Attention to Global Warming

Darwin Choi, Zhenyu Gao and Wenxi Jiang

CUHK Business School

We find that people revise their beliefs about climate change upward when experiencing
warmer than usual temperatures in their area. Using international data, we show that

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

attention to climate change, as proxied by Google search volume, increases when the
local temperature is abnormally high. In financial markets, stocks of carbon-intensive firms
underperform firms with low carbon emissions in abnormally warm weather. Retail investors
(not institutional investors) sell carbon-intensive firms in such weather, and return patterns
are unlikely to be driven by changes in fundamentals. Our study sheds light on peoples’
collective beliefs and actions about global warming. (JEL D83, G12, G14, G15, Q54)

Received December 7, 2017; editorial decision March 27, 2019 by Editor Andrew Karolyi.
Authors have furnished an Internet Appendix, which is available on the Oxford University
Press Web site next to the link to the final published paper online.

Introduction
President Donald Trump, who has called global warming a “hoax” on multiple
occasions, wrote the following message on Twitter on December 28, 2017,
when unusually cold temperatures were expected to hit the Eastern United
States:
In the East, it could be the COLDEST New Year’s Eve on
record. Perhaps we could use a little bit of that good old Global
Warming that our Country, but not other countries, was going to
pay TRILLIONS OF DOLLARS to protect against. Bundle up!
— Donald J. Trump (@realDonaldTrump)

We thank Andrew Karolyi (the Editor), Vikas Agarwal, Laura Bakkensen, Zhi Da, Andrew Ellul, Harrison
Hong, Roger Loh, Pedro Matos, Abhiroop Mukherjee, Adriaan Perrels, Jose Scheinkman, Wing Wah Tham,
Bohui Zhang, and Dexin Zhou; three anonymous referees; and seminar participants at the 2nd International
Conference on Econometrics and Statistics (EcoSta), 8th Helsinki Finance Summit on Investor Behavior, 10th
Annual Volatility Institute Conference: A Financial Approach to Climate Risk, Asian Bureau of Finance and
Economic Research (ABFER) 6th Annual Conference, China International Conference in Finance 2018, Review
of Financial Studies Climate Finance Workshop 2017 and Climate Finance Conference 2018, Society for Financial
Studies (SFS) Cavalcade Asia-Pacific 2018, Cheung Kong Graduate School of Business, Chinese University of
Hong Kong, Hong Kong Baptist University, Hong Kong Polytechnic University, Renmin University of China,
and University of Melbourne for helpful comments. Jingxuan Chen, Haojun Xie, and Hulai Zhang provided
excellent research assistance. We acknowledge the General Research Fund of the Research Grants Council of
Hong Kong (Project Number: 16516616) for financial support. Supplementary data can be found on The Review
of Financial Studies Web site. Send correspondence to Darwin Choi, Department of Finance, CUHK Business
School, The Chinese University of Hong Kong, Shatin, N.T., Hong Kong; telephone: (852) 3943-5301. E-mail:
dchoi@cuhk.edu.hk.

© The Author(s) 2020. Published by Oxford University Press on behalf of The Society for Financial Studies.
All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
doi:10.1093/rfs/hhz086

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1112 1112–1145

 Attention to Global Warming

Pierre-Louis (2017) of The New York Times points out that President
Trump’s tweet made the mistake of confusing local weather with climate.
This misunderstanding about climate change is indeed a common mistake.
Global warming is a long-term trend usually not visible on a personal level.
In contrast, the local temperature in a given month or year is more noticeable,
even though it can be caused by reasons unrelated to global warming, for
example, ocean oscillations, such as the El Niño Southern Oscillation, ENSO
(Intergovernmental Panel on Climate Change, IPCC 2014; Schmidt, Shindell,

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

and Tsigaridis 2014). For example, a record-breaking warm month of July in
New York City provides negligible information about the increase in the average
global temperature in the following decade, but the local temperature in July is
much more visible to New Yorkers than the 10-year global trend.
In this paper, we test how people react to abnormal local temperatures by
examining their attention to climate change and stock prices. Our data cover
seventy-four cities in the world with major stock exchanges. The advantage of
using international attention and financial data is that we can estimate people’s
opinions in different parts of the world at a high frequency (unlike surveys)
and study their follow-up actions, as investors trade on their beliefs and move
stock prices. Humans’ collective belief and effort are important determinants of
the efficacy of climate policies and campaigns. Our study aims to empirically
identify how the general public realizes and responds to the impacts of global
warming.
Because their attention is limited, people are likely to focus on attention-
grabbing weather events and personal experiences. The local weather
conditions are people’s first-hand experience. The impact of local weather
also can be amplified through communication channels and the media (media
attention to climate change appears to be higher in the record-breaking warmest
years than in nonrecord years) (Schmidt 2015). Extreme local temperatures
therefore serve as “wake-up calls” that alert investors to climate change. Our
paper tests this idea in two steps: first, we test whether people pay more attention
to climate change when experiencing abnormally warm weather. The second
set of analyses examines whether this experience affects financial markets;
because of the home bias (see, e.g., the review by Karolyi and Stulz 2003), the
prices of local stocks are affected by local investors’ trading behavior.
Our results show that during abnormally warm months in a particular city,
the volume of Google searches for the topic of “global warming” in that
city increases.1 Our analysis controls for time fixed effects, and therefore the
relationship originates from geographical variation. Not all parts of the world
are equally warm in a given month; people tend to seek more information

1 In this paper, we use the term “abnormally warm” to refer to cases in which a city’s temperature is significantly
higher than the historical average temperature at the same point in the year. Our Google data capture the search
activity in each city and cover different languages. See Section 1 for a list of papers that study the Google search
volume of global warming in the United States.

1113

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1113 1112–1145

 The Review of Financial Studies / v 33 n 3 2020

about global warming if they live in cities that have relatively higher abnormal
temperatures than other cities in that month. This effect is the most prominent
when the local abnormal temperature is in the city’s top quintile, as this weather
experience is more salient.
If investors revise their beliefs about global warming, they may buy stocks
with lower climate sensitivities and sell stocks with higher climate sensitivities
such that the former outperform the latter. We sort stocks into those with high
and low sensitivities using proxies for greenhouse gas emission levels. Firms

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

are classified as high-emission firms if they belong to industries that the IPCC
identifies as major emission sources. These companies tend to be more sensitive
to climate change if their future cashflows are adversely affected by higher
production costs and tighter environmental regulations or if socially responsible
investors avoid holding their stocks.
We find evidence that carbon-intensive firms earn lower stock returns than
other firms when the local exchange city is abnormally warmer in that month.
The effect is again more prominent when the abnormal temperature is in
the city’s top quintile. An increase in the city’s abnormal temperature from
the coolest quintile to the warmest quintile is associated with a reduction
of 48 bps in the long-short emission-minus-clean portfolio. In an alternative
specification, we define high- and low-emission firms according to their MSCI
Carbon Emission Scores, which capture individual companies’ emission levels
relative to their industry peers, and achieve similar results. We do not find
any significant return reversal in the longer term (up to a year). The return
patterns are observed in both energy and nonenergy high-emission sectors and
are robust to size adjustments. Furthermore, we do not obtain the same results
in a “placebo” test that uses an earlier sample period, 1983–2000, when global
warming was less of an issue.
To better understand the mechanism through which temperature affects
prices, we examine proxies for different investors’ trading behavior.2 We
focus on local blockholders, local institutional investors, and retail investors
(the majority of which are local) because they are exposed to the same
temperature. Which of these investors decrease their holdings of high-emission
firms in an abnormally warm quarter? Consistent with our conjecture that
individuals are more prone to limited attention and drawn to notable events,
we find evidence that retail investors sell high-emission firms and buy low-
emission firms. Institutional investors (local and foreign) do not respond
systematically to abnormal temperatures. More interestingly, we find that local
blockholders trade in the opposite direction of retail investors. Therefore, local
abnormal temperatures do not appear to adversely affect all carbon-intensive
firms’ operations in a fundamental manner, as blockholders of these firms are

2 Data on the quarterly equity positions of blockholders (who hold 5% or more of the total number of shares)
and of institutional investors who hold less than 5% of total shares are obtained from DataStream and FactSet,
respectively. The complement of these holdings gives us an estimate of retail investors’ positions.

1114

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1114 1112–1145

 Attention to Global Warming

generally buying shares and household investors should be less informed than
blockholders.3 We conclude that unusually warm weather is a salient event that
affects individual investors, and we run a series of tests and find no evidence
that the price impact is a result of belief updates about firms’ future cashflows.
Rather, people seem to avoid holding high-emission firms as they become more
aware of climate risk, similar in spirit to avoiding “sin” stocks (companies
involved in producing alcohol, tobacco, and gaming).
Recent literature on climate change also examines people’s beliefs and

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

personal experience. Zaval et al. (2014), Akerlof et al. (2013), and Myers
et al. (2012) show that personal experience with global warming, as reported in
surveys, leads to an increased perception of climate risk in the United States;
this finding is confirmed by Broomell, Budescu, and Por (2015) and Howe
et al. (2013) using international surveys. Konisky, Hughes, and Kaylor (2016),
Borick and Rabe (2014), and Joireman, Truelove, and Duell (2010) find a
similar relationship using objective measures of weather experience, such as
outdoor temperature, snowfall, and occurrences of floods and hurricanes. Li,
Johnson, and Zaval (2011) further show that perceived deviations from normal
temperature not only alter beliefs but also are followed by actions: participants
are more likely to donate their earnings to a global warming charity. Surveys
measure beliefs about global warming in all these studies. In contrast, our paper
uses objective proxies for attention to capture the learning process, and we
can examine how updated aggregate beliefs are reflected in prices and trading
behavior. Our findings are related to experiential learning, in which people
begin the learning process based on concrete experience and form abstract
concepts by observing and analyzing information before acting (Boud, Keogh,
and Walker 1985; Kolb 1984). In our context, we can see whether people read
more about global warming (on the Internet) after they are personally affected
by the local weather.
This paper complements previous empirical findings on reactions to climate
and other external conditions. Chang, Huang, and Wang (2018) find that more
health insurance contracts are sold when the air is polluted, but they are more
likely to be canceled if air quality improves shortly afterward. Busse et al.
(2015) and Conlin, O’Donoghue, and Vogelsang (2007) show that the choice
to purchase warm- or cold-weather vehicle types and cold-weather clothing,
respectively, depends on the weather at the time of purchase. Hong, Li, and
Xu (2019) document underreaction of food companies’ stock prices to trends
in droughts that are exacerbated by global warming. Using a comprehensive
database of coastal home sales in the United States, Murfin and Spiegel
(forthcoming) find that real estate prices do not factor in the risk of sea level

3 Although we cannot rule out the possibility that local blockholders revise their beliefs about global warning
downward in an abnormally warm quarter (which would be puzzling), our preferred interpretation is that these
blockholders respond to the decrease in stock prices, similar to the model developed by Hong, Wang, and Yu
(2008). Hong, Wang, and Yu (2008) argue that firms are buyers of last resort for their own stocks. They repurchase
shares when prices drop below their fundamental value.

1115

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1115 1112–1145

 The Review of Financial Studies / v 33 n 3 2020

rise. Our results are also in line with general underreaction to global warming.
Finally, the finding that people pay more attention and our observation of the
differential impacts on the cross-section of stocks distinguish our work from
the literature that links weather-induced investor mood and the stock market
(Goetzmann et al. 2015; Kamstra, Kramer, and Levi 2003, among others).

