No News About Climate Action is Good News for Low-Polluting Firms

Abstract

Media plays a crucial role in shaping public perception of firms, significantly impacting their operations and performance. Firms often attempt to influence media channels to showcase their climate action initiatives, aiming to enhance their public image, reputation, and stakeholder trust. While firms cannot directly control their media coverage, intentional or unintentional dissemination of their climate action efforts can thrust them into the media spotlight. The implications of this media spotlight remain uncertain. While firms may anticipate positive public relations benefits from such exposure, it may also raise environmental expectations for firms and risk highlighting less favorable aspects of their environmental practices. Despite previous research on the impact of media exposure on firms’ financial outcomes, a notable gap exists in understanding how the media spotlight on firms’ climate actions affects their operational and financial performance. A clear hurdle is the lack of a systematic and objective measurement of such a spotlight. Our study addresses this research gap by developing a machine learning-based framework to derive a climate action vocabulary as a novel information artifact. The vocabulary is subsequently used to measure the intensity of media attention on a firm's climate actions. Our research reveals that an increased media spotlight on climate actions, regardless of their sentiment, has an adverse effect on firms’ financial performance, primarily due to the rise in operational costs. Furthermore, we find that the impact of media spotlight is heterogeneous. Low-polluting firms experience negative financial consequences as the downside of the heightened media spotlight. In contrast, high-polluting firms and those operating in polluting industries may witness an overall positive financial impact. We also find that firms’ market orientation (business-to-business vs. business-to-consumer), durability classification, index constituency (i.e., S&P 500), and greenwashing status significantly moderate the relationship. Our research highlights key considerations for corporate leaders with respect to the drawbacks for low-polluting firms in seeking media attention for their climate actions. Given that the media captures society's limited attention, our research suggests that firms should refrain from promoting superficial narratives about climate change that distract society from more impactful sustainability conversations.

1. Introduction

More Companies Decide Silence Is Golden When Going Green. Wall Street Journal's Sustainable Business issue, January 2024 (Uribe, 2024)

Media coverage significantly shapes public perception of firms and serves as a primary, credible channel for communicating firms’ often complex and opaque actions (Deephouse, 2000; Lee and Riffe, 2017). As a strategic asset, media coverage can significantly influence a firm's market valuation (Ahern and Sosyura, 2014; Ba et al., 2013; Chen et al., 2021), operational efficiency and innovation (Lam et al., 2016), resource allocation (Abedi et al., 2022), product management (Astvansh et al., 2022), and customer behavior (Goh et al., 2011). While prior research has examined media's role in sharing firm-related information with external stakeholders, our study addresses an unexplored area related to media attention on firms’ climate actions. ¹ Specifically, we investigate whether and how climate action-related news spotlight affects firms’ financial performance while examining how variations in firms’ environmental performance influence this relationship. Hence, we address a significant gap in the existing literature regarding the financial implications of media attention on firms’ climate actions, a critical dimension of corporate social responsibility (CSR) (Blanco, 2021; Capelle-Blancard and Petit, 2019; Graf-Vlachy et al., 2020). Many firms invest in public relations to promote their climate initiatives. While costly for the firm and with unclear tangible benefits, such public relations efforts can strain society’s limited attention, potentially overshadowing more urgent climate issues. Our study isolates the financial impact of climate-related media coverage, providing business leaders with insights into the potential strategic value and drawbacks of an intentional or unintentional spotlight on their climate actions.

Our analysis covers public US-based firms from 2005 to 2017, allowing us to capture long-term trends and changes in firms’ financial outcomes. We measure financial performance using both accounting-based metrics, including return on assets (ROA), as well as market-based metrics, such as market-to-book (M–B) and Tobin's Q. To understand the mechanisms driving this relationship, we investigate cost of goods sold (COGS) and selling, general, and administrative (SG&A) expenses as potential mediators, evaluating how media spotlight influences these operational cost structures. By studying changes in COGS and SG&A expenses, we aim to capture the financial pressures and adjustments that firms may undertake due to increased public scrutiny of their climate actions.

Environmental actions are known to influence firms’ operational and financial performance (Awaysheh et al., 2020; Flammer, 2015; Jacobs et al., 2016). Related literature has explored how media, especially news, enhances public understanding of corporate social and environmental responsibility, a key issue for policymakers and stakeholders (Ba et al., 2013; Gilley et al., 2000; Hardcopf et al., 2021; Sarkis et al., 2010). Prior studies have focused on how specific environmental, social, and governance (ESG) events, such as disasters, awards, and initiatives, identified in the news, influence firm stock prices, using these extreme events as proxies for environmental behavior (Capelle-Blancard and Petit, 2019; Flammer, 2013; Jacobs, 2014; Jacobs et al., 2010; Klassen and McLaughlin, 1996; Krüger, 2015). However, these studies rely on isolated events, such as oil spills, often within select industries prone to environmental scrutiny, as proxies for firms’ environmental performance without systematically analyzing the financial effects of media attention beyond firms’ actual environmental performance. The focus on events and the lack of emphasis on the media's role may stem from the absence of an automated framework to comprehensively measure the ongoing media spotlight on a firm's climate actions. Addressing this gap, our study introduces a systematic content analysis framework to measure media spotlight on firms’ climate actions and assess its financial impact while controlling for their environmental performance through standard measures such as CO₂ emissions.

