Emissions are also inherently backward-looking, while the impact of climate risk on businesses is forward-looking. These results reinforce those reported in Berkman et al. 2019), which focuses on the regulatory risk component of climate risk disclosures. H1: Market valuations are negatively associated with the extent of climate risk disclosure in companies' 10-Ks.
Second, it can be argued that given the qualitative nature of the required disclosure and diffuse nature of climate risk, managers themselves may choose to (not) disclose climate risk information (see Matsumura et al. 2014). As we argued before, despite the mandatory disclosure of material climate risk, the qualitative nature of the required disclosure and diffuse nature of climate risk may allow managers to self-select to (not) disclose information about climate risk exposure.
Market Response to Major Weather Events
PROPDISCLi,t = ratio of the number of firms in firm i's Fama-French 48 industry classification that provide climate risk disclosure to the total number of firms in firm i's industry in fiscal year t in our sample;. However, it is possible that there will be no change in the value of our climate risk hedging portfolio around our main events. We form portfolios, where we select the one-third of companies with the highest climate risk exposure for inclusion in the long portfolio (HIGH), and the companies with no climate risk score for inclusion in the short portfolio (LOW).
From Cere's website, we get a summary measure based on text analysis of climate risk discussions in companies' 10-K reports. The climate risk data is available on the Ceres website (http://www.ceres.org/resources/tools/sec-climate-disclosure/sec-klimaat-disclosure). Our climate risk measure (ClimateRisk) and the standardized measure (NClimateRisk) are derived from the Ceres/CookESG risk score.
20 The SEC (2010) guidelines have had a significant impact on companies' propensity to provide climate risk disclosures. Prior to our sample period, less than 24% of firms discussed climate risk in their 10-K (Doran and Quinn 2009).
Descriptive statistics
Our standardized measure of climate risk, NClimateRisk, ranges from -1.14 to 4.08.21 Fifty-six percent of the firm-years (untabulated) include reports of some aspect of climate risk. Although many of the correlations between model variables and ClimateRisk and NClimateRisk are statistically significant, with the exception of emissions, none of them are economically large.
Results
Market value and exposure to climate risk
Based on the average share price for our sample of $25.54, this corresponds to 7.8% of the average share price. Because climate risk is industry-related and changes over time, our remaining analyzes use NClimateRisk as our measure of climate risk. Much of the previous literature examining climate risk focuses on the detection of carbon emissions, which primarily affect valuations through their impact on regulatory costs.
The results of this assessment are in column (3) and are consistent with our baseline model results, proving that market valuation of climate risk is not limited to valuing potential regulatory costs. In Table 5, we extend the base model to explore the utility of NClimateRisk relative to other measures used in the literature. A disadvantage of including the environmental performance measures in the model is that our sample size is reduced by approximately half due to missing KLD data.
Because our results are robust and include a measure of environmental performance, we do not control for environmental performance in the remaining models to preserve sample size. We note that these risk measures not only do not capture our measure, but also impose severe limitations on the sample size due to limited data coverage.23 Overall, we conclude that NClimateRisk provides a useful summary measure of the firm-specific climate risk exposure that is available for a wide range of companies and explains the variation in market valuations to a greater extent than measures currently used in the literature.
Self-Selection to Disclose Climate Risk
Based on these results, we conclude that our results on climate risk are not due to our own disclosure choices. As an alternative to Ohlson's (1995) model, we examine the relationship between firm value and climate risk exposure using a valuation model based on Tobin's Q, commonly used in the finance and management literature:24. Qi,t = the sum of the market capitalization of company i and the book value of debt, divided by the book value of total assets in year t;.
ROAi,t = firm's profits divided by the book value of total assets at the end of fiscal year t;. LEVi,t = long-term debt scaled by the market value of common equity for the firm at the end of the fiscal year. CapExpi,t = capital expenditures for buildings and equipment scaled by total assets for firm i in fiscal year t;.
R&Di,t = research and development expenses scaled by total assets for firm i in fiscal year t;.
Event Study Results
Table 7, Panel A reports the results of estimation equation (3), in which we examine the returns of a hedge portfolio based on climate risk for significant climate-related events. Column (1) presents the results where each weather event is included separately in the model and the hedge portfolio is long in the top third of the disclosure scores and short in the companies that have no climate risk disclosure (R0 .6 7RZ e r o) . This is consistent with events increasing the perceived relevance of climate risks to the market, resulting in a more negative change in value for companies with high climate risk exposure than for companies with low climate risk exposure.
