QUANTIFYING THE COSTS OF CLIMATE CHANGE
Abstract
Why is it so difficult to estimate the future costs of climate change? How do economists reach different conclusions using the same scientific data, and how do these conclusions shape their policy recommendations?
There are two broad categories for the key assumptions informing Integrated Assessment Models (IAMs)—models that utilize economic, geophysical, and geographic data—aimed at quantifying the costs of climate change. Structural elements, or facts, refer to phenomena that can be measured and observed, and ethical elements, or values, refer to the roles, rights and obligations of generations now and in the future (Ely, 2008; Dasgupta, 2007). These models are used to make predictions about the future economy as the planet warms.
Structural elements improve with better, more rigorous scientific data collection and modeling. IAMs quantifying the costs of climate change are aggregations of IAMs modeling smaller, more localized phenomena, like damages caused by hurricanes in the U.S, for example (Nordhaus, 2010). Scientific advancement will continue to fine-tune these instruments.
The ethical elements, however, offer some degree of latitude for economists. Using conventional cost benefit analysis (CBA) equations, cost estimates depend heavily upon the values chosen for δ (the discounted rate of time preference), and η (the elasticity of the marginal utility of consumption and risk aversion) functions in the Ramsey equation (Dasgupta, 2008; Heal, 2008; Nordhaus, 2007; Stern, 2006). These values are derived partially by observed market behaviors but still involve some degree of choice for the modeler as they choose their numerical representations of reality.
Some authors suggest that CBA equations may not be the best option to model the economics of climate change. The inherent uncertainty in the non-zero but remote chance of catastrophic consequences of climate change may require an entirely different mathematical structure (Weitzman, 2009). The summation of all current consumption to a single value may need to be replaced with vectors of consumption (Heal, 2008; Tol, 2008). Still others have suggested that the human element of these models needs to be better represented: social institutions, altruistic motivations, cooperation, etc. (Gowdy, 2007). This has prompted a third-wave of modeling that may, in the near future, address these many concerns and more accurately forecast our future (Burke, 2016; Stern, 2016).
The challenges posed by anthropogenic climate change will require interdisciplinary cooperation among economists, scientists, and social scientists to craft new tools that will better prepare humanity for the future.
Introduction
Anthropogenic climate change is the largest threat to biodiversity and human well-being that civilization faces. Excess carbon dioxide in the atmosphere is a negative externality, it is a nonexcludable and nonrival bad, and it is the result of wealthy nations burning disproportionate amounts of fossil fuels. Yet efforts to quantify the future costs of anthropogenic climate change is a relatively new field of study, beginning in the 1970s. Climate change has been dubbed as “possibly the greatest market failure in history” (Stern, 2006, p. 39) and the quest to appropriately quantify the social cost of carbon has prompted economists to decide how much money society should spend now in order to avoid the damages caused by rising temperatures, sea levels, and increased storm and drought intensities in the future (Dasgupta, 2008; Heal, 2008; Nordhaus, 2007; Stern, 2006). Society’s success or failure to act today will benefit or cripple future generations.
Climate change is an inherently difficult problem to solve because it is global, it is long-term with nonlinear relationships between cause and effect, and its consequences are irreversible. It is no surprise that our current economic models fail to account for all of these complexities. Still further, it is no surprise that the various cost estimates produced by these models result in different policy recommendations by the economists analyzing them.
Environmental economists agree that climate change needs to be addressed. Opinions differ regarding how quickly society needs to respond, what form that response should assume, and how much money should be allocated toward solutions. Integrated Assessment Models (IAMs) link the science of climate change to the monetary valuations of ecosystem services and natural resources while accounting for current economic conditions. These models require a great deal of economic and physical science data in order to make predictions about the costs society will incur in the future, but even when data are available, these models are filled with uncertainty.
The various components of IAMs can be grouped into two broad categories: the structural elements or facts, and the ethical elements or values (Ely, 2008; Dasgupta, 2007). Structural elements are stocks and flows that have already been quantified or can be directly measured. These include metrics like temperature, precipitation, wind speed, sea level rise, interest and savings rates, capital assets, GDP, and so on. Economists are using these same empirical measurements in all of their models. So why do different economists reach different conclusions about the future costs of climate change?