Amid rising societal and business concerns about climate change (Sunar and Swaminathan, 2022), stakeholders, including governments, nongovernmental organizations (NGOs), shareholders, and communities, are increasingly scrutinizing firms’ environmental actions (Huang et al., 2020; Sarkis et al., 2010). This demand has intensified media attention on firm climate initiatives (Lee and Carroll, 2011). Our study finds that such media spotlight substantially affects firms’ financial performance and market value by increasing their operational costs, emphasizing the need for firms to manage this attention strategically. To navigate these pressures, firms often rely on public relations departments, which craft environmental messaging, engage in agenda-setting strategies, and respond to media inquiries to strengthen their public image (Lee and Carroll, 2011). The media's perceived impartiality allows firms to use agenda-setting strategies to project a positive environmental image (Aerts and Cormier, 2009; Lee and Riffe, 2017), sometimes even to virtue signal or engage in greenwashing without substantial environmental commitment (Cho et al., 2012; Dorobantu et al., 2024). Institutional pressures (coercive forces from regulatory bodies and NGOs, mimetic forces from competitors, and normative societal expectations) further encourage firms to publicize their climate actions. According to the “mimetic and coercive isomorphism” theory, firms may feel compelled to engage in this visibility-seeking behavior to gain legitimacy and avoid scrutiny (DiMaggio and Powell, 1983; Galbreath, 2019; Villena and Dhanorkar, 2020). Finally, managers may be incentivized to publicize climate actions to bolster their personal reputations among stakeholders and future employers (Krüger, 2015; Liao et al., 2021). Although these theories offer insights into why firms prioritize media coverage of their climate efforts, research remains limited on the financial consequences of such media spotlight, particularly as it may influence both firm costs and market value.

Our research makes three notable contributions. First, we develop a machine learning-based framework to create a climate action vocabulary, a novel information artifact for quantifying climate action spotlights on firms from news sources. We follow best practices in the literature to develop our vocabulary through a fully automated and systematic approach, free from manual researcher bias or intervention (Liu, 2020). Although proprietary databases like FactSet and RavenPack offer firm-related news analytics, their limited transparency and customizability hinder capturing the breadth of the overall climate action news spotlight as we do. Moreover, these databases tend to focus on specific events rather than offering a broad view of overall media attention. Our vocabulary contributes to enhancing the transparent, reproducible measurability of firm climate actions through an interpretable methodology. The vocabulary development framework is domain-adaptable, can be updated over time, and is applicable to diverse text sources, including web pages, blogs, and news articles. Using the vocabulary, we generate quantitative metrics to measure the climate action content in firm-related news. We validate our vocabulary and metrics through four distinct methods: (1) We create ground truth labels for news samples using Amazon Mechanical Turk ratings to determine whether the articles are climate action-related and calculate performance metrics, such as F1 score, to assess the accuracy of our vocabulary-based metrics in capturing the climate action language. (2) We benchmark our approach against Mistral Large 2, a state-of-the-art large language model (LLM) known for its strong text classification capabilities. This comparison allows us to validate that our approach performs on par with advanced models while offering distinct advantages in transparency, interpretability, and replicability. (3) We test the ability of our metrics to rank industries based on their presumed pollution levels. (4) We evaluate the relative predictive utility of our metrics in enhancing the accuracy of firm-level environmental prediction models (Sections A and B in the Supplemental materials). The second contribution of this study is that we apply our metrics to measure firm-level climate action spotlight and analyze its financial impact through a panel longitudinal time-series analysis. We control for firms’ environmental performance using their CO₂ equivalent emissions as a proxy and isolate the news spotlight effects. To establish causality, we use county-level political voting data at firm headquarters as an instrumental variable (IV). We also investigate the mechanisms through which the news spotlight affects financial performance and study the distinct impacts of both positive and negative spotlights. Finally, we explore heterogeneous effects by interacting firms’ climate action spotlight with their environmental performance, pollution status of their industries, market orientation (business-to-business (B2B) vs. business-to-consumer (B2C)), durability classification, index constituency (i.e., S&P 500), and greenwashing status. Our findings offer strategic insights for firms in managing media coverage of their climate actions, advising whether to seek or avoid the spotlight for different firms. Our work contributes to the growing research on automated content analysis, firm environmental practices, media coverage, and public relations.

Our findings indicate that increased media spotlight on climate actions, particularly for low-polluting firms, adversely affects their financial performance, primarily through rising operational costs. Using FinBERT, a pretrained deep learning model for sentiment analysis in finance, we find that both positive and negative spotlight yield similar negative financial impacts. However, there is heterogeneity in the observed effect. Notably, high-polluting firms and those in polluting industries exhibit a positive financial response to increased media attention, contrasting with the negative response observed for low-polluting firms. This disparity may stem from the reputational upside (i.e., starting from a poor reputation) from the media spotlight for high-polluting firms. There are also greater opportunities for environmental improvement (Cordeiro and Tewari, 2015; Gilley et al., 2000; Kassinis et al., 2022), making any operational changes due to added media spotlight more cost-effective for these firms. This observation is notable, considering that existing environmental event study literature primarily focuses on firms in polluting industries, overlooking heterogeneous effects across various industries and firm types. Additionally, we find that B2C firms, with direct consumer interaction, observe a stronger financial impact from climate-related media attention than B2B firms, which are less sensitive to public scrutiny. Also, firms in capital-intensive industries, particularly those in durable goods sectors with complex manufacturing processes, tend to show a less negative reaction, likely due to challenges in swiftly adapting to external pressures. Firms with higher levels of greenwashing show a more pronounced negative reaction to the media spotlight, likely because the increased attention heightens scrutiny and reveals gaps between their claims and actual environmental practices. Finally, larger, more visible firms in the S&P 500 respond more negatively to the climate spotlight, reflecting their higher costs and reputational risks associated with managing public relations (Tables D5–D11 in the Supplemental materials).

From a practical standpoint, firms should carefully consider the implications of the media spotlight when planning public relations strategies related to climate action. Our findings indicate that such a spotlight yields financial benefits primarily for high-polluting firms, which have greater opportunities for operational and reputational improvements, while being financially detrimental for low-polluting firms. Also, firms with higher levels of greenwashing tend to react more negatively to increased media attention. These findings underscore the need to shift from superficial virtue signaling to genuine environmental concerns. Chasing unnecessary media spotlight for climate action solely for virtue signaling can result in financial drawbacks for firms due to the associated costs. Spotlight on climate actions of low-polluting firms with a potentially minimal environmental impact may clutter the news and divert society's attention away from critical sustainability challenges, which are more pressing for high-polluting firms. To ensure the robustness of our findings, we conducted various robustness checks, including alternative measures of the media spotlight, different IVs, various financial performance metrics, and multiple regression specifications.