Results indicate an overall significant decline in value of 2.4 percent of the firms with high climate risk exposure compared to the low climate risk exposure firms. The model in Column (1) examines a hedge portfolio that is long in the top third of firms based on ENV_NET and short in the bottom third of ENV_NET. The models in columns (2) and (3) examine whether our measure of climate risk does a better job of explaining market response to climate-related events than TRI or carbon emissions, respectively.
In addition to providing evidence that our measure captures climate risk beyond measures currently used in the literature, our event study results also provide insight into whether ClimateRisk provides a measure of disclosure quality rather than climate risk. Our finding of negative portfolio returns around extreme weather events is consistent with this argument, providing evidence that ClimateRisk is primarily capturing climate risk.26.
Discussion and conclusions
Our proxy for climate risk depends on both the length of climate risk-related disclosures in company 10-Ks, as well as the relevance of the language used. We find evidence consistent with a lower market value and a higher cost of capital for firms with a higher exposure to climate risk. The primary limitation is our use of a measure of climate risk based on textual analysis.
It is also important to emphasize that using a noisy measure of climate risk exposure is likely to result in an underestimation of the impact of climate risk on firm value. However, consistent with our hypothesis, we find no evidence that self-selection into climate risk disclosure affects our results. Appendix A - Domain mapping methodology and general approach to identifying, categorizing and ranking climate risk disclosures.
Based on keywords and phrases related to climate risks, CERES developed an initial corpus of climate risk disclosures, or “excerpts.” The Climate Risk Disclosure Project provides these narrative disclosures that companies make in their annual reports to shareholders filed with the US. The Climate Risk Disclosure Project currently provides access to four climate-related areas of disclosure.
This rating is a function of the amount of relevant language contained in the abstract, as well as the weighting of that language to reflect how it directly addresses the climate risk domain.
Titan Machinery (FY 2013, reporting date: Apr, 2014) (Low Risk)
MACQUARIE INFRASTRUCTURE (FY 2015, reporting date: Feb, 2016) (High Risk)
EMISSIONS Total level of carbon emissions scaled by sales ASSETS4 Q Tobin's Q, the sum of market value and the book value of debt,. COMPUSTAT BTM The book-to-market ratio is the book value of equity over the market value of. The merged CRSP - COMPUSTAT LEV Leverage is long-term debt scaled by the market value of common equity.
Pooled CRSP - COMPUSTAT Advertising Advertising is ad spend scaled by Total Assets (xad/at) Pooled CRSP. Variable Definitions: ClimateRisk is the result of Ceres/CookESG Research on the extent of total climate risk disclosure in the 10-K base;. NClimateRisk is ClimateRisk, normalized to industry and year; ENV_NET is the net score (constructed as strengths minus concerns) based on the KLD database, normalized by industry and year; TRI is the Sales Scaled Toxic Release Inventory; EMISSIONS is the total level of CO2 emissions scaled by sales; PRICE is the share price three months after the end of the fiscal year; Q is Tobin's Q, calculated as the sum of market capitalization and the book value of debt, divided by the book value of total assets; BVE is the accounting value per share of common capital; EARNINGS is earnings per share to common equity; Earn_Neg is earnings per share to common equity if earnings are ≤0, 0 otherwise.
Variable definitions: ClimateRisk is the Ceres/CookESG Research score for the extent of the total climate risk disclosure in the 10-K-based; NClimateRisk is ClimateRisk, normalized by industry and year; BVE is book value per share of common equity; EARNINGS is earnings per share to common equity; Earn_Neg is earnings per share versus common equity if earnings ≤0, otherwise 0. Variable definitions: PRICE is the share price three months after the end of fiscal year; NClimateRisk is ClimateRisk, normalized by industry and year;. ENV_NET is the net score of environmental strengths (constructed as strengths minus concerns) based on the KLD database; TRI is the Toxic Release Inventory scaled by sales; EMISSIONS is the total level of CO2 emissions by sales; BVE is book value per share of common equity; EARNINGS is earnings per share to common equity; Earn_Neg is earnings per share versus common equity if earnings ≤0, 0 otherwise.
Detection takes the value 1 for ClimateRisk>0 and zero otherwise; NClimateRisk is ClimateRisk, normalized to industry and year; PROPDISCL is the ratio of the number of firms in the Fama-French 48 industry classification that provide climate risk disclosure to the total number of firms in the industry and in the same fiscal year; ASSET is the register of the firm's total assets; MF is the number of management forecasts issued by the firm during the fiscal year; BTM is BVE scaled by PRICE; LEV is total debt scaled by the market value of common equity; II is the percentage of total shares held by institutional investors; accounting value per share of common capital; FRNSALE is the firm's external sales as a percentage of total sales; EARNINGS is earnings per share to common equity; Earn_Neg is earnings per share to common equity if earnings are ≤0, 0 otherwise.