This question can be answered by observing the variations in the ethical elements of the IAMs. Most economists use some variation of the Ramsey Equation, which is a social welfare function meant to compare the costs and benefits of a phenomenon over many generations. It has three components: the pure rate of time preference (δ), a measure of relative risk aversion (η), and a rate of growth of per capita consumption. The pure rate of time preference (also called the social rate of discount) can be thought of as the rate at which we discount or discriminate against future humans purely because they are in the future (Heal, 2008). The risk aversion factor considers more than just the degree to which humans avoid risk; it reflects society’s aversion to interpersonal inequality today and tomorrow (Dasgupta, 2008). The rate of growth of per capita consumption assumes that future generations will be wealthier than current generations, and therefore contributes to the idea that a unit of consumption tomorrow is less valuable than a unit of consumption today (Ely, 2008). Economists choose numerical values for these terms based on observable data, but also on their own judgement. The majority of this paper will explore the ways in which these ethical elements shape the outcomes of IAMs.
Structural Elements: What they are and how they fail.
Quantifying the costs of climate change requires supercomputing capabilities. It is, after all, an aggregation of all available data across the entire planet. To better conceptualize an IAM, it is helpful to look at a smaller subset of data. William Nordhaus used wind speed, temperature, rainfall, storm frequency, elevation, population, capital stocks, and GDP data to forecast future storm damage and had two key findings: 1) damages rise with the ninth power of the maximum wind speed of the hurricane and 2) the cost of damage will increase by 0.08% of U.S. GDP ($10 billion in 2005 USD) due to climate change (Nordhaus, 2010). He also used HURDAT data which clearly show an increase in storm frequency as a consequence of warming temperatures. By linking the physical science of cyclone characteristics to the monetary value of the cities struck by the storm, Nordhaus was able to make predictions about the future ramifications of a warming planet in this particular context. Even though this task seems relatively straightforward, it is anything but simple.
Each of these variables has a certain degree of uncertainty associated with it. Not one single variable can be predicted with complete accuracy or precision. For example, Nordhaus notes that, “the precise relationship between storm intensity and damage will differ for different materials (brittle vs. flexible), for different objects (windows vs. levees), and for different design tolerances” (p. 9). Further, Nordhaus lists several omitted variables, including the value of the vulnerable capital stock at hurricane landfall, storm size, storm surge, and storm lifetime. Including these values may have improved his model.
Increased storm intensity is just one subset of climate change. This hurricane example is a simplification with a distinct advantage over globalized IAMs: The United States manages a great deal of massive federal data sets that are available to the public. These include, but are not limited to, economic data from the Bureau of Economic Affairs, storm characteristic data from the National Oceanic and Atmospheric Administration, emissions data from the Environmental Protection Agency, and energy consumption data from the Energy Information Administration. However, not every nation has the ability to collect, create, and distribute such datasets. Quantifying total planetary damages from climate change requires data that have not yet been recorded. As such, there are enormous gaps in humanity’s collective scientific and economic knowledge.
Even if it were possible to attain and incorporate perfect information, computational representations of reality have their limits. Climate models, for instance, usually only have spatial resolutions of 100 by 100 km cells (Climate Modeling, 2011). Nordhaus does not state the latitudinal/longitudinal resolution of his model, but he claims to have categorized all capital assets 8 meters or less below sea level as “vulnerable to storm damage” and that immobile assets (buildings, infrastructure) are considered more vulnerable than mobile assets. This decision-tree to determine the damages caused by hurricanes is logical and useful, but it is also perpetually outdated. New commercial, residential, and industrial development continuously renders estimates like these irrelevant, and so provides another dimension of ambiguity. When there is a great deal of uncertainty regarding just one city block, that doubt is compounded as the spatial scale becomes larger and more information is aggregated. Therein lies the heart of the problem with structural elements in IAMs: the error bars grow wider as modelers strive toward a mathematical representation of planet Earth.