1. Methods and Hypotheses

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

We would like to identify investor reaction to global warming in times
of unusually warm local weather. Given that climate change is a global
phenomenon, we conduct our study in a broad international setting to
understand people’s collective beliefs and reactions. The international setting
also gives us an identification advantage: climate science research shows that
extreme temperatures rarely occur simultaneously in both the Northern and
Southern Hemispheres (see, e.g., Neukom et al. 2014).
The reaction is first measured by the monthly Google Search Volume Index
(SV I ) of the topic “global warming” in a city, which proxies for people’s
attention. Google offers SV I for topics and search terms. We use topics
instead of search terms because the former addresses misspellings and searches
in different languages, as Google’s algorithms can group different searches
that have the same meaning under a single topic.4 Our idea follows Da,
Engelberg, and Gao (2011), who use SV I of tickers to study investor attention.
Several other papers also examine Google search volume for global warming
and climate change and relate it to local weather conditions: e.g., Lineman
et al. (2015), Cavanagh et al. (2014), Herrnstadt and Muehlegger (2014),
Lang (2014), and Kahn and Kotchen (2011). These studies focus on U.S.
data, whereas our paper covers more than seventy cities worldwide and many
different languages.
To understand the learning process, we decompose local temperatures into
three components, which account for predictable, seasonal, and abnormal
patterns. Specifically, for each city i in month t, we calculate the monthly
Temperatureit by taking the average of daily average temperatures in our data.
Then we define
Temperatureit = Aver_Tempit +Mon_Tempit +Ab_Tempit , (1)
where Aver_T empit is the average monthly local temperature in city i over the
120 months prior to t; Mon_T empit is the average deviation of this month’s
temperature from the average, that is, the average temperature in city i in
the same calendar month over the last 10 years minus Aver_T empit ; and
Ab_Tempit is the remainder. Our focus is how local abnormal temperatures
affect changes in attention (as proxied by the change in SV I , adjusted for

4 See the official Google Search blog for details: https://search.googleblog.com/2013/12/an-easier-way-to-explore

-topics-and.html.

1116

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1116 1112–1145

 Attention to Global Warming

seasonality). Even though a city’s monthly Ab_Temp provides little fundamental

information about the future global climate, it represents new experience and
is a salient event for people in the city.
Then we turn to investor reaction in the stock market. We study monthly size-
adjusted stock returns under abnormally warm weather in the exchange city.
The size-adjusted return is defined as the stock return minus the average return
of stocks in the same size quintile in the exchange in the same month. Exchange
cities are important cities in which many investors are located, and prices are

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

affected by domestic investors (see, e.g., Chan, Hameed, and Lau 2003).5 We
examine the cross-section of firms with different sensitivities to climate change.
If investors begin recognizing the effect of climate on financial markets and buy
low-climate-sensitivity and sell high-climate-sensitivity firms, the former will
earn higher returns than the latter. The short-term and the long-term patterns
are examined. Without reversal or with some continuation in the long run, it is
consistent with belief updating. A reversal indicates that the short-term price
changes overshoot and investors overreact.
The effect of abnormally warm weather on stock prices can occur through
multiple channels. First, climate-unfriendly firms may be fundamentally
damaged. Second, people may update their own valuation of firms when they
revise their beliefs about climate change upward. Investors may think that
high-emission firms’ future cashflows are adversely affected because climate
change can hurt firms’ production functions, impose higher costs for future
emissions, or induce tighter regulations on emissions. Third, and finally, on
recognizing the risk of global warming, socially responsible investors may
stay away from firms that are climate unfriendly, similar to the way in which
“sin” stocks are shunned by some investors (Hong and Kacpercyzk 2009).
Sections 3.4 and 3.5 present a series of tests to distinguish between these
channels.

2. Data
In the following, we describe the various databases we use, as well as the
variables we obtain and examine in our analyses (the databases used in the
Internet Appendix are described there).

2.1 Weather
We obtain daily weather data from the Global Surface Summary of Day
Data, which are produced by the National Climatic Data Center (NCDC). The
input data used in building these daily observations are the Integrated Surface

5 In large countries, such as China, India, Russia, and the United States, population (and therefore local investors)
are more dispersed, and the exchange city’s temperature is a weaker proxy for the effect of weather on investors.
Table 5 shows a weaker relationship between the exchange city’s abnormal temperature and returns in the ten
largest countries in our sample.

1117

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1117 1112–1145

 The Review of Financial Studies / v 33 n 3 2020

Data (ISD), which contain weather records from over 9,000 stations globally
since 1973 (the coverage was considerably lower before 1973). The weather
conditions include temperature, wind speed, cloudiness, precipitation, snow
depth, etc. By identifying the location coordinates, we select the closest weather
station to the address of the exchange. We collect the daily records of seventy-
four cities with major stock exchanges from 1973 to 2017. Our main test period
is from 2001 to 2017, when climate change is a global phenomenon.6 As noted
in Section 1, abnormal temperatures require 10 years of data to calculate. We

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

use the period from 1983 to 2000, when few people recognized climate change,
to conduct a “placebo” test.

2.2 Google Search Volume Index

The data source for internet search activity is Google Trends, which provides
a Search Volume Index (SV I ) of the search topic of “global warming.” We
download the monthly SV I in each of the seventy-four locations from 2004
(when Google Trends began to provide data) to 2017. We examine search
activity at both the city and country levels (except for some small countries for
which the search volume data are available only at the country level).7

2.3 Stock and company information

Stock returns, market capitalization, and industry information are available
from Thomson Reuters DataStream. For U.S. stocks, we use return and market
capitalization data from CRSP (we obtain a list of U.S. stocks from DataStream
and match them to CRSP using ISIN and CUSIP). DataStream covers more than
100,000 equities in nearly 200 countries from 1980 onward. We can observe the
firms’ countries of domicile (from the NATION variable) and their exchange
cities, but not the locations of firms’ establishments. The literature notes that
DataStream may suffer from data errors. We winsorize raw returns at the top
and bottom 2.5% in each exchange in each month. Following Hou, Karolyi,
and Kho (2011) and Ince and Porter (2006), we remove all monthly returns that
are above 300% and reversed within 1 month, as well as zero monthly returns
(DataStream repeats the last valid data point for delisted firms).

6 Our conclusion remains the same if we use other similar test periods in the twenty-first century.
Climate change became a global concern in the early twenty-first century. For example, in its
Third Assessment Report released in 2001, the IPCC claims that “there is new and stronger
evidence that most of the observed warming of the past 50 years is attributable to human activi-
ties” (https://archive.ipcc.ch/graphics/speeches/robert-watson-november-2001.pdf). In 2001, national science
academies of many different countries issued a joint statement stating that “IPCC represents the consensus
of the international scientific community on climate change science” (Science 2001).
7 We also download the monthly SV I for the topic “climate change,” but the search traffic for this topic is much
lower than that of “global warming” in the first few years of our sample period. In more recent years, the SV I s
of the two topics are highly correlated. In the paper, we report the results using the SV I of “global warming.”

1118

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1118 1112–1145

 Attention to Global Warming

2.4 Stock ownership

DataStream provides the aggregate ownership in a stock by domestic and
foreign blockholders (who hold more than 5% of shares outstanding) in
every quarter. Quarterly holdings by institutional investors and their locations
(at the country level) are obtained from FactSet, which covers 33 of our
74 exchange cities. We use the SAS code provided by Ferreira and Matos
(2008), available on WRDS, to calculate the ownership by institutional
investors, excluding blockholders. Then we define retail ownership as (100%

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

− DataStream blockholders’ ownership − FactSet institutional ownership
excluding blockholders).8

2.5 Carbon emission

We identify high-emission firms in two ways. First, we adopt the industry
definitions provided by the Intergovernmental Panel on Climate Change
(IPCC), the leading international body for the assessment of climate change.
Five major industry sectors are identified as major emission sources: Energy;
Transport; Buildings; Industry (such as chemicals and metals); and Agriculture,
Forestry, and Other Land Use (AFOLU). Each sector is further divided into
subcategories (Krey et al. 2014 offers a full list). We hand-match the IPCC
subcategories with the industry names provided by DataStream.9 All firms in
the matched industries are classified as high-emission firms.
Second, we obtain firms’ carbon emission estimates from MSCI ESG
Ratings, which analyzes companies’ environmental, social, and governance
issues. Specifically, MSCI ESG studies greenhouse gas (GHG) emissions of
companies worldwide. A Carbon Emission Score is given to each firm annually
since 2007, on a scale of 0–10. Companies with better performance on this issue
score higher. The score is adjusted by industry and is thus comparable for two
firms from different industries. We define high- (low-) carbon emission firms
as firms whose MSCI Carbon Emission Scores in the previous calendar year
are lower than 3 (higher than 7).
The two definitions identify high-emission firms differently. For example,
Toyota Motor Corporation, listed on the Tokyo Stock Exchange, belongs to
the Automobiles industry in DataStream (which is mapped to the Transport
Equipment industry, IPCC code = 1A2f2). According to the first method, it is
classified as a high-emission firm. The average MSCI score of Toyota Motor
Corporation in our sample period is 9.4, and the second method places it in

8 In other words, we do not directly measure retail ownership, and the proxy is subject to measurement errors.
Several other papers also define the complement of U.S. institutional holdings as a proxy for individual U.S.
investors’ demand: for example, DeVault, Sias, and Starks (2019), Agarwal, Vashishtha, and Venkatachalam
(2018), Malmendier and Shanthikumar (2007), Griffin, Harris, and Topaloglu (2003), and Cohen, Gompers, and
Vuolteenaho (2002).
9 For example, Coal (DataStream Industry Classification Benchmark ICB code = 1771), Gold Mining (DataStream
ICB code = 1777), and General Mining (DataStream ICB code = 1775) are matched with Mining and Quarrying
(IPCC code = 1A2f4). Table IA.1 in the Internet Appendix provides the map.

1119

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1119 1112–1145

 The Review of Financial Studies / v 33 n 3 2020

the low-emission group in all years. One can interpret that the company is a
relatively clean firm in a high-emission industry. Throughout the paper, we
primarily use IPCC definitions because they are available for all firms and for
a longer period. MSCI covers only a subset of firms in a small number of
exchanges. It may have a selection issue and the results should be interpreted
with caution.10

2.6 Price of carbon and environmental regulatory regime index

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

The price of carbon is measured by carbon futures prices. EUA Futures
Contracts, traded on the Intercontinental Exchange (ICE) Futures Europe,
are contracts in which the traders are obliged to make or take the delivery
of 1,000 emission allowances. Each allowance is an entitlement to emit
one ton of carbon-dioxide-equivalent gas. We download the data from
Bloomberg (symbol: MO1 Comdty), beginning in April 2005. We also use the
environmental regulatory regime index developed by Esty and Porter (2001).
The index is a ranking of countries’ regulatory stringency, structure, subsidies,
and enforcement; it represents the quality of the environmental regulatory
system.

3. Empirical Results
Our tests aim to investigate two questions: (1) whether people’s attention varies
with local temperatures and (2) if so, how experiences of local temperatures
affect the stock price of local firms and investors’ trading behavior. Table 1
shows the list of 74 stock exchange cities. It reports the number of unique
stocks, number of foreign firms (whose country of domicile information is
available and is different from that of the exchange city), number of emission
firms and foreign emission firms (defined using the IPCC classification), as
well as average retail and blockholder ownerships in each city in our sample.
In all regressions below, all standard errors are clustered by exchange city and
year-month (or year-quarter for regressions of changes in ownership).

