Environmental damage news and stock returns: Evidence from Latin America

Cavallo, Eduardo A.; Cepeda, Ana; Panizza, Ugo Working Paper Environmental damage news and stock returns: Evidence from Latin America IDB Working Paper Series, No. IDB-WP-1593 Provided in Cooperation with: Inter-American Development Bank (IDB), Washington, DC Suggested Citation: Cavallo, Eduardo A.; Cepeda, Ana; Panizza, Ugo (2024) : Environmental damage news and stock returns: Evidence from Latin America, IDB Working Paper Series, No. IDB-WP-1593, Inter-American Development Bank (IDB), Washington, DC, https://doi.org/10.18235/0012962 This Version is available at: https://hdl.handle.net/10419/299400 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/3.0/igo/ Environmental Damage News and Stock Returns: Evidence from Latin America Eduardo Cavallo Ana Cepeda Ugo Panizza WORKING PAPER No IDB-WP-1593 InterA merican Development Bank Department of Research and Chief Economist May 2024 * InterA merican Development Bank ** International Monetary Fund *** Geneva Graduate Institute and CEPR Environmental Damage News and Stock Returns: Evidence from Latin America Eduardo Cavallo* Ana Cepeda** Ugo Panizza*** InterA merican Development Bank Department of Research and Chief Economist May 2024 Cataloging-in-Publication data provided by the Inter-American Development Bank Felipe Herrera Library Cavallo, Eduardo A. Environmental damage news and stock returns: evidence from Latin America / Eduardo Cavallo, Ana Cepeda, Ugo Panizza. p. cm. — (IDB Working Paper Series ; 1593) Includes bibliographical references. 1. Carbon taxes-Latin America. 2. Climatic changes-Economic aspects-Latin America. 3. Environmental impact charges-Latin America. 4. Carbon dioxide mitigation-Economic aspects-Latin America. I. Cepeda, Ana. II. Panizza, Ugo. III. Inter-American Development Bank. Department of Research and Chief Economist. IV. Title. V. Series. IDB-WP-1593 http://www.iadb.org Copyright © 2024 Inter-American Development Bank ("IDB"). This work is subject to a Creative Commons license CC BY 3.0 IGO (https://creativecommons.org/licenses/by/3.0/igo/legalcode). The terms and conditions indicated in the URL link must be met and the respective recognition must be granted to the IDB. Further to section 8 of the above license, any mediation relating to disputes arising under such license shall be conducted in accordance with the WIPO Mediation Rules. Any dispute related to the use of the works of the IDB that cannot be settled amicably shall be submitted to arbitration pursuant to the United Nations Commission on International Trade Law (UNCITRAL) rules. The use of the IDB's name for any purpose other than for attribution, and the use of IDB's logo shall be subject to a separate written license agreement between the IDB and the user and is not authorized as part of this license. Note that the URL link includes terms and conditions that are an integral part of this license. The opinions expressed in this work are those of the authors and do not necessarily reflect the views of the Inter-American Development Bank, its Board of Directors, or the countries they represent. Environmental Damage News and Stock Returns: Evidence from Latin America Eduardo Cavallo Inter-American Development Bank Ana Cepeda International Monetary Fund Ugo Panizza Geneva Graduate Institute & CEPR May 2024∗ Abstract This paper studies the interplay between environmental performance and financial valuation of firms in Latin America and the Caribbean. We provide insights into how environmental considerations are integrated into financial decision-making and investor behavior by analyzing the stock market reaction to environmental news of firms with different levels of carbon emission intensity. We find that high emission intensity firms tend to underperform after the release of environmental damage news. Our baseline estimates indicate that, after the release of such news, firms at the 75th percentile of the distribution of emission intensity experience stock returns that are 17% lower than those of firms at the 25th percentile of the distribution of emission intensity. These results suggest that investors care about and price carbon risk, but only when this risk is salient. JEL classifications: G12, G14, G18, G32, G38, Q54 Keywords: Carbon emissions, Climate change, Environmental news, Stock returns ∗Contact information: cavallo[email protected]; acepedav[email protected]; [email protected]h. We would like to thank Martha Elena Delgado for outstanding research assistance and Martina Hengge and Richard Varghese for useful discussions on topics related to this paper. The views expressed in this paper are those of the authors and do not necessarily represent the views of the Inter-American Development and the IMF, their Executive Boards, or Management. 1 Introduction We investigate the impact of environmental damages news on the stock returns of firms in Latin America and the Caribbean. We focus on how this impact varies with firm-level carbon emission intensity. We find that high emission intensity firms underperform after the release of environmental damage news and show that this result is principally driven by domestic news. Our findings indicate that environmental factors play an important role in firms’ financial performance and investors’ decision-making. There is near unanimity on the anthropogenic origin of climate change and a better understanding of the link between carbon emissions and financial performance is key for devising policies aimed at creating incentives for reducing emissions. Firms, particularly in certain sectors, are significant contributors to releasing greenhouse gas (GHG).1These emissions are key drivers of global warming and climate change, phenomena that are altering the earth’s natural systems at an unprecedented rate. We focus on environmental damages because GHG emissions contribute to the intensification of extreme weather events, such as hurricanes, floods, and droughts, which have devastating ecological and economic consequences (see, for example, Min et al. (2011)) Moreover, rising global temperatures result in melting polar ice caps, rising sea levels, and destroying habitats. This cascade of effects disrupts biodiversity and poses substantial risks to wildlife, marine life, and natural ecosystems. The alteration in climate patterns also has profound implications for agriculture, water resources, and human health, creating a domino effect of environmental challenges. Even though environmental damages are not necessarily related to climate change (think of an oil spill), media coverage of such events can make the climate change discourse salient and impact investment decisions, either through a preference shift towards greener assets or heightened public awareness of environmental risks, affecting firms’ perceived future financial performance. In fact, there is evidence that investors perceive firms with higher carbon emissions as less environmentally responsible (Dahlmann et al., 2019), and this perception can lead to a more negative reaction after the release of environmental damage news, even if the specific piece of news is not directly directly related to climate change (Lee et al., 2015). To study the relationship between stock market performance, environmental damage news, and carbon emission intensity, we combine firm-level emission intensity data with daily stock return data and a specifically constructed high-frequency dataset of environmental damage news. Our analysis includes a sample of 840 firms based in 16 countries across Latin America and the Caribbean, covering a period of 13 years (2009-2022). By using an econometric specification that builds on Hengge et al. (2023), we find that the release of environmental damage news negatively impacts the stock market performance of firms with high emission intensities. We observe that firms at the 75th percentile of emission intensity distribution experience stock returns that are 1In the rest of the paper, we will use carbon emissions and green house gas (GHG) interchangeably. 