(C) PLOS One
This story was originally published by PLOS One and is unaltered.
. . . . . . . . . .



Assessing corporate climate action: Corporate climate policies and company-level emission reductions [1]

['Lena Klaaßen', 'Climate Finance', 'Policy Group', 'Eth Zurich', 'Zurich', 'Christian Lohmüller', 'Tum School Of Life Sciences', 'Technical University Of Munich', 'Munich', 'Bjarne Steffen']

Date: 2024-12

The resulting sample consists of 1,749 companies and 17,198 observations between 2010 and 2022. As of 2019, the sample’s scope 1 emissions (3.39bn t CO2e) represent 9.2% of global emissions, while the sample’s scope 2 (location-based) emissions (0.63bn t CO2e) account for another 1.7% of global emissions which together is more than the total annual emissions of the European Union [ 61 ]. The sample is an unbalanced panel dataset as not every company reported to CDP every year, and not every observation contains data for every variable (some corporate climate policies have not been surveyed in the earlier years). It is also important to note that our sample does not constitute a random sample. While CDP invites most publicly traded and larger corporations to disclose, reporting is voluntary, and some companies decide not to publish. Thus, we cannot assess whether the act of disclosing itself is associated with subsequent emission reductions, and we cannot rule out that there is misrepresentation in the disclosure (e.g., motivated by greenwashing), although CDP reports has been found to be more comprehensive and accurate compared to corporate sustainability reports as they leave less leeway for own interpretation [ 62 , 63 ]. Nevertheless, the sample should serve well to assess the informative value of disclosed corporate climate policies, as CDP-reporting companies form likely the relevant set for investors aiming to decarbonize their portfolio emissions. On top, the large share of global emissions covered suggests relevant insights for public policymakers concerning the potential role of corporate climate action for capital reallocation. However, it should be noted that potential predictors for future emissions reductions identified for this sample which is focused on publicly traded and larger corporations from OECD countries may not serve as predictors in a similar way for other types of companies. Large companies may face higher pressure to show progress in reducing their emissions once corporate climate policies are communicated and therefore they could be a better predictor for large corporations compared to small ones. At the same time, small corporations might face less pressure to communicate corporate climate policies in the first way and thus might even take them more seriously if they decide to do so which would rather speak for a stronger predictive power of corporate climate policies for small companies.

We define the final sample for our analysis in three steps aiming to maximize its relevance to public policymakers currently implementing (or considering implementing) disclosure mandates (see Fig A in S1 Text for a graphical representation of the sample selection). First, we focus on companies with headquarters based in OECD countries where disclosure mandates are primarily introduced or considered so far (e.g., the Corporate Sustainability Reporting Directive in the European Union or the SEC proposal on enhanced climate-related disclosure in the US). Second, we limit the sample to the companies reporting GHG emission data (scope 1 and 2) for at least 5 years to be able to observe the potential link between corporate climate policies and subsequent emission reductions for our sample companies over a longer time period. Third, we reduce the sample to companies that report an ISIN (International Securities Identification Number) at least once between 2010 and 2022. This enables us to match the CDP data with company-specific data from Refinitv.

We draw on a dataset of the CDP climate change and supply chain program public responses from 15,827 companies between 2010 and 2022, which we complement with other company-specific data extracted from Refinitiv. CDP constitutes the most extensive global database on corporate climate policies and GHG emissions data, with disclosing companies making up more than half of global market capitalization and spanning most regions and industries. Investors constitute one of the main user groups with over 680 financial institutions and over US$130 trillion in assets requesting information from their portfolio companies through CDP [ 57 ]. The CDP database is also widely used in academic research on sustainable finance [ 58 – 60 ].