3.1 Attention and local temperatures

To capture changes in attention, we first calculate the log monthly change in the
Google Search Volume Index, DSV I . DSV Iit is the log change in SV I in city
i in month t, adjusted for seasonality.11 Panel A of Table 2 shows the summary

10 MSCI collects data once a year from the most recent corporate resources, such as annual reports and corporate
social responsibility reports. When direct disclosure is not available, MSCI uses GHG data reported by the Carbon
Disclosure Project or government databases. Note that MSCI does not assign a score to every public firm. The
number of firms with a valid MSCI Carbon Emission Score in our data increases from 1,888 in 2007 to 11,239 in
2017, as shown in Table IA.2 in the Internet Appendix. Although MSCI also issues other climate-change-related
scores to companies, such as the Climate Change Theme Score, the Carbon Emission Score is available for the
longest period.
11 DSV I is defined as the residuals from the regression of the log change in the monthly SV I on month-of-the-year
dummies. The residuals are then winsorized at the top and bottom 2.5% tails. Two cities, Shenzhen and Shanghai,

1120

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1120 1112–1145

 Attention to Global Warming

Table 1
List of exchange cities

# # # #Foreign % %
City Country/area Continent Firms Foreign Emission emission Retail Blockholder
Amman Jordan Asia 228 0 56 0
Amsterdam Netherlands Europe 247 10 68 2 51.47 2.67
Athens Greece Europe 364 0 141 0 59.12 0.69
Bangkok Thailand Asia 796 0 315 0
Berlin Germany Europe 54 4 11 3 72.86 2.16
Bern Switzerland Europe 18 1 2 1

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

Bogota Colombia South America 74 1 27 1
Bratislava Slovakia Europe 25 0 8 0
Brussels Belgium Europe 280 3 63 0 41.55 0.53
Bucharest Romania Europe 272 0 146 0
Budapest Hungary Europe 77 1 23 0
Buenos Aires Argentina South America 97 0 58 0
Busan Korea Asia 1,006 2 466 1
Cairo Egypt Africa 198 0 79 0
Colombo Sri Lanka Asia 294 0 81 0
Copenhagen Denmark Europe 285 4 71 3 48.30 3.27
Dhaka Bangladesh Asia 410 0 123 0
Dublin Ireland Europe 76 4 24 3 54.03 8.30
Dusseldorf Germany Europe 58 1 17 1 76.86 0.01
Frankfurt Germany Europe 1,735 69 462 17 43.06 0.94
Hamburg Germany Europe 65 1 9 0 49.19 1.73
Hanoi Vietnam Asia 400 0 258 0
Harare Zimbabwe Africa 71 0 27 0
Helsinki Finland Europe 208 1 70 1 51.70 2.89
Ho Chi Minh Vietnam Asia 340 0 181 0
Hong Kong Hong Kong Asia 2,064 646 650 251 38.95 0.38
Istanbul Turkey Europe 461 0 167 0
Jakarta Indonesia Asia 592 0 233 0
Johannesburg South Africa Africa 663 14 196 4 55.97 1.57
Karachi Pakistan Asia 410 0 153 0
Kiev Ukraine Europe 83 0 61 0
Kuala Lumpur Malaysia Asia 1,207 12 574 3
Kuwait Kuwait Asia 177 0 39 0
Lagos Nigeria Africa 160 1 55 0
Lima Peru South America 140 3 85 3
Lisbon Portugal Europe 97 1 38 0 36.73 0.73
Ljubljana Slovenia Europe 137 0 51 0
London United Kingdom Europe 3,558 440 940 165 58.34 2.15
Luxembourg Luxembourg Europe 38 3 8 0 58.09 0.45
Madrid Spain Europe 298 2 97 1 45.97 0.33
Manila Philippines Asia 283 0 99 0
Mexico City Mexico North America 183 0 67 0
Milan Italy Europe 519 8 152 4 46.68 0.55
Moscow Russia Europe 349 0 259 0
Mumbai India Asia 4,806 1 1,908 0 46.02 1.12
Munich Germany Europe 90 2 18 0 55.26 3.54
Muscat Oman Asia 98 0 41 0
Nagoya Japan Asia 116 0 47 0 66.50 0.69
New York City United States North America 3,874 324 1,026 121 30.26 11.08
Nicosia Cyprus Europe 143 4 36 0
Osaka Japan Asia 140 0 46 0 62.59 0.04
Oslo Norway Europe 446 52 234 45 39.39 2.74
Paris France Europe 1,578 37 382 6 40.98 0.85
Prague Czechia Europe 71 3 40 0
Riyadh Saudi Arabia Asia 182 0 72 0
Santiago Chile South America 236 0 100 0
Sao Paulo Brazil South America 297 2 136 2
Shanghai China Asia 1,180 0 613 0
(Continued)

1121

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1121 1112–1145

 The Review of Financial Studies / v 33 n 3 2020

Table 1
(Continued)

Shenzhen China Asia 2,024 0 1,093 0

Singapore Singapore Asia 920 140 443 70 40.36 0.62
Skopje Macedonia Europe 40 0 20 0
Sofia Bulgaria Europe 157 0 41 0
Stockholm Sweden Europe 1,102 27 292 7 59.45 2.86
Stuttgart Germany Europe 55 5 12 3 40.86 0.64
Sydney Australia Oceania 2,888 93 1,502 37 72.78 0.59
Taipei Taiwan Asia 1,023 34 556 19
Tel Aviv Israel Asia 785 7 251 0

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

Tokyo Japan Asia 3,656 2 1,465 0 70.39 0.52
Toronto Canada North America 841 39 395 27 49.02 6.54
Vienna Austria Europe 166 2 65 0 31.80 0.47
Warsaw Poland Europe 1,075 23 339 8 29.39 3.49
Wellington New Zealand Oceania 229 5 62 1
Zagreb Croatia Europe 73 0 27 0
Zurich Switzerland Europe 367 15 119 2 63.68 0.68
This table lists the seventy-four exchange cities (and their countries/areas and continents) that we use in analyses
and the number of unique firms, number of foreign firms (whose home country information is available and is
different from that of the exchange city), number of emission firms and foreign emission firms (defined by the
IPCC classification), and average retail and blockholder ownerships in each city during the sample period, from
2001 to 2017.

statistics of DSV Iit , as well as those of Aver_T empit , Mon_T empit , and
Ab_Tempit , the decomposition of temperature in city i in month t according
to Equation (1). The mean DSV I is close to zero (−0.02%), whereas the
mean Aver_T emp, Mon_T emp, and Ab_Temp are 61.9◦ F, 0.16◦ F, and 0.27◦ F,
respectively.
Then we run the following regression:
DSVI it = α +β1 Ab_Tempit +t YearMontht +it , (2)
Our coefficient of interest is β1 . Table 2, panel B, reports the results. In Column
1, the coefficient estimate of Ab_Temp is significantly positive (t-stat = 2.3),
which suggests that people pay more attention to global warming when they
are experiencing an abnormally high temperature. The regression includes
year-month fixed effects, meaning that the relationship is observed from the
geographic variation (in a given month, when a city is abnormally warm relative
to other places, people in that city tend to search about global warming more
than people in other places).
In Column 2, we rank all months into quintiles based on Ab_Tempit in city i
and use these quintile dummies in the regression instead of Ab_Temp. The
coefficients of the quintile dummies indicate that the temperature effect is
nonlinear: the coefficients of quintiles 2, 3, and 4 are not significantly different
from zero, while the coefficient of quintile 5 is 4.84 (t-stat = 2.6). Thus, our

are dropped from the analysis, because Google Trends returns no valid local data for them. Table IA.3 in the
Internet Appendix examines changes in attention at daily, weekly, and quarterly levels. The results are weaker. A
day or week of abnormal temperatures may not shift beliefs, whereas a month is more likely to. It may also take
an extended period of warm weather for the media effect to come into play. (At the quarterly level, the results
are also weaker, but the average extreme temperature within a quarter is attenuated.)

1122

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1122 1112–1145

 Attention to Global Warming

Table 2
Google search volume for “global warming” and abnormal temperature
A. Summary statistics
Variable Obs Mean SD P10 P25 P50 P75 P90
DSVI(city) 11,603 −0.017 66.484 −51.333 −23.129 −0.604 22.652 51.584
DSVI(country) 10,366 0.047 77.578 −57.793 −24.908 −1.506 22.644 59.254
Aver_Temp 11,603 61.861 12.454 48.334 51.655 59.458 72.406 81.695
Mon_Temp 11,603 0.155 10.837 −15.185 −8.069 0.259 8.178 15.506
Ab_Temp 11,603 0.265 2.679 −2.794 −1.201 0.238 1.696 3.419
#Exchange cities 72

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

B. Regression of DSVI on abnormal temperature
(1) (2) (3) (4)
DSVI(city) DSVI(city) DSVI(country) DSVI(country)
Ab_Temp 0.536∗∗ 0.724∗∗
(2.26) (2.43)
Ab_Temp Q2 0.630 −0.279
(0.34) (−0.16)
Ab_Temp Q3 1.220 −1.787
(0.84) (−1.00)
Ab_Temp Q4 1.074 −1.149
(0.58) (−0.47)
Ab_Temp Q5 4.841∗∗ 3.539
(2.57) (1.66)
Year × Month FEs Yes Yes Yes Yes
Obs. 11,603 11,603 10,366 10,366
Adj. R 2 .020 .020 .015 .015
This table reports the results of analyses on the effect of abnormal temperatures on the search volume of the topic
of “global warming” on Google. Panel A presents summary statistics of the variables. DSVI(city) is the monthly
log change of Google’s search volume index (SVI) of the topic “global warming” in the exchange city and adjusted
for seasonality, and DSVI(country) is calculated using the SVI in country of the city. Aver_Temp is the average
monthly temperature (in Fahrenheit degrees) of the exchange’s city over the previous 120 months. Mon_Temp
is the city’s average temperature in the same month of the year over the previous 10 years minus Aver_Temp.
Ab_Temp is the city’s temperature in this month minus Aver_Temp and Mon_Temp. Panel B represents the result
of regressing DSVI(city) (Columns 1 and 2) and DSVI(country) (Columns 3 and 4) on city-level temperature
measures. For each exchange city, months are sorted into quintiles based on Ab_Temp, and Ab_Temp Q2-Q5 are
quintile dummies that equal one if the month belongs to quintiles 2–5, respectively. The sample is from 2004 to
2017. Standard errors are clustered by exchange city and by year-month, and the corresponding t -statistics are
reported in parentheses. *p < .1; **p < .05; ***p < .01.

results suggest that Google search volume increases with the highest abnormal
local temperatures, which are the most salient. This idea is similar in spirit to
the “frog in the pan” hypothesis proposed by Da, Gurun, and Warachka (2014),
who show that investors pay more attention to infrequent dramatic changes than
to frequent gradual changes.
The economic magnitude is worth noting. Based on the estimation in Column
2, compared to the 20% abnormally coolest months, in the 20% abnormally
warmest months people search more about global warming by 4.8%, or about
7.3% of its standard deviation (which is 66.5%, as shown in panel A).
Finally, we repeat all the regressions by replacing SV I in city i with SV I
in the country (for countries with more than one stock exchange, we pick
the exchange city with the largest total market capitalization; the regressions
include sixty-three countries). Panel A of Table 2 shows that the summary
statistics of the two SV I s are similar. Columns 3 and 4 report the regression

1123

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1123 1112–1145

 The Review of Financial Studies / v 33 n 3 2020

results, which are qualitatively similar to those at the city level. In later tests, we
have only country-level information on investor locations. Most exchange cities
are important cities with high populations, concentrated capital, and extensive
media coverage. Because abnormal temperature in the exchange city is strongly
associated with people’s attention at the country level, it seems reasonable to
use the city’s temperature to proxy for people’s experience in the country.
The Internet Appendix examines whether institutional investor and media
attention also change with local abnormal temperatures. We follow Ben-

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

Rephael, Da, and Israelsen (2017) to measure abnormal institutional investor
attention (AI A). AI A tracks how frequently Bloomberg users, who are likely
financial institutions, search and read information about a certain stock. While
we do not have users’ location, we can obtain information at the stock level
(in 45 exchanges). Table IA.4 shows the results. We do not find evidence that
abnormally warm weather leads to different levels of institutional attention
to high-emission and low-emission firms in the exchange. This finding is
in line with our trading results in Section 3.5 and our hypothesis that local
weather mostly affects retail investors. Using data from Raven Pack News
Analytics, Table IA.5 finds that media attention to high-emission firms do not
vary with local abnormal temperatures. However, most data from Raven Pack
News come from English-speaking media (which may not be located in the
exchange city) and are available at the stock level (so that generic stories about
climate change will not be captured). Future research can revisit these questions
if there are better measures of international institutional investor and media
attention.