2 17% lower compared to those at the 25th percentile. Our findings are primarily influenced by the extensive margin (i.e., the presence of at least one piece of news) and are mostly associated with the release of domestic rather than regional or global news. Additionally, our results are not attributed to a specific country, nor do they capture a spurious relationship driven by firm characteristics such as size, profitability, or sector of operation, which could be correlated with both carbon emission intensity and stock returns. Our paper is related to two strands of literature. The first strand focuses on on the link between carbon emissions and stock returns. Two influential papers by Bolton and Kacperczyk (2021,2022) find that the level of carbon emissions is associated with higher stock returns in a cross-section of firms. According to Bolton and Kacperczyk (2021,2022), this negative association is driven by the presence of a carbon risk premium: investors require higher returns to hold stocks of companies that are exposed to carbon pricing and regulation risk. This result has been challenged by Zhang (2023), who suggests that researchers should focus on lagged emissions, and by Aswani et al. (2023), who suggest that carbon emission intensity (instead of the level of emissions) should be the appropriate measure of carbon risk (for a rebuttal, see Bolton and Kacperczyk, 2023). Three papers that, like us, use high-frequency stock prices are Bolton et al. (2022), Millischer et al. (2022), and Hengge et al. (2023). These authors study the impact of carbon price on stock returns within the European Union Emission Trading Scheme (EU ETS). Bolton et al. (2022) and Millischer et al. (2022) show that firm-level availability of carbon allowances is an important driver of the link between carbon price and stock returns. These results indicate that, within the EU ETS, carbon price affects stock returns through an input cost channel. Hengge et al. (2023) use carbon policy surprises and show that transition risk also plays a role because tight carbon policy affects the performance of firms that do not participate in the EU ETS. The presence of a transition risk is also supported by research showing that carbon pricing shocks have a negative effect on stock returns of brown firms (Berthold et al., 2023). This strand of the literature highlights the different effect of carbon emission in equilibrium and in the aftermath of shocks. In equilibrium, green assets are expected to have lower return because they make it possible to hedge climate risk. However, green assets are expected to outperform brown assets in the aftermath of shocks that reveal the risk associated with the latter (for a theoretical model that clarifies this difference see Pástor et al., 2021) . The second strand is related to the effect of news, especially environmental news, on stock returns. Research that analyzes the link between news and stock returns goes back to at least Cutler et al. (1989), who cast doubts on the idea that stock price movements are fully driven by news about future cash flows and discount rates. More recent work uses sentiment analysis of news to show that media tone has an impact on stock prices and that the impact is generally larger for negative news (see, among others, Garcia, 2013, Zhang et al., 2016, Fraiberger et al., 2021). Jeon et al. (2022) show that the sensitivity of stock returns to news has been increasing 3 over time and it is stronger for firms with high media visibility. More closely related to our work are studies that focus on environmental news. CapelleBlancard and Laguna (2010) study stock market reactions to industrial disasters by using data on 64 explosions in chemical plants and refineries. They find that petrochemical firms experience a sharp drop in market value in the two days that follow the report of an explosion. Along similar lines, Carpentier and Suret (2015) use New York Times reporting to study how equity markets respond to 161 major environmental and non-environmental accidents. Contrary to previous studies, they do not find a long-lasting effect of environmental accidents on stock returns. Flammer (2013) uses corporate news related to environmental issues to conduct an event study and finds a positive association between the stance of environmental reporting and stock returns. She also documents that this relationship has become asymmetric over time, with a decreasing positive effect for eco-friendly behavior and an increasingly negative effect for eco-harmful behavior. Faccini et al. (2023) use textual and narrative analysis of climate change-related news and show that climate risk associated with government interventions has an effect on equity returns in the United States. Engle et al. (2020) use text analysis of newspaper coverage of climate change to build climate change risk-hedged portfolios. Ardia et al. (2020) build a climate change concern index based on news published in US newspapers and find that stock returns of firms exposed to this index are associated with firm-level carbon emission intensity. El Ouadghiri et al. (2021) use data on climate-related disasters together with US news on climate change and pollution and Google searches on these terms to show that public attention is positively associated with returns of sustainability-focused stock indices and negatively associated with returns of conventional stock indices. Bessec and Fouquau (2022, 2024) also focus on environmental news coverage in US newspapers and use textual analysis to classify these news items along different measures of tonality and uncertainty. They find that news releases about the environment with either a negative or uncertain tone lead to lower and more volatile stock returns for carbon-intensive firms. To the best of our knowledge, we are the first to study how carbon emission affect the link between environmental news and stock returns in the context of Latin America and the Caribbean, a region that is highly vulnerable to the negative impacts of climate change and environmental degradation. The rest of the paper is organized as follows. Section 2describes the data, with special focus on our purpose-built and novel series of climate damage news. Section 3describes our empirical strategy and presents our baseline estimations together with a battery of robustness checks. Section 4focuses sources of heterogeneity across types of news and countries. Section 5 discusses an alternative approach (based on the event study methodology) for estimating how carbon emission intensity affects the relationship between environmental damage news and stock returns. Section 6concludes. 