Content uploaded by Peter Howard
Author content
All content in this area was uploaded by Peter Howard on Apr 07, 2021
Content may be subject to copyright.
NEW YORK UNIVERSITY SCHOOL OF LAW
March 2021
Peter Howard
Derek Sylvan
Gauging Economic
Consensus on
Climate Change
Copyright © 2021 by the Institute for Policy Integrity.
All rights reserved.
Institute for Policy Integrity
New York University School of Law
Wilf Hall, 139 MacDougal Street
New York, New York 10012
Peter Howard, Ph.D. is the Economics Director at the Institute for Policy Integrity at New York University School of Law,
where Derek Sylvan, MPA, is the Strategy Director. e authors would like to thank Chris Beltrone for valuable research
contributions, and Richard Revesz for useful feedback.
is report does not necessarily reect the views of NYU School of Law, if any.
i
Executive Summary
Thousands of economists have spent years or decades studying the interaction between climate change and the
economic systems that underlie modern life. e views of these experts can help clarify how climate change
will likely aect our society and economy, and how policymakers should approach greenhouse gas emission
reduction eorts.
We conducted a large-sample global survey on climate economics, which we sent to all economists who have published
climate-related research in the eld’s highest-ranked academic journals; 738 responded. To our knowledge, this is the
largest-ever expert survey on the economics of climate change. e results show an overwhelming consensus that
the costs of inaction on climate change are higher than the costs of action, and that immediate, aggressive emissions
reductions are economically desirable.
Respondents expressed striking levels of concern about climate impacts; estimated major climate-related GDP losses
and a reduction in long-term economic growth; and predicted that climate impacts will exacerbate economic inequality
both between countries and within most countries. e economists surveyed also expressed optimism about the viability
and aordability of many zero-emissions technologies. And they widely agreed that aggressive targets to reach net-zero
emissions by mid-century were likely to be cost-benet justied.
Survey Details
is project expands on similar surveys conducted by the Institute for Policy Integrity in 2015 and 2009, but uses a
larger and more geographically diverse sample. Expert-elicitation projects like this one have recently played an inuential
role in climate economics, helping establish consensus on such topics as the appropriate “discount rate” to use when
evaluating climate policies, and the expected magnitude of climate damages.
We invited 2,169 Ph.D. economists to take a 15-question online survey focused on climate change risks, economic
damage estimates, and emissions abatement. Of this pool, 738 participated, a response rate of 34% (not all respondents
submied a response to every survey question, so the sample for some questions is smaller). ese economists have
all published an article related to climate change in a leading economics, environmental economics, or development
economics journal, and their areas of expertise cover a wide range of issues in climate economics. e survey design and
related analysis sought to minimize selection bias, response bias, and anchoring bias.
Growing Concern About Climate Change
When asked about their professional opinions on climate change, an overwhelming majority of respondents (74%)
said that “immediate and drastic action is necessary.” In sharp contrast, less than 1% believe that climate change is “not
a serious problem.” Compared to our 2015 survey, a signicantly larger share of respondents now believe that drastic
action is needed, while far fewer believe that more research is needed before action is taken.
Nearly 80% of respondents also self-report an increase in their level of concern about climate change over the past ve
years, underscoring the high level of overall concern among this group. i s broad majority suggests that even respondents
ii
who have characterized the situation as urgent in the past may feel that the nature of the climate change challenge is
rapidly escalating. When asked to identify items that most aected their views on climate change in recent years, the
most common answer by a signicant margin was “observed extreme weather events aributed to climate change.” e
next most inuential factors identied were new research ndings, both in climate science and in climate economics and
the social sciences.
A Threat to Economic Growth
Economists have traditionally modeled climate damages by focusing on changes to GDP in a specic year (i.e., a level
impact), but some research has suggested evidence of reduced economic growth as a result of current climate impacts. In
total, 76% of survey respondents think it is likely or very likely that climate change will negatively aect global economic
growth rates. Maybe more notable is the dearth of respondents who nd this prospect unlikely (3%) or extremely
unlikely (2%).
Increasing Inequality Between Countries and Within Countries
e vast majority (89%) of respondents believe that climate change will exacerbate income inequality between high-
income and low-income countries (the upper third of countries by per-capita income versus the lower third). is could
create enormous diculties for many countries that already face profound economic challenges and high rates of poverty.
Approximately 70% of economists also believe it is likely or extremely likely that climate change will exacerbate inequality
within most countries (between the lower third of households by household income and the upper third).
The Promise of Zero-Emissions Technologies
Over the last decade, the costs of solar and wind energy technologies have dropped rapidly (-7% annually for solar
photovoltaic systems and -4% annually for onshore wind). When asked whether a similar paern is likely to be replicable
for some other emerging zero-emission and negative-emission technologies, 65% of respondents said this is likely or very
likely, while less than 3% disagreed.
Economists predict rapid expansion of clean energy technologies, estimating that more than 50% of the global energy
mix will consist of zero-emission technologies by 2050—the current share is roughly 10%. ey are also bullish about
negative-emissions technologies eventually becoming viable, with a majority predicting this during the second half of
this century (though a very high percentage of “No Opinion” responses underscores the uncertainty of this projection).
Climate Damages Will Be Very Costly
Respondents were asked to estimate the economic impacts of several dierent climate scenarios. ey project that
economic damages from climate change will reach $1.7 trillion per year by 2025, and roughly $30 trillion per year (5% of
projected GDP) by 2075 if the current warming trend continues. eir damage estimates rise precipitously as warming
intensies, topping $140 trillion annually at a 5°C increase and $730 trillion at a 7°C increase. As expected, experts
believe that the risk of extremely high/catastrophic damages signicantly increases at these high temperatures.
iii
Consistent with the view that society can beer adapt to climate change if the rate of warming is slower or if society is
wealthier, the economists project somewhat lower climate damages in scenarios with slower rates of warming or higher
global income. However, even these damage estimates are high, with a loss of at least 4% of GDP expected in each future
climate scenario presented (and a 6% GDP loss expected in a scenario with faster warming).
To provide context, another survey question asked respondents to estimate the GDP change in 2020 (this information
was not available at the time of the survey, in early February 2021). Respondents’ GDP loss estimates for 2020, when a
pandemic devastated the global economy, are far smaller than their estimates for annual damages from climate change
under a 3°C warming scenario (-3% of GDP vs. -5 %). And unlike the pandemic-related downturn, the climate impacts
projected by survey respondents would recur annually (on average) and worsen over time.
The Costs of Inaction Outweigh the Costs of Reducing Emissions
e survey ndings reveal a clear consensus that ambitious emissions reductions are likely to cost less than the expected
damages from climate change. Respondents overwhelmingly agree that the benets of reaching net-zero emissions
by 2050 would likely outweigh the costs—66% view this as likely or very likely, compared to only 12% who disagree.
Respondents’ abatement cost projections are higher than estimates from some other sources (roughly 3-4% of GDP in
some scenarios). Yet they still clearly indicate that aggressive emissions reduction eorts in line with the Paris Agreement
targets are economically justied, as projected economic damages from climate change are far higher.
Costs are oen cited as a reason to delay or avoid strong action on climate change, but this survey of hundreds of expert
economists suggests that the weight of evidence is on the side of rapid action.
ese results can be useful to both policymakers and economic researchers. In particular, economic modelers who
calibrate “Integrated Assessment Models” (which calculate the Social Cost of Carbon for use in policy analysis) could
use these ndings to help ensure that key model assumptions align with the consensus views of experts. e U.S.
government is currently reviewing the methodology used for this modeling, and these survey ndings could inform
improved calibration of several model parameters.
is survey reveals a clear consensus that immediate and meaningful eorts to reduce emissions are needed to limit the
enormous economic risks of climate change. Policymakers should heed these ndings.
iv
Table of Contents
Why Survey Economists? 1
The Value of Expert Elicitation 2
Survey Methodology 5
Survey Design 5
Respondent Criteria 6
Survey Administration and Response Rate 7
Addressing Possible Biases 7
Survey Results and Analysis 9
Respondents’ Expertise 9
Professional Opinions on Climate Change 11
Climate Change and Economic Growth 15
Distributional Impacts 17
Emissions Abatement 19
Climate Damage and Emissions Abatement Estimates 22
Net-Zero Emission Targets 27
Context from the Pandemic 29
Implications for Economic Modeling 30
Conclusions 32
Appendix A. Survey Questions 33
Appendix B. List of Journals Used to Assemble Survey Respondent Pool 40
Appendix C. Response Data by Survey Question 42
Appendix D. Additional Forecast Analysis 43
References 51
1
Why Survey Economists?
Economists whose research focuses on climate change are uniquely qualied to provide insights on the economic
risks of climate damages and the appropriate policy responses.1 ousands of economists have spent years or
decades developing expertise on the interaction between climate change and the economic systems that underlie
modern life. Since the 1990s, an entire sub-eld of economics has developed to research these issues, with thousands of
articles published in peer-reviewed academic journals. Experts in this area are well suited to provide context on such
issues as:
• e speed, severity, and regional distribution of climate change’s potential economic impacts;
• e dynamics and costs of reducing emissions in key economic sectors;
• e ability of populations to adapt to the impacts of climate change;
• e nature of low-probability climate risks with potentially catastrophic consequences;
• e costs and benets to both current and future generations of climate policies;
• e dynamics of international cooperation related to climate change.
e consensus views of economists with expertise on climate change can provide valuable insights for policymakers
who must weigh the benets and costs of various climate strategies. Expert-elicitation research has been cited in some
important past policy decisions.2 Additionally, the views of experts can also help advance economic research, including
the modeling used to evaluate climate policies.
In order to gauge the views of economists with expertise on climate change, we conducted a large-sample survey of
individuals who have published articles about climate change in highly ranked economics journals since 1994. To our
knowledge, this is the largest-ever survey of economists’ views on climate change. e results of this survey help clarify
consensus on some key economic issues related to climate change.
is project expands on a similar survey we conducted in 2015 as well as a 2009 survey conducted by other researchers at
the Institute for Policy Integrity (Holladay et al., 2009; Howard & Sylvan, 2015; Howard & Sylvan, 2020).ose surveys
revealed widespread consensus that climate change poses dramatic economic risks, that immediate actions to reduce
emissions are economically justied regardless of whether other countries have agreed to act, and that the discount rate
currently used to evaluate climate policies is inappropriately high, among other ndings.
is 2021 survey samples a larger pool of experts than the past surveys. e pool was expanded because many new
articles on climate change have been published since 2015, and because we chose to include authors who published in
highly ranked development economics journals as well as mainstream economics journals and environmental economics
journals. We included development journals in order to help diversify the sample, as these journals generally feature
publications that originate from and focus on a broader range of geographic regions (in part because they are less skewed
toward scholars from rich countries).
2
Beyond the direct policy implications of these survey ndings, the results provide a useful resource for modelers who
estimate climate damages. Economists have developed integrated assessment models (IAMs), which capture various
steps in the climate and economic processes that translate an additional ton of carbon dioxide emissions into economic
damages. ese models are used to estimate the “Social Cost of Carbon” (SCC)—the marginal damage of a ton of
carbon dioxide emissions. e SCC is an essential metric in U.S. government cost-benet analyses of actions that aect
greenhouse gas emissions. However, IAMs and the results derived from them, including the SCC, are highly sensitive
to modelers’ assumptions, which do not necessarily reect the consensus views of experts. Research based on our 2015
survey shows that when an IAM is recalibrated to use the discount rate and damage function preferred by respondents to
an expert survey, the SCC value increases more than tenfold (Howard & Sylvan, 2020).
is survey and other expert elicitations can help establish the appropriate assumptions to be used in IAMs, and could
play a role in the U.S. government’s current review of this modeling methodology. e Biden administration is presently
working to update the models underlying the U.S. federal government’s SCC value (IWG, 2021). is process will rely
in part on recommendations from the National Academies of Sciences, Engineering, and Medicine, including a call to
use expert elicitation in the development of several IAM components (NAS, 2017). e results from this survey and
others like it can be used to help calibrate IAM parameters including climate damage functions; adaptation assumptions;
emissions scenarios; technology availability assumptions; abatement cost functions; and discount rates. Calibrating
these parameters to match the consensus views of experts (as revealed through expert elicitations like this survey) will
likely lead to a more comprehensive account of likely climate impacts. Based on related recalibration eorts, the result
would likely be a signicant increase in the SCC value (Howard & Sylvan, 2020).
e Value of Expert Elicitation
Eliciting the views of experts in a eld can improve understanding of complex topics and highlight prevalent points
of view that might not otherwise stand out. Clarifying these consensus views can help inform the public and provide
insights for policymakers. Expert elicitation is di stinct from public opinion polling, which i s useful for gauging widespread
sentiments and political views.
Policymakers and researchers regularly use expert elicitation to improve understanding of climate change-related topics.
In an eort to determine consensus on climate issues, the United Nations established the Intergovernmental Panel on
Climate Change (IPCC) and tasked it with providing a consensus-based, scientic view on the current understanding of
climate change and related consequences. rough the IPCC’s deliberative review process, thousands of climate experts
from across the globe assess the most recent scientic, technical, and socio-economic information, and then synthesize
their ndings.
e IPCC reviews the research of economists and solicits their expertise to help develop the consensus viewpoint. In
particular, economists participate in the Working Group on “Impacts, Adaptation, and Vulnerability,” which has explored
the consensus view on the SCC and other topics.
However, there are drawbacks to the deliberative process used by the IPCC (and others) to identify consensus. Group
deliberations can lead to “groupthink,” sometimes causing deliberation processes to suer f rom censorship and uniformity
(Sunstein, 2005). Indeed, the IPCC has been criticized for moving too slowly and adopting only the “lowest-common
denominator” conclusions, leading to overly conservative results that ignore more up-to-date viewpoints (McKibben,
2007). In fact, actual measures of sea-level rise have tracked the high end of the IPCC’s projections, and the IPCC’s past
3
temperature predictions were shown to be low (Rahmstorf et al., 2007; Rahmstorf et al., 2012). In other words, the
IPCC has tended to underestimate the rate of climate change, and the results of its deliberative process perhaps only
indicate the minimal consensus in the scientic community—the least we can expect (Oreskes et al., 2019).
Besides deliberation, an alternate method for identifying the consensus opinion of experts is to use surveys and nd a
group’s median or mean answer. Well-developed theories on “the wisdom of crowds” explain why the average answer
from a group is likely to be more accurate than the answers of most individuals in that group, and why large groups
perform beer than small groups.3 For example, groups of experts have been shown to signicantly outperform individual
experts on predicting such uncertain (and climate change-related) quantities as the annual peak rainfall runo of various
countries or changes in the U.S. economy (Armstrong, 2001). By comparison, deliberating groups tend only to do about
as well as their average members on making accurate predictions, and not as well as their best members (Gigone &
Hastie, 1997).
Compared to deliberation, surveys and statistics can oen produce a more nuanced understanding of expert consensus,
and help reveal the full range of opinions in a group. Deliberation tends to reduce variance, since deliberations can
amplify cognitive errors and overemphasize common knowledge, causing a group to converge on a common—though
not necessarily accurate—answer. By showing the diversity of opinion, surveys can indicate where debate still exists on
an issue and where a consensus might emerge in the future. Surveys can also measure the level of uncertainty on a topic,
which can be especially important for policymakers who are risk-averse or who seek to maximize future policy exibility.