2. Literature Review

The literature emphasizes the media's role, particularly news, as a critical source of corporate information and a medium for capturing hard-to-quantify aspects of firms (Goh et al., 2011; Zhang et al., 2022). Media influences corporate decisions (Astvansh et al., 2022), reputation (Deephouse, 2000; Hardcopf et al., 2021; Minichilli et al., 2022), costs (Abedi et al., 2022; Dang et al., 2019; Sodhi and Tang, 2019), as well as operational and financial outcomes (Chen et al., 2021; Foerderer and Schuetz, 2022; Lam et al., 2016). Previous research has studied how news media shapes public perception and assessment of firms’ environmental activities (Capelle-Blancard and Petit, 2019; Lee and Riffe, 2017), with firms often attempting to influence news coverage to gain legitimacy (Ahern and Sosyura, 2014). Although firms cannot directly control their media coverage, prior studies have shown their influence over the attracted media attention through public relations activities (Graf-Vlachy et al., 2020; Lee and Riffe, 2017). Despite firms’ role in affecting their media coverage of environmental issues, the literature has not explored how this media spotlight impacts their financial performance. Our paper addresses this gap.

Several studies have examined the impact of extrafinancial environmental events on short-term market-based financial outcomes, such as stock returns. These studies have primarily focused on environmental disclosure programs, including carbon information disclosure (Matsumura et al., 2014), the US toxic release inventory (Hamilton, 1995), environmental violations and incidents (Capelle-Blancard and Laguna, 2010), and other environmental events and announcements (Capelle-Blancard and Petit, 2019; Jacobs, 2014; Jacobs et al., 2010; Jacobs and Singhal, 2020; Klassen and McLaughlin, 1996; Krüger, 2015; Lyon and Shimshack, 2015). Key contributions in this area include Klassen and McLaughlin's (1996) study of market reactions to 22 negative and 140 positive environmental events that occurred between 1985 and 1991. Jacobs et al. (2010) investigated market responses to 417 Corporate Environmental Initiatives and 363 Environmental Awards and Certifications announcements in major business news sources between 2004 and 2006. Flammer (2013) studied the effects of 273 environmentally significant events reported in the Wall Street Journal from 1980 to 2009. Jacobs and Singhal (2020) studied the impact of the Volkswagen emissions scandal in 2015 on the share values of publicly listed firms within the automotive industry. Notably, the use of news stories (e.g., Wall Street Journal) in these studies is to identify individual environmental events rather than the overall media spotlight on firms (the latter is the focus of this study). Others have also studied market reactions to Newsweek's corporate “Green Rankings” (Lyon and Shimshack, 2015; Yadav et al., 2016), green vehicle innovation (Ba et al., 2013), and ESG rating announcements (Christensen et al., 2022).

These studies rely on environmental disclosures and events as proxies for firms’ actual environmental practices and do not study the effect of the news spotlight itself. We control for firms’ environmental performance in our models to tease out the impact of the overall media spotlight. Existing research discussed above often focuses on environmentally sensitive industries, which are more susceptible to environmental events, without providing a comprehensive cross-sectoral view. In this article, we show that the media spotlight can lead to varying financial outcomes depending on a firm's pollution level, industry pollution status, market orientation, durability classification, index constituency, and greenwashing status. Some prior studies have focused on market reactions to isolated events identified through manual news reading or proprietary datasets. In contrast, our research employs a methodically developed climate action vocabulary, validated through a rigorous machine learning framework (Figures 1 and 2 and Section B in the Supplemental materials). This methodology enables us to leverage a large sample of news and systematically measure their relevance to firm climate actions. Capelle-Blancard and Petit (2019) studied the market effects of “ordinary” ESG events across various industries using events (identified from the news) as a proxy for firms’ ESG performance. Rather than focusing on abrupt and often short-lived events, our study controls for firm environmental performance and teases out the long-term, longitudinal impact of news spotlight on firm financial performance, using more stable firm performance metrics, both accounting and market-based. In a recent study, Serafeim and Yoon (2023) examined the predictive utility of ESG ratings for future ESG news and subsequent market reactions. Unlike our study, they do not focus on climate actions, conduct firm-level analysis, or delineate any causal relationships.

Figure 1.

Steps of the automated framework to create the climate action vocabulary.

Figure 2.

Two-dimensional visualization with t-distributed stochastic neighbor embedding (t-SNE) of the top 100 terms in the climate action vocabulary based on their derived term importance values.

Another relevant research stream investigates the relationship between firms’ environmental and financial performance, yielding mixed findings. Environmental actions are costly (Dowell and Muthulingam, 2017) and, as some studies have shown, may negatively impact firms’ operational and financial performance under certain conditions or have negligible causal effects (Awaysheh et al., 2020; Jacobs et al., 2010; Krüger, 2015; Zhao and Murrell, 2016). Conversely, others argue for a “win–win” scenario where environmental initiatives may enhance financial performance (Ba et al., 2013; Flammer, 2015; Jacobs, 2014; Jacobs et al., 2016; Khuntia et al., 2018; Luffarelli and Awaysheh, 2015; Melnyk et al., 2003). Given this literature, we must control for firms’ environmental performance in our regression analysis to ensure accurate estimations of the financial impacts of climate action spotlight and avoid confounding results.