4 2 Data To study the link between environmental damage news and stock returns we merge four types of data: i) firm-level yearly financial data from Refinitiv Datastream; ii) firm-level daily stock returns, also from Refinitiv Datastream; iii) firm-level yearly data on carbon emission intensity from Urgentem; and iv) country-level daily count data on news related to environmental damage from the Factiva Snapshots API. The Urgentem dataset provides firm-level annual data on emissions categorized under the Greenhouse Gas Protocol’s scopes 1, 2, and 3. Scope 1 emissions measure direct emissions from sources controlled by the firm. Scope 2 emissions measures indirect emissions associated with the purchase of electricity, steam, heat, or cooling. Scope 3 emissions encompass indirect emissions in a firm’s upstream and downstream value chain. Firm emissions are reported annually. Our key variable of interest is firm-level emission intensity, which is defined as the sum of scopes 1 and 2 emissions expressed in tons of carbon dioxide equivalent (tCO2e) per million dollars of revenue (tCO2e/$m revenue). Emissions data span the 2009-2022 period and cover 841 publiclylisted firms across 18 sectors in 16 countries in Latin America and the Caribbean (Figure 1). As emissions are disclosed with a one-year lag, our empirical analysis uses lagged emissions. The Factiva Snapshots Application Programming Interface (API) retrieves specific historical articles and news based on user-defined criteria interfaced with the Dow Jones premium publication archive of over 8,500 licensed news sources. Queries are formulated using Factiva’s code identifiers, which categorize articles into 2,012 regions (i.e., countries, states/provinces, municipalities, cities, economic or political unions, etc), 1,182 industries, 32 languages, 9,363 news sources, and 1,230 subjects. We use version 12.0 (updated on November 8, 2022) of the Dow Jones Intelligent Identifiers (DJID). DJID is a system developed by Dow Jones that uses a proprietary classification and taxonomy system to classify and tag the content of the Factiva business intelligence platform. Thanks to its internal consistency and standardized classification, DJIS allows us to search and retrieve data from sources that contain a vast amount of information. The identifiers are applied to Factiva using a mix of automated technologies (including Artificial Intelligence) and are then manually checked by Dow Jones’ coding specialists, who ensure that the identifiers are accurate and consistent in 28 languages. We targeted news classified under the subjects “Corporate Crime/Legal Action” and “Natural Environment.” The first subject focuses on corporate crimes, legal investigations, lawsuits, and court rulings. It includes cases in which a company is either the defendant or the plaintiff. Natural environment is defined as wildlife, climate, and natural resources. The news stories we examine are about activities affecting the environment and health hazards related to environmental issues. In order to reduce the risk of picking up positive environmental news, we exclude news items which are part of the two subjects mentioned above but are sub-classified under 5 investors, yield results which are essentially identical to those obtained when we only included firm-year fixed effects. If anything, the absolute value of the point estimate of βis slightly larger when we control for firm-quarter fixed effects (compare columns 1-3 of Table 2with columns 4-6 of the same table and also the two panels of Figure 2). As equation 2does not allow us to estimate the main effects of carbon emission and ED News, we also estimate a version of the model that substitute the firm-year (or firm-quarter) and country-day fixed effects, with firm and time fixed effects. The results for the interactive term corroborate the findings of Table 2, while the coefficients of the main effects are imprecisely estimated (see Appendix Table A5). As mentioned, equation 2implicitly assumes that carbon emission intensity is the only firmspecific variable that affects the relationship between ED News and stock returns. We now relax this assumption by allowing for a richer set of interactive effects and estimate the following equation: Ri(c),d(y)= (βCEIi(c),y−1+Xi(c),y−1Γ) ×Newsc,d(y)−1+ϕi(c),y +δd(y),c +εi(c),d(y)(2) where Xi(c),y−1is a matrix of time-varying firm characteristics that are potentially correlated with carbon emission intensity (the main effects of these variables are captured by the firm-year fixed effects ϕi(c),y). We start by augmenting the model with the interaction between ED news and firm size as measured by the log of total assets (in constant USD). We find that the interactive coefficient is negative and statistically significant (column 1 of Table A6). As large firms are likely to be more visible, this result is consistent with the findings of Jeon et al. (2022) that the sensitivity of stock returns is stronger for firms with high media visibility. More important for our purposes is the fact that controlling for the interaction between ED news and firm size has no effect on our parameter of interest: βremains negative, statistically significant and with a magnitude which is basically identical to that of our baseline estimates. Next, we measure firm size with log turnover and again find that large firms tend to underperform in the aftermath of ED News releases. Our parameter of interest is unchanged (column 2 of Table A6). Columns 3 and 4 of Table A6 indicate that more profitable firms (profitability is measured with the operating margin in %) perform better than the market after the release of ED news, while leverage does not matter. Again, our parameter of interest do not change after we include these controls. In column 5, we augment the model with the interaction between ED News and a rich set of sector fixed effects. This exercise sets a particularly high bar because the sector dummies absorb a substantial amount of the cross-firm variance in carbon emission. Nonetheless, our results are robust to augmenting the model with this large set of interaction. Finally, we include all the interactive effects in the same model. As before, we find that large firms tend to outperform after the release of ED News and that parameter of interest does not change (column 6 of Table A6). Results