Past Surveys on Climate Economics
Researchers focusing on climate economics have shown a renewed interest in expert elicitation in recent years. Until
2015, the most prominent surveys on climate damages (Nordhaus, 1994; Schauer, 1995) and discount rates (Weitzman,
2001) were decades old. Many existing estimates suered from shortcomings including small sample size (and related
selection bias); reduced variance due to uniformity or censorship (from using deliberation and consensus building);
and/or respondent bias (from using informal, open web surveys). Beginning in 2015, some scholars began to call for new
expert elicitation eorts, and a number of researchers soon undertook such projects.
Economist Robert Pindyck argued in a 2015 working paper (later published in 2017) that IAMs are over-reliant on the
assumptions of the modeler,4 such that these climate-economic models represent the modeler’s informed opinion (on
climate science, economics, and policy) rather than the scientic consensus. Pindyck proposed using expert opinion from
“a range of economists and climate scientists” to calibrate these models, rather than relying on modelers to independently
set parameter values. In 2016, Oppenheimer et al. (2016) also called for the use of formal expert elicitation in the climate
change context due to the inevitable need for expert judgment in long-run numerical models.
Around this time, several expert elicitations were conducted on climate damages and discount rates, aiming to overcome
past survey shortcomings. From May 2014 to April 2015, Drupp et al. (2018) conducted an expert elicitation on social
discount rates and the related parameters. Critically, the authors found strong support for a median discount rate of 2%,
with a strong consensus for a range of 1% to 3%.
Second, building on a 2009 Policy Integrity survey, Howard and Sylvan (2015; 2020) conducted an expert elicitation on
climate economics and policy. e sur vey revealed high levels of concern about climate change and suppor t for immediate
action using market-based policies. It also showed a strong disparity between the general views of experts publishing on
4
climate economics and the output of IAMs, as theorized by Pindyck. Specically, the survey found substantially higher
estimated climate damages (mean and median estimates of -9.2% and -5% of global GDP, respectively, for a 3°C increase)
and strong support for median discount rate of 2%, consistent with Drupp et al. (2018).
Finally, Pindyck (2019) conducted a survey on climate damages, emissions, and discount rates. His respondents also
provided relatively high climate damage estimates (slightly higher than Howard and Sylvan (2020)), and a low discount
rate.
Taken together, these three studies provided strong evidence that the current expert consensus on key economic
parameters strongly diered from past survey ndings and current IAM assumptions, implying support for lower inter-
generational discount rates and higher climate damages.
In 2016 and 2017 the National Academies of Sciences, Engineering, and Medicine (NAS) published two reports on
the SCC. Critically, NAS (2017) called for the use of expert elicitation in the development of key components of IAMs,
including socio-economic and emission scenarios, though the report cautions against its use for estimating climate
damages. Resources for the Future has begun conducting some of the updates laid out in NAS (2017), including the use
of expert elicitation to develop long-run socio-economic and emission scenarios.5
is 2021 expert elicitation survey builds on these recent eorts and introduces questions on long-run climate damages,
distributional impacts, adaptation, and abatement technology and costs.
5
Survey Methodology
In an aempt to gauge expert consensus on key economic issues related to climate change, we surveyed 2,169 of the
world’s leading experts on climate economics. We sent each respondent a link to a 15-question online survey with
questions focused on climate change risks, economic damage estimates, and emissions abatement. Of this pool, 738
economists participated, for a response rate of 34%. (Not all respondents submied a response to every survey question,
so the sample for some questions is smaller; see Appendix C.)
Survey Design
Our survey was designed to accomplish several objectives. We sought to determine the extent of expert consensus on
critical economic questions related to climate change; understand how experts’ views of climate change have evolved over
time (in part by comparing these ndings with the results of past surveys); and solicit specic estimates of the economic
impacts of climate change and the likely costs of emissions abatement. We surveyed respondents on the following topics:
• eir specic areas of expertise and the subjects on which they have published;
• eir professional views on climate change, including the level of risk that climate change poses to the economy;
the general policy responses that are most appropriate; and how and why their views have shied over the past
ve years;
• e distributional impacts of climate damages, including the eect on inequality both between countries and
within countries;
• Estimated trends and costs for emissions abatement, including projections for some low-emissions technologies,
and the overall costs for various levels of abatement;
• Estimates for the economic impacts of various climate scenarios, and expected levels of adaptation;
• e expected benets/costs of reaching net-zero emissions targets by mid-century;
• Estimates for changes in GDP and emissions during the Covid-19 pandemic; for purposes of comparison.
Because we sought to compare our respondents’ views to the opinions expressed in other surveys, some of our questions
mirrored those from a 2015 Institute for Policy Integrity survey. e full text of our survey is included as Appendix A.
Before distributing the survey, we conducted a series of internal and external tests to help ensure that the questions were
unambiguous, and we made several changes to improve question clarity.
6
Discount Rates and Expert Consensus
We chose not to include a question about the social discount rate in this survey, given that recent studies
(including our previous survey of a similar sample) have already established a clear consensus that a rate of 2%
or less is appropriate in the climate change context.
ere are two views on the best way to determine discount rates: observing market interest rates (the descriptive
approach) and applying ethical/philosophical arguments (the normative approach). Historically, the descriptive
approach has suggested higher rates than the normative approach, particularly in the inter-generational context.
However, this is no longer true due to developments in both approaches.
In the descriptive context, recent research indicates that consumption discount rates are appropriate in the
climate change context, while capital discount rates, which are higher, are not (IWG, 2010; NAS, 2017; Li &
Pizer, 2021). Recent research also demonstrates that the consumption discount rate has declined over the last
several decades to a rate below 2% (U.S. Council of Economic Advisers, 2017; Bauer & Rudebusch, 2020).
In the normative context, recent research by Heal and Millner (2014) demonstrated that a voting procedure
is an ecient, time-consistent way to select a social discount rate when a wide range of views are held over the
appropriate rate. As mentioned above, two recent surveys that elicited these views found median rates of 2%
(Drupp et al., 2018; Howard & Sylvan, 2020). Given this consensus, we decided to focus our survey questions
on other critical climate-economic issues.
Respondent Criteria
We sought to identify a pool of respondents with demonstrated expertise in the economics of climate change. Building
on the approach used in prior surveys by the Institute for Policy Integrity, we compiled a list of all Ph.D. economists who
had published an article related to climate change in a leading economics, environmental economics, or development
economics journal since 1994.6 We included all papers that discussed climate change and had implications for the climate
change debate, even if that was not their primary focus.
We dened leading journals as those ranked in the top 25 economics journals, top seven environmental economics
journals, and top seven development economics journals, according to rankings published in peer-reviewed publications.
Given that the rankings of various journals have changed during this time frame, we used rankings from multiple time
periods and included any journal that met the criteria in any time period. In total, our nal list included 45 economic
journals.7 e list of journals is available in Appendix B.
We conducted a thorough search of each journal for articles that mentioned “climate change” or “global warming” and
signicantly discussed related issues. e articles published by the economists in our sample tended to have an academic
focus on issues related to climate change; they were not political pieces, and most cannot be easily classied as advocating
either for or against climate change policies.
7
Aer removing experts who had died and individuals for whom we could not locate a working email address, our review
revealed 2,169 authors who t our selection criteria.8 From this group, 738 economists participated in the survey, for a
response rate of 34%.
is sample is signicantly larger than our survey from 2015, for which we invited 1,103 experts and had 365 participants
(a response rate of 33.1%).9 e larger sample stems from the publication of hundreds of new relevant articles since
2015, as well as our inclusion of development economics journals and two new environmental and resource economics
journals, which were not included in the prior survey.10
Survey Administration and Response Rate
We conducted our survey online using Qualtrics soware, with the survey open from February 1, 2021 through February
11, 2021. We sent each expert an email message that described the nature of this project, informed them of the reason for
their selection, and requested their participation through an embedded hyperlink to the survey.
e survey included 10 multiple-choice questions and ve quantitative forecasts. Respondents were told that the survey
would take roughly 15 minutes to complete, and that individual responses would be anonymous (the survey did not ask
for any identifying information or track individual responses). Respondents were sent two reminder emails that included
deadline details. ese emails were sent to the entire pool since we could not determine who had already completed the
survey.
Our overall response rate was 34%, which is consistent with the average response rate for online surveys of this ty pe.11 Not
all respondents answered every question. Unsurprisingly, fewer people answered the more complex forecast questions
that asked for multiple quantitative estimates of conditions under various climate scenarios. ese questions had samples
ranging from 212 to 342 responses (see Appendix C).
Addressing Possible Biases
Our methodology for choosing respondents could potentially suer from selection bias, given that highly ranked
academic journals might not publish articles encompassing the entire spectrum of thought on climate change economics.
However, we believe our approach adequately identied a large sample with demonstrated expertise in the economics of
climate change. Furthermore, our respondents were representative of a wide range of opinions, based on the diverse and
oen conicting arguments made in their published articles.12
Response bias could be a concern for our open-ended questions on damage and abatement forecasts, which were
answered by a smaller number of respondents. However, we analyzed response rates for dierent respondent subgroups
and found that the rates were generally balanced and our results are robust across groups.13 Using stratication weights
to adjust for dierent response rates by subgroups in order to ensure our sample is representative of our population,
we found the results to be relatively similar. We also tested whether open-ended responses diered based on timing of
responses relative to our reminder e-mails (this assumes late responses are a phenomenon related to non-responses); we
do not nd evidence of responses diering by timing.14 Finally, given our large sample size, even a smaller response rate
of around 10% for some questions encompasses over 200 economists’ views.
8
Finally, we designed our survey to minimize anchoring bias (whereby respondents rely disproportionately on information
provided recently). In the fourth question of the survey we presented respondents with a list of possible inuences on
their views about climate change. We included a wide range of examples, some of which related to emissions increases
and climate damages while others related to emissions reductions and climate policy development. ese options were
intended in part to encourage respondents to consider a variety of factors in their later responses. Additionally, we asked
respondents to answer forecast questions by providing their 5th and 95th percentile estimates before providing their 50th
percentile estimates, in order to avoid anchoring on these responses. Lastly, just before asking respondents whether mid-
century net-zero-emissions goals are cost-benet justied, we asked them to complete forecasts of both the costs and
benets of emissions abatement.
9
Survey Results and Analysis
Our survey results reveal several areas of consensus with implications for climate change policy. e survey was
divided into ten thematic sections. e ndings from questions in each section are discussed below.
roughout this report, Figure numbers correspond to the question numbers from the survey. e full survey text is
available in Appendix A.
Respondents’ Expertise
In the rst survey question, we asked respondents to indicate their area(s) of expertise and topics on which they had
published, from the following list:
• Estimated damages from climate change
• Climate change uncertainty and risks, including tipping points and fat tails
• Climate change adaptation and system resilience
• Greenhouse gas emissions abatement / mitigation
• Climate scenario modeling or cost-minimization modeling
• Social Cost of Carbon or optimal climate policy modeling
• Global climate strategies / agreements / policies
• Climate change in developing countries / Geographic distribution of climate impacts
• Other climate-related topics
e respondent group reported a diverse mix of expertise, with at least 139 respondents (19% of the full group) identif ying
expertise within each of the listed topic areas. In total, 733 of the 738 respondents answered this question.
10
Figure 1
Respondents’ Areas of Expertise/Publication Topics
Respondents’ Areas of Expertise/Publication Topics
0 5 10 15 20 25 30 35 40 45
% of Respondents
Other climate-related topics
Climate change in developing countries / Geographic distribution of climate impacts
Global climate strategies / agreements / policies
Social Cost of Carbon or optimal climate policy modeling
Climate scenario modeling or cost-minimization modeling
Greenhouse gas emissions abatement / mitigation
Climate change adaptation and system resilience
Climate change uncertainty and risks, including tipping points and fat tails
Estimated damages from climate change
219 Respondents
191 Respondents
243 Respondents
210 Respondents
149 Respondents
332 Respondents
212 Respondents
139 Respondents
197 Respondents 733 total
respondents
Many indicated
multiple topics
11
Professional Opinions on Climate Change
Our questions in this category aempted to gauge general views on climate change, based on respondents’ research. e
rst question was also included in our 2015 survey; it uses wording from an MIT/Harvard public opinion survey that
was conducted on several occasions beginning in 2003 (Ansolabehere & Konisky, 2014).
Figure 2
Which of the following best describes your views about climate change?
A clear consensus is evident on this question, as an overwhelming majority of respondents (nearly three quarters) believe
that “immediate and drastic action is necessary” to address climate change. In sharp contrast, less than 1% (ve total
respondents) believe that climate change is “not a serious problem.”
24%
74%
0 20 40 60 80
Which of the following best describes your views about climate change?
50%
43%
2%
1%
1%
1%
5%
Immediate and drastic action is necessary
Some action should be taken now
More research is needed before action is taken
This is not a serious problem
No response
2021 Survey Responses 2015 Survey Responses
12
Compared to our 2015 survey, a signicantly larger share of respondents now believe that drastic action is needed. Far
fewer respondents believe that more research is needed before action is taken, relative to the 2015 survey. e responses
to the following question appear to conrm this escalating level of concern.
Figure 3
As informed by your research, how has your level of concern
about climate change shifted over the past ve years?
Nearly 80% of respondents self-report an increase in their level of concern about climate change over the past ve years,
underscoring the high level of overall concern among this group. is broad majority suggests that even respondents
who have characterized the situation as urgent in the past may feel that the nature of the climate change challenge is
rapidly escalating.
Respondents whose level of concern remained unchanged over the past ve years still emphasized the need for urgent
action in the prior question. Of the 140 respondents who selected this option, 60% believe that “immediate or drastic
action is necessary” and another 37% believe “some action should be taken now.”
As informed by your research, how has your level of concern
about climate change shifted over the past five years?
0 10 20 30 40 50
No opinion
Strongly decreased
(You view climate change
as a less pressing problem)
Somewhat
decreased
Remained
unchanged
Somewhat
increased
Strongly increased
(You view climate change
as a more pressing problem)
% of Respondents
41%
38%
19%
1%
<.5%
<.5%
13
Figure 4
Which items had the greatest eect on your views
about climate change over the past ve years?
Respondents could select up to three choices
Which items had the greatest effect on your views
about climate change over the past five years?
Respondents could select up to three choices
0 10 20 30 40 50 60
New developments with
negative-emissions technologies
New environmental policies and/or ambitious
policy proposals in other countries
Changes in adaptation technologies and costs
Increased global energy demand
Other(s)
The coronavirus pandemic
Continued expansion of fossil fuel infrastructure
(e.g., North America’s investment in
natural gas; Asia’s investment in coal; etc.)
Insufficient improvements in abatement
technologies and costs
The United States’ withdrawal
from the Paris Agreement
The Paris Agreement
Improvements and/or cost reductions
in low-emissions technologies
The IPCC Special Report on the
impacts of 1.5°C global warming
Increased environmental deregulation within
certain countries (United States, Brazil, etc.)