3.2 Stock returns and local temperatures

Next, we examine whether local weather affects stock prices, focusing on the
differential reactions in the cross-section of firms. We first form two portfolios
according to the IPCC definitions described in Section 2. In each city i from
2001 to 2017, portfolio EMI SSI ONi includes all firms whose DataStream
industry group is mapped with the IPCC sectors. All remaining firms in city
i are assigned to portfolio CLEANi . A long-short portfolio EMCi (which
stands for Emission Minus Clean) is formed by buying EMI SSI ONi and
selling CLEANi . We construct all portfolios using equal weights and value
weights. Panel A of Table 3 shows the summary statistics. Size-adjusted returns
are reported.12 Figure 1 plots the average equal-weighted EMC size-adjusted
returns and the confidence intervals across five temperature quintiles in the

12 We use size-adjusted returns, because the market capitalization data obtained from DataStream have better
coverage than other financial data. Other models can calculate adjusted returns (e.g., a factor model based on
momentum and cashflow-to-price) (Hou, Karolyi, and Kho 2011). One disadvantage is that the sample size
would be greatly reduced when requiring other company information. Specifically, we check the availability of
the following variables in the Worldscope database: book/market, dividend/price, earnings/price, and long-term
debt/common equity. Requiring at least one of these variables will reduce the number of observations in our main
regression by 81%. Twelve of our seventy-four exchange cities have zero observations, and fifty-eight cities have
less than 10% of the observations remaining.

1124

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1124 1112–1145

 Attention to Global Warming

exchange city. We see a general decrease in EMC returns as we move up

the temperature quintiles, with statistically significant underperformance in the
warmest quintile. Summary statistics for raw returns (not adjusted for size) and
longer-term returns (up to 12 months) of EMC are also shown in panel A.
Similar in spirit to Hirshleifer and Shumway (2003) and Saunders (1993),
who examine the relationship between morning sunshine in a city and index
returns, we capture investors’ experience by using local abnormal temperature
in the city. We run the following regression:

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

EMC it = α +β1 Ab_Tempit +t YearMontht +it , (3)
where EMCit is the value-weighted or equal-weighted, size-adjusted or raw
return of the EMC portfolio in city i in month t (from 2001 to 2017), and
Ab_Tempit is the abnormal temperature in city i in month t based on the
decomposition in Equation (1). Year-month fixed effects are included.13
Panels B (equal-weighted) and C (value-weighted) of Table 3 offer the results.
Column 1 of Panel B shows that higher abnormal temperature is associated
with significantly lower EMC size-adjusted returns. A 1-standard-deviation
increase in Ab_Temp corresponds to a decrease of 16 bps in EMC return
(= −0.060×2.676). Column 2 replaces Ab_Temp with the quintile dummies
based on the city’s abnormal temperature. It shows that the negative effect on
EMC returns is the strongest in the highest temperature quintile, consistent
with our Google SV I results in Section 3.1. The economic impact is sizeable,
with a change from temperature quintile 1 (coolest) to quintile 5 (warmest)
corresponding to a drop of 48 bps (t-stat = −4.0) in size-adjusted return.
The results are similar when we consider raw returns (Columns 3 and 4).
Finally, Columns 5 and 6 study EMI SSI ON and CLEAN portfolio size-
adjusted returns, respectively. Relative to the city’s coldest temperature quintile,
EMI SSI ON (CLEAN ) earns significantly lower (higher) returns in the
warmest quintile, at the 1% significance level. Therefore, both portfolios
contribute to the low EMC returns in the warmest months. The results using
value-weighted returns are generally similar, as shown in panel C.
Next, we examine the long-term performance subsequent to an abnormally
warm month:
EMC i,t+1,t+n = α +β1 Ab_Tempit +t YearMontht +it , (4)
where n = {3,6,12} and the returns are measured from month t +1 to month
t +n. Year-month fixed effects are included. If β1 is negative or zero, it is
consistent with slow belief updating; investors with limited attention generally
overlook climate risk but recognize it when reacting to attention-grabbing
weather events. Otherwise, if β1 appears to be positive, it implies that part

13 Table IA.6 in the Internet Appendix runs these regressions at weekly and quarterly levels. We obtain similar
results that are statistically weaker.

1125

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1125 1112–1145

 The Review of Financial Studies / v 33 n 3 2020

Table 3
Emission-minus-clean portfolio return and abnormal temperature
A. Summary statistics
Variable Obs Mean SD P10 P25 P50 P75 P90
Equal-weighted
EMC 12,614 0.044 4.867 −3.528 −1.464 0.000 1.657 3.819
EMC(raw) 12,614 0.060 5.728 −4.189 −1.692 0.112 1.926 4.519
EMISSION 12,614 0.022 3.269 −2.060 −0.829 0.000 0.957 2.251
CLEAN 12,614 −0.022 2.002 −1.391 −0.605 0.000 0.535 1.368
EMCt+1,t+3 12,614 0.057 6.614 −5.605 −2.269 0.142 2.663 5.821

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

EMCt+1,t+6 12,614 0.122 9.089 −8.379 −3.326 0.272 4.157 8.532
EMCt+1,t+12 12,614 0.276 12.908 −12.533 −4.969 0.672 6.601 13.078
Value-weighted
EMC 12,614 0.100 5.999 −5.155 −2.210 0.047 2.507 5.609
EMC(raw) 12,614 0.117 6.710 −5.628 −2.415 0.121 2.713 6.096
EMISSION 12,614 0.032 4.263 −3.545 −1.536 0.001 1.693 3.776
CLEAN 12,614 −0.068 3.108 −2.936 −1.348 −0.019 1.192 2.813
Ab_Temp 12,614 0.307 2.676 −2.776 −1.142 0.306 1.746 3.446
B. Equal-weighted EMC returns
(1) (2) (3) (4) (5) (6)
EMC EMC(raw) EMISSION CLEAN
Ab_Temp −0.060∗∗∗ −0.068∗∗∗
(−3.34) (−2.67)
Ab_Temp Q2 −0.148 −0.297∗ −0.035 0.113∗
(−1.16) (−1.69) (−0.44) (1.90)
Ab_Temp Q3 −0.125 −0.316 −0.041 0.084
(−0.88) (−1.60) (−0.41) (1.47)
Ab_Temp Q4 −0.145 −0.212 −0.094 0.051
(−1.27) (−1.63) (−1.52) (0.90)
Ab_Temp Q5 −0.481∗∗∗ −0.614∗∗∗ −0.285∗∗∗ 0.196∗∗∗
(−4.04) (−3.82) (−3.35) (3.95)
Year × Month FEs Yes Yes Yes Yes Yes Yes
Obs. 12,614 12,614 12,614 12,614 12,614 12,614
Adj. R 2 .020 .020 .018 .018 .014 .022
C. Value-weighted EMC returns
(1) (2) (3) (4) (5) (6)
EMC EMC(raw) EMISSION CLEAN
Ab_Temp −0.055∗∗ −0.066∗∗
(−2.08) (−2.00)
Ab_Temp Q2 −0.211 −0.317 −0.069 0.142
(−1.13) (−1.37) (−0.63) (1.34)
Ab_Temp Q3 −0.337∗ −0.522∗∗ −0.202∗ 0.135
(−1.79) (−2.23) (−1.77) (1.33)
Ab_Temp Q4 −0.310∗ −0.441∗∗ −0.174 0.136
(−1.78) (−2.42) (−1.64) (1.52)
Ab_Temp Q5 −0.476∗∗∗ −0.574∗∗∗ −0.324∗∗∗ 0.152
(−3.06) (−2.81) (−3.10) (1.60)
Year × Month FEs Yes Yes Yes Yes Yes Yes
Obs. 12,614 12,614 12,614 12,614 12,614 12,614
Adj. R 2 .036 .036 .033 .033 .028 .032
At the beginning of month t , EMI SSI ON and CLEAN portfolios are formed based on firms’ industry code.
High-carbon-emission industries are defined following the IPCC’s report. Portfolio return (as a percentage) equals
the average adjusted return of stocks at month t , equal weighted or value weighted. Adjusted return equals raw
return minus the average return of stocks in the same size quintile by each exchange. EMC equals EMI SSI ON
minus CLEAN . EMC(raw) is calculated using raw returns. EMCt+1,t+3 , EMCt+1,t+6 , and EMCt+1,t+12
are calculated using adjusted returns over months t +1 to t +3, t +1 to t +6 and t +1 to t +12, respectively.
Panel A reports summary statistics. Panel B reports the results of regressions of EMC on contemporaneous
temperature variables using equal-weighted portfolio returns, and panel C uses value-weighted returns. The
sample is from January 2001 to December 2017. Standard errors are clustered by exchange city and year-month,
and the corresponding t -statistics are reported in parentheses. *p < .1; **p < .05; ***p < .01.

1126

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1126 1112–1145

 Attention to Global Warming

.4
.2
0
EMC

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

-.2
-.4

1 2 3 4 5
Quintile of Ab_Temp

Figure 1
EMC on abnormal temperature, 2001–2017
The figure presents the average EMC returns (equal-weighted and adjusted for year-month fixed effects, as a %)
by Ab_Temp quintiles with 95% confidence intervals using the sample for 2001–2017.

of the belief update in month t is irrational (overreaction) as the previous

price pattern has reversed. Table 4 presents the results. For brevity, only equal-
weighted EMC size-adjusted returns are reported in the main text. The Internet
Appendix (Table IA.7) reports value-weighted returns.
As shown in Columns 1, 3, and 5, the coefficients of Ab_Temp are statistically
insignificant. The coefficients of temperature quintiles in Columns 2, 4, and 6 do
not show a systematic pattern and are generally statistically insignificant. These
results indicate that there is no strong continuation or reversal in the 3 to 12
months after month t. (It is certainly possible that there is a return reversal after
12 months. With 17 years of data, longer term reversals are statistically difficult
to detect. We encourage future research to test whether the belief updating
process is rational or irrational with longer sample periods.)
A concern about the IPCC industry classification is that we may pick up
some industry effects. Although it is not obvious why such effects would
vary with local abnormal temperatures, we conduct three additional tests to
further confirm our previous results. First, we rerun the return regressions in
Equation (3) using an earlier sample period, 1983 (the beginning year of our
abnormal temperature measures) to 2000. Unlike panel B of Table 3 (in which
the sample period is 2001–2017), Columns 1 and 2 of Table 5 do not show any
systematic difference in EMI SSI ON and CLEAN portfolio returns under
different abnormal temperatures. Climate change was less of a global concern
and the scientific evidence was less conclusive before the 21st century. It is not

1127

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1127 1112–1145

 The Review of Financial Studies / v 33 n 3 2020

Table 4
Long-term EMC returns subsequent to abnormal temperature
(1) (2) (3) (4) (5) (6)
EMCt+1,t+3 EMCt+1,t+6 EMCt+1,t+12
Ab_Temp −0.048 −0.012 −0.001
(−1.40) (−0.38) (−0.01)
Ab_Temp Q2 −0.303 −0.070 −0.005
(−1.41) (−0.26) (−0.01)
Ab_Temp Q3 −0.189 −0.050 −0.126
(−1.44) (−0.21) (−0.31)

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

Ab_Temp Q4 −0.431∗∗ −0.127 −0.138
(−2.64) (−0.50) (−0.33)
Ab_Temp Q5 −0.358 −0.061 0.348
(−1.56) (−0.26) (0.82)
Year × Month FEs Yes Yes Yes Yes Yes Yes
Obs. 12,614 12,614 12,614 12,614 12,614 12,614
Adj. R 2 .034 .034 .048 .047 .053 .053
The table reports the results of regressions of EMCt+1,t+3 , EMCt+1,t+6 , and EMCt+1,t+12 on abnormal
temperature variables at month t . All EMC returns are calculated using the equal-weighted average of adjusted
returns. The sample is from January 2001 to December 2017. Standard errors are clustered by exchange city and
year-month, and the corresponding t -statistics are reported in parentheses. *p < .1; **p < .05; ***p < .01.