are also robust to estimating the equations of Table A6 by substituting firm-year 12 fixed effects with quarter-year fixed effects (see Table A6). Next, we explore whether the result that stock returns of high carbon intensity firm underperform after the release of ED News is driven by the extensive or intensive margin. As a first step, we re-estimate the models of Table 2by replacing the dummy that takes value one if there is at least one piece of ED News with a variable that measures the actual number of ED News in a given day. We find results that qualitatively similar to the baseline estimates (see Table 3). The point estimate of column 1 indicate that one additional ED News, is associated with stock returns for firms at the 75th percentile of carbon emission intensity are 0.12 basis points lower than stock returns for firms at the 25th percentile of the distribution (9% of average daily returns). As expected, the point estimates obtained when using the continuous measure of news are smaller (because the average value is larger), but the standardized coefficients are basically identical (the coefficient of column 1 of Table 2scaled by the standard deviation of the News dummy is 0.0025 and that of column 1 of Table 3scaled by the standard deviation of number of news is 0.0023). We also interact firm-level carbon emission intensity with both the dummy and the continuous measure of ED News. The first interaction captures the extensive margin and the second the intensive margin. We find that both coefficients are negative and that the two coefficients are jointly statistically significant (see the F tests at the bottom of Table 4). However, only the extensive margin is individually statistically significant. Taken at face value, the point estimates of column 1 Table 4indicate that, on days when there is exactly one ED news item, stock returns for firms at the 75th percentile of carbon emission intensity are 22 basis points lower than stock returns for firms at the 25th percentile of the distribution (16% of average daily returns); this difference increases to 29 basis points on days with 2 pieces of ED News and to 46 basis points on days with 5 pieces of ED News. 13 Table 3: Baseline Estimations: Intensive Margin This table reports a set of regressions where the dependent variable is daily stock returns and the main control variable is the interaction between firm-level carbon emission intensity and the release of environmental damage news. Newstmeasures the number of ED News released at time t. Columns 1 and 4 use the specification of equation 2, columns 2-3 and 5-6 have a richer lag structure. All regressions include country-day fixed effects. The models of columns 1-3 include firm-year fixed effects and the models of columns 4-6 include firm-quarter fixed effects. The standard errors are reported in parenthesis and are clustered at the firm level (1) (2) (3) (4) (5) (6) CEIy−1×Newst−1-0.000442*** -0.000958*** -0.00121*** -0.000424*** -0.00104*** -0.00132*** (0.000110) (0.000165) (0.000215) (0.000107) (0.000181) (0.000224) CEIy−1×Newst0.000336* 0.000435 0.000335** 0.000427 (0.000193) (0.000380) (0.000156) (0.000312) CEIy−1×Newst+1 0.000166 -1.55e-06 0.000135 -2.14e-05 (0.000150) (9.96e-05) (0.000115) (7.39e-05) CEIy−1×Newst−26.76e-05 -8.59e-06 (0.000456) (0.000500) CEIy−1×Newst+2 -8.56e-05 -0.000144 (0.000283) (0.000192) Const. -0.0953*** 0.788*** 1.157*** -0.0968*** 0.791*** 1.163*** (0.00886) (0.00980) (0.0120) (0.00864) (0.00657) (0.00750) N. Obs. 856,232 698,293 645,165 856,232 698,293 645,165 R-squared 0.193 0.189 0.187 0.201 0.199 0.199 Time-country FE ✓ ✓ ✓ ✓ ✓ ✓ Firm-year FE ✓ ✓ ✓ x x x Firm-quarter FE x x x ✓ ✓ ✓ *** p<0.01, ** p<0.05, * p<0.1 14 Table 4: Intensive versus Extensive Margin This table reports a set of regressions where the dependent variable is daily stock returns and the main control variable is the interaction between firm-level carbon emission intensity and the release of environmental damage news. We use both a dummy that takes value one after the release of at least one ED News pieces (this variable captures the extensive margin) and a variable that measures the number of ED News pieces (intensive margin). Column 1 includes time-country and firm-year fixed effects, Column 1 includes time-country and firm-quarter fixed effects, and column 3 include firm and time fixed effects. The bottom panel of the table reports the results of a series of F-tests for the joint significance of the extensive and intensive margin. The standard errors are reported in parenthesis and are clustered at the firm level. (1) (2) (3) CEIy−1×Newst−1(Dummy) -0.000607*** -0.000774*** -0.000494*** (9.28e-05) (9.00e-05) (9.20e-05) CEIy−1×Newst−1(Continuous) -0.000215 -0.000132 -0.000240 (0.000142) (0.000123) (0.000146) CEIy−1-5.23e-05*** (1.28e-05) Newst−1(Dummy) -1.598 (1.026) Newst−1(Continuous) 1.169*** (0.436) Const. -0.0844*** -0.0830*** -0.0963 (0.00902) (0.0101) (0.105) N. Obs 856,232 856,232 862,050 R-squared 0.193 0.201 0.102 Firm FE x x ✓ Time FE x x ✓ Time-country FE ✓ ✓ x Firm-year FE ✓x x Firm-quarter FE x ✓x F test 55.66 40.62 26.86 p value 0.000 0.000 0.00 *** p<0.01, ** p<0.05, * p<0.1 4 News and Country Heterogeneity Having established that stocks of firms with high carbon emissions tend to underperform after the release of ED News, we now explore heterogeneity across types of news and whether the results are driven by a particular country. As a first step, we estimate the models of columns 1 and 4 of Tables 2and 3separately for domestic, regional, and global news. We find that the interactive terms are always negative, but that they are statistically significant only for domestic news (see Tables A8,A9, and A10, in the Appendix; one coefficient is marginally significant in one of the four global news regression). The point estimates for the domestic news regression are similar (albeit slightly larger in absolute value) to those obtained when using all news, while the point estimates for regional and global news are marginally smaller. Next, we run a horse race that includes interactions between firm-level carbon emissions and 15 the three types of news. We also find in this case that the interactive terms are always negative but that the only interaction which is statistically significant is that with domestic news (see Table 5). In fact the point estimates and standard errors are similar to those obtained when we estimated the model with one type of news at a time (compare Table 5with Tables A8-A10). Table 5: Horserace Regressions This table reports a set of regressions where the dependent variable is daily stock returns and the main control variable is the interaction between firm-level carbon emission intensity and the release of environmental damage news from three sources: Domestic newspapers and magazines, Regional newspapers and magazines, and Global newspapers and magazines. All regressions include country-day fixed effects. The models of columns 1 and 4 include firm-year fixed effects and the models of columns 