China’s strengthening of its climate policies
New findings in climate economics
and the social sciences
New findings in climate science
Observed extreme weather events
attributed to climate change 52%
31%
29%
27%
26%
23%
19%
12%
12%
9%
9%
8%
5%
5%
5%
6%
3%
14
In an eort to understand some of the key factors driving respondents’ views on climate change, we asked them to select
up to three items that had the greatest eect on their views over the past ve years. We included a wide range of choices,
roughly split between factors related to increasing emissions/climate impacts and those linked to emissions reductions.
We included this range of choices in part to minimize anchoring eects for later questions.
While a wide spectrum of factors seems to have informed respondents’ views, the most common answer by a signicant
margin was “observed extreme weather events aributed to climate change.”
ese empirical observations of climate impacts appear to have had an outsize role in shaping economists’ views, perhaps
due to the high level of damage caused by recent extreme weather events (such as wildres in Australia and the Western
United States, heatwaves in Europe, and historically large numbers of hurricanes). Such events may also have stood
out to economists because many projections anticipated that the current levels of temperature increase and climate-
linked extreme weather would take longer to manifest than they have (Dienbaugh, 2020). Extreme weather events also
frequently elevate the general public’s level of concern about climate change (Sisco et al., 2017).
e next most inuential factors identied by survey respondents were new research ndings, both in climate science
and in climate economics and the social sciences. Relatedly, nearly a quarter of respondents highlighted the inuence
of the IPCC special report on the impacts of 1.5°C global warming. is report discussed research on the high level of
climate damages projected to occur even if global emissions are rapidly reduced and temperature increases are halted
quickly.
Many respondents also said their views were shaped by policy changes in high-emiing countries, including both
environmental deregulation in the United States and Brazil, and the strengthening of climate policies in China. While
the former examples have exacerbated the climate challenge by allowing for increased emissions, China’s policy changes
had the opposite eect. But it is possible that respondents saw China’s policy change as evidence of the enormous risks
presented by climate change. Additionally, respondents may have concerns about the robustness of China’s climate
policies, though few highlighted Asia’s recent and planned expansion of coal power plants as a major inuence on their
views.
15
Climate Change and Economic Growth
Our next survey question focused on one potential channel through which climate change might aect the global
economy.
Figure 5
What is the likelihood that climate change will have a long-term,
negative impact on the growth rate of the global economy?
0 20 40 60 80
What is the likelihood that climate change will have a long-term,
negative impact on the growth rate of the global economy?
3%
2%
Extremely likely
Likely
Not clear
Unlikely
Extremely unlikely
42%
40%
19%
17%
<1%
I anticipate no climate damages
In the 2021 survey, 3% of respondents (24 total) selected “No Opinion.”
They are removed from the sample shown above.
36%
36%
2021 Survey Responses 2015 Survey Responses
16
Economists have traditionally modeled climate damages by focusing on changes to GDP in a specic year (i.e., a level
impact), rather than changes to the growth rate of the economy. Recent empirical research (such as Burke et al. (2015))
has shown evidence of reduced economic growth as a result of current climate impacts, though some researchers have
questioned the assumptions behind these ndings, while others have found growth rate declines only in lower-income
countries. 15
To clarify the nature of the survey question on growth rates, we included a note under the question text that read:
“is is distinct om an impact on the level of GDP in a given year (i.e., climate damages measured
as % of GDP).”
Our survey ndings show signicant consensus for the idea that climate change will negatively aect global economic
growth. Maybe more notable is the dearth of respondents who nd this prospect unlikely (3%) or extremely unlikely
(2%).
Our 2015 survey, which was conducted before many of the major publications on this topic, found very similar results,
suggesting that these views stem from additional factors beyond the inuence of recent empirical research. Despite several
publications and working papers being released since 2015, the results are nearly identical aer dropping respondents
who selected “no opinion” and those who “anticipate no climate damages” (these responses were unavailable in the
previous survey).
Based on the damage estimates provided in other survey questions (discussed below) it appears that many of the
economists in our survey may believe that climate change could damage the global economy both by reducing economic
growth and by reducing GDP through level impacts in specic years.
17
Distributional Impacts
Two survey questions sought to gauge consensus on the relationship between climate impacts and economic inequality,
at both the international and national levels.
Figure 6
What is the likelihood that climate change will increase economic inequality between
low-income and high-income countries (the lower third of countries by per-capita
income versus the upper third of countries by per-capita income)?
e vast majority of respondents believe that income inequality will be exacerbated between high-income and low-
income countries as a result of climate change. is view aligns closely with the economic literature.16 Specically, as
indicated in the literature, poorer countries are generally expected to be more vulnerable to climate impacts due to
their reliance on agriculture and other outdoor activities, initial hoer temperatures, and smaller budgets available for
adaptation (see Tol, 2018).
If this prediction is true, it would counteract recent trends—average incomes in developing countries have increased at a
faster rate over the past 25 years (United Nations, 2021). e results could be highly problematic for many countries that
already face profound economic challenges and high rates of poverty.
What is the likelihood that climate change will increase economic inequality between
low-income and high-income countries (the lower third of countries by per-capita
income versus the upper third of countries by per-capita income)?
No opinion
Extremely unlikely
I anticipate no climate damages
Unlikely
Not clear
Likely
Extremely likely
% of Respondents
1%
<.5%
0 20 40 60 80
55%
34%
8%
1%
2%
18
Figure 7
What is the likelihood that climate change will increase economic inequality
within most countries, between the lower third of households by household
income and the upper third of households by household income?
Approximately 70% of economists also believe it is likely or extremely likely that climate change will exacerbate inequality
within most countries. Given that every country has a unique combination of economic structures, existing inequality
dynamics, and climate-related risk, this nding suggests that many economists expect to see broadly regressive paerns
in climate impacts and/or adaptation measures. e views expressed in our survey are consistent with recent economic
ndings for the United States (Hsiang et al., 2017).
Our ndings regarding inequality both between and within countries have signicant implications, especially given recent
calls to address environmental justice and climate justice concerns. Policymakers may need to devote additional focus
to remedies for inequality. ese anticipated eects could also increase the cost of climate change, including the SCC,
if damage estimates account for inequality aversion via equity weights (Antho et al., 2009). e expansion of national,
regional, and global inequality also raises serious questions about the use of GDP per capita (i.e., a mean estimate where
total GDP is divided by total population) as a proxy for welfare in IAMs.
What is the likelihood that climate change will increase economic inequality within
most countries, between the lower third of households by household income
and the upper third of households by household income?
No opinion
Extremely unlikely
I anticipate no climate damages
Unlikely
Not clear
Likely
Extremely likely
% of Respondents
4%
<.5%
0 20 40 60 80
28%
42%
22%
1%
3%
19
Emissions Abatement
ree of our survey questions focused on predictions related to major technologies that can reduce emissions.
Figure 8
Over the last decade, the costs of solar and wind energy technologies have
dropped rapidly (-7% annually for solar PV and -4% annually for onshore wind).
Do you think a similar pattern is likely to be replicable for some other
emerging zero-emission and negative-emission technologies?
Costs for some low-emissions energy technologies have dropped substantially in recent years (IRENA, 2019; Lazard,
2019). In many cases, these technologies have become cost-competitive with dominant fossil-fuel energy sources
more rapidly than expected (IEA, 2020). Economists have highlighted several factors that have contributed to these
declining costs, including the eects of learning, the increased maturity of supply chains and production processes, and
an increasingly favorable policy landscape (Kavlak et al., 2018).
In total, 65% of respondents believe that these paerns are likely or very likely to be replicable for some other emerging
clean technologies. We chose to phrase this question broadly to focus on general trends rather than specic technologies
or timeframes. It is noteworthy that even when asked to make a broad prediction about a diverse category of technology,
very few respondents (less than 3%) seem to view the recent reductions in solar and wind costs as anomalous.
Over the last decade, the costs of solar and wind energy technologies have
dropped rapidly (-7% annually for solar PV and -4% annually for onshore wind).
Do you think a similar pattern is likely to be replicable for some other
emerging zero-emission and negative-emission technologies?
No opinion
Extremely unlikely
Unlikely
Not clear
Likely
Extremely likely
% of Respondents
2%
0 20 40 60 80
17%
48%
25%
<.5%
7%
20
Compared to our results, DICE (one of the most prominent IAMs, which is used to calculate the U.S. government’s Social
Cost of Carbon) appears to be relatively conservative on the availability of abatement technologies. DICE-2016R2 does
not explicitly model the cost of zero-emission or negative-emission technologies. However, it does show the price of
the backstop technology (i.e., the price of the marginal technology when 100% of emissions are abated), which declines
by approximately 0.5% annually (Nordhaus, 2018). As this technology could be a zero-emission or negative-emission
technology (Nordhaus & Sztorc, 2013, p. 13), this rate may be insucient if price declines in clean technologies can be
maintained near the 4% to 7% annual rate observed recently for wind and solar.
Figure 9
In 2050, what share of the global energy mix do you think will consist of
zero-emission technologies (e.g., solar, wind, nuclear, green hydrogen,
bioenergy with carbon capture and storage, etc.)?
For context, zero-emission sources make up roughly 10% of the
current energy mix according to the International Energy Agency.
On average, respondents estimate that just over half of the global energy mix will consist of zero-emission sources in
2050. e responses to this question essentially follow a normal distribution, with a nearly identical mean and median
of slightly over 50%.
In 2050, what share of the global energy mix do you think will consist of
zero-emission technologies (e.g., solar, wind, nuclear, green hydrogen,
bioenergy with carbon capture and storage, etc.)?
For context, zero-emission sources make up roughly 10% of the
current energy mix according to the International Energy Agency.
Mean response: 53.9%
Median response: 50.5%
0
30
60
90
120
150
90-100%80-89%70-79%60-69%50-59%40-49%30-39%20-29%10-19%0-9%
Number of Responses
Zero-emission share of global energy mix
21
Respondents believe that growth in zero-emissions energy will be prolic, as expanding the share of global primary energy
supplied by these sources from roughly 10% to 50% in three decades would represent a monumental shi. However, even
this enormous increase is likely insucient for a 1.5°C or 2°C temperature pathway, according to prominent projections
(which would require net-zero economywide emissions by 2050 or soon aer).17
Figure 10
During what time period do you believe that net-negative greenhouse gas emission
technologies, such as direct air capture and carbon capture/utilization/storage (CCUS),
will become viable and reliable at a low-enough cost to be adopted on a large scale
(i.e., to signicantly change the global emissions path)?
Respondents appear to be relatively bullish on the viability of negative-emissions technologies, given that few large-scale
projects exist today.18 Roughly 25% of respondents chose the “No Opinion” option, suggesting a high level of uncertainty
for this question. Still, 542 experts did oer a prediction, and a sizeable majority expects negative-emissions technologies
to be viable at a large scale by 2060 or 2080 (the median response is between 2040 and 2060). ese ndings are roughly
consistent with the IPCC (Hansel et al., 2020).19
During what time period do you believe that net-negative greenhouse gas emission
technologies, such as direct air capture and carbon capture/utilization/storage (CCUS),
will become viable and reliable at a low-enough cost to be adopted on a large scale
(i.e., to significantly change the global emissions path)?
0
5
10
15
20
25
30
35
% of Respondents
By
2040
By
2060
By
2080
By
2100
By
2150
After
2150
Net-negative
GHG
technologies
will never
become
viable
No
opinion
22
is result could suggest condence in the ability of research-and-development eorts and other climate policies to
accelerate these technologies and reduce costs. e level of government support and the climate policy landscape could
ultimately determine the level of success and pace of progress for these technologies.
Most scenarios for meeting a 1.5°C or 2°C limit require extensive use of negative-emissions technologies by soon aer
2050.20 According to our survey, economists seem to believe this timeline is viable. However, the optimism of these
results should be tempered somewhat, as engineers and other categories of experts may have equally (or more) relevant
insights on these issues than economists, and their views may dier. Many researchers who focus on the energy transition
advise a “precautionary approach” with respect to negative-emissions technologies, given that they are unproven and
overreliance on these technologies could deter necessary emissions reductions in the near term (Rogelj et al., 2019).
e timeline suggested by our survey ndings is signicantly faster than the one projected by some major IAMs, including
DICE. DICE does not specify when negative-emissions technologies will become economically viable on a large scale.
However, Nordhaus assumes that net-negative emissions (i.e., abatement of more than 100%) will occur in 2240 on
DICE’s business-as-usual emissions path. is only occurs by 2160 on the optimal emissions path (Nordhaus, 2018).
Taken together, the ndings from our series of survey questions on emissions abatement paint a more positive picture for
aordable, eective emissions reductions than does DICE.
Climate Damage and Emissions Abatement Estimates
We next asked questions soliciting forecasts of expected economic impacts under various climate scenarios. e initial
scenarios approximate a business-as-usual warming trend, carried forward to four future time periods.
e scenarios we provided included the year, level of temperature change since the pre-industrial era, average rate of
temperature change over the prior 30 years, and estimated global GDP (based on scenario assumptions from DICE).
e full scenario details provided in the survey are available in Appendix A. e scenario details for our damage estimate
questions originate from the baseline DICE-2016R2 scenario.
We asked respondents to “Please consider the level of global GDP and rate of temperature change, in addition to overall
temperature change (some researchers theorize that society is more capable of adapting to climate change with slower
rates of temperature change and/or higher levels of global economic output).”
In climate economics, economic damages (measured as a percentage of GDP) are represented as functions of temperature
change, the rate of temperature change, and income. Temperature change is used as a proxy variable for other climate
drivers, including increased storm frequency and sea-level rise. As temperatures increase, the economic literature suggests
that there may be initial benets (an empirical question), though the global economy will eventually suer economic
damages at some level of temperature increase (Howard & Sylvan, 2020; Howard & Sterner, 2017; Tol, 2018).
Since we asked experts to provide their projection of net damages (i.e., damages minus adaptation and its costs), slower
rates of temperature increase (that increase the time it takes to reach a particular temperature level) should decrease
damages, as society has more time to adapt. With respect to income, damages could decrease or increase depending on
the relative eects of increased adaptability and relative prices (i.e., the value of non-market goods increases as market
goods become less scarce). Our adaptation scenarios were designed to provide measurements of these eects.
23
To avoid overcondence and anchoring biases in these questions, we asked respondents to provide their 5th and 95th
percentile estimates before providing their median/50th percentile estimates. We also asked respondents to include both
non-market and market impacts, and factor in adaptation to climate change and its corresponding costs.
Figure 11
Climate Damage Estimates
Year 2025 2075 2130 2220
Temperature increase (relative to pre-industrial era) 1.2°C 3°C 5°C 7°C
Economic damages (% of global GDP) - Median estimate -1% -5% -10% -20%
Economic damages (trillions of 2019 USD) - Median estimate -$1.7 -$29.8 -$143.0 -$730.9
Economic damages (% of global GDP) - Mean estimate -2.2% -8.50% -16.10% -25.20%
Economic damages (trillions of 2019 USD) - Mean estimate -$3.8 -$50.6 -$230.3 -$920.9
Standard deviation 2.9 7.6 13.3 20.7
Results above reect the trimming of outlier estimates below the 5th percentile or above the 95th percentile of total responses.