surprising that we observe the return pattern globally only after 2001 if this is
due to the awareness of climate risk.
Second, some high-emission industries’ returns may be correlated with
fluctuations in oil prices. Columns 3 to 6 separately examine all energy firms
(which are in the IPCC Energy sector) and other high-emission firms (which are
in the remaining four IPCC sectors). Both groups underperform when the city
is abnormally warm. The results observed among nonenergy industries confirm
that investors are more likely to react to different carbon emission levels than
to oil prices.
Our third test defines high- and low-emission firms by their MSCI carbon
emission scores. Because these scores are industry-adjusted, it is now possible
to have both high- and low-emission firms in the same industry, and this test
will not be driven by industry effects. Note that this analysis is performed with
a smaller sample (with 14 exchanges), as MSCI scores are only available since
2007 and cover only a subset of exchanges and firms. The results in Columns
7 and 8 are in line with those of our previous tables. EMC earns lower returns
when the city’s Ab_Temp is high, especially when it is in the highest quintile.
In Column 7, a 1-standard-deviation increase in Ab_Temp corresponds to a
decrease of 38 bps in EMC size-adjusted return (t-stat = −2.6).
Columns 9 and 10 of Table 5 conduct another robustness check. Some
exchange cities are small in total market capitalization and contain fewer firms.
The tests in Section 3.5 require equity ownership data from FactSet, which
generally covers larger stock exchanges. The EMC return results using this
subset are similar to those in the full sample. Finally, Columns 11 and 12
include dummy variables that indicate the country of the exchange city is the
ten largest in size in our sample. In our return regressions (Equation (3)),
we use the exchange city’s temperature to proxy for people’s experience in

1128

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1128 1112–1145

[16:27 30/1/2020 RFS-OP-REVF190092.tex]

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

Attention to Global Warming
Table 5
EMC return and abnormal temperature: Robustness tests
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Placebo Energy Nonenergy MSCI FactSet Top-10 areas

Ab_Temp 0.000 −0.084∗∗ −0.048∗∗ −0.143∗∗ −0.058∗∗ −0.075∗∗∗

(0.01) (−2.58) (−2.26) (−2.60) (−2.16) (−3.66)
Ab_Temp Q2 0.117 −0.126 −0.168 −0.415 −0.083 −0.149
(0.54) (−0.55) (−1.25) (−0.68) (−0.79) (−1.17)
Ab_Temp Q3 0.146 −0.282 −0.076 −0.853∗ 0.028 −0.126
(0.60) (−1.30) (−0.51) (−1.92) (0.11) (−0.88)
Ab_Temp Q4 0.115 −0.198 −0.116 −0.151 −0.136 −0.146
(0.65) (−1.07) (−0.94) (−0.36) (−0.92) (−1.28)
Ab_Temp Q5 −0.035 −0.402∗ −0.491∗∗∗ −0.935∗ −0.569∗∗ −0.507∗∗∗
(−0.18) (−1.89) (−3.60) (−1.99) (−2.50) (−3.66)
Top10_Area −0.033 −0.036
(−0.26) (−0.29)
Ab_Temp × Top10_Area 0.079∗∗
(2.36)
Ab_Temp Q5 × Top10_Area 0.156
(0.63)

Year × Month FEs Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Obs. 8,997 8,997 10,744 10,744 12,537 12,537 784 784 5,788 5,788 12,614 12,614
Adj. R 2 .015 .015 .041 .040 .015 .015 .258 .255 .034 .035 .020 .020

This table presents results of several robustness tests of the analysis in Table 3. Columns 1 and 2 are placebo tests: regressions of EMC on contemporaneous temperature variables from
January 1983 to December 2000. EMI SSI ON and CLEAN portfolios are formed based on firms’ industry codes. High-carbon-emission industries are defined following the IPCC’s report.
Portfolio return (as a percentage) is the equal weighted average adjusted return of stocks at month t . Adjusted return equals raw return minus the average return of stocks in the same size
quintile by each exchange. EMC equals EMI SSI ON minus CLEAN . In Columns 3 to 6, the EMI SSI ON portfolio is divided by Energy and Nonenergy firms, and EMC portfolio returns
are calculated for each group. In Columns 7 and 8, EMI SSI ON and CLEAN firms are categorized using MSCI ratings: EMI SSI ON includes stocks with carbon emission scores lower
than 3, whereas the CLEAN portfolio consists of stocks with carbon emission scores higher than 7. This sample is from January 2008 to December 2017 and includes only exchanges with
Page: 1129

more than thirty stocks covered by MSCI. In Columns 9 and 10, the sample includes only exchanges in the FactSet database. Columns 11 and 12 include dummy variables that indicate
the country of the exchange city is the ten largest in size in our sample (the countries are Russia, Canada, China (two exchanges), United States, Brazil, Australia, India, Argentina, and
Saudi Arabia). In all columns, standard errors are clustered by exchange city and year-month, and the corresponding t -statistics are reported in parentheses. *p < .1; **p < .05; ***p < .01.
1129
1112–1145
 The Review of Financial Studies / v 33 n 3 2020

the country. In larger countries where the population is more dispersed, this
proxy will be weaker. We find that the relationship between the city’s abnormal
temperature and emission-minus-clean portfolio returns is indeed weaker in
large countries.14
We explore other variables in our weather data in the Internet Appendix
(Tables IA.9 and IA.10): average wind speed, maximum wind speed,
precipitation, and snow depth. Abnormal weather conditions are defined in
the same way as in Equation (1). There is no evidence that Google DSV I

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

and EMC returns vary with these local conditions in a systematic manner.
While some contemporaneous research establishes a link between institutional
investors’ increased perception of climate risk and extreme weather events
in the United States (Gibson and Krueger 2018; Alok, Kumar, and Wermers
Forthcoming), we acknowledge that our current variables do not perfectly
capture the occurrences of hurricanes (or typhoons), floods, and droughts. We
invite future research to test these links again using better data on international
extreme weather.

3.3 Belief updating process

Overall, our findings suggest that the EMC return is negative in abnormally
warm weather. One might wonder why, after many years, we still see a
reaction in a warm month—why is this update so slow? We offer two potential
explanations. First, there is perhaps some reversal in beliefs. While we do not see
a statistically significant return reversal in the 12 months after the abnormally
warm month, this does not mean that there is no reversal in beliefs at all (it
may simply not be strong enough to be detected in the return data). Figure 1
offers more suggestive evidence: when we sort EMC portfolio returns into
five abnormal temperature quintiles, Quintile 1 (the coolest quintile) shows a
positive return, although it is not significant at the 5% level. We think some
people revise their beliefs downward in an abnormally cool month, for example,
President Donald Trump, as noted in the Introduction.
Additionally, it is possible that the learning process occurs at different times
in different countries, and hence we see a generally slow update when we study
exchange cities all over the world. The impact of climate change could have been
felt in a small subset of countries even before it became a global issue. Following
Hong, Li, and Xu (2019), we focus on countries that experienced increasing
drought trends (that were possibly exacerbated by rising temperatures) in the
second half of the 20th century. For example, in that period, Peru experienced
disrupted water supplies, especially in dry seasons, as the tropical glaciers of
the Cordillera Blanca shrank rapidly (Baraer et al. 2012). We identify these
countries by estimating the time trend of the Palmer Drought Severity Index

14 Table IA.8 in the Internet Appendix runs the return regressions by dropping all firms listed in the United States
(on the New York Stock Exchange), which constitute a large part of our sample and have geographically diverse
investor base. Our main results are robust to dropping firms listed in the United States.

1130

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1130 1112–1145

 Attention to Global Warming

Table 6
EMC return and abnormal temperature: Additional tests
(1) (2) (3) (4)
Early Late Early Late
Ab_Temp 0.035∗ −0.054∗∗
(1.81) (−2.32)
Drought 0.144 0.102 0.329 0.028
(0.96) (0.88) (1.50) (0.28)
Ab_Temp × Drought −0.153∗ 0.015
(−1.84) (0.43)
−0.100

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

Ab_Temp Q2 0.087
(0.35) (−0.93)
Ab_Temp Q3 0.225 −0.087
(0.83) (−0.53)
Ab_Temp Q4 0.192 −0.161
(1.04) (−1.42)
Ab_Temp Q5 0.290 −0.549∗∗∗
(1.47) (−3.29)
Ab_Temp Q5 × Drought −1.237∗∗ 0.416
(−2.02) (1.49)
Year × Month FEs Yes Yes Yes Yes
Obs. 7,804 9,131 7,804 9,131
Adj. R 2 .019 .034 .019 .034
At the beginning of month t, EMISSION and CLEAN portfolios are formed based on firms’ industry codes. High-
carbon-emission industries are defined following the IPCC’s report. Portfolio return (as a percentage) equals the
equal-weighted average adjusted return of stocks. Adjusted return equals the raw return minus the average return
of stocks in the same size quintile by each exchange. EMC equals EMISSION minus CLEAN. See Table 2 for
definitions of Ab_Temp and Ab_Temp Q2-Q5. Drought equals one if the country is ranked in the bottom quintile
based on long-term drought trends. Each country’s long-term drought trend is estimated based on the Palmer
Drought Severity Index (PDSI) from 1900 to 2014 following the method in Hong, Li, and Xu (2019). Columns 1
and 3 labeled as “Early” refers to a regression using the sample from January 1983 to December 2000, whereas
“Late” refers to January 2001 to December 2017. In all regressions, standard errors are clustered by exchange
city and year-month, and the corresponding t -statistics are reported in parentheses. *p < .1; **p < .05; ***p <
.01.

(PDSI) provided by Dai, Trenberth, and Qian (2004). The index measures
drought intensity based on a model developed by Palmer (1965). Regression
Equation (3) is run again with an interaction term of Ab_Temp and Drought,
using Early (1983–2000) and Late sample periods (2001–2017). Drought is
a dummy variable indicating countries that have a PDSI time trend in the lowest
quintile and that experience worsening droughts. (These countries are Austria,
Brazil, Israel, Italy, Japan, Peru, Saudi Arabia, Taiwan, and Venezuela.)
Table 6 shows the results. Column 1 shows that EMC returns react to
Ab_Temp more negatively among Drought countries in the period 1983–2000.
Column 3 presents similar findings in the highest Ab_Temp quintile. In 2001–
2017, the point estimates in Columns 2 and 4 suggest that these countries
show a slightly weaker response in EMC returns to high Ab_Temp. While our
full-sample results in Table 3 show that people typically update their beliefs
under warmer-than-usual temperatures in the period 2001–2017, countries that
have reacted earlier may now show weaker reactions—as most people in these
countries might have already become aware of climate change.15

15 A median of 74% of people across countries in Latin America agree that climate change is a very serious
problem in a 2015 survey conducted by The Pew Research Center (Stokes, Wike, and Carle 2015), compared

1131

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1131 1112–1145

 The Review of Financial Studies / v 33 n 3 2020

3.4 Mechanism of the pricing effect

The above sections suggest that some investors are reacting to local weather
conditions. Google search activity mostly originates from households, so we
expect that retail investors (the majority of whom are local) have limited
attention and react to abnormally warm local temperatures rather than to
fundamental information. Upon recognizing the effect of climate change,
they can update their beliefs about firms’ valuation or stay away from
climate-unfriendly stocks, as discussed in Section 1.

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

We confirm our conjecture in this section and the next section. First, we show
that the return patterns are not entirely attributed to fundamental information
about firms’ valuation. Using high-emission firms in exchange cities with at
least one foreign firm (forty-three cities), Columns 1 and 2 in Table 7, panel A,
run a stock-level regression on local abnormal temperatures and check whether
foreign firms listed on the local exchange show the same results.16 Firms whose
major operations are located in a foreign country will not see their production
harmed by local weather conditions, but their returns are still influenced by local
sentiment (see, e.g., Chan, Hameed, and Lau 2003). Although the stock-level
results are weaker than EMC portfolio results in Table 3, we see high-emission
firms’ stock returns significantly decrease with Ab_Temp. More important, there
is no significant difference between local and foreign firms in their price reaction
to abnormal temperatures. Columns 3 and 4 repeat the tests using ten exchange
cities with the highest proportion of foreign firms (10% of all listed firms
across these exchanges are foreign firms, and therefore the test of the difference
between local and foreign firms has higher statistical power). Again, there is no
significant difference, suggesting that the returns are driven by local investors.
Addoum, Ng, and Ortiz-Bobea (forthcoming) find that high temperature
shocks can negatively affect companies’ earnings in some industries. In
particular, the following industries’ earnings are harmed by extremely
warm temperatures: Electric Utilities, Leisure Products, Construction and
Engineering, Capital Markets, Gas Utilities, and Machinery. Four (Electric
Utilities, Construction and Engineering, Gas Utilities, and Machinery) of these
six industries are classified as high-emission industries according to our IPCC
classifications.
In Columns 1 and 2 of Table 7, panel B, we form the EMI SSI ON
portfolio using only firms in the above four industries (the CLEAN portfolio

with a global median is 54% (and 45% of Americans). Stokes, Wike, and Carle (2015) also note that fears of
drought are particularly prevalent in Latin America. One can link our main findings to the reaction to stale or
redundant information (Tetlock 2011; Huberman and Regev 2001). The scientific evidence of climate change is
abundant, whereas abnormal local temperatures carry little new information. In Drought countries, the reaction
to redundant information is weaker as people generally know more about the impact of climate change.
16 We do not construct EMC portfolios here, because most countries have too few foreign firms. We also run
stock-level regressions with the full sample in Table IA.11 in the Internet Appendix. The results are broadly
consistent with those of the portfolio regressions.