There are three key arenas in modeling the structural elements of climate change. These areas are: the amount of carbon stored in stocks and transported through flows, the relationship between increased CO2 in the atmosphere and temperature—a term called climate sensitivity—and the damages incurred by increased temperatures (Ely, 2008; Heal, 2008; Weitzman, 2009). Stocks and flows of carbon are perhaps the easiest to quantify, although all values assumed will contain inherent uncertainty. Climate sensitivity is even more difficult to account for as it is a nonlinear relationship that depends upon several factors: the radiative forcings of various molecules in the atmosphere, the concentrations of those molecules (in relative and absolute amounts), and albedo (Cleveland and Kaufmann, 2018). Finally, the damages caused by an increase in temperature are extrapolative. Paleoclimate data can be used to try to make predictions about the various ramifications of an abrupt rise in global temperature, but even these ancient proxies cannot give scientists a complete picture because current warming is several orders of magnitude faster than anything ever experienced by the planet through natural means (Weitzman, 2009). This paper will return to the issue of modeling unobservable and unprecedented warming in the final section.
Beyond the real dollar value of physical assets, economists agree that non-monetary values—existence value, bequeath value—need to be considered in order to fully encapsulate the scale of the problem (Dasgupta, 2008; Ely, 2008; Nordhaus, 2008; Tol, 2008). Contingent valuation of ecosystem services and natural resources isn’t given separate consideration in IAMs. However, these non-monetary values are not completely absent from cost estimates; instead they are regarded within the constructs of the ethical elements.
Ethical Elements: Humanity’s moral code, well-being, and consumption in 3 numbers.
Economists are all using the same incomplete and uncertain factual data in their models. That means that the personal assumptions, reasoning, and logic behind the values chosen for the rate of time preference, risk aversion, and the consumption growth rate are responsible for producing dramatically different policy recommendations.
The rate of time preference reflects the fact that humans tend to value benefits received today over benefits received tomorrow due to impatience and risk aversion. To choose a value for δ larger than zero indicates the belief that the felicity, or the well-being, of current generations is more valuable than that of future generations and is thus discounted (Dasgupta, 2008; Gowdy, 2007). A discount rate of zero would indicate intergenerational equity, meaning that the consumption and well-being of humans now is equivalent to that of humans not yet born. In choosing a value for the discount rate, economists also assume that future generations will be wealthier than current generations and can therefore afford to spend more money adapting to climate change than current generations can afford to spend on mitigating climate change (Dasgupta 2008; Nordhaus, 2007; Weitzman, 2009). Some economists disagree with this view and instead assume that the well-being and wealth of future generations will be diminished with natural resource and ecosystem service degradation (Cline, 1996; Stern, 2006 and 2016).
The value chosen for η explicitly models risk aversion and, by extension, encapsulates society’s aversion to interpersonal inequality in the present generation (Dasgupta, 2008). Climate change is the result of fossil fuel burning by wealthy nations that disproportionately affects poor nations who did not contribute to the problem (Ely, 2008; Heal, 2008; Nordhaus, 2007; Stern, 2006). Thus, the distribution of costs borne today needs to reflect the injustice emitters have brought upon sufferers both today and tomorrow. This follows the “polluter pays” principle but is an extremely unpopular concept in the United States. A lot of information about the rights, roles, and responsibilities of all global actors is compressed into just these two numbers, and different assumptions and assignments produce dramatically different cost estimates.
Additionally, economists use a value to represent the consumption growth rate, or the amount of additional consumption that future people will be able to enjoy. This is an assumption that some consider to be fundamentally flawed as perpetual growth is hindered by an ecological ceiling (Stern, 2016). This issue will be addressed in more detail in the final section of the paper.
The best illustration of opposing policy recommendations resulting from different values of δ and η comes from a comparison of William Nordhaus’ DICE model and Nicholas Stern’s PAGE model—their respective IAMs. Nordhaus’ value for δ is 3% declining to 1% per year in 300 years’ time, his value for η is 1, and his assumed growth rate is 4.3% per year. Nordhaus justifies his values by calibrating them to observed market interest rates, values of forecasted consumption, and rates of private and public savings investment. By using observable human behavior about how real people choose to spend or save their money, Nordhaus crafts a policy based on positive economics. He suggests that society should invest in reproducible capital now (in 2007, at the time of his writing) and start implementing carbon controls in several decades in a gradual, ramp-like fashion with increasing abatement over time. In his view, “the highest-return investments today are primarily in tangible, technological, and human capital, including research and development on low-carbon technologies. In the coming decades, damages are predicted to rise relative to output. As that occurs, it becomes efficient to shift investments toward more intensive emissions reductions” (Nordhaus, 2007, p. 688).
