Auffhammer QuantifyingEconomicDamages 2018
Auffhammer QuantifyingEconomicDamages 2018
Auffhammer QuantifyingEconomicDamages 2018
REFERENCES
Linked references are available on JSTOR for this article:
https://www.jstor.org/stable/10.2307/26513495?seq=1&cid=pdf-
reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
American Economic Association is collaborating with JSTOR to digitize, preserve and extend access
to The Journal of Economic Perspectives
Maximilian Auffhammer
C
limate scientists have spent billions of dollars and eons of supercom-
puter time studying how increased concentrations of greenhouse gases
and changes in the reflectivity of the earth’s surface affect dimensions
of the climate system relevant to human society: surface temperature, precipita-
tion, humidity, and sea levels. Recent incarnations of physical climate models have
become sophisticated enough to be able to simulate intensities and frequencies
of some extreme events, like tropical storms, under different warming scenarios.
The current consensus estimates from what may be the most heavily p eer-reviewed
scientific publication in human history, the 5th Assessment Report of the Inter-
governmental Panel on Climate Change, are that the average global surface
temperature has increased by 0.85° Celsius (1.5° F) since the industrial revolution.
Estimates of future warming by the end of the current century range from 0.9 to
5.4°C (1.6 - 9.7°F) (IPCC 2013; Hsiang and Kopp in this issue of the journal).
In a stark juxtaposition, the efforts involved in and the public resources targeted
at understanding how these physical changes translate into economic impacts are
disproportionately smaller, with most of the major models being developed and
maintained with little to no public funding support. This is concerning, because
optimal policy design in the context of addressing the biggest environmental market
failure in human history requires an understanding of the external cost imposed
Figure 1
Sample of Social Cost of Carbon Estimates Used in Federal Rulemakings
2016 IWG
2014 EPA
2010 EPA/NHTSA/DOE
2009 EPA
2009 NHTSA
2008 EPA
2008 DOE
0 20 40 60 80 100
Social cost of carbon (2010, 2007$/tCO2)
Sources: Rose (2012); Rose et al. (2014); IWG (2016); EPA (2018).
Note: Estimates for the social cost of carbon are for emissions of a ton of CO2 in 2010 in 2007 dollars.
NHTSA is National Highway Traffic Safety Administration; IWG is Interagency Working Group; EPA
is Environmental Protection Agency; DOE is Department of Energy. The black diamond indicates the
“central estimate,” if one was identified. The grey bars indicate selected upper and lower bounds used
in regulatory analyses.
more ton of emissions then becomes the social cost of carbon—essentially the
external cost of one ton of additional CO2 emissions at a point in time. There is a
nascent literature calculating social costs of other greenhouse gases (for example,
methane is a more potent greenhouse gas, but with a shorter atmospheric lifetime).
A tremendous number of modeling assumptions need to be made to calculate
the social cost of carbon for use in rulemaking. The modeler needs to decide on
the time horizon to be considered, the approach to discounting and the rate to be
used, the reflection of uncertainties, the changes to risks, which impacts can be
included, the choice of reference conditions, whether one should equity weight
across countries, and what recent literature should be incorporated (Rose 2012).
Among these, the three factors of possibly biggest consequence are the choice of
discount rate, which sectors are omitted (for example, ecosystem services), and
whether one should consider only domestic or global damages. The latter decision
is really a legal question, as the externality is global and hence, from an economic
point of view, the global number is the correct estimate of the externality. Figure 1
shows the evolution of the social cost of carbon for a ton emitted in 2010 (measured
in 2007 US dollars) in federal rulemaking for a sample of rules.
The first official estimates of the social cost of carbon in 2008 were made under
the Bush administration. The 2008 National Highway and Traffic Safety Adminis-
tration (NHTSA) number was an estimate of global damages used for setting fuel
economy standards. The 2008 Department of Energy (DOE) number was a global
social cost of carbon used for setting air conditioner equipment and gas range stan-
dards. The 2008 Environmental Protection Agency (EPA) estimates were used in
the proposed rulemaking for regulating greenhouse gas emissions under the Clean
Air Act. The bar here indicates the distribution of the central number used. The
actual analysis also considered an additional range from −$7 to $781. It is note-
worthy that this first round of proposed rulemaking under the Bush administration
stated that CO2 is a global pollutant and that “economic principles suggest that
the full costs to society of emissions should be considered in order to identify the
policy that maximizes the net benefits to society, i.e., achieves an efficient outcome
(Nordhaus, 2006).” The document further acknowledges that “domestic estimates
omit potential impacts on the United States (for example, economic or national
security impacts) resulting from climate change impacts in other countries” (US
EPA 2008).
President Obama convened an Interagency Working Group, which was charged
with calculating an official social cost of carbon to be used across the board in
federal rulemaking (Greenstone, Kopits, and Wolverton 2013). Three prominent
Integrated Assessment Models—Nordhaus’s DICE model,1 Anthoff and Tol’s FUND
model,2 and Hope’s proprietary PAGE model—were used to calculate a distribution
of the social cost of carbon across time and scenarios for a set of common socioeco-
nomic assumptions, discount rates, and uncertainty over a number of parameters.