2 and 4 include firm-quarter fixed effects. Columns 1 and 2 use a dummy that takes value one after the release of at least one ED News. Columns 3 and 4 use a measure of the number of ED news. The standard errors are reported in parenthesis and are clustered at the firm level (1) (2) (3) (4) CEIy−1×DomesticNewst−1-0.000963*** -0.000978*** -0.000425*** -0.000398** (0.000307) (0.000315) (0.000159) (0.000156) CEIy−1×RegionalNewst−1-0.000611 -0.000682 -0.000637 -0.000722 (0.000839) (0.000780) (0.000685) (0.000637) CEIy−1×W orldNewst−1-0.000783 -0.000892 -0.000727 -0.000814 (0.000494) (0.000608) (0.000469) (0.000558) Const. -0.0841*** -0.0828*** -0.0947*** -0.0960*** (0.0112) (0.0121) (0.00972) (0.0101) N. Obs. 856,232 856,232 856,232 856,232 R-squared 0.193 0.201 0.193 0.201 Time-country FE ✓ ✓ ✓ ✓ Firm-year FE ✓x✓x Firm-quarter FE x ✓x✓ News is Dummy Dummy Continuous Continuous Robust standard errors clustered at firm level in parentheses *** p<0.01, ** p<0.05, * p<0.1 One possible source of concerns is that our sample is dominated by a small number of large countries. Brazilian firms account for 43% of the observations, and the top three countries (Brazil, Mexico, and Chile) account for 77% of observations. To make sure that the observations are not driven by an individual country or by a small set of countries, we re-estimate the baseline regressions by weighing each observation by the inverse of the number of observations in that specific country. Thus, each observation involving Brazilian firms has a weight of 1 368,215, and each observation involving Panamanian firms has a weight of 1 5925. With this weighting scheme, each country has exactly the same weight. The results, reported in Table 6, are essentially identical to the baseline findings of Tables 2and 3. They thus confirm that our findings are not driven by a particular country or group of countries with a large number of firms. As a further robustness check we re-estimate the baseline model (with and without inverse country weights) by dropping one country at a time. The results confirm that the findings are not driven by just one influential country (see Appendix Table A11). The point estimates become smaller (but still statistically significant at the 1 percent confidence level) when we 16 exclude Brazil and larger when we exclude Mexico (this is the country with the second largest number of observations, corresponding to 19% of the total). We also estimate country-by-country regressions for the 6 countries with the largest number of observations (Argentina, Brazil, Chile, Colombia, Mexico, and Peru). We find that the coefficient of interest is negative and statistically significant (or very close to being statistically significant in the case of Argentina) in four countries, negative but far from being statistically significant in Colombia and positive but close to zero in Chile (Appendix Table A12). Table 6: Weighted Regressions This table reports a set of regressions where the dependent variable is daily stock returns and the main control variable is the interaction between firm-level carbon emission intensity and the release of environmental damage news. All regressions include country-day fixed effects. The models of columns 1 and 4 include firm-year fixed effects and the models of columns 2 and 4 include firm-quarter fixed effects. Columns 1 and 2 use a dummy that takes value one after the release of at least one ED News piece. Columns 3 and 4 use a measure of the number of ED News pieces. Each observation is weighted by the inverse of the number of observations in its specific country. The standard errors are reported in parenthesis and are clustered at the firm level (1) (2) (3) (4) CEIy−1×Newst−1-0.000868*** -0.000887*** -0.000372** -0.000352** (0.000208) (0.000236) (0.000144) (0.000142) Const. 1.184*** 1.184*** 1.178*** 1.178*** (0.00417) (0.00471) (0.00463) (0.00455) N. Obs 856,232 856,232 856,232 856,232 R-squared 0.287 0.294 0.287 0.294 Time-country FE ✓ ✓ ✓ ✓ Firm-year FE ✓x✓x Firm-quarter FE x ✓x✓ News is Dummy Dummy Continuous Continuous Robust standard errors in parenthesis clustered at the firm level *** p<0.01, ** p<0.05, * p<0.1 5 Event Study As an alternative method to study how environmental damage news affect carbon intensive firms, we use a two-step event-study approach. In the first step, we conduct a classic event study and compute abnormal returns around the release of environmental damage news for all firms in our sample. In the second step, we look at the relationship between these abnormal returns and carbon emission intensity. To estimate abnormal returns associated with environmental damages news released on day T, we start by estimating the following regression for a 57 trading days window that starts at time T−60 and ends at time T−3: ri(c),d =α+βmc,d +ui(c),d (3) 17 where ri(c),d is the daily return of firm i(based in country c) on day dand mc,d is the market return in country con day d. Unlike most event studies, we have several events which are close to each other (as discussed in the data section, on average, we have more than two events per month). The problem with having events close to each other is that they contaminate the estimation window (in the typical event study there should no events in the estimation window). Consider, for instance, a situation in which there is one event on April 15 and one event on May 20. The estimation window for the first event starts in mid January (approximately 12 weeks before the event) and ends on April 13, and the estimation window for the second event starts in mid February and ends on May 17. There is thus a substantial overlap between the two estimation windows and the event of April 15 could affect the parameter estimates used to compute the excess returns for the May 20 event. In our case, the problem is even worse, as we have many cases with estimations windows that include multiple events. One way to address this problem is to exclude all cases in which there are events which overlap with another event’s estimation window. The problem with this strategy is that we would end up with a very small number of events. As an intermediate strategy, we allow for some overlap. Specifically, we include in our sample events with a maximum overlap of 10 days and exclude events with longer overlaps.5. This strategy yields a sample of 28,461 firm-events. We then use the parameter estimates of equation 3to obtain excess (“abnormal”) returns as out-of-sample forecast error over a 5-day event window that starts at T−2and ends at T+ 2: ari(c),d =ri(c),d −(ˆα+ˆ βmc,d)(4) and compute average accumulated average abnormal returns for event E as: AARi(c),E =1 5 T+2 X d=T−2 ari(c),d.