driven by heterogeneity in the social UDRs, we plot the former against the latter in Figure 4 for all countries. There appears to be no systematic pattern, suggesting that estimates of the social UDR are not explaining a substantial share of the heterogeneity in the social discount rates.12 This is not surprising because the very 10 https://www.imf.org/external/pubs/ft/weo/2015/01/weodata/index.aspx (Derived from NGDPRPC indicator) NOTE: We should provide more details about exactly how we derived this calculation and how it compares to the way others calibrate parameters in the Ramsey equation in the broader macro literature. 11 Although we focus on the mean aggregation rule, a potential alternative using the median for the United States may be of interest. Given the near zero rate of median income growth in the United States (DeNavas-Walt and Proctor 2015), a social discount rate based on the median aggregation would be close to 2.3 percent, which is close to the OMB’s lower bound. 12 Some countries have negative social discount rates as a result of negative per capita GDP growth projections. 10 different demographic patterns can lead to the same UDR, and it is unlikely that demographic patterns can be used to explain GDP growth. 5. Tests of Some Assumptions Our estimates of the social UDRs have been based on the WHO life tables that provide statistics in binned age classes. But for some countries, e.g., the United States, more detailed life tables are available such as the U.S. Social Security life tables. The Social Security Life Table treats every birth year as distinct (e.g., age 0,1,2,3,…). For the United States, therefore, we can therefore explore the sensitivity of our results to the level of detail captured in the life tables. Using the data available for every year of life, we estimate social UDRs of 4.1 and 2.4 percent for the mean and median aggregation rules, respectively. These are comparable to the previous U.S. estimates of 3.3 and 2.3 percent using the WHO life tables. The mean rate differ by 0.8 percentage points, and the median rate differs by only 0.1 percentage point. The mean rates are more sensitive to the older end of the age distribution, and the US Social Security Administration is much more detailed for the extreme elderly. We also explore the empirical basis for our assumption of a stable population structure. We do this by estimating country specific UDRs based on the WHO life tables for 1990, 2000, and 2012 for all countries in existence in 1990. Then to evaluation whether they vary to large degree, we illustrate the standard deviations of the estimates in Figure 5. The top panel maps the standard deviations for the mean aggregation, and the bottom panel does so for the median aggregation. We find that for most countries the UDRs are quite stable over this period. The mean standard deviation for mean and median aggregated social UDRs is less than 0.25 percentage points (0.20 and 0.14 percent, respectively). Moreover, more than 78 percent of countries’ standard deviations for the mean aggregation are less than a quarter of a percentage point, and the same holds for more than 88 percent of the countries regarding the median aggregation. The countries that do not have relatively stable rates experienced war, major political-economic change (e.g., fall of socialism in central and eastern Europe), or rapid development associated with major political or economic reforms and institutional change (e.g., Chili and China). 6. Discussion and Conclusion 11 Using age structure and life expectancy at age, we construct utility discount rates within the range, but slightly greater than the central tendency measure, that economists generally discuss as reasonable. However, difference in the aggregation approach can lead to difference is in the UDR that are on the order of 1 percentage point, which is quite substantial. Moreover, in comparison with statements of economists (Drupp et al. 2015; Weitzman 2001) it likely some economists are making various adjustments when stating the pure rate of time preference for a representative agent. Indeed, it is interesting that Drupp et al. (2015) find the older economists suggest high social utility discount rates. While they provide a few hypotheses, this result is not surprising if personal preferences are creeping into survey responses. Moreover, when combined with per capita GDP growth rate forecasts, our approach produces consumption discount rates in the range generally considered appropriate for business as usual consumption rates and rates observed in markets. Ramsey’s discounting rule remains the cornerstone of models of economic dynamics. Ethical arguments by economists often conclude that when applying the Ramsey rule the UDR, , should be equal to or near equal to zero (Ramsey 1928; Pearce and Ulph 1995; Asheim 2010; Kula 1997; Stern 2007).13 Ramsey developed, what in modern terminology, would be referred to as a representative agent model. The discounting literature is replete with references to such a representative agent in the Ramsey framework (e.g., Dasgupta 2008; Weitzman 2009). Ramsey’s work has provided great insight into how to discount, but Ramsey’s simplifying assumption of an infinitely lived representative agent, often interpreted as a sequence of nonoverlapping generations, provides little guidance on parameterizing his discounting rule. Ramsey’s assumption of an infinitely lived representative agent is fully consistent with our approach when all individuals have zero mortality risk. But, this can hardly represent any conceivable aggregation of agents alive today. The representative agent must die. Global climate change motivated the recent debates about discount rates, but infrastructure, conservation, and public investments of all types are evaluated using a social discount rate. The choice of appropriate discount rates for long lived public projects is complicated, and it is more complicated when these projects span countries, as in the case of climate change. 13 Kula (1997) reviews debates about discounting among classical economists and early neo-classical economists. 