The central and often-cited estimate of the social cost of carbon, which is the mean
number across 50,000 simulations for each model at a 3 percent discount rate,
is $42 (in 2007 dollars) for one ton of emissions made in the year 2020.3 If one
uses a 5 percent discount rate, this value drops to $12; if one uses a 2.5 percent
discount rate, it increases to $62. The Interagency Working Group also ran a
so-called “high-impact scenario,” which is the 95th percentile number at a 3 percent
discount rate and valued at $123. The central estimate of the social cost of carbon
was projected to rise to $50/ton in 2030 and $69 in 2050.
The Obama administration later commissioned the National Academies of
Sciences to assess the Interagency Working Group modeling exercise and suggest
improvements. The National Academies of Sciences (2017) recommended substan-
tial revisions to the way the social cost of carbon is estimated. President Trump,
however, disbanded the Interagency Working Group, which could have imple-
mented these changes. Two current proposed rulemakings under the Trump
administration use a social cost of carbon that only considers domestic damages and
discount rates of 3 percent and 7 percent.
1
The DICE model is at https://sites.google.com/site/williamdnordhaus/dice-rice.
2
The FUND model is at http://www.fund-model.org.
3
Of course, 42 is also the Answer to the Ultimate Question of Life, the Universe, and Everything,
according to the The Hitchkiker’s Guide to the Galaxy.
The top bar in Figure 1 indicates the range of the domestic social cost of
carbon using the 3 percent and 7 percent discount rates currently proposed by the
National Highway Traffic Safety Administration for the “revision” of the Corporate
Average Fuel Economy (CAFE) standards for fuel economy of cars and light trucks,
which clearly represent a drastic decrease in the estimated externality to between $1
and $7 for a ton emitted in 2020.
The estimates also do not incorporate any of the major updates suggested by
the National Academies of Sciences (2017) report, which implies that the 2018
estimates do not represent best available science. For example, the National Acad-
emies of Sciences made suggestions relating to how one constructs a baseline future
economy out to the year 2300, assumptions made in the climate modeling, and the
discounting approach taken. Maybe most importantly for the purposes of this paper,
the National Academies of Sciences report points a stern finger at the damage func-
tions used in all three Integrated Assessment Models.
The damage functions in the Integrated Assessment Models, which are used to
calculate the social cost of carbon, are outdated. Greenstone (2016) points out that
the most recent studies in the FUND model stem from 2009, with the majority of the
literature cited stemming from the early and m id-1990s. For example, the damage
function for agriculture in the FUND model implies that warming up to roughly 5°C
produces benefits for the sector (Rose et al. 2014). This is not consistent with the
recent literature on agricultural impacts, which for example, points at the signifi-
cant negative impact of extreme heat days. Moore, Baldos, Hertel, and Diaz (2017)
updated the FUND damage function by incorporating the most recent empirical
estimates for agriculture and find a doubling of the social cost of carbon by simply
updating this sector alone. The literature underlying the DICE damage function also
mostly comes from studies conducted in the 1990s. None of the cites for the PAGE
model are from after 2010. As Greenstone (2016) shows, this ignores more than 100
studies published since 2010—which use more up-to-date econometric techniques
and exploit the explosion in data availability.
Figure 2
Mapping Weather into Impacts—The Importance of Accounting for Adaptation
20 20
With adaptation
15 Pre-climate-change 15
20
40
Time
60
80
100
0 5 10 15
Outcome (for example, electricity in kWh
Source: Author.
Note: The top left panel shows the weather pattern of temperature generated in two climate regimes. The
light gray time series depicts a pre-climate-change world and the dark series shows a post-climate-change
world, with a temperature series displaying higher mean and variance. The top right panel displays two
damage functions (the parabolas) which map weather into an outcome, in this case temperature into
household electricity consumption (measured in kilowatt-hours). The effect can be seen in the bottom
panel.
discuss below, a rapidly growing empirical literature uses weather variation to iden-
tify response functions that partially or fully allow for adaptation.
So how does one go about calibrating these damage functions and using them
to project damages? The question asked of any empirical economist these days is
“what would the perfect counterfactual be?” In this context, a researcher actually
needs to be concerned about two counterfactuals: 1) the counterfactual future
climate; and 2) the counterfactual for identifying the appropriate damage function.
The first counterfactual, the climatic one, asks the question: What level of climate
change will occur? Given our metaphysical inability to experiment by randomly
imposing different levels of greenhouse gases on a large sample of otherwise iden-
tical Planet Earths, researchers instead resort to computational counterfactuals of the
climate system, which are referred to as “global circulation models” (GCMs). These
models use different scenarios of greenhouse gas emissions and physical represen-
tations of the climate system to predict changes in the climate system (IPCC 2013;
Auffhammer, Hsiang, Schlenker, and Sobel 2013). They provide projections of, for
example, surface temperatures, precipitation, and sea-level rise at a reasonable level of
disaggregation and make these freely available through public depositories (Climate
Impact Lab 2018; NASA 2018). A companion paper in this symposium by Hsiang and
Kopp discusses these models and their limitations in more detail.