1132

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1132 1112–1145

 Attention to Global Warming

Table 7
EMC return and abnormal temperature: Additional tests
A. Local and foreign firms
(1) (2) (3) (4)
Dep. var.: Return All exchanges with foreign firms Top-10 exchanges
Ab_Temp −0.021∗ −0.029∗
(−1.79) (−2.09)
Foreign −0.493∗ −0.496∗ −0.390 −0.407
(−1.97) (−1.95) (−1.47) (−1.46)
Ab_Temp × Foreign −0.024 −0.008

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

(−0.91) (−0.32)
Ab_Temp Q2 −0.036 −0.139
(−0.43) (−1.00)
Ab_Temp Q3 −0.195∗∗ −0.149
(−2.46) (−1.36)
Ab_Temp Q4 −0.021 −0.033
(−0.28) (−0.25)
Ab_Temp Q5 −0.135 −0.228
(−1.29) (−1.48)
Ab_Temp Q5 × Foreign −0.008 0.070
(−0.06) (0.62)
Year × Month FEs Yes Yes Yes Yes
Obs. 808,211 808,211 515,466 515,466
Adj. R 2 .006 .006 .009 .009
B. High-emission firms that may be harmed by extremely warm temperatures
(1) (2) (3) (4)
Dep. var.: EMC Harmed Not harmed
Ab_Temp −0.086∗∗∗ −0.048∗∗∗
(−2.94) (−2.76)
Ab_Temp Q2 −0.356 −0.146
(−1.47) (−1.08)
Ab_Temp Q3 −0.309∗∗ −0.113
(−2.55) (−0.73)
Ab_Temp Q4 −0.398∗ −0.081
(−1.92) (−0.65)
Ab_Temp Q5 −0.590∗∗∗ −0.441∗∗∗
(−3.10) (−3.68)
Year × Month FEs Yes Yes Yes Yes
Obs. 11,789 11,789 12,513 12,513
Adj. R 2 .006 .006 .019 .019
(Continued)

remains the same).17 Columns 3 and 4 form the EMI SSI ON portfolio using
firms in all the remaining high-emission industries according to our IPCC
classifications. While the first set of high-emission firms may suffer from
negative earnings shocks, the second set is unlikely to be significantly affected
by high temperatures. We obtain statistically significant results in both tests.

17 They correspond to the following ICB industry codes in our paper: 7535 (Conventional Electricity), 2353
(Building Materials & Fixtures), 2357 (Heavy Construction), 3728 (Home Construction), 7573 (Gas Distribution),
and 2757 (Industrial Machinery).

1133

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1133 1112–1145

 The Review of Financial Studies / v 33 n 3 2020

Table 7
(Continued)
C. Carbon prices and environmental regulation scores
(1) (2) (3) (4)
EMC EMC EMC EMC
Ab_Temp −0.038 −0.096∗
(−1.32) (−1.86)
Ab_Temp × High_Price −0.015
(−0.27)
Reg_High 0.041 0.064

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

(0.30) (0.50)
Ab_Temp × Reg_High 0.045
(0.74)
Ab_Temp Q2 −0.186 −0.045
(−1.26) (−0.32)
Ab_Temp Q3 −0.148 −0.090
(−0.84) (−0.55)
Ab_Temp Q4 −0.090 −0.171
(−0.66) (−1.45)
Ab_Temp Q5 −0.325∗∗ −0.435∗∗
(−2.07) (−2.29)
Ab_Temp Q5 × High_Price −0.086
(−0.42)
Ab_Temp Q5 × Reg_High −0.075
(−0.22)
Year × Month FEs Yes Yes Yes Yes
Obs. 9,541 9,541 10,590 10,590
Adj. R 2 .016 .016 .023 .022
Panel A of this table reports results of regressions of individual stocks’ adjusted returns on contemporaneous
abnormal temperate. Adjusted return equals raw return minus the average return of stocks in the same size
quintile by each exchange. Foreign equals one if it is a foreign firm listed on the local exchange, and zero if it is
a local firm (firms with missing home country information are dropped). See Table 2 for definitions of Ab_Temp
and Ab_Temp Q2-Q5. Columns 1 and 2 include exchanges with foreign firms, and Columns 3 and 4 use ten
exchange cities with the highest proportion of foreign firms (Hong Kong, Singapore, London, Oslo, New York
City, Toronto, Frankfurt, Taipei, Sydney, and Stockholm). In panel B, at the beginning of month t, EMISSION
and CLEAN portfolios are formed based on firms’ industry codes. In Columns 1 and 2, high-carbon-emission
industries include Electric Utilities, Construction and Engineering, Gas Utilities, and Machinery. In Columns
3 and 4, high-carbon-emission industries include all the remaining high-emission industries according to the
IPCC’s report. Portfolio return (as a percentage) equals the equal-weighted average adjusted return of stocks.
EMC equals EMISSION minus CLEAN. In Panel C, High_Price is a dummy variable that equals one if the
carbon price is in the upper half of all months. Reg_High is a dummy variable that equals one if the regulation
environment score in the country is positive (twenty-seven countries), and zero if it is negative (twenty-five
countries); countries with missing scores are dropped. The sample is from January 2001 to December 2017. In
all regressions, standard errors are clustered by exchange city and year–month, and the corresponding t -statistics
are reported in parentheses. *p < .1; **p < .05; ***p < .01.

Taken together, panels A and B of Table 7 suggest that our overall findings are
not purely driven by adverse earnings news.18
However, investors may still update their private beliefs about firms’
valuation as they become more aware of climate risk, even if in the absence

18 The reaction is net of firms’ hedging activities. Addoum, Ng, and Ortiz-Bobea (forthcoming) design a test using
city-specific weather derivative introduction and discontinuation dates from the CME Group. They find that
hedging activities using weather derivatives in the United States have a small impact on earnings sensitivities
to temperatures. We believe that hedging activities of international firms have an even smaller impact, given
the absence of weather derivatives in most countries. (It is possible to hedge climate change risk using other
securities. For example, Engle et al. (forthcoming) propose to use a large panel of equity returns to build climate
change hedge portfolios.)

1134

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1134 1112–1145

 Attention to Global Warming

of any real change in firms’ valuation. While such beliefs are not observable,
we examine cases in which future cashflows of high-emission firms are more
harmed by climate change. If investors revise their estimates about future
cashflows, the revisions would be more prominent under these situations, and
we would observe a stronger negative link between high-emission firms’ returns
and local abnormal temperatures.
Table 7, panel C, analyzes two examples of these situations. Columns 1 and
2 study periods when the carbon futures price is above the sample median.

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

When the price of carbon is high, the production costs of carbon-intensive
firms increase. Columns 3 and 4 use the environmental regulatory regime index
developed by Esty and Porter (2001). We define a dummy variable, Reg_H igh,
to denote countries in which the index is positive (the index is positive in 35
exchange cities from 27 countries; the quality of the environmental regulatory
system is good in these countries). Tighter regulations in these places make
it more costly to emit carbon. We do not observe stronger reactions in EMC
returns to Ab_Temp when the carbon price is high and when the quality of
regulations is good.19 Therefore, there is no evidence that investors update
their beliefs about firms’ future cashflows in unusually warm months.

3.5 Trading behavior

Next, we turn to additional data to help us study the trading activity of
different types of investors. As described in Section 2, we obtain data on
blockholders’ ownership and institutional ownership from DataStream and
FactSet, respectively. Retail investors’ ownership is defined as (100% −
Blockholders’ ownership − Institutional ownership excluding blockholders).
Note that we do not observe retail ownership directly and the estimate is subject
to measurement errors.
Trading activity is the change in ownership between two quarters. For
example, if the retail ownership in a particular stock increases from 75% to
77% (of total shares outstanding), we infer that retail investors buy 2% in this
quarter. Similar to returns, we calculate the EMC_ for each type of investors,
defined as the average change in ownership in high-emission firms minus that
in low-emission firms. Panel A of Table 8 reports the summary statistics.
We run regressions similar to Equation (3):
EMC_it = α +β1 Ab_Tempit +t YearQuarter t +it , (5)
where EMC_it is the average net buy across all high-emission firms (in
exchange city i) minus the average net buy across all low-emission firms in

19 The index is calculated based on information available in 2001. Investors may react to expected changes in
future regulations instead. Table IA.12 in the Internet Appendix studies some of the most important international
events, namely, the establishment of the Copenhagen Agreement in December 2009 and the Paris Agreement in
November 2015 and the release of the IPCC Reports in February 2007 and September 2013, which might trigger
tighter future regulations because of coordinated efforts and advances in scientific research. We do not observe
a stronger link between EMC and Ab_Temp after these events.

1135

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1135 1112–1145

[16:27 30/1/2020 RFS-OP-REVF190092.tex]

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

1136

The Review of Financial Studies / v 33 n 3 2020

Table 8
EMC of trading and abnormal temperature
A. Summary statistics
% Obs Mean SD P10 P25 P50 P75 P90
EMC_Retail 2,008 −0.007 2.345 −0.942 −0.315 −0.015 0.298 0.924
EMC_Institution 2,008 0.006 0.317 −0.229 −0.070 0.000 0.076 0.245
EMC_DomInstitution 2,008 −0.004 0.222 −0.114 −0.032 0.000 0.025 0.107
EMC_ForInstitution 2,008 0.009 0.199 −0.138 −0.038 0.000 0.052 0.161
EMC_Blockholder 2,008 −0.003 2.344 −0.917 −0.271 0.000 0.279 0.884
EMC_DomBlockholder 2,008 0.000 2.098 −0.761 −0.201 0.000 0.194 0.698
EMC_ForBlockholder 2,008 −0.012 1.320 −0.214 −0.025 0.000 0.034 0.174
#Exchange cities 33
B. Stock trading on abnormal temperature
(1) (2) (3) (4) (5) (6)
EMC_Retail EMC_Institution EMC_Blockholder
Ab_Temp −0.080∗ 0.003 0.077∗∗
(−2.01) (0.84) (2.08)
Ab_Temp Q2 −0.102 0.012 0.064
(−0.67) (0.89) (0.40)
Ab_Temp Q3 −0.165 −0.001 0.132
(−1.18) (−0.05) (0.96)
Ab_Temp Q4 −0.316∗∗ 0.005 0.303∗∗
(−2.22) (0.31) (2.32)
Ab_Temp Q5 −0.396 0.037∗ 0.362
(−1.62) (1.76) (1.50)
Year × Quarter FEs Yes Yes Yes Yes Yes Yes
Page: 1136

Obs. 2,008 2,008 2,008 2,008 2,008 2,008

Adj. R 2 .006 .003 .021 .021 .003 .001
(Continued)
1112–1145
[16:27 30/1/2020 RFS-OP-REVF190092.tex]

