A Service of

Leibniz-Informationszentrum econstor Wirtschaft Leibniz Information Centre Make Your Publications Visible. zbw for Economics

Nordhaus, William

Article Integrated assessment models of climate change

NBER Reporter

Provided in Cooperation with: National Bureau of Economic Research (NBER), Cambridge, Mass.

Suggested Citation: Nordhaus, William (2017) : Integrated assessment models of climate change, NBER Reporter, National Bureau of Economic Research (NBER), Cambridge, MA, Iss. 3, pp. 16-20

This Version is available at: http://hdl.handle.net/10419/178753

Standard-Nutzungsbedingungen: Terms of use:

Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Documents in EconStor may be saved and copied for your Zwecken und zum Privatgebrauch gespeichert und kopiert werden. personal and scholarly purposes.

Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle You are not to copy documents for public or commercial Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich purposes, to exhibit the documents publicly, to make them machen, vertreiben oder anderweitig nutzen. publicly available on the internet, or to distribute or otherwise use the documents in public. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, If the documents have been made available under an Open gelten abweichend von diesen Nutzungsbedingungen die in der dort Content Licence (especially Creative Commons Licences), you genannten Lizenz gewährten Nutzungsrechte. may exercise further usage rights as specified in the indicated licence. www.econstor.eu Integrated Assessment Models of Climate Change

William Nordhaus

Many areas of the natural and lenge to natural and social scien- social sciences involve complex tists, who must incorporate a wide systems that link multiple areas variety of geophysical, economic, and disciplines. This is particu- and political disciplines into their larly true for the science, econom- diagnoses and prescriptions. ics, and policy of climate change, Integrated assessment mod- is Sterling Professor of which involve a wide variety of els of climate change grew organi- Economics at Yale University and a research fields from atmospheric chemistry cally from energy models. One of associate in four NBER Programs: Economic to game theory.1 As understand- the earliest careful comparisons of Fluctuations and Growth; Environment and ing progresses across the differ- energy models was the Modeling Energy Economics; Productivity, Innovation, ent fronts, it is increasingly neces- Resource Group (MRG) analysis of and Entrepreneurship; and Public Economics. sary to link the different areas to the 1970s.2 This project, chaired He is a pioneer in analyzing economic conse- develop models and policies that by Nobel Prize-winning economist quences of climate change and one of the lead- reflect the complex interactions. Tjalling Koopmans, analyzed sev- ing global authorities on climate policy. A full analysis would reflect that eral energy models that projected Nordhaus developed the DICE and RICE economic activity drives emissions, energy demands and technologies models of the economics of climate change, which affect atmospheric concen- over a long time horizon. The ear- which are widely used in research on climate- trations, thence climate and the lier work of Koopmans on the lin- change economics and policies. He also has hydrological cycle, which in turn ear programming approach to pro- studied a range of additional topics in both affect human and natural systems, duction, as well as the Samuelson micro- and macroeconomics, including wage which ultimately contribute to the principle of “markets as maximiza- and price behavior, health economics, aug- determination of climate policies. tion,” formed the intellectual core mented national accounting, the political busi- Integrated assessment analyses of much of the energy modeling ness cycle, and the dynamics of productivity. He and models play a key role in put- starting at that time and proceed- is the author of three treatises on climate pol- ting the pieces together. Integrated ing to the present.3 icy, The Climate Casino, Managing the Global assessment models (IAMs) can be The first IAMs in climate Commons, and Warming the World, and is a co- defined as approaches that inte- change were basically energy models author, with the late Paul Samuelson, of the clas- grate knowledge from two or more with an emissions model included, sic textbook, Economics. domains into a single framework. and later with other modules such Nordhaus served as president of the These are sometimes theoreti- as a model and a small American Economic Association (AEA) in 2015. cal but are increasingly comput- climate model. My early approaches He is a member of the National Academy of erized, empirical, dynamic, non- were partial equilibrium energy Sciences and a Fellow of the American Academy linear models of varying levels of models with exogenous output.4 of Arts and Sciences, as well as a Distinguished complexity. A. S. Manne’s model, the first to Fellow of the AEA. He also has been a mem- The challenge of coping with imbed an energy system in a full ber of the President’s Council of Economic global warming is particularly economic-growth model, was an Advisers (1977–9), provost of Yale University daunting, because it spans many important landmark.5 The earliest (1986–8), a member of the Brookings Panel on disciplines and parts of society. versions of my models adopted a Economic Activity since 1972, and a director Ecologists may see it as a threat to growth-theoretic framework simi- and chair (2013–14) of the board of directors of ecosystems, marine biologists as a lar to Manne’s and extended it to the Federal Reserve Bank of Boston. problem leading to ocean acidifi- geophysical variables.6 A native of Albuquerque, New Mexico, who cation, and coastal communities as IAMs are increasingly used in received his B.A. from Yale and his Ph.D. from a lottery with intense hurricanes, analyses by national governments MIT, Nordhaus lives in New Haven with his while ski resorts may view it as and in international assessments. wife, Barbara, who works at the Yale Child Study a mortal danger to their already Among the most important appli- Center. short seasons. It also poses a chal- cations are:

16 NBER Reporter • No. 3, September 2017 • Making projections that Climate and the Economy (RICE) specified using the Negishi approach, have consistent inputs and out- models have gone through several in which regions are aggregated using puts of the different components revisions since their initial develop- time- and region-specific weights sub- of the system so that, for example, ment around 1990. The latest pub- ject to budget constraints. the GDP projections are consistent lished versions are the RICE-2010 This sketch of a pair of IAMs in with the emissions projections. and DICE-2016R2 models. The latest the DICE and RICE models makes it • Calculating the impacts of DICE model is available in GAMS, clear that they are highly simplified alternative assumptions on impor- a fine mathematical software system, representations of complex economic tant variables such as output, emis- and a full description of the earlier and geophysical realities — what sions, temperature change, and version is available.7 might be called geo-macroeconomics. the effect of economic activity on DICE is a globally aggregated While small and comprehensive mod- climate. model. RICE is essentially the same, els have many advantages, they also • Tracing the effects of alter- except that output and abatement have major shortcomings because of native policies on all variables in a have structures for 12 regions. This their simplifications. consistent manner as well as esti- discussion will use the term “DICE A useful analogy here is to return mating the costs and benefits of model,” and for most modules the to the animal kingdom. Each model is alternative strategies. analysis applies equally to the RICE like an animal that has its fruitful niche • Estimating the uncertainties model. in the analytical ecosystem. Small mod- associated with alternative variables The DICE model views the eco- els can be fleet and can adapt easily to and strategies along with the value nomics of climate change from the a changing environment or new data, of research and new technologies. perspective of neoclassical economic while large models take many years to There are dozens of IAMs today, a growth theory. In this approach, mature but are able to handle much handful of which have a track record economies make investments in cap- larger and more complex tasks. of at least a decade. Models range ital, education, and technologies, from small ones like the DICE model thereby reducing consumption today Illustrative Results described below to enormous ones in order to increase consumption in that may have as many as a half-mil- the future. The DICE model extends Here are some representative results lion variables. Different IAMs are like this approach by including the “natu- using the DICE model. These results different animals in terms of com- ral capital” of the as an are from the most recent version of the parative strengths and weaknesses in additional kind of capital stock. By model DICE-2016R2.8 One applica- tackling the various questions above. devoting output to emissions reduc- tion is to compare the economic and Small comprehensive models can yield tions, economies reduce consump- climate trajectories associated with dif- a full cost/benefit analysis, but are tion today but prevent economically ferent policy approaches. Here are four weak on regional or industrial detail. harmful climate change and thereby different policy options: Larger species provide great detail, but increase consumption possibilities in Baseline: No climate-change poli- may be unable to trace impacts and the future. cies are adopted. damages, are less transparent, and are The DICE model has 12 behav- Optimal: Climate-change policies unable to do full uncertainty analy- ioral equations, two variables to be maximize economic welfare, with full ses. Some models are able to trace the optimized, and several identities. participation by all nations starting in impacts of policies on land use. Others In the GAMS version, the simplest 2020. can investigate a wide range of tech- model has about 240 lines of opera- Temperature-limited: The optimal nologies. A few have full damage func- tional code. A run of 1,000 years takes policies are undertaken subject to a tions, while others include a limited five seconds, so it can be used for proj- further constraint that global tempera- number of technologies and engineer- ects with multiple states of the world ture does not exceed 2.5 °C above the ing variables. The great diversity of and Monte Carlo experiments. 1900 average. (The international goal the modeling ecosystem allows most The RICE model has the same of 2 °C is not feasible with current important questions to be addressed. basic economic and geophysical struc- DICE estimates without technologies ture, but contains a regional elabo- that allow negative emissions by mid- The DICE and RICE ration. The specification of prefer- 21st century.) Models as Examples ences in RICE is different because Stern discounting: These are results it must encompass multiple regions. associated with an extremely low dis- The Dynamic Integrated model The general preference function is count rate as advocated by The Stern of Climate and the Economy (DICE) a Bergson-Samuelson social welfare Review on the Economics of Climate and Regional Integrated model of function over regions. The model is Change.9