For the second counterfactual, we need to identify how agents in a given loca-
tion respond to weather generated from a different climate regime. As a thought
experiment, what is the right counterfactual for climate change in the United
States by end of century? The US average historical (1986–2005) June/July/August
temperature is 74°F. By end of century (under the aggressive RCP8.5 scenario), this
temperature is projected to be 84° (Climate Impact Lab 2018).
One could contemplate a number of counterfactuals that might be used. If one
has a set of units that are similar on observables and unobservables, but with different
weather due to different local climate regimes, one might use a cross-sectional
comparison. If one has long time series over a period of time where climate has
changed, one might exploit time-series variation, possibly across units, to get econo-
metric identification. But these approaches become questionable when we are
comparing places that are far apart in characteristics space. Neighboring counties
in California might possibly serve as counterfactuals for each other. However, using
the economies of Pakistan, India, Mali, and Thailand as “hotter counterfactuals”
for the United States or Europe, on the grounds that current mean temperatures in
these countries are close to 84 degrees, is a stretch.
The econometric approaches discussed below all suffer from this issue of a
fundamental lack of comparability, and I am afraid that there is no perfect way to
overcome it. Indeed, the problem is even more severe than thinking about counter-
factuals based on geographic and time-series variation would suggest. Comparing
any current day or preindustrial society to a climate-changed world 100 years from
now will be an imperfect comparison.
Many of the econometric studies I will describe below, including ones I have
authored, use a counterfactual where we impose e nd-of-century climate on today’s
economy, which is a suboptimal way to circumvent the challenge of characterizing
to the y value at point A. If faced with a significantly hotter climate, the farmer
becomes indifferent between growing crop 1 and crop 2 at point B. If climate warms
further still, the farmer would be much better off at point C, that is, switching to
crop 2, rather than at point D where the farmer continues to grow crop 1. Because
the cross-sectional regression observes optimizing farmers across the climate spec-
trum, this approach estimates the envelope of the individual crop-specific payoff
functions and allows for climate adaptation. As a result, this approach both esti-
mates a response that allows for adaptation to climate change and relies on data that
are readily available in many regions in both the developed and developing world.
It uses hotter locations as a counterfactual for the response of cooler location to
climate change.
Three main criticisms of this method have been raised. First, this c ross-sectional
approach to damage function estimation is vulnerable to omitted variables bias,
hence putting in question whether the estimates are plausibly causal. Any drivers
of land values (or net profits) that are correlated with the climate indicators and
outcome and are excluded from the model will confound the estimates of the
marginal value of climate. As one vivid illustration, Schlenker, Hanemann, and
Fisher (2005) reexamined the analysis of Mendelsohn, Nordhaus, and Shaw (1994),
and point out that irrigation is an important driver of farm profits. This was omitted
from the original regression model. When correcting for this by limiting the anal-
ysis to agricultural land east of the 100th meridian (the 100th meridian runs down
through the middle of North and South Dakota and down through the middle
of Texas) where agriculture is mostly non-irrigated, the marginal value of climate
changed significantly. The estimated impacts of climate change went from being
slightly beneficial to robustly negative.
Second, this Ricardian approach essentially assumes costless adaptation to
climate change. But switching crops is not costless (Quiggin and Horowitz 1999).
The fixed costs to switching from growing one crop to another may include invest-
ment in new harvesting equipment, irrigation infrastructure, and the acquisition of
technical know-how. If these costs are big enough, it may be optimal for the farmer
to delay or avoid change—in Figure 3, to continue farming crop 1 at point D rather
than changing to crop 2 at point C. Hence, this method may provide biased esti-
mates of the effect of climate change depending on how costly it is for farmers to
switch from one crop to the next.
Third, this framework is applied retrospectively under the assumption that only
historical climate matters. This assumption may no longer be tenable, as the climate
has been changing since the 1960s. If agents know this, they should base their actions
on expected rather than historical climate. Severen, Costello, and Deschenes (2016)
provide an interesting extension of the Ricardian method by incorporating climate
expectations. They show evidence that farmers already incorporate this information,
suggesting that failing to incorporate expectations leads to a significant underesti-
mation of projected impacts of climate change.
This cross-sectional framework has been applied in a number of other sectors.