We believe the median estimate best reects expert consensus due to the skewness of the results, which arises from
outlier responses (see below for more detail). When expert survey responses include signicant outliers or focus on a
belief rather than a testable forecast, the median result can oen be more useful than the mean (Lorenz et al., 2011). Our
analysis of the forecast questions therefore focuses primarily on the median results.
Respondents project that economic damages from climate change will reach $1.7 trillion per year by 2025, and roughly
$30 trillion per year (5% of projected GDP) by 2075 if the current warming trend continues. Damage estimates rise
precipitously as warming intensies, topping $140 trillion annually at a 5°C increase and $730 trillion at a 7°C increase.
ese damage estimates exceed those in DICE and other commonly cited IAMs, though they are consistent with past
surveys (Howard & Sylvan, 2020; Pindyck, 2019). We also asked questions about impacts at higher temperatures and
income levels than some past surveys. Like Nordhaus (1994), we found that climate damages do not appear to follow a
quadratic path in the long run, providing some support for the earlier DICE damage function that limits climate damages
to 100% of GDP.
Trimming and the Median Response
We use a two-pronged approach to ensure that outliers in our data set do not overly inuence the results. First, we
trim the overall results to eliminate responses below the 5th percentile and above the 95th percentile. Second, we focus
primarily on the median rather the mean estimate. We selected this approach in part because the near-term forecasts
that result from this method approximate the range of estimates in the literature, while the inclusion of outlier responses
produced vastly larger damage estimates.
24
In our 2015 survey and the corresponding academic paper (Howard & Sylvan, 2020), we used a dierent two-part
approach to address outliers in damage estimates. First, we trimmed responses at the 99th percentile; we selected the 99th
percentile to eliminate only the most extreme responses. Second, in addition to reporting the mean damage estimates,
we conducted sensitivity analysis by reporting median damage results. e mean is generally the appropriate central
estimate (i.e., wisdom of the group) if experts’ responses are forecasts, as the mean minimizes forecast error. However,
if results are highly skewed and/or if elicited damages are more representative of beliefs than forecasts, then the median
may beer reect expert consensus (Lorenz et al., 2011; Freeman & Groom 2015; Colson & Cooke, 2018). We believe
the median is appropriate in this context.
Our 2021 survey responses on this topic are highly skewed, as the mean diers considerably from the median. In this
survey, we asked respondents for estimates of impacts in the near future (i.e., a 1.2°C increase in 2025) when uncertainty
is considerably lower relative to a 3°C increase in 2075 (given that it is much closer to our current state). Using our
previous method of trimming at the 99th percentile would imply a range of damages of +2% to -45% of GDP for a 1.2°C
in 2025. e upper end of this range is far above the range in the literature: DICE-2016R’s 0% to -0.7% (Nordhaus,
2017), Howard and Sterner (2017)’s range of -0.5% to -2.7% for non-catastrophic and total impacts; and Burke et al.
(2015)’s range of approximately -1.7% to -6.8% for market-only impacts.21
To address outlier estimates and avoid puing our ngers on the scale, we apply a 95th percent condence interval
trimming methodology, implying a damage range of 0% to -18% in 2025. However, as the upper end of this range implies
a permanent catastrophic event with a magnitude akin to the Great Depression occurring in under ve years, we also
conducted 90th percentile trimming to analyze the median estimate. As the median is relatively stable across various
trimming ranges, we focus our analysis primarily on the median estimate aer 95th percentile trimming.
Comparing Our Results to Other Estimates
Economists’ primary focus in the climate damage literature is on the economic impact of a doubling of CO2 emissions
from pre-industrial levels. Based on the central estimate of the “equilibrium climate sensitivity parameter” (i.e., the
amount of long-run warming from a doubling of CO2 emissions), the bulk of climate damage estimates fall around 3°C.
Focusing on our 95th percentile trimming results, we nd that mean and median damage estimates for a 3°C increase in
2075 are -8.5% and -5% in GDP losses, respectively. is estimate is in line with our 2015 survey results.22 Both estimates
are signicantly higher than DICE-2016R2, which implies a central estimate of only -2.1% of GDP for damages under a
3°C increase in 2075 (Nordhaus, 2017). Our survey ndings are however within the range characterized by Howard and
Sterner (2017), which forms the basis of the alternative damage function in Nordhaus (2019). Compared to Burke et al.
(2015), our survey ndings appear to be relatively conservative.
For a 1.2°C increase by 2025, we nd mean and median losses of -2.2% and -1%, respectively, indicating that most
economists believe that initial benets from climate change are no longer (or maybe never were) present; this is
consistent with our 2015 survey. Again, this estimate is far above DICE-2016R and is more consistent with Howard and
Sterner (2017). However, only the median estimate is consistent with the DICE 95th percent condence interval for non-
catastrophic damages, though the mean estimate is relatively consistent with Burke et al. (2015).
25
We al so asked respondents to provide damage estimates for higher temperature increases of 5°C and a 7°C. ese estimates
imply a damage function with less curvature than the standard quadratic damage function of DICE, potentially more
consistent with earlier DICE damage functions that limit damages to below 100% of GDP; these results are consistent
with Nordhaus (1994)’s survey results (Roughgarden & Schneider, 1999).
Figure 12
Climate Damage Estimates: Adaptation Scenarios
Changes in Economic Damage Due to Dierent Rates of Temperature Change or Income
Scenario Scenario 1:
Baseline
Scenario 2:
10% Higher
GDP
Scenario 3:
Faster
Warming
Scenario 4:
Slower
Warming
Year 2075 2075 2050 2100
Temperature increase (relative to pre-industrial era) 3°C 3°C 3°C 3°C
Economic damages (% of global GDP) - Median estimate -5% -5% -6% -4%
Economic damages (trillions of 2019 USD) - Median estimate -$29.8 -$32.7 -$20.8 -$36.1
Economic damages (% of global GDP) - Mean estimate -8.5% -7.40% -9.90% -6.60%
Economic damages (trillions of 2019 USD) - Mean estimate -$50.6 -$48.4 -$34.3 -$59.5
Standard deviation 7.6 7.3 8.8 6.4
Results above reect the trimming of outlier estimates below the 5th percentile or above the 95th percentile of total responses.
Comparing our “Adaptation Scenario” results to the prior question, we nd evidence that higher income and slower rates
of warming result in lower climate damages. is is consistent with the widespread view that society can beer adapt
to climate change if the rate of warming is slower or if society is wealthier. However, even in the scenarios with slower
warming or higher GDP, damage estimates are high, with a loss of at least 4% of GDP expected in each scenario (and a
6% loss expected if warming is faster than the baseline). It is also worth noting that respondents expect higher GDP to
provide only a minor adaptation benet: the median damage estimate for this scenario is nearly identical to the baseline,
though the mean estimate is less severe than the baseline mean estimate (the higher GDP scenario results in losses of
7.4% of GDP rather than 8.5%). In terms of actual dollars, the monetary impacts look quite similar, indicating that a
wealthier society might save slightly in relative terms but the costs of climate change remain consistently high.
As DICE ignores potential impacts from divergent GDP levels and warming rates in its specication of the damage
function, our results further indicate that it likely underestimates the benets of climate action. is is particularly true
when the DICE model deviates from its central calibration point (which is around 3°C for most of the damage estimates
that underlie its calibration), as when it solves for the socially optimal temperature increase.
26
Figure 13
Estimated Cost of Emissions Abatement
Scenario
Abatement
Scenario A -
2075
Abatement
Scenario A -
2130
Abatement
Scenario B -
2075
Year 2075 2130 2075
Temperature increase (relative to pre-industrial era) 2.5°C 3.5°C 2°C
Abatement level (% of emissions controlled) relative
to RCP8.5-like pathway 68% 97% 88%
Abatement cost (% of global GDP) - Median estimate 3% 4% 4%
Abatement cost (trillions of 2019 USD) - Median estimate $17.5 $55.8 $25.3
Abatement cost (% of global GDP) - Mean estimate 4.3% 5.9% 6.2%
Abatement cost (trillions of 2019 USD) - Mean estimate 25.0 82.4 39.2
Standard deviation 4.5 6.2 6.7
Results above reect the trimming of outlier estimates below the 5th percentile or above the 95th percentile of total responses.
is question focuses on the cost of reducing emissions below a business-as-usual path (such as the RCP 8.5 scenario)
at dierent times and levels of abatement. In the question text, we asked respondents to factor in all costs, including
opportunity and general-equilibrium costs.
ese abatement cost projections are higher than estimates from the IPCC and some other sources. However, the survey
ndings reveal a clear belief that reducing emissions by a meaningful amount is likely to cost less than the expected
damages from climate change. ese abatement cost estimates cannot be directly balanced against the damage estimates
from the prior questions since they measure dierent outcomes under dierent scenarios and represent dynamic paths,
but the underlying message is clear. Respondents already estimated that warming of 3°C by 2075 would lead to a loss
of roughly 5% of GDP and a risk of even more catastrophic damages. But investing roughly 3-4% of GDP in emissions
abatement by 2075 could cut that warming signicantly, to 2.5°C or even 2°C, avoiding a great deal of damages in 2075
and later. A s discussed below, the same respondents who made these abatement costs estimates also expressed widespread
agreement that the costs of reaching net-zero emissions targets by mid-century were outweighed by the benets.
Additionally, abatement costs are generally assumed to encompass investments in clean technologies, which entail
upfront costs but oer benets over a longer duration. By contrast, climate damage estimates are oen projected to recur
annually (on average) and worsen over time. As a reference point, respondents estimated that a continuation of the same
3°C warming trend until it reached 5°C would result in an annual GDP loss of 10% ($143 trillion in damages in 2130).
ese abatement cost estimates are on the high end of the literature, though they are still roughly consistent with some
IAM results.23 Several factors could be contributing to the disparity between our survey estimates and the literature. Un-
27
like responses to the prior questions, the median abatement cost estimate is sensitive to 95th percentile trimming, indi-
cating that a lower cost estimate of 2.75% of GDP (to achieve a target of 2.5°C by 2075) may in fact be more accurate.24
Additionally, if we focus on respondents with potentially more relevant expertise, i.e., those self-identifying as publishing
on emissions abatement or those with multiple publications that meet the survey criteria, the resulting abatement cost
projections drop to be more in line with the IPCC estimates (see Appendix D). Even then, the central estimates of these
subgroups exceeds both DICE and the median IPCC estimates.25
On the other hand, this disparity is partially the result of the IPCC potentially underrepresenting abatement cost
uncertainty, possibly leading to an underestimate of abatement costs. Specically, the IPCC uses cost estimates from
“idealized implementation scenarios,” which assume a ubiquitous price on carbon and other GHGs is applied across the
globe in every sector of every country and rises over time.”26
A clear advantage of expert elicitation is that experts can account for dicult-to-model factors. When estimating damages,
survey respondents can account for non-market, socially contingent, and tipping point impacts of climate change in
ways that models might not. e same likely applies for abatement costs, and our relatively high estimates might reect
assumptions that emissions reductions will occur in a manner that is not perfectly ecient, due to either technological or
political reasons. If we assume that our results represent the wisdom of the crowd (i.e., are asymptotically ecient), then
a key advantage of our expert elicitation is that we implicitly capture these considerations to the extent that respondents
consider and weigh them.
Despite the higher cost estimates from our survey, the cost trends for more aggressive mitigation are lower than some
other projections. We nd that the percentage increase in costs from Scenario A to Scenario B is lower than that of DICE-
2016R2, implying that an extrapolation of our results would eventually show equal or lower costs to reach even more
aggressive mitigation targets than DICE.27
Net-Zero Emission Targets
Aer earlier survey questions asked respondents to think through both the costs and benets of climate action and
inaction, we directly asked whether a net-zero emissions target by mid-century was likely to be cost-benet justied.
Mid-century net-zero targets have recently been adopted by some European countries and a small number of U.S.
states (Energy & Climate Intelligence Unit, 2021; Center for Climate and Energy Solutions, 2021). ese are the most
aggressive emissions-reduction targets adopted by governments to date, and they align generally with the aims of the
Paris climate agreement, to limit temperature increases to less than 2°C.
28
Figure 14
Many government entities have set goals to reach net-zero GHG emissions by
roughly mid-century (this would be consistent with a global average surface
temperature limit of 1.5° to 2°C according to many projections).
Are the expected benets of mid-century net-zero GHG targets
likely to outweigh the expected costs?
Please account for any relevant co-benets and co-costs
in your implicit present-value estimates.
e survey results suggest that economists with climate expertise believe that aggressive mitigation measures are
warranted given the expected damages from climate change. Before answering this question, economists were asked to
make specic forecasts about the costs of emissions abatement and the expected damages from climate impacts, ideally
leading them to think through this high-level cost-benet analysis in a somewhat detailed manner. Given this context,
the ndings are especially striking.
Nearly two-thirds of respondents believe that these net-zero goals are “likely” or “very likely” to be cost-benet justied,
despite the aggressive timeline and the anticipated abatement costs highlighted in the prior question. Meanwhile, only
12% of respondents think the goals are “unlikely” or “extremely unlikely” to be net-benecial.
Many government entities have set goals to reach net-zero GHG emissions by
roughly mid-century (this would be consistent with a global average surface
temperature limit of 1.5° to 2°C according to many projections).
Are the expected benefits of mid-century net-zero GHG targets
likely to outweigh the expected costs?
Please account for any relevant co-benefits and co-costs
in your implicit present-value estimates.
No opinion
Extremely unlikely
Unlikely
Not clear
Likely
Extremely likely
% of Respondents
9%
0 20 40 60
31%
18%
3%
4%
35%
29
Likely due to the uncertainties involved, more than 18% chose the “Not Clear” option while an additional 4% expressed
“No Opinion.” But of the remaining respondents, economists that believed strong climate action was economically
justied outnumbered their counterparts almost six to one.
ese ndings stand in contrast to models such as DICE, which estimates an “optimal” temperature (where benets
and costs are balanced) of 3.5°C in 2100 and a maximum of 4.1°C in 2165. ese results and other past research suggest
that DICE does not align with the consensus views of experts on a number of key issues, including expected damages/
damage functions, abatement costs, discount rates, and negative-emission technology availability.
Context from the Pandemic
Our nal question focused on estimated changes to GDP and greenhouse gas emissions in 2020, in order to provide
some context about how estimated climate impacts might compare to the impacts from the Covid-19 pandemic.
At the time of the survey, in early February 2021, ocial estimates of GDP and greenhouse gas emissions for 2020 were
not yet available, so respondents would not have been able to nd reliable anchors for their estimates (the question also
stipulated not to use outside sources like the internet).
We included this question partially as a “seed question”—once respondents’ answers can be empirically veried, we can
use their relative accuracy to weight responses to other questions. (is analysis will be undertaken in future research. As
of this publication, ocial estimates of the change in GDP and greenhouse gas emissions remain unavailable.)
Figure 15
Given the unprecedented events of 2020, please estimate the % change
in global GDP and global greenhouse gas emissions (CO2e) from 2019 to 2020,
without using outside sources like the internet.