NBER Reporter • No. 3, September 2017 17 Figures 1–3 show outcomes associated with the four Projected CO Emissions in Dierent Scenarios policy options. A few key results emerge. First, emissions Global industrial emissions of CO (Gt / year) differ sharply, with major cuts in emissions in cases with 80 Baseline ambitious policies. Second, the temperature in the uncon- 70 trolled scenario continues to rise sharply over the current 60 century. 50 Figure 3 is the outcome that is generated uniquely by

40 IAMs: the carbon prices associated with each policy. This Optimal 30 is a concept that measures the marginal costs of reductions of emissions of greenhouse gases. In a market environment 20 Optimal with low Optimal with maximum temperature increase of 2.5 °C such as a cap-and-trade regime, the carbon prices would be 10 discount rate (Stern discounting) the trading price of carbon emission permits. In a carbon- 0 2010 2025 2040 2055 2070 2085 2100 tax regime, they would be the harmonized carbon tax among participating regions. Carbon prices in the baseline scenario Source: W.D. Nordhaus, NBER Working Paper No. 22933 are the current average prices in world markets, roughly $2 per ton of CO2. Prices under the optimal and temperature- Figure 1 limited scenarios at first rise to $35 and $229 per ton of CO2, respectively, by 2020. The carbon prices associated with the low-discounting scenario are close to those of the Temperature Change in Di erent Scenarios temperature-limited policy. The carbon price is closely linked to an important pol- Global mean temperature increase since 1900 (°C) icy instrument, the , or SCC. This con- Baseline cept represents the economic cost of an additional ton of 4 carbon dioxide emissions (or, more succinctly, carbon) or its Optimal equivalent. The advantage of IAMs is that they can calculate 3 Optimal with maximum the shadow price of carbon emissions along a reference path temperature increase of 2.5 °C of output, emissions, and climate change. In an optimized 2 Optimal with low discount rate climate policy, abstracting away from distortions, the social (Stern discounting) cost of carbon will equal the carbon price or the carbon tax. 1 Estimates of the SCC are a critical ingredient in cli- mate-change policy. They provide policymakers a guidepost 0 2010 2025 2040 2055 2070 2085 2100 to aim for if they are seeking an economically efficient pol- icy for carbon pricing. Another application is for rulemak- Source: W.D. Nordhaus, NBER Working Paper No. 22933 ing where countries do not have comprehensive policies cov- Figure 2 ering all greenhouse gases. In this context, regulators might use the SCC in a calculation of social costs and benefits of policies involving energy or climate-affecting decisions. For example, the U.S. government has undertaken rulemaking Carbon Prices in Di erent Scenarios proceedings to determine the SCC for use in such areas as subsidies for the installation of low-carbon energy sources, Price per ton of CO (2010 U.S. dollars) 500 regulations requiring energy efficiency standards in build- ings and motor vehicles, and for power plants. Current regu- 400 lations using the SCC have more than $1 trillion in benefits, Optimal with low discount rate 10 (Stern discounting) according to the U.S. Environmental Protection Agency. 300 Estimates of the SCC vary by model and approach. Optimal with maximum temperature increase of 2.5 °C Table 1, on the following page, shows estimates from the 200 most recent DICE model. Two points emerge here. The Optimal 100 first is the critical importance of discounting. Looking at 2020, the SCC ranges from $22 to $133 per ton of CO2 as Baseline 0 the real discount rate ranges from 2.5 to 5 percent per year. 2010 2025 2040 2055 2070 2085 2100 Second, the SCC is extremely high — around $200 per ton Source: W.D. Nordhaus, NBER Working Paper No. 22933 of CO2 — for damage functions that would justify the tem- perature-limited objective (2 °C) that has been adopted at Figure 3 international meetings in Copenhagen and Paris.

18 NBER Reporter • No. 3, September 2017 economic impacts. tant feature of damage studies is that they Social Cost of Carbon (2010 U.S. Dollars / Ton of CO) Take the “simple” are generally limited to global temperature example of tropical increases of up to 3 °C, with the upper limits Scenario 2015 2020 2025 2030 2050 cyclones — hurricanes shown in Figure 2 not well-studied.