Our research also contributes to the growing field of text analysis and natural language processing in business and operations management research by introducing a systematic methodology for measuring climate action spotlight in the news. We detect climate action language using machine learning-based text mining and content analysis techniques. Recent studies in this area have applied various machine learning and unstructured data analysis methods, including text mining, to monitor firm activities. These methods have been used to examine topics such as strategic manipulation (Lee et al., 2018), organizational culture (Gupta et al., 2022), financial disinformation (Zhang et al., 2022), service quality (Altman et al., 2021; Mankad, 2016; Mejia et al., 2021), competitor identification (Pant and Sheng, 2015), platform seller behavior (Huang et al., 2023), pricing strategy (Keskin et al., 2024), and environmental performance (Blanco, 2022). Vocabulary-based content analysis methods involve creating a vocabulary of keywords to extract meanings and measure distinct concepts in a systematic and interpretable manner (Gupta et al., 2022). Operations management researchers have used vocabulary-based content analysis methods for various management-oriented functions, including consumer sentiment analysis (Lau et al., 2018), physician review analysis (Lu and Rui, 2018), social media emotion generation (Gour et al., 2022), employee culture identification (Gupta et al., 2022), news coherence analysis (Zhang et al., 2022), and fraud detection (Liu et al., 2023). These methods have also been applied to detect topics and monitor activities such as disaster management (Yan and Pedraza-Martinez, 2019), restaurant hygiene (Mejia et al., 2019), and hospital quality assessment (Salge et al., 2022). Our work builds on these studies by developing a novel climate action vocabulary to measure news spotlight on firms’ climate actions. Previous studies in environmental content analysis have often relied on ad hoc, author-defined terms (Flammer, 2013; Jacobs et al., 2010; Klassen and McLaughlin, 1996). Addressing this gap, we employ a validated topical vocabulary developed using information retrieval and machine learning methods to capture the climate action language in news media.

Our vocabulary creation methodology has several key advantages. First, our vocabulary is free from researcher bias as the term/keyword selection is entirely automated, ensuring an objective framework without manual intervention. Our developed vocabulary is intensity-based, meaning each term is associated with a data-driven intensity value that determines its semantic relatedness to climate action. Additionally, our vocabulary has high scalability, ease of updating, and straightforward interpretability, which facilitates replication by other researchers. These features enable our vocabulary to successfully capture the complexities of sustainability and climate action language in textual documents. While we have focused on climate action, the methodology we propose for vocabulary generation is versatile and easily adaptable to other topics, extending its utility across various domains.

3. Hypothesis Development

Despite rising media attention on corporate climate actions and the known impact of news on various aspects of firm operations (Chen et al., 2021; Foerderer and Schuetz, 2022; Lam et al., 2016), the effect of climate action-related media spotlight on firms’ financial performance remains unknown.

While heightened media focus on a firm's climate actions may enhance its reputation and social image (Graf-Vlachy et al., 2020), it may also adversely affect its financial performance through several cost sources. Media attention can pressure firms to adopt costly environmental practices, such as transitioning to sustainable energy or redesigning products, which can divert resources from more profitable investments (Erhun et al., 2021). For instance, Starbucks gained significant media coverage for its global initiative to eliminate plastic straws from its stores by 2020. This led to Starbucks making bold operational changes to expand its ecological initiatives, redesign and produce new lids, find suppliers for alternative straws, and retrain employees, efforts that required substantial capital investment and operational changes across 30,000 stores (Caron, 2018). These efforts would translate into an increased COGS for the firm. Increased news attention can incur substantial operational costs through compliance measures such as hiring auditors (Marinovic and Varas, 2016), accreditation (Batista, 2023), preparing disclosures (Lee and Riffe, 2017; Sodhi and Tang, 2019), and conducting internal audits (Di Giuli and Kostovetsky, 2014). Such costs may manifest as increased SG&A expenses, reflecting the financial impact of maintaining regulatory and stakeholder expectations under media scrutiny. Media spotlight can also prompt increased stakeholder oversight and impose costly expectations (Lu et al., 2021; Sarkis et al., 2010). When firms are spotlighted for sustainability improvements, it often brings visibility and scrutiny to their entire operations, potentially fueling consumer skepticism. For example, Coca-Cola introduced “PlantBottle” as an eco-friendly alternative to its plastic bottles, claiming it contains 30% plant-based materials. Such a claim put Coca-Cola under the spotlight, prompting the question of why the entire bottle could not be made from eco-friendly materials. This sparked scrutiny of the firm, revealing that the bottle is primarily composed of nonrenewable fossil fuels and that less than 10% of its packaging operation utilizes plant-based materials (Batista, 2023). Neutrogena's promotion of “Natural” skincare as being “100% natural” led to scrutiny, revealing that the skincare contains synthetic, environmentally harmful preservatives (Bucher, 2017). These examples illustrate how a spotlight on selective environmental initiatives can raise scrutiny of superficial environmental actions, thereby eroding consumer trust and leading to reputational costs (Darendeli et al., 2022; Jacobs, 2014). Also, unfavorable media coverage suggesting noncompliance with environmental regulations can result in legal expenses and penalties (Hardcopf et al., 2021; Kölbel et al., 2017).

From these considerations, we postulate:

H1: Firms with higher news spotlight on their climate actions have lower financial performance.

H2: Firms with higher news spotlight on their climate actions have higher operational costs.

H3: The operational costs mediate the effects of news spotlight on firms’ financial performance.

Recognizing the significant influence of news sentiment on readers’ perceptions (Lee et al., 2018; Zhang et al., 2022), we separately investigate the impact of positive and negative news spotlights on firms’ financial performance and market value. A positive spotlight on firm climate actions could contain information about firms’ climate action initiatives, environmental commitments, or proactive adoption of energy-efficient practices. While positive news spotlight may laud a firm's environmental initiatives, enhance its reputation, reduce environmental noncompliance costs, and potentially lead to increased customer loyalty and competitive advantage (Graf-Vlachy et al., 2020), it could also inflate operational costs as firms seek to live up to these expectations (Di Giuli and Kostovetsky, 2014). Positive attention could prompt firms to expedite their sustainability initiatives, incurring immediate increased costs. It might also necessitate firms to ensure their supply chain's alignment with sustainability values, thoroughly vetting suppliers and shifting to more costly sustainable suppliers (Hendricks and Singhal, 2005). Additionally, positive attention could lead to demands for greater transparency, which can increase proprietary costs (Kraft et al., 2022), with regular sustainability reporting and third-party audits adding to operational expenses (Marinovic and Varas, 2016; Sodhi and Tang, 2019). Also, firms may escalate marketing efforts to capitalize on the positive spotlight, potentially diverting resources from other areas (Abedi et al., 2022). While the long-term financial benefits of proactive climate actions are hard to quantify (Krüger, 2015), the immediate financial implications of these actions, often perceived as non-revenue-generating costs, have historically impeded progress in international climate treaty negotiations (Hsu and Wang, 2013). We propose that the immediate inflated operational costs might not be quickly offset by any financial benefits, aligning with the overall spotlight effect posited in hypothesis 1.