(5) The ratio AAR σar√5(where σar is the standard deviation of ˆui(c),d in the estimation window) is a t-test on AAR. As some of our estimates yield very large excess return, we trim our data at 5 percent of the abnormal return variable, and we are left with a sample of 25,615 firm-events with average excess returns that range between -1,500 and 1,500 basis points. Nearly 20,000 (77 percent of the total) of these estimated excess returns are not statistically significant, 2,869 (11 percent of the total) are positive and statistically significant at the 5 percent confidence level, and 3,077 (12 percent of the total) are negative and statistically significant. Having built our sample of firm-event abnormal returns, we are now ready to test whether carbon-intensive firms are more likely to experience negative abnormal returns around release 5We explore with different maximum overlaps and obtain similar results 18 of environmental damages news.6We start by regressing abnormal returns over the log of carbon emission intensity and find a negative but not statistically significant coefficient (column 1 of Table 7). Lack of significance could be due to the fact that our dependent variable is imprecisely estimated (77 percent of observations are not statistically significant, often because of large confidence interval). One possible solution would be to concentrate the analysis on statistically significant returns. However, it would be arbitrary to include an observation with a t-test of, say, 1.97 and exclude an observation with a t-test of 1.95. As an alternative, we rescale our t-test to range between 0 and 1 and weight each observation by its own t-test. In this way we give more weight to precisely estimated abnormal returns and less weight to abnormal returns with a large confidence interval. The weighted regressions show a much stronger and statistically significant negative relationship between carbon emission intensity and abnormal returns. The point estimate indicate that a 1 percent increase in carbon emission intensity is associated with a 242 basis points decrease in abnormal returns in the days that surround the release of environmental damage news (column 2 of Table 7). The results are robust to substituting the log of carbon emissions with the level of carbon emissions (column 3 of Table 7). We find qualitatively similar results when we augment the model with country and event fixed effects (columns 4 and 5 of Table 7). However, controlling for fixed effects reduces the magnitude of the effect of carbon emission intensity on abnormal returns. The point estimates now suggest that a one percent increase in carbon emissions is associated with a 65 basis points decrease in abnormal returns on the 5 days that surround the release of environmental damages news. To probe further, we use two non-parametric approaches to allow for non-linearities in the relationship between abnormal returns and carbon emission intensity. We first compute average abnormal returns at different points in the distribution of carbon emissions. Specifically, we compute t-test weighted average abnormal returns for all firms in the bottom 10 percent of the distribution of carbon emissions, and we then move to firms in the bottom 20, 30, 40 and 50 percent of the distribution of carbon emissions. We also compute average abnormal returns for firms in the top 50, 60, 70, 80 and 90 percent of the distribution of carbon emissions.7Figure 4 plots the average values with their corresponding 95 percent confidence intervals and shows that firms in the bottom part of the distribution of carbon emission tend to have positive abnormal returns, while firms in the upper part of the distribution (and this is especially the case for firms in the top 20th and 10th percentile) tend to have negative abnormal returns. 6In order to avoid focusing on firms with extreme values of carbon emissions, we also trim carbon emissions at 5 percent. We are left with a sample of 23,119 firm-events. Of these firm-events, 5,327 (23 percent of the total) have statistically significant abnormal returns, 2,560 firm-events (11 percent of the total) have positive and statistically significant abnormal returns, and 2,767 (12 percent of the total) have negative and statistically significant abnormal returns. 7In practice, we run t-test-weighted regressions with no controls on different sub-samples and report the constant and its 95 percent confidence interval. 19 Table 7: Abnormal Returns and Carbon Emissions This table plots the result of a set of regression in which abnormal returns are regressed on firm-year level carbon emissions. The regressions of columns 2-5 are weighted by the absolute value of abnormal return t-statistics rescaled to range between 0 and 1. The regressions of columns 4-5 include country and event fixed effects. (1) (2) (3) (4) (5) ln(CEI)-0.209 -2.419** -0.649** (0.240) (1.057) (0.220) CEI -0.017*** -0.005** (0.004) (0.0018) Constant -2.478** 12.30*** 6.082*** 5.112*** 3.556*** (1.021) (3.990) (1.621) (0.883) (0.382) N. Obs 23,119 23,119 23,119 23,058 23,058 R2 0.000 0.018 0.030 0.289 0.289 Weights x ✓ ✓ ✓ ✓ Country Fixed Effects x x x ✓ ✓ Event Fixed effects x x x ✓ ✓ Robust standard errors in parenthesis clustered at country and event level *** p<0.01, ** p<0.05, * p<0.1 Figure 4: Average Abnormal Returns at Different Levels of Carbon Emissions This figure plots average abnormal returns with 95% confidence intervals for subsamples of firms at different levels of carbon emissions. CEI<P10 plots average abnormal returns for firms in the bottom 10% of the distribution of carbon emission intensity; CEI<P20 plots average abnormal returns for firms in the bottom 20% of the distribution of carbon emission intensity, and so on for CEI<P30, CEI<P40, and CEI<P50. CEI>P50 plots average abnormal returns for firms in the top 50% of the distribution of carbon emission intensity, CEI>P60 plots average abnormal returns for firms in the top 60% of the distribution of carbon emission intensity, and so on for CEI>P70, CEI>P80, CEI>P90. All averages are weighted by the absolute value of abnormal return t-statistics rescaled to range between 0 and 1. -15 -12.5 -10 -7.5 -5 -2.5 0 2.5 5 7.5 10 CEI<P10 CEI<P20 CEI<P30 CEI<P40 CEI<P50 CEI>P50 CEI>P60 CEI>P70 CEIP80 CEI>P90 We also estimate a series of regressions using dummies that take value one at different points of the distribution of carbon emissions. For instance, the regression reported in column 1 of Table 20 8includes a dummy that take value one for all firms in the bottom 10 percent of the distribution of carbon emission and value 0 for all other firms. Similarly, the regression of column 2 uses a dummy that takes value one for all the firms in the bottom 20 percent in the distributions. Columns 3 and 4 focus on the different extreme of the distribution and use dummies that take value 1 for firms in the top 20 and 10 percent of the distribution of carbon emissions. Columns 5 and 10 jointly include the bottom 10 percent and top 10 percent dummies. The interpretation of the regressions of Table 8is different from that of the average values of Figure 4. In the latter case, we are looking at average values in different points of the distribution of carbon emission intensity. In the former, we are testing for the difference of average values at different points of the distribution. For