Base parameters in the United States. A difficulty in assessing IAMs is the Baseline 30.0 35.7 42.3 49.5 98.3 Basic physics indicates inability to use standard statistical tests Optimal controls 29.5 35.3 41.8 49.2 99.6 that a warmer ocean because of the lack of a probabilistic struc- is likely to increase ture. One concern, alluded to above, is 2.5 degree maximum the intensity of hurri- the major uncertainties associated with Maximum 184.1 229.0 284.0 351.0 1,008.4 canes; more precisely, the results of IAMs. The uncertainties are Max. for 50 years 147.2 183.2 227.2 280.4 615.6 the probability distri- particularly pronounced because of the bution of damage from long time periods — literally hundreds discounting Uncalibrated 256.5 299.6 340.7 381.7 615.5 wind speed will shift to of years — required for estimates of opti- the right. IAMs need mal policies and the SCC. A recent study

Alternative discount rates to integrate this find- using the DICE-2016R2 model examined 2.5% 111.1 133.4 148.7 162.3 242.6 ing with economic the uncertainties of major outcomes from 3% 71.6 85.3 94.4 104.0 161.7 impacts. A major sur- parametric uncertainty of five major vari- 4% 34.0 39.6 44.5 49.8 82.1 5% 18.9 21.7 24.8 28.1 48.4 prise of research here ables: equilibrium temperature sensitivity, was that the economic productivity growth, the damage function, impacts of hurricanes the carbon cycle, and the rate of decarbon- Source: W.D. Nordhaus, NBER Working Paper No. 22933 are a high-power func- ization. Table 2, on the following page, tion of maximum wind shows the results. The best guess is the stan- Table 1 speed; normalized by dard DICE model where parameters are economic variables set at their expected values, while the oth- Calibration in IAMs such as local capital, damage is estimated ers are the distribution of outcomes. For to be wind speed to the ninth power.11 the important SCC calculation, the mean IAMs belong to a class of models, both But that is only a small slice of damage. value with full uncertainty is about 15 per- in economics and more generally in applied To construct damage functions, research- cent above the best guess. The change in sciences, that rely on calibration rather than ers need to aggregate across sectors, regions, temperature for 2100 is only slightly higher econometric estimation. Calibration involves levels of development, and climate change than the best guess. On the other hand, determination of system param- output is much higher because of eters and behavior using external Damage and Temperature Estimates from Major Studies the large estimated uncertainty of evidence rather than statistical productivity growth. Estimated damage (percentage of global output) systems estimation. Calibration 5 Another approach to esti- gained widespread use with the mating uncertainty, illustrative introduction of computable gen- 0 rather than statistical, is to exam- eral equilibrium models, of which ine model revisions. For this pur- -5 DICE and RICE are examples, pose, I looked at revisions of the and more recently with real busi- -10 DICE model over its quarter- ness cycle models. Outside eco- century history. The study found nomics, important examples of -15 that the major revisions have calibrated models are earth-sys- come primarily from economic -20 tem models such as the familiar 0 2 4 6 8 10 aspects of the model, whereas the climate models, transportation Temperature increase (°C) revisions to environmental mod- Area of the circles represents weights assigned to each study models, and engineering stud- Source: W.D. Nordhaus and A. Moƒat, NBER Working Paper No. 23646 ules have been much smaller. ies. Generally, while estimation is Particularly sharp revisions have desirable, calibration is necessary Figure 4 occurred for global output, dam- when the model reaches a certain complexity scenarios. This has proven the most difficult ages, and the social cost of carbon. These or when there are no relevant data to use for part of IAMs. In aggregate models, damage results indicate that the economic projec- estimation. Both of these are true for IAMs is often represented as a quadratic function tions are the least precise parts of IAMs of future climate change. of global mean temperature. Andrew Moffat and deserve much greater study than has As an example, one of the most con- and I recently did a systematic review of been the case up to now, especially careful troversial aspects of IAMs is the damage damage estimates; Figure 4 summarizes the studies of economic growth prospects to function, which relates climate change to findings of different studies.12 One impor- 2100 and beyond.13

NBER Reporter • No. 3, September 2017 19 DICE_Manual_100413r1. Uncertainty Analysis in Baseline Scenario pdf

DICE 50th Standard Interquartile Coef. of Models available at Variable (All except SCC refer to the year 2100) Mean best guess percentile deviation range variation www.econ.yale. edu/~nordhaus/ Social cost of carbon, 2015 (2010 U.S. dollars) 34.5 30.0 25.3 32.5 39.3 0.94 homepage/ Temperature increase from 1990 (°C) 4.22 4.10 4.08 1.12 2.01 0.22 DICEmodels09302016.