On the other hand, a negative climate action spotlight might highlight a firm's environmentally unfriendly policies, resistance to sustainability, or lack of environmental commitment. Such a spotlight can result in significant costs associated with social irresponsibility (Fang and Cho, 2020; Kölbel et al., 2017; Krüger, 2015), including reputational damage, regulatory and legal risks, reduced customer loyalty, increased operational expenses, divestment from investors, and lost business opportunities (Foerderer and Schuetz, 2022; Kölbel et al., 2017; Krüger, 2015). Thus, we propose the following:

H4: Firms with higher positive news spotlight on climate actions have lower financial performance.

H5: Firms with higher negative news spotlight on climate actions have lower financial performance.

Firms’ environmental performance and the pollution status of their industries are likely to influence how climate action news spotlight affects their financial performance due to varying exposure to stakeholder pressure and scrutiny (Arora et al., 2020; Cordeiro and Tewari, 2015). We hypothesize two potential interaction directions. First, firms with poor environmental performance are more likely to be targeted by regulators and activists and may face more reputational damage from environmental inaction (Flammer, 2013; Hsu and Wang, 2013; Klassen and McLaughlin, 1996). Thus, an increased climate action spotlight on these firms could indicate a favorable response to regulatory demands, enhancing their reputation, customer loyalty, and investor confidence. Previous research suggests that polluting firms or those with adverse sustainability incidents often react more positively to ESG announcements (Flammer, 2013), benefiting from reduced cost of debt, enhanced reputation, and improved transparency (Arora et al., 2020; Gao et al., 2022). Conversely, news spotlight on these firms might lead to higher operational costs due to the high expenses of increased exposure and pollution control in dirtier firms (García-Marco et al., 2020). Also, an intensified climate action spotlight on high-polluting firms could be perceived skeptically by stakeholders (Arora et al., 2020). Therefore, while we can expect that the environmental performance of firms will interact with the media spotlight to affect financial performance, the direction of the overall impact is uncertain and will depend on the relative strength of the pathways discussed above. Thus, we propose:

H6: Firms’ environmental performance moderates the effect of news spotlight on financial performance.

While we expect a similar moderating effect from polluting firms and those within polluting industries, we study them separately to account for any underlying differences. This distinction is crucial as stakeholders may view the spotlight on firms within polluting industries with greater skepticism (Arora et al., 2020), or they may not differentiate between the environmental performance of individual firms within an industry. Thus, a separate analysis may be needed to assess these distinct effects. We propose:

H7: Pollution status of firms’ industries moderates the effect of news spotlight on financial performance.

Our study investigates an underexplored area, examining the interplay between media coverage of firms’ climate actions and their financial outcomes. We present hypotheses that may seem counterintuitive to some, suggesting that increased media attention could negatively impact firms’ financial performance under specific conditions. By testing these hypotheses, we uncover the complex financial effects of media attention on climate actions, deepening our understanding of this relationship.

4. Data Description

Our climate action vocabulary is developed using text data from the climate action plans of 365 US cities, covering all municipalities known to have committed to and developed such plans. These plans, rich in environmental terms, are critical sources of information on city-level efforts to regulate emissions and promote environmental well-being. Tailored specifically to climate action, they reflect the interests of public and nonprofit entities in sustainability. Choosing this data source ensures our vocabulary is grounded in topical data about climate action, free from potential for-profit biases that may be present in firms’ own narratives. This approach allows us to identify the language and themes prevalent in the public domain, ensuring that our subsequent news analysis is grounded in capturing the social and societal perceptions of climate action rather than being influenced by the potentially biased language of for-profit entities. The automated methodology for constructing this vocabulary is detailed in Figure 1 and Section A in the Supplemental materials. We accessed firm news data from LexisNexis News Sources, “the media archive of choice,” which is a premier news source comprising the top 50 US-based newspapers, as listed in the Editor & Publisher Yearbook. We used the news data to create our main variable of interest (i.e., News Spotlight). Our study period spans from 2005 to 2017, yielding a corpus of 316,692 news articles about US public firms.

For firms’ environmental performance data, we used Refinitiv, a globally recognized provider of financial market data. Refinitiv is known for its comprehensive and reliable corporate environmental data, encompassing over 100 environmental metrics such as emissions, waste, and resource consumption (Refinitiv, 2021). The environmental data provided by Refinitiv is sourced from publicly available, reliable sources, including annual reports, news, CSR reports, firm and NGO websites, and stock exchange filings. This data is further validated by a team of hundreds of ESG specialists (Refinitiv, 2021), ensuring its accuracy and reliability for our analysis. We used Refinitiv's CO₂ equivalent emissions data as an objective proxy for firms’ actual environmental performance. Unlike rating-based metrics, this proxy is an impartial measure, providing a reliable basis for our analysis.

We sourced firms’ fundamental and financial information from Compustat, accessed via Wharton Research Data Services. Compustat is a leading database management system offering comprehensive data on approximately 24,000 firms listed on North American public exchanges. To consolidate our analysis, we used firm tickers as identifiers to merge data from LexisNexis, Refinitiv, and Compustat. To ensure accuracy in the merging process, we carefully addressed any discrepancies in tickers by examining ticker changes, as well as mergers and acquisitions. Additionally, we used firm names as a supplementary matching method to further validate the accuracy of our data integration. Our final data set, obtained by merging these sources and removing instances with missing values, comprises 452 firms across different industries. For the number of firm-year observations for each variable, refer to Table 1. We also accessed county-level US presidential election data from MIT Election Data and Science (2022) to construct IVs. These variables are crucial for our IV regression analysis, enabling us to examine the causal relationships and mitigate potential endogeneity issues.

Table 1.

Descriptive statistics of variables used in the study.