Global GDP
(% change from
2019 to 2020)
Global Greenhouse
Gas Emissions
(% change from
2019 to 2020)
Median estimate -3% -3.5%
Mean estimate -2.9% -3.1%
Standard deviation 4.1 4.7
Results above reect the trimming of outlier estimates below the 5th percentile or above the 95th percentile of total responses.
ese ndings underscore the severity of respondents’ climate damage forecasts from earlier in the survey. Respondents’
GDP loss estimates for 2020, when a pandemic devastated the global economy, are far smaller than their estimates for
annual damages from climate change under a 3°C warming scenario (-3% of GDP vs. -5 %). And unlike the pandemic-
related downturn, which will presumably be followed by a rebound, the climate impacts are projected to recur and worsen.
30
Implications for Economic Modeling
e ndings from this survey and other expert elicitation projects can help improve policy-relevant modeling of climate
damages. One of the Biden administration’s rst acts was to reconvene the Interagency Working Group on the Social
Cost of Greenhouse Gases (IWG) that had developed the ocial U.S. social cost of carbon estimates during the Obama
administration (IWG, 2021). e Biden administration asked the IWG to develop interim estimates for the social cost
of greenhouse gases within one month and then to comprehensively update the estimates by 2022 to reect the best
available scientic information.
Much of this update will likely correspond to recommendations from the National Academies of Sciences, Engineering,
and Medicine, which published two reports on updating the SCC modeling methodology in 2016 and 2017. NAS (2017)
called for the use of expert elicitation in the development of several IAM components, including socio-economic and
emission scenarios. Expert elicitation can help modelers understand the current consensus, or lack thereof, on a number
of key climate-economic parameters that are highly uncertain, so that they can make appropriate changes.
e results from this survey and others like it can be used to help calibrate IAMs. e methodology laid out in Howard
and Sylvan (2020) demonstrates how to calibrate key parameters in a prominent IAM (DICE). Specically, ndings
from our 2021 and 2015 surveys can be used to calibrate:
• Climate damage functions – Modelers can use elicited damage estimates from the survey to calibrate impacts
under various potential temperature increases. As the baseline version of DICE-2016R2 models temperature
increases of 1°C to 7.2°C, our survey elicits damage estimates for 1.2°C, 3°C, 5°C, and 7°C to allow for a exible
functional form. We ask about general damages and specically ask respondents to consider non-market eects,
to allow for a full consideration of impacts. More generally, by asking for 5th and 95th percentile estimates, we
capture the overall risk of climate damages, nding a higher likelihood of severe damages than small damages
(i.e., positive skewness).
• Adaptation and its costs – Modelers can use survey estimates to calibrate the impact of the rate of temperature
change and per-capita income on net damages, using alternative damage scenarios. We ask respondents to
consider adaptation and its costs when estimating damages, in order to reect these values. By asking for the
5th and 95th percentile estimates, we also represent the impact of adaptation on the overall risk prole of climate
damages.
• Technological availability and adoption – When calibrating the emissions scenario and the availability of
negative-emissions technology, modelers can use survey estimates of the percent of zero-emissions energy in
2050 and the period when negative-emissions technology will become scalable.
• Mitigation costs – When calibrating the abatement cost function, modelers can use survey estimates for
abatement costs in several climate scenarios. is will help capture the magnitude of abatement costs, how
they change over time, and how they change with mitigation levels. Our question design allows respondents to
consider imperfect implementation and technological constraints. Again, by asking for 5th and 95th percentile
estimates, we also represent the small but real probability of higher-than-expected abatement costs.
• Discount rate – Modelers can calibrate the discount rate using the median rate provided by survey respondents
(representing the normative quality of discount rates). is is based on a voting procedure. Other methods are
also available, as discussed in Howard and Sylvan (2020).
31
By using survey data to calibrate existing IAMs or a modular IAM (as recommended by NAS (2017)), the IWG can
provide the public with transparent estimates of these parameters along with a full representation of their likely values.
is will allow the IWG to fully characterize the range of SCC estimates.
Expert elicitation may or may not be the proper tool for all these parameters. One NAS (2017, p. 149) recommendation
advises against using “top-down” expert elicitation for climate damage function calibration, instead focusing on its use
to quantify specic, hard-to-measure impacts (such as the impact and probability of crossing specic tipping points).
However, modelers could generate more robust results by developing several damage function modules that use dierent
approaches, as the magnitude of damage estimates varies signicantly by estimation methodology (Howard & Sterner,
2017; Howard, 2019).
Boom-up approaches that include expert elicitation on focused impacts may be beer suited to tracing damages to
specic pathways, as noted by NAS (2017). However, this comes at a cost. Specically, boom-up estimates generally
omit certain impacts, such as inter-sectoral impacts and feedbacks.
As risk and ambiguity (i.e., unknown unknowns) are critical parts of climate change economics, it is important to account
for a wide range of impacts and possibilities, particularly to meet the NAS (2017) goal of fully representing the SCC
range. Given the multitude of potential climate impacts (Howard, 2014), formal expert elicitation cannot be applied for
all categories. However, the top-down expert elicitation approach is critical for fully understanding the risk that society
faces, and it is a good barometer of the potential biases inherent to other damage-estimation strategies. Furthermore,
developing a full set of boom-up damage estimates will likely take decades. Top-down elicitation can be useful as the
IWG conducts its review in a one-year timeframe.
We believe the respondent criteria used for this survey is appropriate for informing model calibration. NAS (2017, p.
225) interprets the literature to suggest that elicitation should focus on a handful of experts (ve to 10) identied by
their number of citations or recognized expertise. But research suggests that reputation, peer rankings, and individual
citations are uncorrelated with an individual’s predictive performance (Aspinall, 2010; Colson & Cooke, 2018). Instead,
the application of consistency checks (Howard & Sylvan, 2020) and performance weights, based upon responses to “seed
questions”—related questions whose responses can be veried aer the survey is conducted—can improve predictive
performance relative to equal weighting (Cooke & Goossens, 2008; Howard & Sylvan, 2020). e application of these
techniques to a large sample oers some clear benets over small-sample expert surveys, which can suer from selection
bias and whose limited sample size can reduce estimate eciency.
In future research, we plan to re-calibrate DICE using our survey results as discussed above. Specically, we plan to
recalibrate the DICE damage and abatement cost functions (including technological availability). In doing so, we will
also allow the damage function to vary with the rate of temperature change and income per capita, in addition to overall
temperature change, to capture the impacts of adaptation. We will also recalibrate the DICE discount rate to match the
expert consensus found in Drupp et al. (2018) and Howard and Sylvan (2020). Finally, we will recalibrate the DICE
climate model, which currently suers from some calibration issues (Howard & Sylvan, 2020). is recalibration will
beer represent the full range of climate uncertainty. In doing so, we will aim to elucidate the full range of SCC estimates.
32
Conclusions
Economists with expertise on climate change can provide unique insights into the risks from climate impacts and
the appropriate policy strategies. Our survey ndings suggest that economic experts are increasingly alarmed
about the threat that climate change poses to the global economy, even compared to signicant levels of concern
expressed ve years ago.
Respondents in our survey predict a number of problematic impacts from climate change, including increased inequality
both between and within countries, a reduction in the global economic growth rate, and signicant economic damages
under all climate scenarios presented.
ese experts also believe that low-emissions technologies have signicant economic promise, projecting that some
emerging technologies are likely to replicate the dramatic cost reductions seen in solar and wind energy, and that
negative-emissions technologies are likely to be viable in the coming decades. is survey reveals a strong consensus that
bold climate mitigation strategies (including mid-century net-zero GHG targets) are likely to be economically justied.
ese ndings help clarify the level of consensus on key climate issues among economic experts. e results can be useful
to both policymakers and economic researchers; both groups should heed the clear call for immediate and meaningful
eorts to reduce emissions and limit the enormous economic risks of climate change.
Appendix A. Survey Questions
Page 1 of Survey
SURVEY ON CLIMATE CHANGE ECONOMICS AND POLICY
e Institute for Policy Integrity at New York University School of Law is conducting a survey to examine the professional
opinions of expert economists on climate change economics and policy. is survey is being sent only to economists who
have published a climate change-related article in a top economic journal. e survey should take roughly 15 minutes
to complete and consists of 10 multiple-choice questions and ve forecasts. e aggregate results of this survey will be
used in academic research and potentially distributed to journalists, but individual responses will be anonymous and
condential.
Page 2 of Survey
Q1. You have published and consider yourself an expert on the following topics (check all that apply):
° Estimated damages from climate change
° Climate change uncertainty and risks, including tipping points and fat tails
° Climate change adaptation and system resilience
° Greenhouse gas emissions abatement / mitigation
° Climate scenario modeling or cost-minimization modeling
° Social Cost of Carbon or optimal climate policy modeling
° Global climate strategies / agreements / policies
° Climate change in developing countries / Geographic distribution of climate impacts
° Other climate-related topics
Page 3 of Survey
Professional Opinions on Climate Change
Q2. Which of the following best describes your views about climate change?
° is is not a serious problem
° More research is needed before action is taken
° Some action should be taken now
° Immediate and drastic action is necessary
Q3. As informed by your research, how has your level of concern about climate change shied over the past
ve years?
° Strongly increased (You view climate change as a more pressing problem)
° Somewhat increased
° Remained unchanged
° Somewhat decreased
° Strongly decreased (You view climate change as a less pressing problem)
° No opinion
33
34
Q4. Please check up to three items that had the greatest eect on your views about climate change over the
past ve years.
° e coronavirus pandemic
° e Paris Agreement
° e United States’ withdrawal from the Paris Agreement
° e IPCC Special Report on the impacts of 1.5°C global warming
° Increased global energy demand
° Continued expansion of fossil fuel infrastructure (e.g., North America’s investment in natural gas; Asia’s
investment in coal; etc.)
° Increased environmental deregulation within certain countries (United States, Brazil, etc.)
° China’s strengthening of its climate policies
° New environmental policies and/or ambitious policy proposals in other countries
° Observed extreme weather events aributed to climate change
° New ndings in climate science
° New ndings in climate economics and the social sciences
° Improvements and/or cost reductions in low-emissions technologies
° New developments with negative-emissions technologies
° Insucient improvements in abatement technologies and costs
° Changes in adaptation technologies and costs
° Other(s) ________________________________________________
Page 4 of Survey
Climate Change and Economic Growth
Q5. What is the likelihood that climate change will have a long-term, negative impact on the growth rate of
the global economy? is is distinct om an impact on the level of GDP in a given year (i.e., climate damages
measured as % of GDP).
° Extremely likely (80% to 100% probability)
° Likely (60% to 80%)
° Not clear (40% to 60%)
° Unlikely (20% to 40%)
° Extremely unlikely (0% to 20%)
° I anticipate no climate damages
° No opinion
Page 5 of Survey
Distributional Impacts
Q6. What is the likelihood that climate change will increase economic inequality between low-income and
high-income countries (the lower third of countries by per-capita income versus the upper third of
countries by per-capita income)?
° Extremely likely (80% to 100% probability)
° Likely (60% to 80%)
° Not clear (40% to 60%)
° Unlikely (20% to 40%)
° Extremely unlikely (0% to 20%)
° I anticipate no climate damages
° No opinion
35
Q7. What is the likelihood that climate change will increase economic inequality within most countries,
between the lower third of households by household income and the upper third of households by
household income?
° Extremely likely (80% to 100% probability)
° Likely (60% to 80%)
° Not clear (40% to 60%)
° Unlikely (20% to 40%)
° Extremely unlikely (0% to 20%)
° I anticipate no climate damages
° No opinion
Page 6 of Survey
Emissions Abatement
Q8. Over the last decade, the costs of solar and wind energy technologies have dropped rapidly (-7%
annually for solar PV and -4% annually for onshore wind). Do you think a similar paern is likely to be
replicable for some other emerging zero-emission and negative-emission technologies?
° Extremely likely (80% to 100% probability)
° Likely (60% to 80%)
° Not clear (40% to 60%)
° Unlikely (20% to 40%)
° Extremely unlikely (0% to 20%)
° No opinion
Q9. In 2050, what share of the global energy mix do you think will consist of zero-emission technologies
(e.g., solar, wind, nuclear, green hydrogen, bioenergy with carbon capture and storage, etc.)?
For context, zero-emission sources make up roughly 10% of the current energy mix according to the
International Energy Agency.
% of global energy mix
0 10 20 30 40 50 60 70 80 90 100
Zero-GHG Technologies
36
Q10. During what time period do you believe that net-negative greenhouse gas emission technologies, such
as direct air capture and carbon capture/utilization/storage (CCUS), will become viable and reliable at
a low-enough cost to be adopted on a large scale (i.e., to signicantly change the global emissions path)?
° By 2040
° By 2060
° By 2080
° By 2100
° Between 2100 and 2150
° Aer 2150
° Net-negative GHG technologies will never become viable
° No opinion
Page 7 of Survey
Climate Damage Estimates
Q11. Please provide your best estimates for how the following climate scenario would aect global GDP
over time.
Please enter your median/50th percentile estimate and your 95th percent condence interval of the impact
on global output. To avoid (over)condence and anchoring biases, please ll out the 95th percent condence
interval before selecting the 50th percentile.
Please include non-market and market impacts, and factor in adaptation to climate change and its corresponding
costs. Please also consider the level of global GDP and rate of temperature change, in addition to overall
temperature change (some researchers theorize that society is more capable of adapting to climate change with
slower rates of temperature change and/or higher levels of global economic output).
Please provide your answer as a % of global GDP. If you believe these impacts will increase GDP rather
than decrease it, please indicate this with a (+).
Scenario Scenario 1 -
2025
Scenario 1 -
2075
Scenario 1 -
2130
Scenario 1 -
2220
Year 2025 2075 2130 2220
Temperature increase
(relative to pre-industrial era) 1.2°C 3°C 5°C 7°C
Average annual temperature
increase over previous 30 years 0.03°C 0.04°C 0.03°C 0.01°C
Estimated global GDP without climate
change (trillions in 2019 USD) 173.3 595.1 1430.4 3654.5
Scenario 1 - 2025
Climate Damage
(% of GDP)
Scenario 1 - 2075
Climate Damage
(% of GDP)
Scenario 1 - 2130
Climate Damage
(% of GDP)
Scenario 1 - 2220
Climate Damage
(% of GDP)
5th Percentile
50th Percentile
95th Percentile
37
Adaptation Estimates
Q12. Please provide your best estimates of how adaptation could alter the level of climate damages, based on
dierent rates of temperature change and anticipated levels of adaptation.
Some researchers theorize that society is more capable of adapting to climate change with slower rates of
temperature change and/or higher levels of global economic output.
We are interested in your best guess (median/50th percentile estimate) and your 95th percent condence
interval of the impact on global output, as a percentage of GDP, of three alternative climate scenarios to Scenario
1 - 2075 (provided earlier in column 2 of Question 11). To avoid (over)condence and anchoring biases, please
ll out the 95th percent condence interval before selecting the 50th percentile.
Please include non-market and market impacts, and factor in adaptation to climate change and its corresponding
costs. Please provide your answer as a % of global GDP. If you believe these impacts will increase GDP
rather than decrease it, please indicate this with a (+).