For example, Albouy, Graff, Kellogg, and Wolff (2016) back out the marginal value
of climate in a cross-sectional study looking at residential home values. Mansur,
Figure 3
Crop Choice and Profits in the Long and Short Run
Crop 1
Crop 2
A
Value of economic activity
C
Crop 3
B
Crop 4
D
E
Mendelsohn, and Morrison (2008) use this approach to study the effects of
impacts of climate change on energy consumption, where the adaptation is not
crop-switching, but rather fuel-switching.
the distance between weather stations, the bigger measurement error concerns
become. The United States and Europe have tens of thousands of weather stations,
but many locations in in sub-Saharan Africa do not have a weather station within
hundreds of miles. If the measurement error is classical, this is likely to attenuate
the response towards zero.
into two periods and estimates the response function separately. One can then use
statistical tests to search for evidence of adaptation between the two periods. For
example, Barreca, Clay, Dechenes, Greenstone, and Shapiro (2016) examine the
mortality response to weather over time in the United States and show a massive
decrease in the effect of a hot day on mortality over time, which is due to the signifi-
cant rollout of air conditioning in the hot and often humid areas of the United
States. One example of this approach is Roberts and Schlenker (2011).
A second approach along these lines represents a marriage of the panel data esti-
mation approach using short-run weather fluctuations and the Ricardian approach.
The concept here is that if one observes a large number of units (like counties, house-
holds, or firms) over a significant number of periods covering a spatial area with large
heterogeneity in climate, one can estimate separate response functions for subgroups
of the individual units using observed short-run weather fluctuations (for example,
use within-household variation to identify a short-run response function by zip code).
By controlling for unit- and time-fixed-effects, it is possible to obtain plausibly causal
estimates of local short-run dose response functions. One can then either in a second
step regress the slopes of the dose response on climate (for example, long run average
summer temperature) across subgroups, or, through an interaction term in a single
regression, estimate how the slope of the dose response function varies across areas
with different climates, incomes, and other observables that vary across space. Sight-
ings of this approach include Bigano, Hamilton, and Tol (2007), Auffhammer and
Aroonruengsawat (2012), Hsiang and Narita (2012), Butler and Huybers (2013), Davis
and Gertler (2015); Heutel, Miller, and Molitor (2017), and Carleton et al. (2018).
This approach offers two important forward steps beyond the panel studies
discussed above. First, it explicitly models climate adaptation by exploiting
cross-sectional differences in the slopes of dose response functions. Second, it allows
us to model explicitly the effects of income and population on the damage functions.
While this approach has significant appeal, it does not overcome some of the
shortfalls of the Ricardian and panel methods. The econometrician is always limited
by using historical observations in order to parameterize equations. The best we can
do is simulate how income, population, and climate have affected short-run dose
response functions historically and to assume that this relationship remains stable.
We can approximate a future San Francisco with the climate of Fresno by assigning
the appropriate climate, income, and population, but none of these approaches
properly address the fact that Fresno may be structurally very different from a future
San Francisco—even if we assign the right income and population. We simply lack
the crystal ball that lets us look to 2100 and beyond. But this issue has plagued social
science broadly, because predicting what the world looks like 100 years out is, well,
rather difficult.
(2013, 2016, 2017). Ultimately, these criticisms raise the possibility that for studying
climate change, conventional econometric studies may need to be supplemented
with a healthy dose of “expert elicitation.”
Pindyck’s first criticism is that in Integrated Assessment Models, the functional
form of relationships and their parameterization—including those in damage
functions—are “arbitrary.” Second, he expresses concern that many of the studies
cited above “are limited to short time periods and small fluctuations in tempera-
ture and other weather variables,” which is effectively the same as pointing out that
econometricians rely on observed data and technology to parameterize their dose
response functions. In whichever way one phrases this concern, the bottom line is
that existing studies may not account well for long-term adaptation and in partic-
ular for the possibility of very significant changes in technology. Third, the biggest
impacts of climate change may result from extreme and catastrophic events, which
can be thought of as low-probability events with possibly massive economic conse-
quences. Examples would include the shutdown of the Thermohaline Circulation
that gives Europe its lovely climate, the melting of the West Antarctic ice sheet, and
the possible rapid release of significant amounts of methane from the tundra. We
have (fortunately!) not observed these events in the measured historical record and
hence econometric estimation cannot provide estimates of the economic damages
from such events. A final concern is that there is little agreement over the correct
approach to discounting and which discount rate to apply in placing a value on
future damages from climate change.
In response to these concerns, Pindyck has strongly argued for “expert elicita-
tion.” For example, in response to estimating the risks and costs of extreme climate
events, one can imagine that teams of scientists with an understanding of the phys-
ical and economic consequences might be able to provide coarse estimates of the
damages resulting from such large events. There are well-established procedures for
such expert elicitation, and this may be a fruitful avenue forward to make progress
on this topic. However, experts in this arena have to rely on “process understanding,”
as there are no data here to help. Similarly, one can imagine a group of experts
who might tackle the question of what discount rate is most appropriate to use,
which is what Drupp, Freeman, Groom, and Nesje (forthcoming) did. The median
answer for the risk-free social discount rate is 2 percent in their study, which is quite
different from the 3 percent and 7 percent rates applied in the most recently used
social cost of carbon in proposed US government rulemakings for automotive fuel
economy (CAFE) standards.