Variables	Mean	SD	25th	50th	75th	No. of observations
(1) News Spotlight	61.202	204.463	0.000	0.000	24.568	12,977
(2) Env. Perform. (CO2 Eq. Emiss.)	7,427,661	2.02 × 107	150,415	760,043	4,418,147	3,069
(3) Firm Size (Market Cap. USD millions)	14,200	39,600	1,080	3,010	10,300	12,775
(4) Leverage	0.246	0.223	0.079	0.215	0.359	12,900
(5) Revenue Growth	0.127	0.971	−0.011	0.057	0.138	11,893
(6) Labor Productivity	771.433	3,898.425	213.830	342.568	671.084	12,870
(7) Capital Expenditure (CapEx)	0.041	0.049	0.010	0.027	0.055	12,859
(8) R&D Expenditure	0.026	0.170	0.000	0.000	0.020	11,401
(9) Dividend	288.885	1,035.225	0.000	24.055	149.601	12,837
(10) Liquidity	0.135	0.151	0.028	0.079	0.188	12,946
(11) Fixed Asset	0.243	0.233	0.054	0.165	0.370	12,231
(12) Age	36.441	68.551	16.000	26.000	44.000	12,833
(13) Return on Assets (ROA)	0.047	0.112	0.014	0.046	0.086	12,977
(14) Return on Equity (ROE)	0.025	0.150	0.028	0.049	0.069	12,841
(15) Return on Sales (ROS)	0.076	0.137	0.030	0.069	0.129	12,834
(16) Net Profit Margin (NPM)	0.065	0.126	0.029	0.065	0.115	11,352
(17) Return on Investment (ROI)	0.079	0.133	0.035	0.077	0.133	12,839
(18) Market-to-Book (M–B)	3.072	4.166	1.421	2.253	3.698	12,825
(19) Tobin's Q	1.937	1.330	1.154	1.524	2.196	12,844
(20) Debt-to-EBITDA	2.332	3.160	0.532	1.620	3.204	11,307
(21) Cost of Goods Sold (COGS)	0.601	0.226	0.449	0.641	0.776	12,834
(22) SG&A Expenses	0.211	0.196	0.063	0.157	0.303	10,755

See Table C2 in the Supplemental materials for detailed descriptions of variable operationalizations.

5. Research Methodology

5.1. Regression Specification

The primary regression model for testing hypothesis 1 is:

$$\begin{array}{rl}{Y}_{i,t+1}& ={\beta }_0+{\beta }_{1}NewsSpotligh{t}_{i,t}+{\beta }_2\frac{1}{3}\sum _{k=0}^{2}En{v}_{i,t-k}\\ & +{\beta }_3{X}_{i,t}+{\alpha }_i+{\lambda }_t+{ϵ}_{i,t}\end{array}$$

Here, Y_i,t+1 represents firm i's financial performance and market value at time t + 1. News Spotlight_i,t (i.e., treatment variable) is measured for firm i at time t to address any reverse causality concerns. We control for firms’ environmental performance (Env), given its significant impact on financial performance, as indicated in previous studies (Awaysheh et al., 2020; Ba et al., 2013; Flammer, 2015; Jacobs, 2014; Jacobs et al., 2010; Khuntia et al., 2018; Montabon et al., 2007). To capture firms’ environmental performance, we average the data over the preceding 3 years to account for potential time lags and address look-ahead bias, as CO₂ emissions data are generally published after the fiscal year ends. Our machine learning-based climate action vocabulary, developed from the climate action plans of 365 US cities, is used to construct the News Spotlight variable. We adopted a design science perspective and developed a systematic framework using information retrieval, machine learning, and natural language processing methods, including word embeddings, to develop a semantically coherent vocabulary. A summary of this vocabulary development process is illustrated in Figure 1. Refer to Section A in the Supplemental materials for more detailed information on the methodology. This vocabulary, consisting of 338 terms including unigrams and bigrams such as “sustainability,” “emissions,” “greenhouse gas,” “energy efficiency,” and “climate change,” effectively captures the essence of climate action language in the news. Figure 2 shows a two-dimensional representation of the top 100 climate action terms from the vocabulary. The complete list of terms and their degree of relevance to the climate action topic is detailed in Table C1 (Supplemental materials). We validated the effectiveness of our vocabulary in capturing climate action semantics by conducting comprehensive analyses at the news, industry, and firm levels, using the text data from LexisNexis News Sources. Our validation approach involved manually labeling a sample of 1,000 news articles by Amazon Mechanical Turk workers, which provided ground truth labels for measuring the accuracy of our vocabulary and vocabulary-based climate action metrics. Through this benchmarking, we found that climate action news classification based on our metrics achieved an extremely high area under the ROC curve (AUC) of 0.97 and an F1 score of 0.93. In other words, the vocabulary we created through machine learning techniques and the metrics we built on top of it are, for all practical purposes, as good as a committee of human workers in identifying climate action news. However, unlike a committee of human workers, our proposed method has no marginal cost, is transparent, and can scale to large numbers of news stories. We also benchmarked our results against Mistral Large 2, a state-of-the-art LLM, confirming that our vocabulary achieves comparable performance. We also conducted firm- and industry-level validation by showing the vocabulary's ability to consistently rank polluting and clean industries and their relative predictive utility in predicting firms’ environmental performance. These results confirm the robustness of the vocabulary in capturing climate action content relevant to each firm, providing a reliable basis for subsequent econometric analyses. The details of our validation process are elaborated in Section B in the Supplemental materials.

We used our developed vocabulary to systematically measure the news spotlight by identifying and quantifying firms’ news related to climate action. The News Spotlight for each firm at the firm-year level is calculated as the weighted sum of the number of climate action-related news articles. An article is classified as climate action-related if it contains at least one term from our vocabulary. The weight for each article is based on the average importance of the climate action terms it contains. These importance values measure each term's relevance to climate action and are computed as the term's frequency in our source corpus (the climate action plans of US cities) relative to its general prevalence in Google search results. This metric is detailed in step 5 of our vocabulary development framework (Figure 1, and Table C1 and Section A in the Supplemental materials). We recreated the News Spotlight metric by rating news articles with Mistral Large 2, a state-of-the-art LLM known for its advanced natural language processing capabilities. Using one-shot learning, we prompted Mistral to rate each article's relevance to firm climate actions on a scale from 0 to 100. In this way, we ensure robustness in our econometric findings, regardless of the text analytics method used to create News Spotlight (refer to EC.B4 in the Supplemental materials for further details). Our parameter of interest is β₁, with a positive (negative) β₁ indicating improved (deteriorated) firm performance with increased news spotlight. According to hypothesis 1, we anticipate that firms with a higher News Spotlight have lower financial performance (i.e., negative β₁).