By contrast, Stern’s discount rate is 0.1% per year, his value for η is 1, and his consumption growth rate is 1.4% per year. With these parameters in place, Stern advocates for immediate and strong action on climate change. Abatement, now. He estimates the costs of damages incurred from warmer temperatures to be 5% of global GDP—with somewhat conservative estimates about the structural elements in the model—but that they could be up to 20% of global GDP every year, indefinitely. He suggests that wealthy nations invest 2% of their GDP to avoid these future costs.
Stern’s model has drawn keen criticism from many economists primarily because the risk aversion factor adheres to revealed preferences in the observable market, but the discount rate of time preference does not. It is a number chosen based on the philosophical argument that all humans should be weighted equally rather than discriminated against due to their birthdates. This is an example of normative economics, or recommendations made for an idealized world, rather than the actual world we live in.
Economists do not dissent to the idea of intergenerational equity, but rather to Stern’s inconsistency in choosing values. When using 0.1% for δ and 1 for η, the model dictates that 97.5% of aggregate output needs to be saved for the future to avoid the consequences of climate change—resulting in starvation now for the sake of consumption in the future (Dasgupta, 2007). Criticized another way, this model produces, “a one-time consumption hit of approximately $30,000 billion today” to address speculative damages that occur in about two centuries’ time (Nordhaus, 2007, p. 696). Clearly the outcome of Stern’s model is extreme, daunting, and highly unlikely to convince policymakers and the general public to invest so heavily in climate change mitigation.
Partha Dasgupta (2008) has suggested that values for δ and η should not be chosen according to current market trends, but rather according to forecasted trends and society’s conception of distributive justice. He argues that it is ethical to use a near-zero value for δ, but notes that all of the other moral implications then rely upon the value chosen for η. He took Stern’s model, kept all other values the same, and changed η to 3, thus producing a savings rate of 25% today to avoid the most expensive and extensive damages from climate change. This is still a large chunk of consumption forgone, but comes much closer to current savings rates seen in the market. Dasgupta concludes his paper by stating, “it is possible to believe that Humanity should invest a lot more in reducing climate change than the 2% of the GDP of rich countries proposed by Stern (2006). One can hold such a belief even while being unable to justify it from formal modelling.”
Mitigation Recommendations: Quantity versus price.
The ultimate goal of quantifying the costs of climate change is to craft policies that will dramatically reduce CO2 emissions—either with a carbon tax or cap-and-trade market—and to invest in renewable energy sources. Some economists, like Richard Ely (2008), recommend quantity-based mechanisms, i.e. a cap-and-trade market, to ensure that emissions are appropriately cut to stabilize atmospheric concentrations of CO2. Other economists focus on calculating the social cost of carbon—which is defined as the net present value of the damages caused by increasing carbon dioxide emissions—and taxing accordingly (Nordhaus, 2007; Tol, 2008). Increasing the price of carbon should dissuade consumers from using carbon-intensive goods, but the amount of abatement remains unclear. Economic theory suggests the outcomes produced by a carbon tax and a cap-and-trade market are extremely similar. The true barrier to implementation of mitigating practices is political will, which is beyond the scope of this text and will not be granted further consideration.
The Third Wave: The various ways we can improve our models.
Every climate change economist who has authored academic literature on the subject has explicitly mentioned the various ways in which traditional cost-benefit analysis fails to achieve a robust picture of the future, a warmer planet, and its associated costs.
Martin Weitzman (2009) rails against the CBA by proving that it is a mistake to assume a thin-tailed probability distribution for catastrophic events to take place due to climate change. He aggregated 22 climate models used by the IPCC and found that there was a 5% chance of a 7°C increase in temperature over 200 years, using a very strict climate sensitivity parameter. This is a conservative estimate that suggests Earth may warm far beyond the recommended 2°C stabilization target required to avoid the worst damages. With a more elastic climate sensitivity parameter, he finds a 5% chance of a 10°C increase! He notes that such an increase in temperature would make catastrophic events like melting permafrost and methane hydrates much more likely, ultimately representing the end of civilization as we know it. In other words, the tails of the probability distribution functions in our models should have fat tails, not thin ones. He likens this idea to the value of a statistical life (VSL) but instead calls it, “the value of statistical life on Earth” and introduces lambda into the equation to account for this massive uncertainty. Running a Monte Carlo simulation with λ > 1,000 resulted in a future simulation where 99.9% of welfare-equivalent consumption is destroyed in the EU due to climate change. None of our current IAMs have produced such a dramatic number and yet the possibility of complete and utter disaster is non-zero. Future models need to take this into consideration.