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

Attention to Global Warming
Table 8
(Continued)
C. Stock trading for domestic and foreign investors
(1) (2) (3) (4) (5) (6) (7) (8)
EMC_DomInstitution EMC_ForInstitution EMC_DomBlockholder EMC_ForBlockholder
Ab_Temp 0.002 0.001 0.047∗ 0.023
(0.99) (0.46) (1.94) (0.94)
Ab_Temp Q2 0.007 0.002 0.003 0.036
(1.08) (0.24) (0.02) (0.39)
Ab_Temp Q3 0.017 −0.017 0.147 −0.040
(1.25) (−1.19) (1.69) (−0.48)
Ab_Temp Q4 −0.010 0.018∗∗ 0.277∗ −0.006
(−1.11) (2.07) (1.83) (−0.08)
Ab_Temp Q5 0.039 −0.001 0.128 0.172
(1.66) (−0.07) (1.31) (1.03)
Year × Quarter FEs Yes Yes Yes Yes Yes Yes Yes Yes
Obs. 2,008 2,008 2,008 2,008 2,008 2,008 2,008 2,008
Adj. R 2 .013 .016 .012 .013 −.006 −.007 .005 .005
At the beginning of quarter t, EMISSION and CLEAN portfolios are formed based on firms’ industry codes. High-carbon-emission industries are defined following the IPCC’s report.
EMC_Retail refers to the difference in the average changes in retail investors’ ownership over the quarter between emission and clean firms (as a percentage). EMC_Institution
and EMC_Blockholder refer to changes in ownership by institutional investors and blockholders, respectively. Blockholders are those who own more than 5% of shares outstanding,
whereas institutional investors include mutual funds, banks, and others but exclude those who are blockholders. Institutional investors and blockholders are further divided into
domestic and foreign investor categories. Panel A reports the summary statistics of key variables, and panels B and C report the results of regressions of the EMC of ownership
changes on abnormal temperature variables, which are defined in Table 2. In all regressions, year-quarter fixed effects are also included. The sample is from January 2001 to
December 2016. Standard errors are clustered by exchange city and year-quarter, and the corresponding t -statistics are reported in parentheses. *p < .1; **p < .05; ***p < .01.
Page: 1137

1137
1112–1145
[16:27 30/1/2020 RFS-OP-REVF190092.tex]

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

1138

The Review of Financial Studies / v 33 n 3 2020

Table 9
EMC of trading and abnormal temperature: Robustness test
A. Using MSCI carbon emission scores
(1) (2) (3) (4) (5) (6)
EMC_Retail EMC_Institution EMC_Blockholder
Ab_Temp −0.075 0.006 0.061
(−1.65) (0.24) (1.36)
Ab_Temp Q2 −0.215 0.037 0.244
(−0.79) (0.33) (1.09)
Ab_Temp Q3 0.028 −0.140 0.189
(0.12) (−1.03) (0.91)
Ab_Temp Q4 −0.175 −0.054 0.156
(−0.81) (−0.48) (0.79)
Ab_Temp Q5 −0.289∗∗∗ −0.067 0.296
(−3.21) (−0.42) (1.74)
Year × Quarter FEs Yes Yes Yes Yes Yes Yes
Obs. 312 312 312 312 312 312
Adj. R 2 .019 .006 −.003 −.007 .018 .005
(Continued)
Page: 1138
1112–1145
[16:27 30/1/2020 RFS-OP-REVF190092.tex]

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

Attention to Global Warming
Table 9
(Continued)
B. Stock trading for energy/nonenergy firms
Energy Nonenergy
EMC_Retail EMC_Institution EMC_Blockholder EMC_Retail EMC_Institution EMC_Blockholder
Ab_Temp −0.046∗ 0.019∗∗ 0.028 −0.082∗ 0.001 0.080∗
(−1.77) (2.55) (1.02) (−1.89) (0.22) (2.03)
Ab_Temp Q2 −0.030 0.026 −0.016 −0.085 0.006 0.053
(−0.16) (0.48) (−0.09) (−0.57) (0.45) (0.33)
Ab_Temp Q3 −0.287∗∗ 0.008 0.265 −0.145 −0.010 0.113
(−2.06) (0.19) (1.62) (−1.05) (−0.44) (0.79)
Ab_Temp Q4 −0.127 0.058 0.052 −0.308∗∗ −0.003 0.302∗∗
(−0.53) (1.19) (0.22) (−2.29) (−0.15) (2.46)
Ab_Temp Q5 −0.393∗ 0.067 0.337 −0.366 0.029 0.338
(−1.72) (1.43) (1.56) (−1.51) (1.23) (1.43)
Year × Quarter FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Obs. 1,708 1,708 1,708 1,708 1,708 1,708 2,008 2,008 2,008 2,008 2,008 2,008
Adj. R 2 .016 .016 .006 .003 .019 .020 .003 .000 .018 .018 −.000 −.003
In panel A, EMI SSI ON and CLEAN firms are categorized using MSCI ratings: EMI SSI ON includes stocks with carbon emission scores lower than 3, whereas the
CLEAN portfolio consists of stocks with carbon emission scores higher than 7. EMC_Retail refers to the difference in average changes of retail investors’ ownership
over the quarter between emission and clean firms (as a percentage). EMC_Institution and EMC_Blockholder refer to changes in ownership by institutional investors and
blockholders, respectively. Blockholders are those who own more than 5% of shares outstanding, whereas institutional investors include mutual funds, banks, and others but
exclude those who are blockholders. In panel B, EMI SSI ON and CLEAN firms are categorized based on firms’ industry codes. High-carbon-emission industries are defined
following the IPCC’s report. The EMI SSI ON portfolio is divided into energy and nonenergy firms, and the EMC portfolio is calculated for each group. In both panels, the
EMC of ownership changes is regressed on abnormal temperature variables, which are defined in Table 2, with year-quarter fixed effects. The sample is from January 2001 to
December 2016. Standard errors are clustered by exchange city and year–month, and the corresponding t -statistics are reported in parentheses. *p < .1; **p < .05; ***p < .01.
Page: 1139

1139
1112–1145
 The Review of Financial Studies / v 33 n 3 2020

quarter t, Ab_Tempit is the abnormal temperature in city i in quarter t (defined

in the same way as monthly Ab_Temp), and Y earQuartert are year-quarter
fixed effects. A separate regression is run for each type of investors. We also
run regressions by replacing Ab_Temp with abnormal temperature quintile
dummies.
Panel B of Table 8 reports the results for retail investors (Columns 1 and
2), institutional investors excluding blockholders (Columns 3 and 4), and
blockholders (Columns 5 and 6). In line with our expectation, retail investors

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

reduce their holdings in high-emission firms under abnormally warm weather;
in Column 1, the coefficient of Ab_Temp is negative and statistically significant
(t-stat = −2.0). Column 2 shows that retail investors reduce their EMC holdings
by 0.40% in the warmest temperature quintile compared to the coldest quintile.
We do not find evidence that institutional investors alter their EMC holdings
systematically according to local abnormal temperatures.20
Blockholders seem to respond to the decrease in stock prices and buy high-
emission firms as Ab_Temp increases. This finding is similar to the idea that
firms are buyers of last resort for their own stocks, and they repurchase shares
when prices drop below their fundamental value (Hong, Wang, and Yu 2008).21
Panel C of Table 8 shows the regressions for domestic institutions (Columns
1 and 2), foreign institutions (Columns 3 and 4), domestic blockholders
(Columns 5 and 6), and foreign blockholders (Columns 7 and 8). Only domestic
blockholders significantly increase their EMC ownership when the abnormal
temperature increases.
Table 9 repeats the tests using MSCI scores (instead of IPCC industries) to
define emission levels (panel A) and separately studies energy and nonenergy
high-emission sectors (panel B). Similarly, retail investors reduce their holdings
of high-emission firms in unusually warm quarters.22 Because retail investors
should not be more informed than domestic blockholders, we do not think

20 Possibly, some institutions are net buyers and some are net sellers, and they appear constant when we sum
them. In a survey to institutional investors around the world (Krueger, Sautner, and Starks forthcoming), most
respondents believe that climate risks have financial implications for their portfolios. Examining whether this
subset of investors reacts to local weather conditions will be interesting.
21 Blockholders’ trading patterns can be potentially explained two ways. First, high-emission firms are shunned by
some retail investors; like sin stocks, they should earn high expected returns (Hong and Kacpercyzk 2009). Table 3,
panel A, shows that EMI SSI ON firms earn higher size-adjusted returns than CLEAN firms unconditionally
(2.2 bps per month vs. −2.2 bps (equal-weighted) and 3.2 bps vs. −6.8 bps (value-weighted)). Second, attention-
induced price pressure reverses in the long run (although we cannot identify significant reversals in our tests).
To examine the second explanation, Table IA.13 in the Internet Appendix checks whether the price reaction to
abnormal temperatures is stronger under high investor attention. The evidence is mixed.
22 While retail investors may not fully understand the nuance of industry-adjusted MSCI scores, we believe that
they can identify some clean and emission firms, at least within some industries. For example, Toyota Motor
Corporation (Virgin America) is a CLEAN (EMI SSI ON ) firm according to MSCI scores. Toyota is widely
recognized for its efforts to reduce vehicle CO2 emissions, and it ranked top in Carbon Clean 200, a list of world’s
top clean companies. On the other hand, Virgin America has been criticized for its fuel efficiency by the media.
A study by the International Council for Clean Transport shows that Virgin America produces more greenhouse
gas emissions per passenger than other domestic U.S. carriers. Perhaps unsurprisingly, Japanese and U.S. retail
investors can recognize these firms relative to their industry peers, given the media reports. Nevertheless, we
urge the reader to interpret the results with caution.

1140

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1140 1112–1145

 Attention to Global Warming

retail investors are reacting to fundamental changes (confirming our argument

in Section 3.4: the return results are not entirely driven by negative cashflow
shocks). We interpret this as evidence that retail investors’ changes in beliefs
about climate risk move stock prices—they avoid holding high-emission firms
as they become more aware of global warming. Institutional investors and
blockholders, on the other hand, do not appear to update their beliefs in
abnormally warm weather.

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

4. Conclusion
Surveys of the scientific literature show a 97%–98% consensus among scientists
that humans are causing global warming (Cook et al. 2016, 2013; Anderegg et al.
2010; Oreskes 2004). Anthropogenic influence is evident from the emission
of greenhouse gases such as CO2 from human activities. Despite all these
scientific facts, not everyone treats climate risk seriously and reacts to it—a
U.S. survey (Marlon et al. 2016) estimates that only 70% of adults believe that
global warming is happening, and 40% think it will harm them personally.23
Global warming is an important long-term issue that requires collective action
from humans, not just from climate scientists, to address. Our paper aims to
understand how people update their beliefs about climate change.
One reason for the discrepancy between scientific findings and aggregate
beliefs is that people have limited attention. The effects of climate risk are
usually overlooked in normal times because they focus on attention-grabbing
weather events and personal experiences. Consistent with this idea, we show
that people revise their beliefs upward when the local temperature is abnormally
warm. Google search activity on the topic “global warming” is greater. In
financial markets, carbon-intensive firms underperform in the month in which
the exchange city is warmer than usual. We find that retail investors, rather
than institutional investors and blockholders, shun climate-unfriendly stocks
and seem to be responsible for these price patterns. While climate change is
a long-term trend, local temperatures are more noticeable even though they
contain negligible information about the global trend. Retail investors react
to salient but uninformative weather events, and their beliefs and actions are
reflected in prices and trading activity.
We document evidence that people in countries where the impact of climate
was more prominent in the past suffer less from limited attention. To increase
public awareness and the efficacy of climate campaigns, policies that reduce the
information gaps between the scientific community and the general public will

23 Opinions vary across different parts of the United States, which are reflected in housing prices in the neighborhood
(Baldauf, Garlappi, and Yannelis Forthcoming). Global surveys also suggest that climate change deniers are
present all over the world and that the proportion of deniers varies by country. Only 42% of surveyed adults
worldwide (53% in the United States, 57% in the United Kingdom, and 50% in France) see global warming as
a serious threat (results come from Gallup surveys in 111 countries in 2010). The Gallup report is available at
https://news.gallup.com/poll/147203/Fewer-Americans-Europeans-View-Global-Warming-Threat.aspx.

1141

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1141 1112–1145

 The Review of Financial Studies / v 33 n 3 2020

be helpful. For example, people are more concerned about flood risk after the
disclosure of high-resolution flood maps in Finland, resulting in a price drop
in coastal properties (Votsis and Perrels 2016). Although governments and
environmental organizations are not able to alter local weather conditions, they
can educate the public on climate risk. The findings in our paper suggest that
methods relating to personal and salient experiences (e.g., simulated extreme
weather events, maps of potential sea-level rise) will be more effective. When
aggregate beliefs are closer to the scientific consensus, we expect to see weaker

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

links between local abnormal temperatures and attention and stock prices, but
a more organized global effort to fight climate change.