However, expert elicitation seems less useful in coming up with better esti-
mates of damages in order to overcome the first two of Pindyck’s critiques. I
would argue that the recent literature has made significant headway in esti-
mating plausibly causal damage functions incorporating adaptive response from
partially cross-sectional variation. The formulations doubtless can be critiqued
and questioned, but they are not arbitrary. I question whether experts would
come up with “better” estimates than the cutting-edge papers in this literature.
Maybe more fundamentally, a group of experts called upon to participate in an
expert elicitation exercise concerning the functional form of damage models and
Cline (1992) put forth a list of important sectors for which we require a better
understanding of their climate sensitivity. Table 1 below replicates his table and I
have subjectively filled in where this literature currently stands in terms of published
and ongoing efforts. A glance shows that there is a lot of work to do.
Yet it is clear that the literature on the econometric estimation of damage func-
tions of climate change is rapidly expanding—both in terms of methods as well as
sectoral and spatial coverage. The previously stagnant state of affairs where most
of the damage functions in Integrated Assessment Models had not been updated
significantly in over a decade has changed dramatically. Economists need to push
forward in improving sectoral and spatial coverage of the damage functions provided
to modelers, using methods that allow us to parameterize plausibly causal damage
functions, which account for adaptation and allow us to estimate welfare impacts of
climate change. The current frontier is probably best described by work using the
“Ricardo meets panel data” approach.
Moore et al. (2017) is one published attempt to incorporate the most recent
estimates of damage functions for the agricultural sector into an Integrated Assess-
ment Model (the FUND model) and this one-sector exercise doubles the social cost
of carbon (SCC), which underlines the importance of these efforts.
Those interested in this area will want to keep an eye on two major efforts that
involve ambitious ongoing collaborations between climate scientists and economists.
The Climate Impacts Lab, managed jointly by researchers at the University of Chicago,
UC Berkeley, Rutgers University, and the Rhodium Group, produces damage func-
tions for mortality, migration, energy consumption, agricultural yields, and conflict
which satisfy the characteristics laid out above and have global coverage. At the same
time, a group at Resources for the Future has undertaken the task of implementing
the changes suggested by the National Academies of Sciences in the modeling of the
social cost of carbon. The governments of Mexico and Canada have pledged their
support of these efforts, as all US federal government development of the modeling
behind the social cost of carbon has been halted—a fact which is deeply concerning.
As these and other researchers dig deeper, three key areas require especially
deep thinking. First, we need to improve how we incorporate damages from cata-
strophic events, which may well require abandoning the econometric toolkit and
relying on cross-disciplinary expert solicitation. Second, we need to think about
general equilibrium effects across space and spillover effects across sectors in our
models. Collaborations between trade and climate economists (Dingel, Meng, and
Hsiang 2018), as well as academics working on supply chains (for example, Seetharam
2018), will likely yield fruitful insights. Finally, it is shocking how little work has been
done on the effects of climate change on nonmarket goods other than mortality. It is
Table 1
Coverage of the Damage Function Literature
Plausibly
causal Adaptation Global
Sector estimates addressed coverage Examples
Morbidity and mortality Yes Yes Yes Deschênes and Greenstone (2011);
Carleton et al. (2018)
Crime and conflict Yes No Maybe Burke, Hsiang, and Miguel (2015b)
Productivity Yes No No Peng, Deschênes, Meng, and Zhang (2018)
Water consumption No No No
Pollution Yes Maybe No Bento, Mookerjee, and Severenini (2017)
Storms Yes Yes No Hsiang and Narita (2012); Deryugina,
Kawano, and Levitt (2018)
Source: Cline (1992) put forth a list of important sectors for which we require a better understanding
of their climate sensitivity. Table 1 below replicates this list and I have subjectively filled in where this
literature currently stands in terms of published and ongoing efforts.
paramount that we begin developing approaches that will allow us to quantify damages
from species loss, ecosystem services—as well as effects on human morbidity—and
incorporate these into the models that estimate costs of climate change.
■ I thank the Berkeley Climate Economics Lunch group, the members of the National Academy
of Sciences panel on the Social Cost of Carbon, David Anthoff, Lint Barrage, Marshall Burke,
Tamma Carleton, Chris Costello, Olivier Deschênes, Tony Fisher, Michael Greenstone, Michael
Hanemann, Sol Hsiang, David Lobell, Rob Mendelsohn, Pierre Mérel, Frances Moore, Michael
Roberts, Wolfram Schlenker, Joe Shapiro, Andy Solow, Richard Tol, Fiona Wilkes, Gary Yohe,
and many others for numerous conversations that have informed my thinking on the topics
discussed in this survey over the years. All misguided thinking is solely mine.
References
Adams, Richard M. 1989. “Global Climate Change on Domestic and International Tourism:
Change and Agriculture: An Economic Perspec- A Simulation Study.” Integrated Assessment Journal
tive.” American Journal of Agricultural Economics 7(1).