Scenario
Scenario 2 -
10% Higher GDP
Relative to
Scenario 1 - 2075
Scenario 3 -
Faster Warming
Relative to
Scenario 1 - 2075
Scenario 4 -
Slower Warming
Relative to
Scenario 1 - 2075
Year 2075 2050 2100
Temperature increase
(relative to pre-industrial era) 3°C 3°C 3°C
Average annual temperature increase
over previous 30 years 0.04°C 0.06°C 0.02°C
Estimated global GDP without climate
change (trillions in 2019 USD) 654.61 346.6 901.5
Scenario 2
Climate Damage
(% of GDP)
Scenario 3
Climate Damage
(% of GDP)
Scenario 4
Climate Damage
(% of GDP)
5th Percentile
50th Percentile
95th Percentile
38
Page 8 of Survey
Abatement Cost Estimates
Q13. Please provide your best estimates of global abatement costs (as a percentage of GDP) for dierent
climate scenarios.
We are interested in your best guess (median/50th percentile estimate) and your 95th percent condence
interval of the total cost of greenhouse gas abatement, as a percentage of GDP, for three emission scenarios.
Please factor in all costs, including opportunity and general-equilibrium costs. If you believe that low-carbon
options are cheaper than conventional technologies, such that abatement costs are negative rather than positive,
please indicate this with a (-).
To avoid (over)condence and anchoring biases, please ll out the 95th percent condence interval before
selecting the 50th percentile.
Scenario Abatement
Scenario A - 2075
Abatement
Scenario A - 2130
Abatement
Scenario B - 2075
Higher Mitigation
Year 2075 2130 2075
Temperature increase
(relative to pre-industrial era) 2.5°C 3.5°C 2°C
Abatement level
(% of emissions controlled)
relative to RCP8.5-like pathway
68% 97% 88%
Abatement
Scenario A - 2075
Abatement Cost
(% of GDP)
Abatement
Scenario A - 2130
Abatement Cost
(% of GDP)
Abatement
Scenario B - 2075
Abatement Cost
(% of GDP)
5th Percentile
50th Percentile
95th Percentile
39
Page 9 of Survey
Net-Zero Emission Targets
Q14. Many government entities have set goals to reach net-zero GHG emissions by roughly mid-century (this
would be consistent with a global average surface temperature limit of 1.5° to 2°C according to many
projections).
Are the expected benets of mid-century net-zero GHG targets likely to outweigh the expected costs?
Please account for any relevant co-benets and co-costs in your implicit present-value estimates.
° Extremely likely (80% to 100% probability)
° Likely (60% to 80%)
° Not clear (40% to 60%)
° Unlikely (20% to 40%)
° Extremely unlikely (0% to 20%)
° No opinion
Page 10 of Survey
Recent Impacts
Q15. Given the unprecedented events of 2020, please estimate the % change in global GDP and global
greenhouse gas emissions (CO2e) from 2019 to 2020 without using outside sources like the internet.
Answers to this question will provide context for survey results on climate change. To avoid (over)condence
and anchoring biases, please ll out the 95th percent condence interval before selecting the 50th percentile.
Indicate a decline with a (-) and an increase with a (+).
Global GDP
(% Change from
2019 to 2020)
Global Greenhouse
Gas Emissions
(% Change from
2019 to 2020)
5th Percentile
50th Percentile
95th Percentile
Page 11 of Survey
We thank you for your time spent taking this survey.
Your response has been recorded.
40
Appendix B. List of Journals Used to Assemble
Survey Respondent Pool
40
Journals Included in 2015 Survey
Economics
American Economic Review Yes
Econometric Theory Yes
Econometrica Yes
Economic Journal Yes
Economic Theory Yes
Economics Letters Yes
European Economic Review Yes
Games and Economic Behavior Yes
International Economic Review Yes
Journal of Applied Econometrics Yes
Journal of Business and Economic Statistics Yes
Journal of Development Economics Yes
Journal of Econometrics Yes
Journal of Economic Dynamics and Control Yes
Journal of Economic Literature Ye s
Journal of Economic Perspectives Yes
Journal of Economic Theory Yes
Journal of Financial Economics Yes
Journal of Human Resources Yes
Journal of International Economics Yes
Journal of Labor Economics Yes
Journal of Monetary Economics Ye s
Journal of Money, Credit, and Banking Yes
Journal of Political Economy Yes
Journal of Public Economics Yes
Journal of the European Economic Association Yes
NBER Macroeconomics Annual Ye s
Quarterly Journal of Economics Ye s
Rand Journal of Economics Yes
Review of Economic Studies Yes
41
Journals Included in 2015 Survey
Environmental and Resource Economics
American Journal of Agricultural Economics Yes
Ecological Economics Yes
Environment and Resource Economics Yes
Journal of Environmental Economic Management Yes
Land Economics Yes
Resource and Energy Economics Yes
Review of Environmental Economics and Policy No
Journal of the Association of Environmental and Resource Economists No
Development Economics
Journal of Development Economics Yes
World Development No
World Bank Economic Review No
World Bank Research Observer No
Journal of Development Studies No
Economic Development and Cultural Change No
Development Policy Review No
Appendix C. Response Data by Survey Question
42
Question Responses
Response Rate
(Based on 2,169
Survey Invitations)
% of Total Respondents
(Based on 738
Submitted Surveys)
1 733 33.8% 99.3%
2 738 34.0% 100.0%
3 736 33.9% 99.7%
4 727 33.5% 98.5%
5 733 33.8% 99.3%
6 731 33.7% 99.1%
7 730 33.7% 98.9%
8 726 33.5% 98.4%
9 712 32.8% 96.5%
10 725 33.4% 98.2%
11* 276 to 301 12.7% to 13.9% 37.4% to 40.8%
12* 224 to 229 10.3% to 10.6% 30.4% to 31%
13* 212 to 222 9.8% to 10.2% 28.7% to 30.1%
14 571 26.3% 77.4%
15* 340 to 342 16.7% to 15.8% 46.1% to 46.3%
Question numbers correspond to the numbering used in the survey questions
(see Appendix A) as well as the numbering of gures in the report.
*ese questions solicited multiple forecasts under various scenarios. Some respondents submied
estimates for only a subset of the scenarios, so we present a range of response rate data.
Appendix D. Additional Forecast Analysis
43
e analysis below provides additional context for the survey questions that asked for quantitative forecasts of climate
damages and abatement costs.
Examining 95th Percentile Results
ere are several relevant takeaways from respondents’ 95th percent condence intervals of damages and abatement
costs. First, net-negative climate damages are decisively predicted this decade, as 0% damages lies outside the 95th percent
condence interval for a 1.2°C increase in 2025. is is consistent with ndings from our 2015 survey of 0% climate
damages at approximately a 1°C increase in 2020 (Howard & Sylvan, 2020). Second, we nd a 5% probability of a GDP
loss of -10% (median) or -17% (mean) for a 3°C increase in 2075. is implies slightly lower catastrophic risks compared
to our previous survey results, which found probabilities of 10% (median) and 22% (mean) that a -25% GDP loss would
occur.28 ird, we nd that there is approximately a 5% chance of a 40% or more loss of GDP for a 7°C increase. is is
unsurprising, as we nd that damage uncertainty increases with temperature. Last, we nd some evidence that increases
in income and slower rates of temperature change decrease the risk of catastrophic climate change. Since the GDP path
and equilibrium climate sensitivity are highly uncertain in the long run (though temperature and income are specied
as certain in our scenario), catastrophic risks for 5°C and 7°C by 2130 and 2220 are greatly understated. Calibrating a
model like DICE to our results, similar to the approach used in Howard and Sylvan (2020), will be necessary to fully
specify the implied catastrophic risks.
Focusing on the 95th percentile results for abatement, we nd that respondents generally believe that emissions mitigation
cannot be accomplished at a negative cost (i.e., “free lunch” where companies act irrationally) as implied by some research
(Gillingham & Stock, 2018). Similar to damages, we nd that abatement costs are positively skewed. is corresponds to
the empirical ndings of Lemoine and McJeon (2013). Finally, we nd that the range of mitigation costs increases with
mitigation levels as indicated by the empirical literature (IPCC, 2014).
Topic Expertise and Forecasts
While the criteria for our survey sample is designed to identify experts in climate economics, one could potentially
argue that knowledge of more specic topics is needed to make accurate forecasts. To address this concern, in the rst
question of our survey we requested that respondents self-identify the topics on which they have published and consider
themselves an expert. us, we can test whether economists with self-identied expertise on damages provide dierent
climate damage forecasts than others, and we can conduct similar tests for abatement cost experts on abatement forecasts,
etc. Additionally, we sent separate but identical surveys to economists based on the number of qualifying publications
they authored (one or multiple) and the type of journal in which they published (economics, environmental, and
development). us, we can test whether economists with multiple qualifying publications provide dierent estimates
than other groups.
44
Damage forecasts are relatively similar between the full sample and those who self-identify as publishing on climate
damages. At low levels of warming, these experts believe that damage estimates are higher than the full sample, but aer
temperatures rise past 3°C the damage experts predict lower levels of damage than the full group. is may stem from
a belief that adaptation is particularly sensitive to both income and the rate of temperature increase, given that global
income grows substantially over time along with temperature (the laer of which slows down over time).
However, damage forecasts are higher for economists who self-identify as publishing on climate adaptation, relative
to the full sample. ese economists appear to be more pessimistic about society's ability to adapt to climate change,
particularly for lower temperature increases.
Finally, economists with multiple qualifying publications forecast a slightly lower magnitude of damages than the full
sample. us, there is no consistent direction by which relevant expert subgroups dier in their damage forecasts from
the full sample.
A brief analysis of subgroup dierences for our abatement cost forecasts can be found in the main report text.
45
Table D1. 95th Percent Condence Intervals for Damage, Abatement, and Pandemic Questions
Question Scenario
5th percentile 50th percentile 95th percentile
Obs Mean Median Std.
Dev. Obs Mean Median Std.
Dev. Obs Mean Median Std.
Dev.
11 S.1 - 2025 268 -1.0 -0.5 2.3 279 -2.2 -1 2.9 265 -5.2 -3 5.5
11 S.1 - 2075 260 -3.5 -2 3.8 270 -8.5 -5 7.6 265 -17.4 -10 15.2
11 S.1 - 2130 244 -7.5 -5 7.6 253 -16.1 -10 13.3 250 -31.0 -22 24.0
11 S.1 - 2220 240 -12.2 -7 13.5 251 -25.2 -20 20.7 249 -45.1 -40 30.7
12 S.2 - 2075 202 -3.0 -2 3.6 210 -7.4 -5 7.3 201 -14.6 -10 13.2
12 S.3 - 2075 205 -4.9 -3 5.8 203 -9.9 -6 8.8 201 -18.7 -12 16.1
12 S.4 - 2075 199 -2.4 -1.3 2.7 204 -6.6 -4 6.4 199 -13.2 -8.5 12.5
13 S.A - 2075 203 1.5 1 2.7 203 4.3 3 4.5 197 8.2 5 7.5
13 S.A -2130 192 2.4 1 3.7 192 5.9 4 6.2 193 12.2 8 11.8
13 S.B - 2075 197 2.9 2 4.6 197 6.2 4 6.7 193 11.2 8 10.4
15 GDP 297 -0.1 -1 4.0 310 -2.9 -3 4.1 310 -6.0 -6 5.6
15 GHG 280 0.1 -1 4.2 307 -3.1 -3.5 4.7 297 -6.6 -7 6.1
Table D2. Responses to Damage, Abatement, and Pandemic Questions by Self-Reported Expertise
Question Scenario
Damage Expertise Adaptation Expertise Abatement Expertise
Obs Mean Median Std.
Dev. Obs Mean Median Std.
Dev. Obs Mean Median Std.
Dev.
11 S.1 - 2025 90 -2.7 -1 4.5 90 -3.7 -2 5.1 134 -2.0 -1 2.6
11 S.1 - 2075 89 -9.2 -5 10.1 90 -11.3 -7.5 10.6 134 -8.4 -5 8.6
11 S.1 - 2130 83 -14.0 -8 12.8 84 -17.9 -15 13.3 126 -18.0 -10 17.9
11 S.1 - 2220 80 -21.9 -15 19.3 82 -26.8 -20 20.5 124 -27.0 -20 24.2
12 S.2 - 2075 66 -7.2 -5 7.0 71 -10.1 -6.7 10.0 104 -6.7 -3.75 7.3
12 S.3 - 2075 65 -10.6 -6 9.7 70 -12.7 -8 11.4 100 -8.7 -5.5 8.2
12 S.4 - 2075 64 -5.6 -3 5.5 71 -9.2 -5 10.0 101 -5.3 -3 5.8
13 S.A - 2075 63 6.3 2.5 8.3 60 5.6 3 5.9 110 3.8 2 4.1
13 S.A -2130 58 7.1 3 9.6 61 8.3 5 8.9 105 5.3 3 5.5
13 S.B - 2075 57 7.6 4 9.2 60 7.8 4.5 8.8 108 5.7 3 6.6
15 GDP 88 -2.6 -3 4.3 86 -2.1 -3 4.6 155 -2.1 -3 4.2
15 GHG 87 -2.3 -3 4.0 84 -1.7 -2 4.9 152 -2.4 -3 4.6
Results above reect the trimming of outlier estimates below the 5th percentile or above the 95th percentile of total responses.
46
Table D2. Responses to Damage, Abatement, and Pandemic Questions by Self-Reported Expertise (Continued)
Question Scenario
Climate Uncertainty Expertise Scenario Modeling Expertise Optimal Policy Modeling Expertise
Obs Mean Median Std.
Dev. Obs Mean Median Std.
Dev. Obs Mean Median Std.
Dev.
11 S.1 – 2025 63 -2.0 -1 2.5 64 -1.8 -1 2.5 90 -1.9 -1 2.1
11 S.1 – 2075 62 -9.1 -7 7.3 66 -8.3 -5 7.8 88 -7.4 -6 5.7
11 S.1 – 2130 57 -20.7 -15 16.5 61 -18.8 -12 17.9 81 -15.4 -14 10.9
11 S.1 – 2220 55 -34.9 -30 25.5 60 -32.0 -22.5 27.9 79 -26.4 -20 19.7
12 S.2 – 2075 55 -9.4 -5 11.4 56 -7.5 -5 8.5 66 -7.2 -5 6.7
12 S.3 – 2075 51 -12.2 -10 12.2 53 -10.4 -7 9.7 66 -10.0 -6 8.8
12 S.4 – 2075 52 -7.4 -4.5 8.3 54 -6.6 -3.5 8.0 65 -6.1 -4 5.4
13 S.A – 2075 50 4.4 3 5.1 56 5.1 3 6.2 69 3.8 2 4.5
13 S.A -2130 45 5.2 3 6.2 54 5.7 5 5.8 66 4.9 3.25 5.2
13 S.B – 2075 47 6.6 4 7.9 54 7.1 5 8.3 67 5.4 4 5.8
15 GDP 65 -2.4 -3 4.5 76 -1.6 -2 4.8 105 -3.3 -4 4.9
15 GHG 66 -2.9 -3.25 4.3 76 -1.9 -3 5.4 103 -3.4 -4 4.7
Results above reect the trimming of outlier estimates below the 5th percentile or above the 95th percentile of total responses.
47
48
Table D2. Responses to Damage, Abatement, and Pandemic Questions by Self-Reported Expertise (Continued)
Question Scenario
Climate Policy Expertise Developing Country/Distribution Expertise
Obs Mean Median Std. Dev. Obs Mean Median Std. Dev.