Variables and regression model

For the dependent variable, we operationalize the climate performance as a continuous variable by using absolute emissions as well as emission intensities. Emission intensities are calculated by dividing absolute emissions by the company’s total revenue, following previous studies [19,20,42,49,53]. The carbon footprint of a company can be divided into three categories of emissions: scope 1 refers to direct emissions from a company’s own activities, scope 2 refers to emissions from the production of purchased energy (especially electricity), and scope 3 refers to emissions from up- and downstream activities along the value chain [64]. In our regression model, we rely on scope 1 and scope 2 location-based emissions as dependent variable. We solely rely on scope 2 location-based emissions and do not consider scope 2 market-based emissions in the dependent variable, as the latter have only been introduced by CDP in 2016 and have been criticized for not reflecting real emission reduction as they largely rely on the purchase of renewable energy certificates [65]. We do not include scope 3 emissions in our dependent variable following [20,49,51] since companies have only limited direct influence on those emissions through corporate climate policies, and they are subject to large inconsistencies and incompleteness [62,66,67]. The lack of comparability of scope 3 data prevails not only across companies but also within one company over time. Even if we looked at only one company over time, there would be a lack of comparability of scope 3 emissions levels as many companies increased their effort on scope 3 emission measuring and reporting over time (e.g., moving from initially only including business travel, to later covering the full scope of value chain emission including purchased goods and services and used of sold products). Additionally, scope 3 emissions are often estimated based on secondary data (e.g., by using spend-based emission factors–see [68])–and thus do not represent “actual”/primary data which could be influenced by corporate climate policies. To our knowledge, there is no paper with a similar analysis that includes scope 3 emissions in the dependent variable (see Table A in S1 Text for a detailed overview of existing literature).

For the independent variables, we extract corporate climate policies from CDP adhering to the four key areas as discussed above: targets, governance, implementation and MRV. The CDP dataset covers a wide range of corporate climate policies. We focus on questions with binary or categorical answers since long, individual answers are unlikely to serve in large-N analyses conducted by investors to decide whether to include a company in a climate-oriented portfolio. The main strategies to integrate ESG criteria used by institutional investors revolve around exclusion and avoidance, norm-based and inclusionary screening or best-in-class approaches. These are often done following a rule-book-based approach applied to criteria of good comparability in order to decide on exclusion or inclusion or by looking at objective metrics such as the emission intensity compared to investors [69]. This approach results in 13 corporate climate policies spread across the four key areas (see Table 1). We then operationalize the raw CDP data in three steps: First, the responses to all CDP questions belonging to the same corporate climate policies are collected across all years and CDP programs. Second, the responses are standardized: Since most CDP questions (and answer options) changed multiple times between 2010 and 2022, we match them thematically with near-same questions in other years. Third, the answers are operationalized for the statistical analysis by converting answers to a binary format (see section 3 in S1 Text for more details).

Beyond the single policies shown in Table 1, we also use two different measures for comprehensiveness as independent variables: The comprehensive policy mix I requires the introduction of at least one policy from each of the four areas shown in Table 1 (targets, governance, implementation, MRV). Absolute target, board-level oversight, strategic integration and scope 3 disclosure or science-based target, monetary incentives, value-chain engagement and scope 1 verification are two examples for how a comprehensive policy mix I could look like. The comprehensive policy mix II requires the introduction of all corporate climate policies that have been included in the CDP questionnaire in a given year. We do not require the introduction of a science-based target or an internal carbon price for the completion of the comprehensive policy mix II. This is due to their unique characteristics (described in Section 4.1) which result in much lower adoption rates compared to other corporate climate policies especially in the first years of introduction (leading to a drop in companies with comprehensive policy mix II close to zero which would distort the effect). All other policies must be existent as soon as they are included in the CDP questionnaire to meet the requirements of the comprehensive policy mix II. For example, in 2013, value chain engagement needs to be introduced by a company to have a comprehensive policy mix II. Importantly, we also differentiate in the policy mix definition between our dependent variables as we exclude intensity targets for both types of mixes for regressions with absolute emissions as depedent variable and exclude absolute targets for both types of mixes for regressions with absolute emissions as dependent variable. The rationale behind this approach is that absolute targets aim to reduce absolute emissions, while intensity targets aim to reduce emission intensities. Since a substantial share of companies only introduced one of the two targets (see Section 4.1 for descriptive results), allowing intensity targets for policy mix I and requiring them for policy mix II would distort the effect on absolute emissions (or vice versa for absolute targets for emission intensities).

We incorporate several control variables into our analysis (see Table E in S1 Text for the rationale behind each control variable and the related data source). These include the sector and regional location of the companies. In addition, we include additional company-specific and country-specific controls. For the company-specific controls, we consider total revenues as a proxy for company size, the debt ratio, and a binary variable to indicate whether a company was publicly traded during a specific year. To further account for variations, we calculate a company’s baseline emissions in relation to those of its sectoral and regional peers by determining the percentile of emissions for each sector, region, and year (see Table D in S1 Text for sector definitions). As country-specific control, we use the green financial policy density of the country of the company’s headquarter to account for the potential influence of green finance regulation, including disclosure mandates [7].