71(5): 1272–79. Bohra-Mishra, Pratshya, Michael Oppenheimer,
Adams, Richard M., Bruce A. McCarl, Daniel J. and Solomon M. Hsiang. 2014. “Nonlinear Perma-
Dudek, and J. David Glyer. 1988. “Implications of nent Migration Response to Climatic Variations But
Global Climate Change for Western Agriculture.” Minimal Response to Disasters.” PNAS 111(27):
Western Journal of Agricultural Economics 13(2): 9780–85.
348–56. Burke, Marshall, and Kyle Emerick. 2016.
Albouy, David, Walter Graf, Ryan Kellogg, and “Adaptation to Climate Change: Evidence from US
Hendrik Wolff. 2016. “Climate Amenities, Climate Agriculture.” American Economic Journal: Economic
Change, and American Quality of Life.” Journal of Policy 8(3): 106–40.
the Association of Environmental and Resource Econo- Burke, Marshall, Solomon M. Hsiang, and
mists 3(1): 205–46. Edward Miguel. 2015a. “Global Non-Linear Effect
Auffhammer, Maximilian. 2018. “Climate of Temperature on Economic Production.” Nature
Adaptive Response Estimation: Short and Long 527(7577): 235–39.
Run Impacts of Climate Change on Residential Burke, Marshall, Solomon M. Hsiang, and
Electricity and Natural Gas Consumption Using Edward Miguel. 2015b. “Climate and Conflict.”
Big Data.” NBER Working Paper 24397. Annual Review of Economics 7(1): 577–617.
Auffhammer, Maximilian, and Anin Aroon- Butler, Ethan E., and Peter Huybers. 2013.
ruengsawat. 2012. “Hotspots of Climate-Driven “Adaptation of US Maize to Temperature Varia-
Increases in Residential Electricity Demand: A tions.” Nature Climate Change 3(1): 68–72.
Simulation Exercise Based on Household Level Carleton, Tamma, Michael Delgado, Michael
Billing Data for California.” White Paper from the Greenstone, Solomon M. Hsiang, Andrew
California Energy Commission’s Climate Change Hultgren, Amir Jina, Robert Kopp, Ishan Nath,
Center. James Rising, Ashwin Rode, Samuel Seo, Justin
Auffhammer, Maximilian, Solomon M. Hsiang, Simcock, Arvid Viaene, Jiacan Yuan, and Alice
Wolfram Schlenker, and Adam Sobel. 2013. “Using Zhang. 2018. “Valuing the Global Mortality
Weather Data and Climate Model Output in Consequences of Climate Change Accounting
Economic Analyses of Climate Change.” Review of for Adaptation Costs and Benefits.” In The
Environmental Economics and Policy 7(2): 181–98. Economic Consequences of Environmental Change:
Auffhammer, Maximilian, V. Ramanathan, and Essays on Climate and Water. University of Cali-
Jeffrey R. Vincent. 2006. “Integrated Model Shows fornia Berkeley, Dissertation.
that Atmospheric Brown Clouds and Greenhouse Carleton, Tamma A., and Solomon M. Hsiang.
Gases Have Reduced Rice Harvests in India.” PNAS 2016. “Social and Economic Impacts of Climate.”
103(52): 19668–672. Science 353(6304): aad9837.
Barreca, Alan, Karen Clay, Olivier Deschenes, Climate Impact Lab. 2018. Climate
Michael Greenstone, and Joseph S. Shapiro. 2016. Impact Map. https://www.impactlab.
“Adapting to Climate Change: The Remarkable org/map/#usmeas=absolute&usyear=
Decline in the US Temperature–Mortality Rela- 1981-2010&gmeas=absolute&gyear=1 986-
tionship over the Twentieth Century.” Journal of 2005&tab=global (accessed April 15, 2018).
Political Economy 124(1): 105–159. Cline, William R. 1992. The Economics of Global
Baylis, Patrick. 2015. “Temperature and Warming. Institute for International Economics.
Temperament: Evidence from a Billion Tweets.” d’Arge, Ralph C. 1975. Economic and Social
Energy Institute at Haas Working Paper 265. Measures of Biologic and Climatic Change, vol. 6.
November. Institute for Defense Analyses.
Bento, Antonio, Mehreen Mookerjee, and Davis, Lucas W., and Paul J. Gertler. 2015.