11 S.1 - 2025 89 -2.4 -1 3.8 76 -3.3 -2 5.0
11 S.1 - 2075 88 -8.2 -5 8.1 75 -9.7 -5 10.3
11 S.1 - 2130 81 -15.6 -10 13.6 70 -14.3 -9 12.5
11 S.1 - 2220 80 -24.6 -20 21.4 69 -19.7 -15 17.3
12 S.2 - 2075 72 -8.4 -4 9.7 61 -7.7 -5 8.2
12 S.3 - 2075 71 -10.8 -6 11.4 58 -10.7 -6 9.6
12 S.4 - 2075 74 -8.2 -4 10.6 58 -6.3 -3 7.1
13 S.A - 2075 76 3.9 2 4.4 54 4.2 2.75 4.5
13 S.A -2130 69 5.1 3 5.6 53 6.2 4 7.7
13 S.B - 2075 72 5.0 3 5.5 52 6.1 4 6.5
15 GDP 111 -1.7 -3 4.5 79 -2.4 -3 4.8
15 GHG 108 -2.3 -3.5 5.2 77 -1.9 -2 5.1
Results above reect the trimming of outlier estimates below the 5th percentile or above the 95th percentile of total responses.
49
Table D3. Responses to Damage, Abatement, and Pandemic Questions, by Number of Qualifying Publications
Question Scenario
Respondents with Multiple Qualifying Publications Respondents with One Qualifying Publication
Obs Mean Median Std. Dev. Obs Mean Median Std. Dev.
11 S.1 - 2025 95 -1.5 -1 2.2 187 -2.9 -2 3.8
11 S.1 - 2075 91 -7.1 -5 6.4 181 -9.1 -6 8.1
11 S.1 - 2130 86 -14.5 -9 13.0 171 -17.6 -14 14.5
11 S.1 - 2220 85 -20.9 -12 17.3 168 -27.1 -20 22.0
12 S.2 - 2075 62 -6.8 -3 8.1 148 -7.8 -5 7.2
12 S.3 - 2075 62 -10.3 -6 10.6 145 -10.1 -7 8.6
12 S.4 - 2075 61 -6.1 -4 6.7 145 -6.7 -5 6.3
13 S.A - 2075 70 3.3 2 3.6 133 4.8 3 4.8
13 S.A -2130 69 6.0 3 9.1 128 6.6 5 6.3
13 S.B - 2075 70 4.7 3 4.8 128 6.9 4 7.5
15 GDP 110 -2.8 -3 4.0 201 -2.9 -3.5 4.2
15 GHG 105 -2.8 -3 4.5 201 -3.2 -4 4.7
Results above reect the trimming of outlier estimates below the 5th percentile or above the 95th percentile of total responses.
50
Table D4. Responses to Damage, Abatement, and Pandemic Questions by Category of Qualifying Journal Publication
Question Scenario
Respondents from
Economics Journals
Respondents from Environmental
Economics Journals
Respondents from Development
Economics Journals
Obs Mean Median Std.
Dev. Obs Mean Median Std.
Dev. Obs Mean Median Std.
Dev.
11 S.1 - 2025 56 -1.2 -1 1.2 206 -2.8 -1.5 4.0 19 -2.7 -2 2.8
11 S.1 - 2075 56 -5.3 -4 4.7 204 -10.2 -7 9.6 19 -8.5 -5 9.2
11 S.1 - 2130 51 -11.8 -7 11.0 192 -18.7 -15 16.0 18 -12.6 -10.5 10.6
11 S.1 - 2220 48 -19.0 -12 15.6 189 -28.1 -20 23.1 18 -17.0 -17 12.7
12 S.2 - 2075 39 -4.8 -4 3.9 158 -8.8 -5 8.7 16 -9.9 -4.25 16.3
12 S.3 - 2075 36 -7.0 -5 5.7 156 -11.5 -7 10.8 16 -10.7 -6 10.6
12 S.4 - 2075 36 -4.6 -4 3.0 156 -6.7 -4.5 6.6 16 -8.1 -2.6 12.1
13 S.A - 2075 38 2.8 2 2.7 155 4.8 3 5.1 12 3.6 2 3.6
13 S.A -2130 35 3.4 3 3.1 148 6.8 4.75 7.3 12 5.5 3 6.3
13 S.B - 2075 36 3.4 3 2.3 152 7.5 4.25 8.6 12 5.4 3.75 5.9
15 GDP 64 -2.2 -3 4.3 223 -3.1 -3 4.1 22 -3.0 -4 4.4
15 GHG 59 -0.9 -2 5.2 215 -3.6 -4 4.3 21 -3.4 -4 3.8
Results above reect the trimming of outlier estimates below the 5th percentile or above the 95th percentile of total responses.
Respondents om Economics Journals - Had at least one qualifying publication in a top economics journal
Respondents om Environmental Economics Journals - Had at least one qualifying publication in a top environmental economics journal and NO qualifying
publications in a top economics journal.
Respondents om Development Economics Journals - Had at least one qualifying publication in a top development economics journal and NO qualifying
publications in a top economics or environmental economics journal.
51
References
Anderson, M., Richardson, J., McKie, J., Iezzi, A., & Khan, M. (2011). e relevance of personal characteristics in health
care rationing: what the Australian public thinks and why. American journal of economics and sociology, 70(1), 131-151.
Ansolabehere, S., & Konisky, D. M. (2014). Cheap and clean: how Americans think about energ y in the age of global warming.
MIT Press.
Antho, D., Hepburn, C., & Tol, R. S. (2009). Equity weighting and the marginal damage costs of climate change.
Ecological Economics, 68(3), 836-849.
Armstrong, J.S. (2001). Combining Forecasts. In J.S. Armstrong (Ed.). Principles of forecasting: A handbook for researchers
and practitioners (pp. 417-440). Springer Science & Business Media.
Bauer, M., & Rudebusch, G. D. (2020). e Rising Cost of Climate Change: Evidence om the Bond Market (Working Paper
Series). Federal Reserve Bank of San Francisco. hps://papers.ssrn.com/sol3/papers.cfm?abstract_id=3649958.
[BOEM,2015] Industrial Economics, Inc. (2015). Consumer Surplus and Energy Substitutes for OCS Oil and Gas
production: e 2015 Revised Market Simulation Model (MarketSim). U.S. Department of the Interior, Bureau of
Ocean Energy Management. OCS Study BOEM 2015-054.
Burke, M., Hsiang, S. M., & Miguel, E. (2015). Global non-linear eect of temperature on economic production. Nature,
527(7577), 235-239.
Burke, M., Davis, W. M., & Dienbaugh, N. S. (2018). Large potential reduction in economic damages under UN
mitigation targets. Nature, 557(7706), 549-553.
U.S. Council of Economic Advisers. (2017, January). Discounting for public policy: eory and recent evidence on the
merits of updating the discount rate. hps://obamawhitehouse.archives.gov/sites/default/les/page/les/201701_cea_
discounting_issue_brief.pdf.
Carbon180. (2021). e Direct Air Capture Map of Actors, Plants, and Projects [map]. hps://carbon180.org/dac-
mapp.
Center for Climate and Energy Solutions (2021). U.S. State Greenhouse Emissions Targets [map]. hps://www.c2es.
org/document/greenhouse-gas-emissions-targets/.
Climate Action Tracker. (2021). 2100 warming projections (December 2020 update) [Data set]. hps://
climateactiontracker.org/documents/830/CAT_2020-12_PublicData_EmissionPathways.xlsx.
Colson, A. R., & Cooke, R. M. (2018). Expert elicitation: using the classical model to validate experts’ judgments. Review
of Environmental Economics and Policy, 12(1), 113-132.
Dienbaugh, N. S. (2020). Verication of extreme event aribution: Using out-of-sample observations to assess changes
in probabilities of unprecedented events. Science advances, 6(12), eaay2368.
Dienbaugh, N. S., & Burke, M. (2019). Global warming has increased global economic inequality. Proceedings of the
National Academy of Sciences, 116(20), 9808-9813.
52
Drupp, M. A., Freeman, M. C., Groom, B., & Nesje, F. (2018). Discounting disentangled. American Economic Journal:
Economic Policy, 10(4), 109-34.
Energy & Climate Intelligence Unit (2021). Net Zero Tracker: 2020 Scorecard. hps://eciu.net/netzerotracker.
Fan, W., & Yan, Z. (2010). Factors aecting response rates of the web survey: A systematic review. Computers in Human
Behavior, 26(2), 132-139.
Freeman, M. C., & Groom, B. (2015). Positively gamma discounting: combining the opinions of experts on the social
discount rate. e Economic Journal, 125(585), 1015-1024.
Gillingham, K., & Stock, J. H. (2018). e cost of reducing greenhouse gas emissions. Journal of Economic Perspectives,
32(4), 53-72.
Gigone, D., & Hastie, R. (1997). Proper analysis of the accuracy of group judgments. Psychological Bulletin, 121(1), 149.
Hänsel, M. C., Drupp, M. A., Johansson, D. J., Nesje, F., Azar, C., Freeman, M. C., Groom, B., & Sterner, T. (2020).
Climate economics support for the UN climate targets. Nature Climate Change, 10(8), 781-789.
Heal, G. M., & Millner, A. (2014). Agreeing to disagree on climate policy. Proceedings of the National Academy of Sciences,
111(10), 3695-3698.
Holladay, J.S., Horne, J., & Schwartz, J.A. (2009). Economists and climate change: Consensus and open questions. Institute
for Policy Integrity. hp://policyintegrity.org/les/publications/EconomistsandClimateChange.pdf.
Howard, P. H. (2019). e social cost of carbon: capturing the costs of future climate impacts in US policy. In Managing
Global Warming (pp. 659-694). Academic Press.
Howard, P. H., & Sterner, T. (2017). Few and not so far between: a meta-analysis of climate damage estimates.
Environmental and Resource Economics, 68(1), 197-225.
Howard, P. H., & Sylvan, D. (2015). e Economic Climate: Establishing Expert Consensus on the Economics of Climate
Change. Institute for Policy Integrity.
Howard, P. H., & Sylvan, D. (2020). Wisdom of the experts: Using survey responses to address positive and normative
uncertainties in climate-economic models. Climatic Change, 162(2), 213-232.
Hsiang, S., Kopp, R., Jina, A., Rising , J., Delgado, M., Mohan, S., Rasmussen, D.J., Muir-Wood, R ., Wilson, P., Oppenheimer,
M., Larsen, K., & Houser, T. (2017). Estimating economic damage from climate change in the United States. Science,
356(6345), 1362-1369.
International Energy Agency, Nuclear Energy Agency, & Organisation for Economic Co-Operation and Development.
(2020). Projected Costs of Generating Electricity. IEA. hps://www.iea.org/reports/projected-costs-of-generating-
electricity-2020.
Intergovernmental Panel on Climate Change Working Group III. (2014). Climate Change 2014: Mitigation of Climate
Change. In O. Edenhofer, R. Pichs-Madruga, Y. Sokona, E. Farahani, S. Kadner, K . Seyboth, A. Adler, I. Baum, S. Brunner,
P. Eickemeier, B. Kriemann, J. Savolainen, S. Schlömer, C. von Stechow, T. Zwickel & J.C. Minx (Eds.). Fih Assessment
Report of the Intergovernmental Panel on Climate Change. Cambridge University Press.
53
[IPCC, 2018] Intergovernmental Panel on Climate Change (2018). Global Warming of 1.5°C. An IPCC Special Report
on the impacts of global warming of 1.5°C above pre-industrial levels and related global greenhouse gas emission
pathways, in the context of strengthening the global response to the threat of climate change, sustainable development,
and eorts to eradicate poverty. Masson-Delmoe, V., P. Zhai, H.-O. Pörtner, D. Roberts, J. Skea, P.R. Shukla, A. Pirani,
W. Moufouma-Okia, C. Péan, R. Pidcock, S. Connors, J.B.R. Mahews, Y. Chen, X. Zhou, M.I. Gomis, E. Lonnoy, T.
Maycock, M. Tignor, and T. Watereld (Eds.).
[IRENA, 2019] Taylor, M., Ralon, P., Anuta, H., & Al-Zoghoul, S. (2020). IRENA Renewable Power Generation Costs in
2019. International Renewable Energy Agency.
[IWG, 2010] Interagency Working Group on Social Cost of Carbon. (2013, May) Technical Support Document: Technical
Update of the Social Cost of Carbon for Regulatory Impact Analysis. hps://obamawhitehouse.archives.gov/sites/default/
les/omb/assets/inforeg/technical-update-social-cost-of-carbon-for-regulator-impact-analysis.pdf.
[IWG, 2021] Interagency Working Group on Social Cost of Carbon. (2021, February). Technical Support Document:
Social Cost of Carbon, Methane, and Nitrous Oxide Interim Estimates under Executive Order 13990. hps://www.
whitehouse.gov/wp-content/uploads/2021/02/TechnicalSupportDocument_SocialCostofCarbonMethaneNitrou-
sOxide.pdf
Kalaitzidakis, P., Mamuneas, T. P., & Stengos, T. (2003). Rankings of academic journals and institutions in economics.
Journal of the European Economic Association, 1(6), 1346-1366.
Kalaitzidakis, P., Mamuneas, T. P., & Stengos, T. (2011). An updated ranking of academic journals in economics.
Canadian Journal of Economics/Revue canadienne d'économique, 44(4), 1525-1538.
Kavlak, G., McNerney, J., Trancik, J. (2018). Evaluating the causes of cost reduction in photovoltaic modules. Energy
Policy (123), 700-710.
Lazard. (2019). Lazard’s Levelized Cost of Energy Analysis—Version 14.0. hps://www.lazard.com/media/451419/
lazards-levelized-cost-of-energy-version-140.pdf.
Lemoine, D., & McJeon, H. C. (2013). Trapped between two tails: trading o scientic uncertainties via climate targets.
Environmental Research Leers, 8(3), 034019.
Lea, M., & Tol, R. S. (2019). Weather, climate and total factor productivity. Environmental and Resource Economics,
73(1), 283-305.
Li, Q., & Pizer, W. A. (2021). Use of the consumption discount rate for public policy over the distant future. Journal of
Environmental Economics and Management, 102428.
Lorenz, J., Rauhut, H., Schweitzer, F., & Helbing, D. (2011). How social inuence can undermine the wisdom of crowd
eect. Proceedings of the national academy of sciences, 108(22), 9020-9025.
Manfreda, K. L., Bosnjak, M., Berzelak, J., Haas, I., & Vehovar, V. (2008). Web surveys versus other survey modes: A
meta-analysis comparing response rates. International journal of market research, 50(1), 79-104.
McKibben, B. (2007). Warning on warming [Review of the IPCC Fourth Assessment Report]. The New York Review of
Books, 15, 44-45. hp://www.nybooks.com/articles/19981.
54
[MIT, 2008] Ansolabehere, S. (2008). Public Aitudes Toward America’s Energy Options: Report of the 2007 MIT
Energy Survey. MIT Center for Energy and Environmental Policy Research. hp://dspace.mit.edu/bitstream/
handle/1721.1/45068/2007-002.pdf?sequence=1.
National Academies of Sciences, Engineering, and Medicine. (2017). Valuing climate damages: Updating estimation of the
social cost of carbon dioxide. National Academies Press.