References

Addoum, J., D. Ng, and A. Ortiz-Bobea. Forthcoming. Temperature shocks and earnings news. Review of
Financial Studies.

Agarwal, V., R. Vashishtha, and M. Venkatachalam. 2018. Mutual fund transparency and corporate myopia.
Review of Financial Studies 31:1966–2003.

Akerlof, K., E. Maibach, D. Fitzgerald, A. Cedeno, and A. Neuman. 2013. Do people “personally experience”
global warming, and if so how, and does it matter? Global Environmental Change 23:81–91.

Alok, S., N. Kumar, and R. Wermers. Forthcoming. Do fund managers misestimate climatic disaster risk? Review
of Financial Studies.

Anderegg, W., J. Prall, J. Harold, and S. Schneider. 2010. Expert credibility in climate change. Proceedings of
the National Academy of Sciences 107:12107–9.

Baldauf, M., L. Garlappi, and C. Yannelis. Forthcoming. Does climate change affect real estate prices? Only if
you believe in it. Review of Financial Studies.

Baraer, M., B. Mark, J. McKenzie, T. Condom, J. Bury, K. Huh, C. Portocarrero, J. Gomez, and S. Rathay. 2012.
Glacier recession and water resources in Peru’s Cordillera Blanca. Journal of Glaciology 58:134–50.

Ben-Rephael, A., Z. Da, and R. Israelsen. 2017. It depends on where you search: Institutional investor attention
and underreaction to news. Review of Financial Studies 30:3009–47.

Borick, C., and B. Rabe. 2014. Weather or not? Examining the impact of meteorological conditions on public
opinion regarding global warming. Weather Climate and Society 6:413–24.

Boud, D., R. Keogh, and D. Walker. 1985. Reflection: Turning experience into learning. London: Routledge
Falmer.

Broomell, S., D. Budescu, and H. Por. 2015. Personal experience with climate change predicts intentions to act.
Global Environmental Change 32:67–73.

Busse, M., D. Pope, J. Pope, and J. Silva-Risso. 2015. The psychological effect of weather on car purchases.
Quarterly Journal of Economics 130:371–414.

Cavanagh, P., C. Lang, X. Li, H. Miao, and J. Ryder. 2014. Searching for the determinants of climate change
interest. Geography Journal 2014:1–11.

Chan, K., A. Hameed, and S. Lau. 2003. What if trading location is different from business location? Evidence
from the Jardine Group. Journal of Finance 58:1221–46.

Chang, T., W. Huang, and Y. Wang. 2018. Something in the air: Pollution and the demand for health insurance.
Review of Economic Studies 85:1609–34.

Cohen, R., P. Gompers, and T. Vuolteenaho. 2002. Who underreacts to cashflow news? Evidence from trading
between individuals and institutions. Journal of Financial Economics 66:409–62.

1142

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1142 1112–1145

 Attention to Global Warming

Conlin, M., T. O’Donoghue, and T. Vogelsang. 2007. Projection bias in catalog orders. American Economic
Review 97:1217–49.

Cook, J., D. Nuccitelli, S. Green, M. Richardson, B. Winkler, R. Painting, R. W., P. Jacobs, and A. Skuce. 2013.
Quantifying the consensus on anthropogenic global warming in the scientific literature. Environmental Research
Letters 8:1–7.

Cook, J., N. Oreskes, P. Doran, W. Anderegg, B. Verheggen, E. Maibach, J. C., S. Lewandowsky, A. Skuce, S.
Green. 2016. Consensus on consensus: A synthesis of consensus estimates on human-caused global warming.
Environmental Research Letters 11:1–7.

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

Da, Z., J. Engelberg, and P. Gao. 2011. In search of attention. Journal of Finance 66:1461–99.

Da, Z., U. Gurun, and M. Warachka. 2014. Frog in the pan: Continuous information and momentum. Review of
Financial Studies 27:2171–218.

Dai, A., K. Trenberth, and T. Qian. 2004. A global dataset of Palmer Drought Severity Index for 1870–2002:
Relationship with soil moisture and effects of surface warming. Journal of Hydrometeorology 5:1117–30.

DeVault, L., R. Sias, and L. Starks. 2019. Sentiment metrics and investor demand. Journal of Finance 74:985–
1024.

Engle, R., S. Giglio, H. Lee, B. Kelly, and J. Stroebel. Forthcoming. Hedging climate change news. Review of
Financial Studies.

Esty, D., and M. Porter. 2001. Ranking national environmental regulation and performance: A leading indicator
of future competitiveness? In The Global Competitiveness Report 2001, eds. M. Porter and J. Sachs. New York:
Oxford University Press.

Ferreira, M., and P. Matos. 2008. The colors of investors’ money: The role of institutional investors around the
world. Journal of Financial Economics 88:499–533.

Gibson, R., and P. Krueger. 2018. The sustainability footprint of institutional investors. Working Paper, University
of Geneva.

Goetzmann, W., D. Kim, A. Kumar, and Q. Wang. 2015. Weather-induced mood, institutional investors, and
stock returns. Review of Financial Studies 28:73–111.

Griffin, J., J. Harris, and S. Topaloglu. 2003. The dynamics of institutional and individual trading. Journal of
Finance 58:2285–320.

Herrnstadt, E., and E. Muehlegger. 2014. Weather, salience of climate change and congressional voting. Journal
of Environmental Economics and Management 68:435–48.

Hirshleifer, D., and T. Shumway. 2003. Good day sunshine: Stock returns and the weather. Journal of Finance
58:1009–32.

Hong, H., and M. Kacpercyzk. 2009. The price of sin: The effects of social norms on markets. Journal of Financial
Economics 93:15–36.

Hong, H., F. Li, and J. Xu. 2019. Climate risks and market efficiency. Journal of Econometrics 208:265–81.

Hong, H., J. Wang, and J. Yu. 2008. Firms as buyers of last resort. Journal of Financial Economics 88:119–45.

Hou, K., G. Karolyi, and B. Kho. 2011. What factors drive global stock returns? Review of Financial Studies
24:2527–74.

Howe, P., E. Markowitz, T. Lee, C. Ko, and A. Leiserowitz. 2013. Global perceptions of local temperature change.
Nature Climate Change 3:352–56.

Huberman, G., and T. Regev. 2001. Contagious speculation and a cure for cancer: A nonevent that made stock
prices soar. Journal of Finance 56:387–96.

Ince, O., and R. Porter. 2006. Individual equity return data from Thomson DataStream: Handle with care! Journal
of Financial Research 29:463–79.

1143

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1143 1112–1145

 The Review of Financial Studies / v 33 n 3 2020

Intergovernmental Panel on Climate Change. 2014. Climate Change 2014 Synthesis Report. Intergovernmental
Panel on Climate Change, Geneva, Switzerland.

Joireman, J., H. Truelove, and B. Duell. 2010. Effect of outdoor temperature, heat primes and anchoring on belief
in global warming. Journal of Environmental Psychology 30:358–67.

Kahn, M., and M. Kotchen. 2011. Business cycle effects on concern about climate change: The chilling effect
of recession. Climate Change Economics 2:257–73.

Kamstra, M., L. Kramer, and M. Levi. 2003. Winter blues: A SAD stock market cycle. American Economic
Review 93:324–43.

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

Karolyi, G., and R. Stulz. 2003. Are financial assets priced locally or globally? In Handbook of the economics of
finance, eds. G. M. Constantinides, M. Harris, and R. M. Stulz, 975–1020. Amsterdam, the Netherlands: Elsevier.

Kolb, A. 1984. Experiential learning: Experience as the source of learning and development. Englewood Cliffs,
NJ: Prentice Hall.

Konisky, D., L. Hughes, and C. Kaylor. 2016. Extreme weather events and climate change concern. Climatic
Change 134:533–47.

Krey V., O. Masera, G. Blanford, T. Bruckner, R. Cooke, K. Fisher-Vanden, H. Haberl, E. Hertwich, E. Kriegler,
D. Mueller, S.Paltsev, L. Price, S. Schlömer, D. Ürge-Vorsatz, D. van Vuuren, and T. Zwickel. 2014. Annex II:
Metrics & methodology. In Climate change 2014: Mitigation of climate change. Contribution of Working Group
III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change, eds. 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, and J. C. Minx. Cambridge, UK: Cambridge
University Press.

Krueger, P., Z. Sautner, and L. Starks. Forthcoming. The importance of climate risks for institutional investors.
Review of Financial Studies.

Lang, C. 2014. Do weather fluctuations cause people to seek information about climate change? Climate Change
125:291–303.

Li, Y., E. Johnson, and L. Zaval. 2011. Local warming: Daily temperature change influences belief in global
warming. Psychological Science 22:454–59.

Lineman, M., Y. Do, J. Kim, and G. Joo. 2015. Talking about climate change and global warming. PLoS ONE
10:e0138996.

Loewenstein, G., T. O’Donoghue, and M. Rabin. 2003. Projection bias in predicting future utility. Quarterly
Journal of Economics 118:1209–48.

Malmendier, U., and D. Shanthikumar. 2007. Are small investors naive about incentives? Journal of Financial
Economics 87:457–89.

Marlon, J., P. Howe, M. Mildenberger, and A. Leiserowitz. 2016. Yale climate opinion maps — U.S. 2016.
https://climatecommunication.yale.edu/visualizations-data/ycom-us-2016/?est=happening&type=value&geo=
county.

Murfin, J., and M. Spiegel. Forthcoming. Is the risk of sea level rise capitalized in residential real estate? Review
of Financial Studies.

Myers, T., E. Maibach, C. Roser-Renouf, K. Akerlof, and A. Leiserowitz. 2012. The relationship between personal
experience and belief in the reality of global warming. Nature Climate Change 3:343–7.

Neukom, R., J. Gergis, D. Karoly, H. Wanner, M. Curran, J. Elbert, F. Gonzalez-Rouco, B. Linsley, A. Moy, I.
Mundo, C. Raible, E. Steig, T. van Ommen, T. Vance, R. Villalba, J. Zinke, and D. Frank. 2014. Inter-hemispheric
temperature variability over the past millennium. Nature Climate Change 4:362–7.

Oreskes, N. 2004. The scientific consensus on climate change. Science 306:1686.

Palmer, C. 1965. Meteorological drought. Research Paper, U.S. Department of Commerce.

Pierre-Louis, K. 2017. It’s cold outside. Cue the Trump global warming Tweet. New York Times, December 28.

1144

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1144 1112–1145

 Attention to Global Warming

Saunders, E. 1993. Stock prices and Wall Street weather. American Economic Review 83:1337–45.

Science. 2001. The science of climate change. Science 292:1261.

Schmidt, G. 2015. Thoughts on 2014 and ongoing temperature trends. RealClimate, January 22.
http://www.realclimate.org/index.php/archives/2015/01/thoughts-on-2014-and-ongoing-temperature-trends/.

Schmidt, G., D. Shindell, and K. Tsigaridis. 2014. Reconciling Warming Trends. Nature Geoscience 7:158–60.

Stokes, B., R. Wike, and J. Carle. 2015. Global concern about climate change, broad support for limiting emissions.
Report, Pew Research Center, Washington, DC.

Downloaded from https://academic.oup.com/rfs/article-abstract/33/3/1112/5735304 by University College Cork user on 26 May 2020

Tetlock, P. 2011. All the news that’s fit to reprint: Do investors react to stale information? Review of Financial
Studies 24:1481–512.

Votsis, A., and A. Perrels. 2015. Housing prices and the public disclosure of flood risk: A difference-in-differences
analysis in Finland. Journal of Real Estate Finance and Economics 53:450–71.

Zaval, L., E. Keenan, E. Johnson, and E. Weber. 2014. How warm days increase belief in global warming. Nature
Climate Change 4:143–7.

1145

[16:27 30/1/2020 RFS-OP-REVF190092.tex] Page: 1145 1112–1145