Edson Severnini. 2017. “A New Approach to “Contribution of Air Conditioning Adoption to
Measuring Climate Change Impacts and Adapta- Future Energy Use under Global Warming.” PNAS
tion.” Unpublished paper. http://econweb.ucsd. 112(19): 5962–67.
edu/cee/papers/Bentoetal_ClimatePenalty.pdf. Dell, Melissa, Benjamin F. Jones, and Benjamin
Bigano, Andrea, Jacqueline Hamilton, and A. Olken. 2012. “Temperature Shocks and
Richard S. J. Tol. 2007. “The Impact of Climate Economic Growth: Evidence from the Last Half
Century.” American Economic Journal: Macroeco- Heutel, Garth, Nolan H. Miller, and David
nomics 4(3): 66–95. Molitor. 2017. “Adaptation and the Mortality
Dell, Melissa, Benjamin F. Jones, and Benjamin Effects of Temperature Across U.S. Climate
A. Olken. 2014. “What Do We Learn from the Regions.” NBER Working Paper 23271.
Weather? The New Climate–Economy Litera- Houser, Trevor, Solomon Hsiang, Robert
ture.” Journal of Economic Literature 52(3): 740–98. Kopp, and Kate Larsen. 2015. Economic Risks of
Deryugina, Tatyana, Laura Kawano, and Steven Climate Change: An American Prospectus. Columbia
Levitt. 2018. “The Economic Impact of Hurricane University Press.
Katrina on Its Victims: Evidence from Individual Hsiang, Solomon. 2016. “Climate Economet-
Tax Returns. American Economic Journal: Applied rics.” Annual Review of Resource Economics 8: 43–75.
Economics 10(2): 202–33. Hsiang, Solomon M., Marshall Burke, and
Deschênes, Olivier, and Michael Greenstone. Edward Miguel. 2013. “Quantifying the Influence
2007. “The Economic Impacts of Climate Change: of Climate on Human Conflict.” Science 341(6151):
Evidence from Agricultural Output and Random 1235367
Fluctuations in Weather.” American Economic Review Hsiang, Solomon M., and Daiju Narita. 2012.
97(1): 354–85. “Adaptation to Cyclone Risk: Evidence from the
Deschênes, Olivier, and Michael Greenstone. Global Cross-Section.” Climate Change Economics
2011. “Climate Change, Mortality, and Adaptation: 3(2): 1250011.
Evidence from Annual Fluctuations in Weather Interagency Working Group (IWG) on Social
in the US.” American Economic Journal: Applied Cost of Greenhouse Gases, United States
Economics 3(4): 152–85. Goverment. 2016. Technical Support Docu-
Diaz, Delavane, and Frances C. Moore. 2017. ment: Technical Update of the Social Cost of
“Quantifying the Economic Risks of Climate Carbon for Regulatory Impact Analysis Under
Change.” Nature Climate Change 7: 774–82. Executive Order 12866. https://www.epa.gov/
Dingel, Jonathan I., Kyle C. Meng, and sites/production/files/2 016-12/documents/
Solomon Hsiang. 2018. “The Spatial Structure sc_co2_tsd_august_2016.pdf.
of Productivity, Trade, and Inequality: Evidence Intergovernmental Panel on Climate Change
from the Global Climate.” https://www.dropbox. (IPCC). 2013: “Summary for Policymakers.” In:
com/s/2dibmgy94kn7any/DingelMengHsiang. Climate Change 2013: The Physical Science Basis.
pdf?dl=1. Contribution of Working Group I to the Fifth Assess-
Drupp, Moritz A., Mark C. Freeman, ment Report of the Intergovernmental Panel on Climate
Ben Groom, and Frikk Nesje. Forthcoming. Change, edited by T. F. Stocker, D. Qin, G.-K.
“Discounting Disentangled.” American Economic Plattner, M. Tignor, S.K. Allen, J. Boschung, A.
Journal: Economic Policy. Nauels, Y. Xia, V. Bex, and P.M. Midgley. Cambridge
Environmental Protection Agency (EPA). 2018. University Press.
Social cost of carbon spreadsheet provided by EPA Kokoski, Mary F., and V. Kerry Smith. 1987. “A
to OMB, July, 16, 2018. https://www.regulations. General Equilibrium Analysis of Partial-Equilibrium
gov/contentStreamer?documentId=E PA-HQ- Welfare Measures: The Case of Climate Change.”
OAR-2018-0283-0453&attachmentNumber=16&co American Economic Review 77(3): 331–41.
ntentType=excel12book. Livingstone, David N. 2011. “Environmental
Fisher, Anthony C., W. Michael Hanemann, Determinism.” In The SAGE Handbook of Geograph-
Michael J. Roberts, and Wolfram Schlenker. 2012. ical Knowledge, edited by John A. Agnew, 368–380.
“The Economic Impacts of Climate Change: SAGE publishing.
Evidence from Agricultural Output and Random Mansur, Erin T., Robert Mendelsohn, and
Fluctuations in Weather: Comment.” American Wendy Morrison. 2008. “Climate Change Adapta-
Economic Review 102(7): 3749–60. tion: A Study of Fuel Choice and Consumption
Greenstone, Michael. 2016. “A New Path in the US Energy Sector.” Journal of Environmental
Forward for an Empirical Social Cost of Carbon.” Economics and Management 55(2): 175–93.
Presentation to the National Academies of McIntosh, Craig T., and Wolfram Schlenker.