Our regression model includes a vector X of firm attributes recognized in the literature to be associated with firms’ financial performance (Barker et al., 2022; Darendeli et al., 2022; Flammer, 2015; Hsu and Wu, 2023; Li et al., 2022; Lo et al., 2022; Lu et al., 2021), which could also affect its relationship with the news spotlight. First, to control for persistence in firm revenue growth, we include Revenue Growth, the annual percentage increase in firm i's revenue. Firm Size (i.e., market capitalization) and Firm Age (i.e., years since initial public offering) are also included, as smaller and younger firms often exhibit greater diversity in growth and performance. We include Labor Productivity to capture the impact of workforce efficiency. We control for Leverage to account for how much a firm's operations are financed through debt. We also include R&D expenditures normalized by sales to account for firms’ innovation and knowledge generation efforts. We include Fixed Asset ratio to account for firms’ asset tangibility and capital intensity. We also control for Dividends to account for fluctuations in anticipated financial performance. We further control for Asset Liquidity, the ratio of a firm's cash flow to its total assets. Finally, we include CapEx as a control for firm investments in tangible assets.

We use various measures of firm financial performance (Y), including ROA, return on sales (ROS), return on equity, return on investment (ROI), and net profit margin, following prior studies (Arora et al., 2020; Chang et al., 2022; Flammer, 2015; Lo et al., 2022; Mani et al., 2015). We measure financial leverage using the Debt-to-EBITDA ratio, indicating a firm's debt-paying capacity relative to its earnings before interest, taxes, depreciation, and amortization (EBITDA). Beyond accounting-based metrics, we include M–B ratio and Tobin's Q as measures of firm valuation, reflecting a firm's market value relative to its book value and asset replacement cost. Our dataset comprises observations from various firms across different periods, allowing us to use panel data analysis to test our hypotheses. We regress firm i's financial performance (Y) on News Spotlight, control variables (X), and introduce Firm and Year-fixed effects (α and λ) to account for unobserved heterogeneity and time-series variations. We use robust standard errors to account for potential heteroscedasticity in the analysis.

5.2. IV Analysis

Despite controlling for various factors and Firm and Year-fixed effects, endogeneity may persist due to unobserved factors associated with News Spotlight that could affect firms’ financial performance (i.e., omitted variable bias). For instance, firms may engage in activities that attract media attention and influence their financial outcomes for reasons such as greenwashing (Section 1). To address these endogeneity concerns, we conduct an IV analysis using county-level presidential voting data from firms’ headquarters as the IV. Prior studies (Albuquerque et al., 2019; Darendeli et al., 2022) have shown that political preferences are correlated with environmental attitudes (Di Giuli and Kostovetsky, 2014). Firms in counties with higher Democratic voting ratios face more CSR pressure (Albuquerque et al., 2019; Darendeli et al., 2022) and receive more news spotlight (fulfilling the IV relevance condition). We construct our IV, “Democratic Voting” ratio, using Securities and Exchange Commission (SEC) filings on firms’ headquarters and the proportion of Democratic votes in the 2000 presidential election. This year predates our sample period by 5 years, reducing the risk of a direct IV effect on firm financial performance. We estimate the following two-stage least squares (2SLS) regression model:

First stage:

$$\begin{array}{rl}NewsSpotligh{t}_{i,t}& ={\pi }_0+{\pi }_{1}IV+{\pi }_2\frac{1}{3}\sum _{k=0}^{2}En{v}_{i,t-k}+{\pi }_3{X}_{i,t}\\ & +{\delta }_{I(i)}+{\gamma }_{s(i)}+{\lambda }_t+{\nu }_{i,t}\end{array}$$

Second stage:

$$\begin{array}{rl}{Y}_{i,t+1}& ={\beta }_0+{\beta }_1\widehat{NewsSpotligh{t}_{i,t}}+{\beta }_2\frac{1}{3}\sum _{k=0}^{2}En{v}_{i,t-k}+{\beta }_3{X}_{i,t}\\ & +{\delta }_{I(i)}+{\gamma }_{s(i)}+{\lambda }_t+{ϵ}_{i,t}\end{array}$$

$\widehat{NewsSpotligh{t}_{i,t}}$ is the predicted value of $NewsSpotligh{t}_{i,t}$ from the first stage. Given the cross-sectional variation in historical voting across counties, we include Industry-fixed effects (i.e., two-digit Standard Industrial Classification (SIC) codes) ( ${\delta }_{I(i)}$ ) in the IV estimations. We also include State and Year-fixed effects $({\gamma }_{s(i)},{\lambda }_t)$ to account for location and time-specific shocks. These fixed effects help mitigate potential correlations between financial performance and political ideologies that may stem from geographic industry clustering or variations in state-level policies, such as taxation in Democratic-leaning regions (Albuquerque et al., 2019), which likely satisfies the exclusion restriction condition. ${\nu }_{i,t}$ and ${ϵ}_{i,t}$ are the error terms that account for unobserved effects. All remaining model components are defined in the main specification (Section 5.1). We use standard Stata packages to perform the 2SLS analyses in a panel data setting.

6. Main Results

In Figure 3, we present the distribution of the News Spotlight ratio ² across various industries. Industries with the highest ratios, as per two-digit SIC codes, include 10–14 (Mining), 40–49 (Transportation & Public Utilities), and 20–39 (Manufacturing), which aligns with their status as environmentally sensitive industries. Conversely, industries like 60–67 (Finance, Insurance, Real Estate), 50–51 (Wholesale Trade), 70–89 (Services), and 52–59 (Retail Trade) have lower News Spotlight ratios.