12 A core reason to worry about computing a utility discount rate or a consumption rate constructively, i.e., from the utility discount rate, is in models with endogenous growth or per capita consumption. In these cases the utility discount rate anchors the analysis. Such analyses are essential for the economics of climate change. While our results lend support to Nordhaus’s (2007) aggregate results. It is the regional and country level analysis and comparison that are critical. It is quite possible that when our country specific UDRs are used in place of a single UDR for all countries (e.g., Nordhaus and Yang 1996) the country level and aggregate results could be substantially different from analysis where the UDR is the same for all countries. Therefore, we hope the Supplemental Table with results for all countries is a useful resource for economists conducting such analyses. An interesting question emerges if the policy being evaluated will meaningfully impact the life expectance or demographic structure, making endogenious. Disease, addressable with healthcare investment, continue to be a major concern, especially in developing countries (Chakraborty, Papageorgiou, and Sebastian 2010). The practical implication is that the assumption a stable age distribution and UDR is probably much more tenable for developed countries. However, empirically our approach to calculating UDRs yields fairly stable estimates even for much of the developing world, at least for the last 30 years. This suggests that endogenous demographic change may be more on interesting academic problem that a practical policy concern. Within country heterogeneity is important for establishing social UDRs, but the use of social discount rates across countries is also not straight forward. Our work suggests that countries with young populations, which are overwhelming poorer, would place greater weight on future utility than countries with older populations, which are generally richer. Such younger countries may argue for the world to place greater weight on future utility provided by public, non-market goods. However, there does not appear to be a relationship between the rank ordering of UDRs and consumption discount rates. Though it is possible that poorer, less stable countries may have greater consumption rates, perhaps because they have expectations of higher growth rates of income or incur greater risk (perhaps as a result of poorly functioning institutions). This could result in a situation where countries with generally younger populations, which are generally poorer and subject to risk, articulate a desire for greater weight on the future, e.g., more global climate mitigation, conservation, and provision of other public goods, but 13 prefer to consume at high rates today relative to older (wealthier) countries. While older wealthier countries may be willing to forgo a greater amount of present consumption, but simply value future utility less since a larger fraction of their populations will not enjoy that future. This paper is certainly not the final word on discounting. Rather, we hope it is useful starting point. A starting point that leads to a more data-driven approach to establishing UDRs, which are critical for considering intertemporal allocation decisions. 14 References Asheim, Geir B. 2010. "Intergenerational Equity." Annual Review of Economics no. 2:197‐222. Chakraborty, Shankha, Chris Papageorgiou, and Fidel Perez Sebastian. 2010. "Disease, infection dynamics, and development." Journal of Monetary Economics no. 57:859‐872. doi: 10.1016/j.jmoneco.2010.08.004. Dasgupta, Partha. 2007. Human well‐being and the natural environment. New York: Oxford University Press. ———. 2008. "Discounting climate change." Journal of Risk and Uncertainty no. 37:141‐169. DeNavas‐Walt, Carmen, and Bernadette D. Proctor. 2015. Income and poverty in the United States: 2014. Washington, DC: United States Census Bureau. Drupp, Moritz, Mark Freeman, Ben Groom, and Frikk Nesje. 2015. Discounting disentangled: an expert survey on the determinants of the long‐term social discount rate. Centre for Climate Change Economics and Policy and Grantham Research Institute on Climate Change and the Environment. Eckstein, Otto. 1961. "A Survey of the Theory of Public Expenditure Criteria." In Public Finances: Needs, Sources, and Utilization, edited by Universities‐National Bureau, 439‐504. Princeton, NJ: Princeton University Press. Fredrick, Shane, George Loewenstein, and Ted O'Donoghue. 2002. "Time Discounting and Time Preference: A Critical Review." Journal of Economic Literature no. 40 (2):351‐401. Goulder, Lawrence H, and Roberton C. III Williams. 2012. The choice of discount rate for climate change policy evaluation. Washington DC: Resources for the Future. Kula, Erhun. 1997. Time Discounting and Future Generations: The harmful effects of an untrue economic theory. Westport, CT: Quorum Books. Millner, Antony, and Geoffrey Heal. 2014. "Resolving intertemporal conflicts: economics vs politics." NBER working paper no. 20705. doi: 10.3386/w20705. Nordhaus, William D. 2007. "A review of the Stern Review on the economics of climate change." Journal of Economic Literature no. 45 (3):686‐702. ———. 2014. "Estimates of the social cost of carbon: Concepts and results from the DICE‐2013R model and alternative approaches." Journal of the Association of Environmental and Resource Economics no. 1:273‐312. Nordhaus, William D, and Zili Yang. 1996. "A regional dynamic general‐equilibrium model of alternative climate‐change strategies." The American Economic Review no. 86 (4):741‐765. OMB. 2003. Circular A‐4: Regulatory analysis. edited by Office of Management and Budget. Washington, D.C.: Executive Office of the President. Pearce, David, and David Ulph. 1995. A social discount rate for the United Kingdom. University College London and University of East Anglia: Centre for Social and Economic Research on the Global Environment. Ramsey, F. P. 1928. "A mathematical theory of saving." The Economic Journal no. 38 (152):543‐559. Smith, V. Kerry, Donald H Jr Taylor, and Frank A Sloan. 2001. "Longevity expectations and death: Can people predict their own demise?" American Economic Review no. 91:1128‐1134. Stern, Nicholas. 2007. The Economics of Climate Change: The Stern Review. New York: Cambridge University Press. Weitzman, M.L. 2001. "Gamma Discounting." The American Economic Review no. 91 (1):260‐271. Weitzman, Martin L. 2009. "On modeling and interpreting the economics of catastrophic climate change." The Review of Economics and Statistics no. 91 (1):1‐19. 15 Figure 1: Mean (top) and median (bottom) country-specific social utility discount rates 16 Figure 2: Mean country-specific social utility discount rates by fraction of population over the age of 65 and life expectancy at birth 17 Figure 3: Country-specific social (consumption) discount rates 18 Figure 4: Social (consumption) discount rate against social utility discount rate for all countries using the mean aggregation rule 19 Figure 5: Standard deviation of mean (top) and median (bottom) social utility discount rates from 1990-2012 for Countries That Existed Prior to 1990 20 Appendix Table 1: Country-specific social utility discount rates and consumption discount rates 21 Appendix Table 1: (Continued) 22
© Copyright 2026 Paperzz