Sciences, https://sites.nationalacademies.org/ 2006. “Identifying N on-Linearities in Fixed Effects
cs/groups/dbassesite/documents/webpage/ Models.” Unpublished paper, https://gps.ucsd.
dbasse_172599.pdf. edu/_files/faculty/mcintosh/mcintosh_research_
Greenstone, Michael, Elizabeth Kopits, and identifying.pdf.
Ann Wolverton. 2013. “Developing a Social Cost Mendelsohn, Robert, William D. Nordhaus,
of Carbon for US Regulatory Analysis: A Method- and Daigee Shaw. 1994. “The Impact of Global
ology and Interpretation.” Review of Environmental Warming on Agriculture: A Ricardian Analysis.”
Economics and Policy 7(1): 23–46. American Economic Review 84(4): 753–71.
Missirian, Anouch, and Wolfram Schlenker. of Climate Change: Adaptations Past and Present, pp.
2017. “Asylum Applications and Migration Flows.” 225–51, edited by Gary D. Libecap and Richard H.
American Economic Review 107(5): 436–40. Steckel. University of Chicago Press.
Moore, Frances C., Uris Baldos, Thomas Roberts, Malcolm J., Pier Luigi Vidale, Matthew
Hertel, and Delavane Diaz. 2017. “New Science of S. Mizielinski, Marie-Estelle Demory, Reinhard
Climate Change Impacts on Agriculture Implies Schiemann, Jane Strachan, Kevin Hodges, Ray
Higher Social Cost of Carbon.” Nature Communica- Bell, and Joanne Camp. 2015. “Tropical Cyclones
tions 8: 1607. in the UPSCALE Ensemble of High-Resolution
NASA. 2018. “NASA Earth Exchange Global Global Climate Models.” Journal of Climate 28(2):
Daily Downscaled Projections ( NEX-GDDP).” 574–96.
https://cds.nccs.nasa.gov/nex-gddp/ (accessed Rose, Steven K. 2012. “The Role of the Social
June 2018). Cost of Carbon in Policy.” Wiley Interdisciplinary
National Academies of Sciences, Engineering, Reviews: Climate Change 3(2): 195–212.
and Medicine. 2017. Valuing Climate Damages: Rose, S., D. Turner, G. Blanford, J. Bistline, F.
Updating Estimation of the Social Cost of Carbon de la Chesnaye, and T. Wilson. 2014. Understanding
Dioxide. Washington, DC: The National Academies the Social Cost of Carbon: A Technical Assessment.
Press. https://doi.org/10.17226/24651. EPRI Report 3002004657.
Nordhaus, William. 2006. “Paul Samuelson and Sanderson, Marie. 1999. “The Classification of
Global Public Goods.’’ Chap. 5 in Samuelsonian Climates from Pythagoras to Koeppen.” Bulletin of
Economics and the Twenty-First Century, edited by the American Meteorological Society 80(4): 669–73.
Michael Szenberg, Lall Ramrattan, and Aron A. Schlenker, Wolfram, W. Michael Hanemann,
Gottesman. Oxford University Press. and Anthony C. Fisher. 2005. “Will US agriculture
Peng, Zhang, Olivier Deschênes, Kyle C. Meng, really benefit from global warming? Accounting
and Junjie Zhang. 2018. “Temperature Effects on for irrigation in the hedonic approach.” American
Productivity and Factor Reallocation: Evidence Economic Review 95:1, 395–406.
from a Half Million Chinese Manufacturing Schlenker, Wolfram, and Michael J. Roberts.
Plants.” Journal of Environmental Economics and 2009. “Nonlinear Temperature Effects Indicate
Management 88(3): 1–17. Severe Damages to US Crop Yields under Climate
Pindyck, Robert S. 2013. “Climate Change Change.” PNAS 106(37): 15594–98.
Policy: What Do the Models Tell Us?” Journal of Seetharam, Ishuwar. 2018. “The Indirect Effects
Economic Literature 51(3): 860–72. of Hurricanes: Evidence from Firm Internal
Pindyck, Robert S. 2016. “The Social Cost of Networks.” https://web.stanford.edu/~ishuwar/
Carbon Revisited.” NBER Working Paper 22807. JMP.pdf.
Pindyck, Robert S. 2017. “The Use and Misuse Severen, Christopher, Christopher Costello,
of Models for Climate Policy.” Review of Environ- and Olivier Deschenes. 2016. “A Forward Looking
mental Economics and Policy 11(1): 100–114. Ricardian Approach: Do Land Markets Capitalize
Quiggin, John and Horowitz, John K. 1999. Climate Change Forecasts?” NBER Working Paper
“The Impact of Global Warming on Agriculture: A 22413.
Ricardian Analysis: Comment.” American Economic US Environmental Protection Agency (EPA).
Review, 1999, 89(4), pp. 1044–45. 2008. “Regulating Greenhouse Gas Emissions
Roberts, Michael J., and Wolfram Schlenker. Under the Clean Air Act; Proposed Rule.” Federal
2011. “The Evolution of Heat Tolerance of Corn: Register, Wednesday, July 30, 73(147).
Implications for Climate Change.” In The Economics