Table 2 shows the correlation between all independent variables shown in Table 1 and control variables which is reasonably low with correlation coefficients never exceeding 0.4 except for incentives and monetary incentives (0.811), scope 1 verification and scope 2 verification (0.885), and percentile of absolute emissions and percentile of emission intensities (0.889) (see Table F-R in S1 Text for all correlation coefficients on a yearly level). From a corporate management perspective, this seems plausible given that monetary incentives represent a specific kind of Incentives and scope 1 and scope 2 emissions are likely to be verified together. To avoid multicollinearity issues, we, therefore, exclude the variables incentives and scope 2 verification from the regression analyses, given that monetary incentives are the more stringent policy and scope 1 emissions are almost five times higher in our sample compared to scope 2 emissions. The high positive correlation of the percentile of absolute emissions and percentile of emission intensities also seems plausible given that high emitting sectors usually come with high absolute emissions as well as high emission intensities. However, this does not pose an issue for our model as they are never used in the same specification as only one of each is included in accordance with the dependent variable.

We use a fixed effects (FE) model to evaluate the link between corporate climate policies and subsequent GHG emissions. Previous studies suggest a time lag exists between adopting corporate climate policies and potential reductions in GHG emissions (Dahlmann et al., 2019; Doda et al., 2016; Qian & Schaltegger, 2017). Following the findings of those studies, the model lags the emission data by one year, so emissions data of year t+1 are regressed on corporate climate policies of year t. To test H1, we run multiple specifications: We study each corporate climate policy separately (as did the vast majority of previous studies). Thus, we aim to establish benchmark specifications, which come closest to previous model setups. To test H2, we regress a dummy which indicates whether a company discloses a comprehensive corporate climate policy mix. We only include one type of comprehensive policy mix per regression specification. While all model results are presented in Section 4, S1 text provides additional sample descriptives. To summarize our regression model with Eq 1 being linked to H1 and Eq 2 being linked to H2: (1) (2) where Climate performance i,t+1 are the absolute emissions or emission intensity of company i in year t+1. a denotes the Corporate climate policy (Eq 1) disclosed by company i in year t, with Climate management practices i,t,a ,. and Comprehensive practices combination I/II i,t being the respective dummy indicating if the policy or the mix is present. account for the company size, the capital structure, the ownership structure and the emission baseline of a company i in year t, sector c and region d. Green financial policy density i,b represents the number of green finance regulations, including disclosure mandates, in country b in year t while ε i,t,b,c,d denotes the error term. To control for additional confounders, all equations also include FEs at the year (α t ), sector (γ c ) and region (δ d ) level—see Table C and Table D in S1 Text for a detailed overview of the regions and sectors applied (defined regions: Asia, Europe, North America, Ozeanien, South America). This is in line with previous studies such as [20] who also use year, region and sector FEs but goes beyond most previous studies which only applied sector and year FEs [70–73]. As more recent studies with larger datasets increasingly resort to year as well as entity-level FEs–either at the corporate or facility level [18,19,50,53], we also calculate each specification with company and year FEs to test the robustness of the main specification (while leaving out region and sector fixed effects in these specifications to avoid multicollinearity issues). For all our models we use robust standard errors which computed using the Huber-White heteroskedasticity-consistent estimator (see section 4.3 in S1 Text for Breusch-Pagan test results indicating the need for robust standard errors).

With our empirical design, we address potential biases that could affect the relationship between disclosed corporate climate policies and subsequent emission reductions and take measures to mitigate their impact on our results. Firstly, to address omitted variable bias, we employ company-level fixed effects next to year fixed effects in our analysis. By doing so, we also account for company-specific variables that may influence both the disclosure of corporate climate policies and emission reductions. This approach surpasses previous studies that have solely relied on sector-fixed effects, allowing us to capture a more nuanced understanding of the relationship between disclosure and emissions. Secondly, regarding reverse causality, we acknowledge that our focus lies on the predictive value of corporate climate policies for firm-level emission reductions rather than establishing causal relationships. Understanding the correlation between disclosure and emissions reduction can be valuable for investors and policymakers, irrespective of the direction of the effect. Lastly, concerning collider bias, we note that our sample comprises the relevant population for investors, namely large and publicly traded companies. We incorporate numerous control variables, including regulatory pressure represented by green financial policy density, to account for potential confounding factors. Even if unobservable factors influence company-level emissions, our analysis remains relevant for understanding the link between corporate climate policies and emissions reduction within this specific population. By addressing these potential biases and providing context for their relevance to our study, we aim to offer a comprehensive assessment of the relationship between disclosed corporate climate policies and subsequent emission reductions.

[END]
---
[1] Url: https://journals.plos.org/climate/article?id=10.1371/journal.pclm.0000458

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.