instance, the point estimates of column 1 in Table 8tell us that average abnormal returns around the release of environmental damage news of firms in the bottom 10 percent of the distribution of carbon emission intensity are 75 basis points higher than abnormal returns in the remaining 90 percent of firms, but the difference is not statistically significant. Columns 3 and 4, instead, indicate that abnormal returns are significantly lower for firms in the top 20 and 10 percent of the distribution of carbon emission intensity (the difference is 1,200 and 1,500 basis points, respectively). Column 5 shows that abnormal returns for firms in the bottom 10 percent of the distribution of carbon emission intensity are higher (but not significantly higher) than those of firms at the 10th-90th percentile of the distribution of carbon emission intensity and abnormal returns for firms in the top 10 percent are significantly lower than those of firms at the 10th-90th percentile of the distribution of carbon emission intensity. Columns 6-10 of Table 8are qualitatively similar when we control for country and event fixed effects. However, controlling for these variables results in lower point estimates. For instance, the difference in abnormal returns between firms in the top 10th percentile and firms in the bottom 90th percentile goes from 1,500 to 400 basis points. 21 Table A5: Estimations with Main Effects This table reports a set of regressions where the dependent variable is daily stock returns and the main control variables are firm-year level carbon emissions, country-day-levels ED News, and the interaction between these two variables. Columns 2 and 3 also include a richer lag structure. All regressions include firm and time fixed effects. The standard errors are reported in parenthesis and are clustered at the firm level. (1) (2) (3) CEIy−1×Newst−1-0.000866*** -0.000834*** -0.00123*** (0.000199) (0.000143) (0.000196) CEIy−1-5.20e-05*** -6.00e-05 -5.39e-05 (1.29e-05) (7.58e-05) (5.45e-05) Newst−10.434 -0.933 -4.242*** (0.725) (1.112) (1.304) CEIy−1×Newst0.000601 0.000558 (0.000464) (0.000499) CEIy−1×Newst+1 0.000382 0.000337 (0.000456) (0.000458) Newst5.553*** 3.833*** (1.150) (1.271) Newst+1 4.899*** 3.527*** (1.024) (1.169) CEIy−1×Newst−20.000206 (0.000746) CEIy−1×Newst+2 8.13e-05 (0.000552) Newst−2-6.661*** (1.330) Newst+2 -3.560*** (1.373) Const. -0.0975 0.223* 1.431*** (0.105) (0.127) (0.158) N. Obs 862,050 704,073 650,929 R-squared 0.102 0.100 0.098 Firm FE ✓ ✓ ✓ Time FE ✓ ✓ ✓ *** p<0.01, ** p<0.05, * p<0.1 28 Table A6: Multiple Firm-Level Interactions (firm-year FE) The sectors are: Accommodation and food (S1); Administration and Support (S2); Agriculture, forestry and fishing (S3); Construction (S4); Education (S5); Electricity, gas and steam (S6); Entertainment (S7); Financial and insurance (S8); Health and social work (S9); Information and communication (S10); Manufacturing (S11); Mining (S12); Professional and technical (S13); Public administration (S14); Real estate (S15); Transport and storage (S16); Water, sewage, and water management (S17); Wholesale and retail (excluded group). (1) (2) (3) (4) (5) (6) CEIy−1×Newst−1-0.0009*** -0.0009*** -0.0009*** -0.0009*** -0.0009*** -0.0009*** (0.00017) (0.00017) (0.00017) (0.00018) (0.00018) (0.00020) ln(Assetsy−1)×Newst−1-1.414*** -2.474*** (0.352) (0.868) ln(T urny−1)×Newst−1-0.982*** 0.977 (0.350) (0.850) Op. Marg.y−1×Newst−10.00019*** 1.61e-05 (0.00001) (0.0006) Leveragey−1×Newst−1-0.00306 -0.00290 (0.0114) (0.0118) S1×Newst−11.290 0.166 (4.525) (4.980) S2×Newst−13.855 7.802 (4.492) (4.965) S3×Newst−1-9.716* -7.952 (5.688) (5.119) S4×Newst−12.446 3.546 (3.362) (3.530) S5×Newst−15.538 6.296 (5.547) (5.791) S6×Newst−1-2.156 -0.658 (2.711) (3.035) S7×Newst−1-1.339 -2.835 (5.986) (5.978) S8×Newst−1-2.197 2.443 (2.771) (3.398) S9×Newst−1-2.837 -4.510 (4.263) (5.075) S10 ×Newst−12.253 4.201 (3.511) (3.756) S11 ×Newst−1-0.216 -0.233 (2.666) (2.742) S12 ×Newst−11.590 2.837 (3.469) (3.884) S13 ×Newst−1-4.192 -3.135 (4.568) (5.791) S14 ×Newst−1-3.801 -4.936* (2.678) (2.809) S15 ×Newst−14.061 5.797 (3.215) (4.143) S16 ×Newst−10.392 2.785 (3.853) (4.200) S17 ×Newst−11.763 5.934* (3.586) (3.355) N. Obs 794,286 785,416 782,125 761,958 856,232 753,871 R-squared 0.205 0.205 0.205 0.210 0.193 0.209 Time-country FE ✓ ✓ ✓ ✓ ✓ ✓ Firm-year FE ✓ ✓ ✓ ✓ ✓ ✓ Robust standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1 29 Table A7: Multiple Firm-Level Interactions (firm-quarter FE) The sectors are: Accommodation and food (S1); Administration and Support (S2); Agriculture, forestry and fishing (S3); Construction (S4); Education (S5); Electricity, gas and steam (S6); Entertainment (S7); Financial and insurance (S8); Health and social work (S9); Information and communication (S10); Manufacturing (S11); Mining (S12); Professional and technical (S13); Public administration (S14); Real estate (S15); Transport and storage (S16); Water, sewage, and water management (S17); Wholesale and retail (excluded group). (1) (2) (3) (4) (5) (6) CEIy−1×Newst−1-0.0009*** -0.001*** -0.001*** -0.001*** -0.001*** -0.0009*** (0.0002) (0.0002) (0.0002) (0.00019) (0.00021) (0.00022) ln(Assetsy−1)×Newst−1-1.243*** -2.372*** (0.363) (0.883) ln(T urny−1)×Newst−1-0.840** 1.048 (0.353) (0.876) Op. Marg.y−1×Newst−10.00014** -0.000165 (0.00001) (0.00067) Leveragey−1×Newst−1-0.00614 -0.00572 (0.0111) (0.0115) S1×Newst−11.180 0.184 (4.850) (5.306) S2×Newst−12.103 6.421 (4.582) (5.136) S3×Newst−1-8.903 -6.845 (5.480) (4.802) S4×Newst−12.725 4.104 (3.358) (3.483) S5×Newst−13.822 4.776 (5.555) (5.833) S6×Newst−1-1.748 -0.176 (2.719) (3.024) S7×Newst−1-3.334 -4.286 (5.111) (5.178) S8×Newst−1-1.789 2.936 (2.751) (3.328) S9×Newst−1-2.580 -3.875 (4.760) (5.561) S10 ×Newst−13.828 5.928 (3.408) (3.631) S11 ×Newst−10.587 0.738 (2.643) (2.707) S12 ×Newst−13.877 5.547 (3.525) (3.982) S13 ×Newst−1-4.058 -3.553 (4.495) (6.231) S14 ×Newst−1-6.109** -7.181** (2.695) (2.803) S15 ×Newst−14.728 6.754 (3.221) (4.126) S16 ×Newst−10.544 3.276 (3.935) (4.225) S17 ×Newst−11.536 5.610 (3.796) (3.859) N. Obs 794,286 785,416 782,125 761,958 856,232 753,871 R-squared 0.213 0.214 0.214 0.218 0.201 0.217 Time-country FE ✓ ✓ ✓ ✓ ✓ ✓ Firm-quarter FE ✓ ✓ ✓ ✓ ✓ ✓ Robust standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1 30 Table A8: Domestic News This table reports a set of regressions where the dependent variable is daily stock returns and the main control variable is the interaction between firm-level carbon emission intensity and the release of environmental damage news in domestic newspapers and magazines. All regressions include country-day fixed effects. The models of columns 1 and 4 include firm-year fixed effects and the models of columns 2 and 4 include firm-quarter fixed effects. Columns 1 and 2 use a dummy that takes value one after the release of at least one ED News piece. Columns 3 and 4 use a measure of the number of ED News pieces. The standard errors are reported in parenthesis and are clustered at the firm