Weitzman’s concern for the VSL on Earth is echoed by Marshall Burke (2016) and others who note that there is no geologic analog for scientists to directly study in order to make more accurate forecasts. The rate of current climate change has never been seen in the geologic record before, and so society is left to examine past mass extinctions to glean some semblance of understanding, even though these extinctions occurred over millennia, rather than centuries. Weitmzan and Burke agree that the risk aversion parameter, η, is not calibrated for the most extreme levels of disaster posed by climate change. More terms will need to be included in the equations of IAMs to make more useful predictions.
Geoffrey Heal (2008) and Richard Tol (2008) highlight another shortcoming of our current models: all Earthly goods and services are aggregated to a single value representing consumption, even though these goods and services are affected by local and regional conditions and therefore deteriorate at different rates. Both economists suggest using vectors to represent these goods and services to better capture these differences. Nicholas Stern (2016) builds off this idea and notes that the variable deterioration of natural goods and ecosystem services means that the ecological ceiling prevents perpetual economic growth. He argues that using BAU (“business as usual”) as the baseline for projections of economic welfare is inherently flawed because it assumes there is no limit to how much humans can mine and extract from the land in order to generate more usable items. “Business as usual” will cause economic output to decline with rising temperatures, rather than stay constant. These criticisms of current IAMs have resulted in just about every climate change economist agreeing that consumption will vary over time and space, and should not be assumed to be a fixed product of whichever equation is being used to calculate it.
Some condemnations of current approaches to climate change economics extend beyond the technical limitations of the models and breach moral ground. The debate over whether or not it is ethical to reduce nature to a dollar value has existed before climate change came front and center to the global stage, but John Gowdy (2007) goes a bit deeper with his assessment. He laments that current economic theory is based on the idea that each actor is acting rationally in his/her best interest to maximize total welfare. He argues that climate change will diminish the well-being of everyone if we fail to act today, and therefore our models should better represent the altruistic, cooperative, and self-preserving behaviors of human beings by incorporating a value representing our social behaviors and institutions. On the other side of this coin, Burke (2016) further notes that new IAMs will need to account for social conflicts, like mass migrations and wars over scarce resources. The models from the early 2000s do not consider these costs in their analyses.
Finally, the big unknown that every economist has pondered extensively is the role of technology. If carbon sequestration technology, improvements in energy efficiency, and transitions to renewable energy develop quickly, society may be able to prevent the most catastrophic damages. However, this is a monumental economic investment; essentially every scholar acknowledges that it may not come forth quickly enough to stabilize atmospheric concentrations of CO2 at reasonable levels. Human ingenuity and capital can be substituted for natural resources, but only to a certain point. Society will need to engineer its way out of the problem before ecosystems completely collapse.
Conclusion
Quantifying the costs of climate change is extremely difficult due to the planet-sized spatial scale of the problem, the infinite time horizon over which damages will unfold, and the numerous uncertainties that envelop our attempts to forecast the future. Scientific uncertainty regarding the carbon cycle stocks and flows, climate sensitivity, and the unknown local and regional expressions of a warmer Earth permeates IAMs and forces academics of all disciplines to pause before stating definitively what can and should be done today to ensure that the world is still habitable 200 years from now.
Economists using the same available scientific data can still make radically different policy proposals based on the values they choose for the discounted rate of time preference, risk aversion, and consumption growth rate. William Nordhaus chose to keep all of his values consistent with observed market trends, while Nicholas Stern opted for values that emphasized the moral obligations of this generation to future generations. Neither of their strategies— immediate abatement nor gradual abatement—has been adopted by the United States, and it has already been twelve years since the publication of these economic papers. The lack of clear, definitive cost estimates, and political stubbornness have paralyzed some U.S. policymakers at a time when every day we emit excessive amounts of CO2 tips the scale further toward utter planetary devastation.