Newell, R. G., Prest, B. C., & Sexton, S. (2018). e GDP-temperature relationship: implications for climate change damages
(RFF Working Paper 18-17). Resources for the Future.
Nordhaus, W. D. (1994). Expert opinion on climatic change. American Scientist, 45-51.
Nordhaus, W. D. (2017). Revisiting the social cost of carbon. Proceedings of the National Academy of Sciences, 114(7),
1518-1523.
Nordhaus, W. (2018). Evolution of modeling of the economics of global warming: Changes in the DICE model, 1992–
2017. Climatic change, 148(4), 623-640.
Nordhaus, W. (2019). Climate change: the ultimate challenge for economics. American Economic Review, 109(6),
1991-2014.
Nordhaus, W., & Sztorc, P. (2013). DICE 2013R: Introduction and user’s manual. Home Page of William D. Nordhaus.
Accessed October, 11, 2017.
Oppenheimer, M., Lile, C. M., & Cooke, R. M. (2016). Expert judgement and uncertainty quantication for climate
change. Nature climate change, 6(5), 445-451.
Oreskes, N., Oppenheimer, M., & Jamieson, D. (2019). Scientists have been underestimating the pace of climate change.
Scientic American, 19(08). hps://blogs.scienticamerican.com/observations/scientists-have-been-underestimating-
the-pace-of-climate-change/.
Pindyck, R. S. (2015). e Use and Misuse of Models for Climate Policy (NBER Working Paper 21097). National Bureau
of Economic Research.
Pindyck, R. S. (2017). e use and misuse of models for climate policy. Review of Environmental Economics and Policy,
11(1), 100-114.
Pindyck, R. S. (2019). e social cost of carbon revisited. Journal of Environmental Economics and Management, 94, 140-
160.
Rahmstorf, S., Cazenave, A., Church, J. A., Hansen, J. E., Keeling, R. F., Parker, D. E., & Somerville, R. C. (2007). Recent
climate observations compared to projections. Science, 316(5825), 709-709.
Rahmstorf, S., Foster, G., & Cazenave, A. (2012). Comparing climate projections to observations up to 2011.
Environmental Research Leers, 7(4), 044035. hp://iopscience.iop.org/article/10.1088/1748-9326/7/4/044035/pdf.
[RFF, 2015] Resources for the Future, e New York Times, & Stanford University. (2015). Global Warming National
Poll (Part III) [Data set]. Resources for the Future. hps://media.r.org/documents/RFF-NYTimes-Stanford-global-
warming-poll-Jan-2015-topline-part-3.pdf?_ga=2.68744663.791000388.1616068841-1400195777.1606942801.
55
Rogelj, J., Huppmann, D., Krey, V. et al. (2019). A new scenario logic for the Paris Agreement long-term temperature
goal. Nature 573, 357–363. hps://doi.org/10.1038/s41586-019-1541-4
Roughgarden, T., & Schneider, S. H. (1999). Climate change policy: quantifying uncertainties for damages and optimal
carbon taxes. Energy Policy, 27(7), 415-429.
Rousseau, S. (2008). Journal evaluation by environmental and resource economists: A survey. Scientometrics, 77(2),
223-233.
Rousseau, S., Verbeke, T., & Rousseau, R. (2009). Evaluating environmental and resource economics journals: A TOP-
curve approach. Review of Environmental Economics and Policy, 2009, 270-287.
Schauer, M. J. (1995). Estimation of the greenhouse gas externality with uncertainty. Environmental and Resource
Economics, 5(1), 71-82.
Sheehan, K. B. (2001). E‐mail survey response rates: A review. Journal of Computer‐Mediated Communication, 6(2), 0-0.
Sisco, M. R., Bosei, V., & Weber, E. U. (2017). When do extreme weather events generate aention to climate change?.
Climatic change, 143(1), 227-241.
Surowiecki, J. (2004). e Wisdom Of Crowds. Knopf Doubleday Publishing Group.
Sunstein, C. R. (2005). Group judgments: Statistical means, deliberation, and information markets. NYU Law Review,
80, 962. hp://www.nyulawreview.org/sites/default/les/pdf/NYULawReview-80-3-Sunstein.pdf.
Tol, R. S. (2009). e economic eects of climate change. e Journal of Economic Perspectives, 29-51.
Tol, R. S. (2018). e economic impacts of climate change. Review of Environmental Economics and Policy, 12(1), 4-25.
United Nations (2021). Inequality – Bridging the Divide. hps://www.un.org/en/un75/inequality-bridging-divide.
Weitzman, M. L. (2001). Gamma discounting. American Economic Review, 260-271.
Weyant, J. (2017). Some contributions of integrated assessment models of global climate change. Review of Environmental
Economics and Policy, 11(1), 115-137.
56
Endnotes
1 Parts of this discussion and the section that follows are
adapted (with updates) from the Institute for Policy Integ-
rity’s publications on past surveys of economists. See Howard
and Sylvan (2015; 2020) and Holladay et al. (2009).
2 For example, New York State’s Department of Environmental
Conservation cited Drupp et al. (2018) as primary evidence
to justify its use of a 2% discount rate in the calculation of its
central Value of Carbon. Similarly, the U.S. government’s In-
teragency Working Group on the Social Cost of Greenhouse
Gases (2021) cited Drupp et al. (2018) and Howard and Syl-
van (2020) as evidence supporting their interim recommen-
dation to allow agencies to apply discount rates below 2.5%
when valuing climate impacts. Outside of the climate context,
expert elicitation has been used by several agencies: the U.S.
Environmental Protection Agency developed concentration-
response functions for mortality from particulate maer
exposure using expert elicitation (Howard, 2019); and the
Department of Interior relied on the expert opinion of Dr.
Stephen Brown of UNLV to develop a multitude of parame-
ters for its MarketSim model when estimates were unavailable
in the literature (BOEM, 2015).
3 In particular, the Condorcet Jury eorem states that the
probability of a correct answer by a majority of the group
increases toward certainty as the size of the group increases,
if each individual person is more likely than not to be correct
(Surowiecki, 2004).
4 Specically, Pindyck (2015) states that “the ad hoc equations
that go into most IAMs are no more than reections of the
modeler’s own ‘expert’ opinion…determining plausible
outcomes and probabilities, and the emission reductions
needed to avert these outcomes, would mean relying
on ‘expert’ opinion. For an economist, this is not very
satisfying…But remember that the inputs to IAMs (equations
and parameter values) are already the result of ‘expert’
opinion; in this case the modeler is the ‘expert’…In eect, we
would use expert opinion to determine the inputs to a simple,
transparent and easy-to-understand model.”
5 See Resources for the Future. e Social Cost of Carbon Initia-
tive. hps://www.r.org/topics/scc/social-cost-carbon-initia-
tive/.
6 We use 1994 as the rst year for inclusion as Tol (2009)
notes that modern climate economics began to take shape
in approximately that year. Given that a starting date of 1994
implies the inclusion of some publications 25 years or older,
in future work we may analyze authors based on the timing of
their publications.
7 For our 2015 survey, we included economics journals ranked
in the top 25 according to Kalaitzidakis et al. (2003) or Kalait-
zidakis et al. (2011), as well as the top six environmental eco-
nomics journals according to Rousseau (2008) or Rousseau et
al. (2009). For our 2021 survey, we update these 37 journals
to include papers published from 2015 to 2020. Additionally,
we expanded our environmental economics journal category
to include the Journal of the Association of Environmental and
Resource Economists and Review of Environmental Economics
and Policy due to their rise in prominence since 2015. Finally,
we included seven development economics journals based on
Kalaitzidakis et al. (2011), the World Bank ranking, and Google
Scholar.
8 We restricted our sample for this survey to economists, remov-
ing the authors of qualifying articles who did not have a Ph.D.
in economics or work as an economist/economics professor.
We conducted this screening to the best of our ability, though
we were not able to nd credentials for every respondent. In our
2015 survey, we chose to include all those who had authored a
qualifying article in an economics journal, even if their creden-
tials were in another discipline or they had not received a Ph.D.
9 e Holladay et al. (2009) survey used a smaller respondent
pool, focusing only on the top 25 “standard” economics
journals. at survey was sent to 289 experts, receiving 144
responses.
10 We expanded our selection criteria to include development
journals in part based on concerns that our 2015 survey may
have overly reected the views of scholars in wealthy countries
and, subsequently, climate impacts in North America and
Europe.
11 ough the academic literature does not clearly dene what
constitutes an “acceptable” response rate (Anderson et al.,
2011), our general response rate was roughly in line with the
average rate for online surveys in recent periods. Our overall
eective response rate (RR6) is slightly lower than the 37% av-
erage found across 31 studies summarized in Sheehan (2001).
However, there is strong evidence that e-mail survey response
rates have been declining over time (Sheehan, 2001; Fan &
Yan, 2010). For example, Sheehan (2001)’s response rates over
the 1998 and 1999 period average to 31%; these numbers are
slightly lower than our response rates in this survey. Similarly,
Manfreda et al. (2008) nd that the average response rate for
45 web surveys was 11% (Fan & Yan, 2010). Our response rates
are in line with or above the averages from these studies.
12 For example, the group that received the survey included
authors who proposed an economic model that predicted a
potentially positive eect on global agriculture from climate
change, and others who subsequently criticized that model and
approach.
13 We identied authors by whether they had published one or
multiple articles that met our selection criteria, and whether
they published in economics, environmental economics, or de-
velopment economics journals. We sent these groups dierent
links to identical surveys so that we could test for dierences
between groups.
57
14 Following Howard and Sylvan (2020), we regress the re-
sponse to each question on two indicator variables: the rst
equal to one if the respondent replied between the rst e-mail
reminder and the second e-mail reminder, and the second
equal to one if the respondent replied aer the second e-mail
reminder. Only two respondents submied responses aer
the close date, so they were grouped with those who respond-
ed aer the second e-mail reminder. e tests for whether
the coecients for each indicator variable are individually
and jointly signicant imply no response bias when we fail to
reject the null hypothesis of equaling zero. We reject response
bias in all cases for the standard signicance rate of 5%.
15 Newell et al. (2018) questions the identication assump-
tions in Burke et al. (2015; 2018), nding that climate change
has no statistical impact on economic growth. Lea and Tol
(2019) analyze the impact of climate change on total factor
productivity nding that climate change has a negative eect
on the growth rate of poor countries only.
16 See Dienbaugh and Burke (2019); Tol (2018).
17
Based on Climate Action Tracker (2021) data, emission cuts
of 70% to 86% (relative to business-as-usual) are necessary by
2050 to meet a 2°C target, while cuts of 86% to 95% are neces-
sary to reach 1.5°C. While Climate Action Tracker projections
align closely with IPCC (2018, p. 12) in the short-run (i.e.,
2030), IPCC predicts that net-zero emissions are necessary by
2050 and 2070 to reach the 1.5°C and 2°C targets, respective-
ly. Climate Action Tracker projects net-zero deadlines of 2069
and post-2100 for these targets. Specically, IPCC (2018, p.
15) argues with high condence that renewables will need to
supply 70–85% (interquartile) of electricity in 2050 to meet
the 1.5°C target (with a full range of 59% to 97%).
18 As a point of reference, only three direct air capture facilities
are operational globally beyond the small-scale demonstration
scale as of March 2021 (Carbon180, 2021).
19 Hansel et al. (2020) states, “A key dierence between DICE
and the IPCC Special Report is the stance regarding the avail-
ability of carbon removal technologies leading to net-negative
emissions. While the scenarios considered by the IPCC make
use of negative-emissions technologies roughly by the year
2050, the DICE-2016R2 model assumes that this will only be
feasible from 2160 onwards.”
20 For instance, Climate Action Tracker (2021) shows that
negative-emissions technologies must be deployed at a large
scale by 2064 to 2088 to reach a 1.5°C limit and by 2089 to
reach 2°C limit in the worst-case scenario. Similarly, the IPCC
assumes negative-emissions technologies will be used by
2050.
21 is range appears to far exceed Burke et al. (2015)’s range
as represented in their Figure 5a. However, the authors state
that “At levels of Warming below 2°C, our estimates are at
least 3 times higher than IAM estimates, typically 5-20 times
higher, and sometimes up to 100X higher. At higher levels of
warming, our estimates are again at least 2.4 times larger than
the highest IAM estimate and typically are much larger.” eir
range of 5 to 20 times the DICE-2016R central estimate, as
indicated by this statement, suggests a wide range of -1.7%
to -6.8%.
22 Our untrimmed mean and median from the 2021 survey
are -10.7% and -5%, which match quite closely the values of
-10.2% and -5.5%, respectively, from our 2015 survey (How-
ard & Sylvan, 2020). If we use 99th percentile trimming on
the 2021 data, we nd damage estimates of -9.8% and -5%,
which again matches quite closely our corresponding 2015
survey’s estimate of -9.2% and -5%.
23 Calibrating DICE-2016R2’s carbon price to replicate these
temperature limits, we nd that DICE-2016R2’s abatement
cost estimates are far below our survey results: 0.9% and
0.8% to reach 2.5°C in 2075 and 3.5°C in 2130, respectively,
consistent with Scenario A; and 1.7% to reach 2°C by 2075.
However, our median estimate of the 5th percentile response
is roughly consistent with DICE-2016R: 1% of GDP for
2.5°C in 2075, 1% of GDP for 3.5°C in 2130, and 2% for 2°C
by 2075. As total abatement cost estimates vary signicantly
by model (IPCC WGIII, 2014, p. 39), our cost estimates
are consistent with the upper end of the IPCC range for
corresponding abatement scenarios. One confounding issue
is that our survey respondents were quite optimistic about
mitigation technology. However, our cost estimates are more
consistent with the unavailability of negative-emissions
technologies.
24 When we trim at the 95th percentile, our trimming is not
symmetric. e 2.75% estimate reects the median of the
untrimmed data.
25 For climate damages, the results are less clear, as experts
with the most relevant expertise (publishing on climate
damages, publishing on adaptation, and those identied
based on multiple publications) are inconsistent in the
direction by which they believe the full sample is potentially
biased with respect to their damage forecasts. See Appendix
D.
26 Similarly, the failure of the models to capture the slow pace
of climate policy implementation to date, combined with
potential technology limits (particularly resulting from
insucient R&D) could lead to a further downward bias in
the IPCC’s mitigation cost estimates. is partially explains
why Weyant (2017) argues that a target of “550 ppm [this
century] will likely cost somewhere between 0.1 percent
and 10 percent of gross world product (GWP) per year”
puing our survey estimates well within this wide potential
range.
27 Future modeling is necessary to determine this assertion,
as calibration of DICE-2016R2’s climate model currently
makes reaching 2°C impossible, contrary to scientic con-
sensus (Howard & Sylvan, 2020).
28 Using this data, Howard and Sylvan (2020) found the prob-
ability of a 25% loss of GDP for a 3°C increase to be 9.2%
(mean) and 5.0% (median). is is still slightly lower than
our 2021 survey ndings.
Institute for Policy Integrity
New York University School of Law
Wilf Hall, 139 MacDougal Street, New York, New York 10012
policyintegrity.org
NEW YORK UNIVERSITY SCHOOL OF LAW