Carbon concentrations (ppm) 969.6 826.6 829.8 413.3 488 0.12 htm. Return to text World output (trillions 2010 U.S. dollars) 1,433 759 766 1,660 1,056 0.67 8 W. D. Nordhaus, CO emissions (Gt / year) 109.4 70.9 71.1 105.2 114.5 0.63 “Projections and Damages (percent of output) 4.3% 3.8% 3.2% 3.7% 4.9% 0.67 Uncertainties about Climate Change in an Era of Real interest rate (percent / year) 3.6% 3.6% 3.6% 1.6% 2.2% 0.33 Minimal Climate Policies,” NBER Working Paper No. Source: W.D. Nordhaus, NBER Working Paper No. 22933 22933, December 2016, revised Table 2 September 2017. Conclusion 2 Report of the Modeling Resource Group, Return to text Synthesis Panel of the Committee on Nuclear 9 The Economics of Climate Change: IAMs are important tools for under- and Alternative Energy Systems, National The Stern Review, Cambridge, England: standing the implications and policy aspects Research Council, Energy Modeling for an Cambridge University Press, 2007; W. of climate change. They have fundamentally Uncertain Future, Washington, D.C.: The D. Nordhaus, “The ‘Stern Review’ on the transformed the way economists and envi- National Academies Press, 1978. Economics of Climate Change,” NBER ronmentalists approach climate policy, shift- Return to text Working Paper No. 12741, December ing from a pure engineering approach — “do 3 P. A. Samuelson, “Market Mechanisms 2006; and “A Review of the Stern Review this and don’t do that” — to approaches like and Maximization,” Santa Monica, CA: on the Economics of Climate Change,” cap-and-trade or carbon taxes that empha- Rand Corporation, P-69, March 28, 1949. Journal of Economic Literature, 45, size market mechanisms. Return to text September 2007, pp. 686–702. As hurricanes Harvey and Irma remind 4 W. D. Nordhaus, “Can We Control Return to text us, the impacts of weather events can be Carbon Dioxide?” International Institute 10 W. D. Nordhaus, “Revisiting the Social extremely large. So climate change is likely for Applied Systems Analysis Working Paper, Cost of Carbon,” Proceedings of the National to continue growing as an economic prob- WP 75-63, Laxenburg, Austria, June 1975; Academy of Sciences, 114(7), 2017, pp. lem. Improving integrated assessment mod- W. D. Nordhaus, “Economic Growth and 1518–23. els is therefore an important research area for Climate: The Case of Carbon Dioxide,” Return to text economists — full of puzzles, challenges, and American Economic Review, 67(1), 1977, 11 W. D. Nordhaus, “The Economics of policy applications. pp. 341–6. Hurricanes in the United States,” NBER Return to text Working Paper No. 12813, December 2006, 1 This research report draws upon W. 5 A. S. Manne, “ETA: A Model for Energy and Climate Change Economics, 2010, 1 D. Nordhaus, The Climate Casino: Technology Assessment,” Bell Journal of (1), pp. 1–20. Risk, Uncertainty, and Economics for a Economics, 7(2), 1976, pp. 379-406. Return to text Warming World, New Haven, CT: Yale Return to text 12 W. D. Nordhaus and A. Moffat, “A University Press, 2013; W. D. Nordhaus, 6 W. D. Nordhaus, “An Optimal Transition Survey of Global Impacts of Climate “Integrated Economic and Climate Path for Controlling Greenhouse Gases,” Change: Replication, Survey Methods, Modeling,” in Handbook of Computable Science, 258, 1992, pp. 1315–9; W. D. and a Statistical Analysis,” NBER General Equilibrium Modeling, P. Dixon Nordhaus, Managing the Global Working Paper No. 23646, August 2017. and D. Jorgenson, eds.: Elsevier, 2013, pp. Commons: The Economics of Climate Return to text 1069–131; W. D. Nordhaus, “Projections Change, Cambridge, MA: MIT Press, 1994. 13 W. D. Nordhaus, “Evolution of and Uncertainties About Climate Change in Return to text Assessments of the Economics of Global an Era of Minimal Climate Policies,” NBER 7 W. D. Nordhaus, P. Sztorc, “DICE Warming: Changes in the DICE Model, Working Paper No. 22933, December 2016, 2013R: Introduction and User’s Manual,” 1992–2017,” NBER Working Paper No. revised September 2017. October 2013. http://www.econ.yale. 23319, April 2017. Return to text edu/%7Enordhaus/homepage/documents/ Return to text

20 NBER Reporter • No. 3, September 2017