A third wave of modeling has tried to address the copious concerns regarding uncertainty, probability, and regional/local variations including social costs, and at the time of this paper’s writing, the latest cost estimate is hundreds of billions of dollars by the end of this century, potentially shrinking the economy by 10% (U.S. Global Change Research Program, 2018). Economists certainly do not agree on how large an avoidance cost society should pay now, but they all maintain that doing nothing is not an option. Unfortunately, the United States seems to have chosen exactly that, to the detriment of all life on Earth.
References
Burke, M., Craxton, M. et al. (2016). ‘Opportunities for Advances in Climate Change Economics’, Science, 352(6268) [online]. Available at: http://science.sciencemag.org/content/352/6283/292 (Accessed: 11 November 2018)
Cleveland, C.J., Kaufmann, R.K. (2018). ‘Introduction to Environmental Science.’ Boston University, Boston, Massachusetts
Cline, W. (1996). ‘The Impact of Global Warming of Agriculture: Comment’, The American Economic Review, 86(5), 1309-1311 [online]. Available at: http://www.jstor.org/stable/2118295 (Accessed: 10 December 2018)
‘Climate Modeling.’ 2011. Available at: https://scied.ucar.edu/longcontent/climate-modeling. (Accessed 7 December 2018)
Dasgupta, P. (2008) ‘Discounting Climate Change’, Journal of Risk and Uncertainty, 37(2-3) [online]. Available at: https://link.springer.com/article/10.1007/s11166-008-9049-6(Accessed: 13 November 2018)
Dasgupta, P. (2007) ‘The Stern Review's Economics of Climate Change’, National Institute Economic Review, 199(1) [online]. Available at: https://doi.org/10.1177%2F0027950107077111(Accessed: 13 November 2018)
Ely, R.T. (2008) ‘The Economics of Climate Change’, American Economic Review: Papers and Proceedings, 98(2) [online]. Available at: https://pubs.aeaweb.org/doi/pdf/10.1257/aer.98.2.1(Accessed: 28 October 2018)
Gowdy, J.M. (2007) ‘Behavioral Economics and Climate Change Policy’, Rensselaer Working Papers in Economics, 701(1) [online]. Available at: http://www.economics.rpi.edu/workingpapers/rpi0701.pdf(Accessed: 12 November 2018)
Heal, G. (2008) ‘Climate Economics: A Meta-Review and Some Suggestions’,
Review of Environmental Economics and Policy, 3(1) [online]. Available at: https://www.nber.org/papers/w13927.pdf(Accessed: 29 September 2018)
Nordhaus, W.D. (2007) ‘A Review of the Stern Review on the Economics of Climate Change’,Journal of Economic Literature,45(3)[online]. Available at: https://pubs.aeaweb.org/doi/pdfplus/10.1257/jel.45.3.686(Accessed: 6 November 2018)
Nordhaus, W.D. (2010) ‘The Economics of Hurricanes and Implications of Global Warming’, Climate Change Economics, 1(1) [online]. Available at: https://www.worldscientific.com/doi/pdf/10.1142/S2010007810000054(Accessed: 9 November 2018)
Stern, N.H. (2006) The Economics of Climate Change: The Stern Review. Cambridge: Cambridge University Press.
Stern, N. (2016) ‘Economics: Current Climate Models are Grossly Misleading’, Nature,530(7591) [online]. Available at: https://www.nature.com/news/economics-current-climate-models-are-grossly-misleading-1.19416(Accessed: 11 November 2018).
Tol, R.S.J. (2008) ‘The Economic Impact of Climate Change’,Perspektiven der Wirtschaftspolitik, 11(Supplement) [online]. Available at: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.417.9912&rep=rep1&type=pdf(Accessed: 2 October 2018)
USGCRP. (2018). ‘Fourth National Climate Assessment: Summary Findings.’ NCA4, [online]. Available at: https://nca2018.globalchange.gov
Weitzman, M.L. (2009) ‘On Modeling and Interpreting the Economics of Catastrophic Climate Change’, The Review of Economics and Statistics, 91(1) [online]. Available at: https://www.mitpressjournals.org/doi/pdf/10.1162/rest.91.1.1(Accessed: 7 November 2018)