level. (1) (2) (3) (4) CEIy−1×Newst−1-0.000977*** -0.000992*** -0.000427*** -0.000399*** (0.000254) (0.000262) (0.000148) (0.000145) Const. -0.0882*** -0.0875*** -0.100*** -0.102*** (0.0111) (0.0115) (0.0108) (0.0106) N. Obs 856,232 856,232 856,232 856,232 R-squared 0.193 0.201 0.193 0.201 Time-country FE ✓✓✓✓ Firm-year FE ✓x✓x Firm-quarter FE x ✓x✓ News is Dummy Dummy Continuous Continuous *** p<0.01, ** p<0.05, * p<0.1 Table A9: Regional News This table reports a set of regressions where the dependent variable is daily stock returns and the main control variable is the interaction between firm-level carbon emission intensity and the release of environmental damage news in regional newspapers and magazines. All regressions include country-day fixed effects. The models of columns 1 and 4 include firm-year fixed effects and the models of columns 2 and 4 include firm-quarter fixed effects. Columns 1 and 2 use a dummy that takes value one after the release of at least one ED News piece. Columns 3 and 4 use a measure of the number of ED News pieces. The standard errors are reported in parenthesis and are clustered at the firm level. (1) (2) (3) (4) CEIy−1×Newst−1-0.000771 -0.000832 -0.000675 -0.000756 (0.000863) (0.000818) (0.000724) (0.000676) Const. -0.128*** -0.128*** -0.128*** -0.127*** (0.00373) (0.00354) (0.00355) (0.00331) N. Obs 856,232 856,232 856,232 856,232 R-squared 0.193 0.201 0.193 0.201 Time-country FE ✓ ✓ ✓ ✓ Firm-year FE ✓x✓x Firm-quarter FE x ✓x✓ News is Dummy Dummy Continuous Continuous Robust standard errors clustered at firm-level in parentheses *** p<0.01, ** p<0.05, * p<0.1 31 Table A10: Global News This table reports a set of regressions where the dependent variable is daily stock returns and the main control variable is the interaction between firm-level carbon emission intensity and the release of environmental damage news in global newspapers and magazines. All regressions include country-day fixed effects. The models of columns 1 and 4 include firm-year fixed effects, and the models of columns 2 and 4 include firm-quarter fixed effects. Columns 1 and 2 use a dummy that takes value one after the release of at least one ED News piece. Columns 3 and 4 use a measure of the number of ED News pieces. The standard errors are reported in parenthesis and are clustered at the firm level. (1) (2) (3) (4) CEIy−1×Newst−1-0.000690* -0.000817 -0.000672 -0.000780 (0.000417) (0.000524) (0.000421) (0.000514) Const. -0.129*** -0.129*** -0.129*** -0.129*** (0.00113) (0.00142) (0.00133) (0.00162) N. Obs. 856,232 856,232 856,232 856,232 R-squared 0.193 0.201 0.193 0.201 Time-country FE ✓ ✓ ✓ ✓ Firm-year FE ✓x✓x Firm-quarter FE x ✓x✓ News is Dummy Dummy Continuous Continuous Robust standard errors clustered at firm-level in parentheses *** p<0.01, ** p<0.05, * p<0.1 Table A11: Regression Results with Country Exclusions This table reports the coefficients and t-statistics of the interaction between firm-level carbon emission intensity and the release of environmental damage news using the same models as column 4 of Table 2and column 2 of Table 6but by dropping one country at a time. Excluded Not Weighted Weighted Country Coefficient t statistics Coefficient t statistics N. Obs Argentina -0.0010 4.691 *** -0.0008 3.292 *** 4,102,850 Bahamas -0.0010 4.975 *** -0.0009 3.768 *** 4,441,666 Brazil -0.0005 3.303 *** -0.0005 2.608 ** 2,869,348 Chile -0.0010 5.262 *** -0.0009 3.900 *** 3,679,330 Colombia -0.0010 4.947 *** -0.0009 3.680 *** 4,251,082 Costa Rica -0.0010 4.974 *** -0.0009 3.767 *** 4,441,666 Dominican Republic -0.0010 4.973 *** -0.0009 3.766 *** 4,446,960 Ecuador -0.0010 4.983 *** -0.0009 4.105 *** 4,436,372 Honduras -0.0010 4.973 *** -0.0009 3.766 *** 4,446,960 Jamaica -0.0010 4.973 *** -0.0009 3.766 *** 4,304,022 Mexico -0.0012 18.62 *** -0.0012 10.15 *** 3,626,390 Panama -0.0010 4.976 *** -0.0009 3.787 *** 4,394,020 Peru -0.0010 4.876 *** -0.0009 3.584 *** 4,092,262 Trinidad and Tobago -0.0010 4.971 *** -0.0009 3.741 *** 4,372,844 Uruguay -0.0010 4.973 *** -0.0009 3.766 *** 4,446,960 Venezuela -0.0010 4.974 *** -0.0009 3.774 *** 4,436,372 32 Table A12: Country by Country Regressions This table reports the coefficients and t-statistics of the interaction between firm-level carbon emission intensity and the release of environmental damage news based on country-by-country regressions using the same models as column 4 of Table 2but with time fixed effects instead of country-time fixed effects. Country Coefficient p value Argentina -0.0131 0.108 Brazil -0.0012 0.000 Chile 0.0001 0.909 Colombia -0.0030 0.479 Mexico -0.0004 0.013 Peru -0.0025 0.099 33 Appendix B: Factiva Professional News Examples B1: Review of the Top 15 News Results Obtained from our Selected Query Code in Factiva Professional Figure A1: Domestic News for Chile The selected query topic filter is: ((‘Natural Environment’ and ‘Corporate Crime/Legal Action’) and not (‘Carbon Sequestration’ or ‘Energy Efficiency’ or ‘Environmental Protection’)) Filter criteria (including selected query topic): Sources: All domestic news sources from Chile, Related country article content: Chile, Date range: 01/01/2009 – 12/31/2022, Language: Spanish 34 Figure A2: Domestic News for Panama The selected query topic filter is: ((‘Natural Environment’ and ‘Corporate Crime/Legal Action’) and not (‘Carbon Sequestration’ or ‘Energy Efficiency’ or ‘Environmental Protection’)). Filter criteria (including selected query topic): Sources: All domestic news sources from Panama, Related country article content: Panama, Date range: 01/01/2009 – 12/31/2022, Language: Spanish 35 B2: Review of News Articles Identified through API News Counts using Factiva Professional Figure A3: Filtered news search based on API count results: domestic sources for Chile The selected query topic filter is: ((‘Natural Environment’ and ‘Corporate Crime/Legal Action’) and not (‘Carbon Sequestration’ or ‘Energy Efficiency’ or ‘Environmental Protection’)). Filter criteria (including selected query topic): Source: Domestic news source “La Tercera,” Related country article content: Chile, Date: 11/07/2022 Note: This news article was translated from Spanish (original) to English, powered by Google Translate. 36 Figure A4: Filtered News Search Based on API Count Results: Domestic Sources for Honduras The selected query topic filter is: ((‘Natural Environment’ and ‘Corporate Crime/Legal Action’) and not (‘Carbon Sequestration’ or ‘Energy Efficiency’ or ‘Environmental Protection’)) Filter criteria (including selected query topic): Source: Domestic news source “Criterio,” Related country article content: Honduras, Date: 07/28/ 2022 Note: This news article was translated from Spanish (original) to English, powered by Google Translate. Figure A5: Filtered News Search Based on API Count Results: Regional Sources for Argentina Filter criteria (including selected query topic): Source: Regional news source “El Mercurio (Chile),” Related country article content: Argentina, Date: 10/23/2015 Note: This news article was translated from Spanish (original) to English, powered by Google Translate. 37