Figure 3.

Distribution of average News Spotlight ratio across various industries. Firms in construction, agriculture, forestry & fishing, and public administration are excluded, as they account for less than 2% of firms in the data (Table B2 in the Supplemental materials).

Table 1 shows the descriptive statistics of the variables used in our study. A detailed description of the variables can be found in Table C2 in the Supplemental materials. The average News Spotlight is 61.2, which indicates that, on a yearly basis, the number of news articles related to a firm's climate actions, weighted by the average term importance value of the terms contained within, is approximately 61.2. Univariate analysis of mean differences (Table C4 in the Supplemental materials) shows that firms with higher News Spotlight tend to be larger, older, and more polluting, with a higher fixed asset ratio and lower R&D expenditure.

6.1. Firm Financial Performance and Value

In this section, we examine the impact of News Spotlight on firms’ financial performance and value, as outlined in hypothesis 1. Initial ordinary least squares regression analysis reveals a statistically significant negative effect of News Spotlight on financial performance (Table C5 in the Supplemental materials) while controlling for potential confounders (variables 2 to 12 in Table 1) and Firm and Year-fixed effects. Notably, when we apply the “Democratic Voting” ratio from the 2000 presidential election as an IV for News Spotlight (Table 2), the negative effect on financial performance becomes more significant, and the economic impact is magnified. On average, one additional article per year, which equates to a 38.36-unit ³ increase in News Spotlight, leads to a 0.69 percentage point (=38.36 × 1.81 × 10⁻⁴) reduction in ROA (median ROA = 0.046 or 4.6%) (column 1, Table 2). To confirm the robustness of our findings, we substitute ROA with other financial performance metrics, such as ROS, and observe consistent results (Table 2), corroborating hypothesis 1. The first-stage regression of News Spotlight on the IV and other exogenous variables yields a high Cragg-Donald Wald statistic recommended by Stock and Yogo (2005). This indicates the strength and relevance of our chosen instrument, thereby bolstering the reliability of our IV analysis.

Table 2.

IV regression coefficients: News Spotlight impact on firm financial performance.

	$RO{A}_{t+1}$	$RO{S}_{t+1}$	$RO{E}_{t+1}$	$\text{Tobin’s }{Q}_{t+1}$ (Ln)	$NP{M}_{t+1}$
${\text{Independent variables}}_t$	(1)	(2)	(3)	(4)	(5)
News Spotlight	−1.81***	−4.39***	−1.97***	−5.07**	−4.33***
	(0.40)	(0.84)	(0.65)	(2.17)	(0.76)
Instrumented Voting 2000: Bush vs. Gore	Yes	Yes	Yes	Yes	Yes
Control variables	Yes	Yes	Yes	Yes	Yes
Industry-fixed effects	Yes	Yes	Yes	Yes	Yes
State- and Year-fixed effects	Yes	Yes	Yes	Yes	Yes
No. of observations	2,453	2,453	2,453	2,401	2,447
Cragg-Donald Wald statistic	36.91	36.91	36.91	34.79	36.86

Note. IV = instrumental variable; ROA = return on assets; ROS = return on sales; ROE = return on equity; NPM = net profit margin.

Controls: Var. 2 to 12 (Table 1). The coefficients and the robust standard errors (in parenthesis) are multiplied by 10⁴ to reflect the scale difference between News Spotlight and the dependent variables as seen in Table 1. Refer to Tables D1 and C2 (Supplemental materials) for the full table and variable definitions.

Statistical significance is denoted as ***p < 0.01, **p < 0.05, and *p < 0.1.

6.2. Potential Mechanism and Robustness

To explore the mechanisms underlying the effect of News Spotlight on firm financial performance (hypotheses 2 and 3), we analyze its impact on COGS and SG&A (Table 3). COGS is measured as the ratio of the cost of goods sold to net sales, while SG&A expenses represent the ratio of selling, general, and administrative expenses to total revenue. Our results reveal a significant positive effect of News Spotlight on both COGS and SG&A expenses, suggesting that an increase in News Spotlight leads to higher operational expenses and costs, corroborating hypothesis 2. Specifically, each additional news article per year, equivalent to a 38.36-unit increase in News Spotlight, corresponds to a 1.10 and a 0.63 percentage point rise in COGS and SG&A, respectively (median COGS = 0.64 and SG&A = 0.16). Given the median sales/revenue of $2,270m in the data, the estimated effects translate to a $24.97m ($14.30m) loss in sales/revenue due to increased operational costs through COGS (SG&A) per firm-year. Also, we examine the impact of News Spotlight on the Debt-to-EBITDA ratio as a measure of firms’ financial leverage and debt management. The estimated coefficient is highly significant, indicating that higher News Spotlight leads to increased debt relative to earnings. This shows that due to increased News Spotlight, firms may need to finance capital expenditures related to environmental initiatives and public relations activities with no immediate returns. Growing operational costs (COGS and SG&A expenses) could reduce EBITDA even if the debt remains constant, thereby increasing the Debt-to-EBITDA ratio.

Table 3.

IV regression coefficients: News Spotlight impact on firm cost and debt metrics.

	$COG{S}_{t+1}$	$SG\mathrm{\&}{A}_{t+1}$	${\text{Debt-to-EBITDA}}_{t+1}$
${\text{Independent variables}}_t$	(1)	(2)	(3)
News Spotlight	2.90***	1.63***	40.40***
	(0.94)	(0.60)	(14.79)
Instrumented Voting 2000: Bush vs. Gore	Yes	Yes	Yes
Control variables	Yes	Yes	Yes
Industry-fixed effects	Yes	Yes	Yes
State-fixed effects	Yes	Yes	Yes
Year-fixed effects	Yes	Yes	Yes
No. of observations	2,453	1,001	2,443
Cragg-Donald Wald statistic	36.91	57.19	37.03

Note. IV = instrumental variable; COGS = cost of goods sold; SG&A = selling, general, and administrative.

Controls: Var. 2 to 12 (Table 1). The coefficients and the robust standard errors (in parenthesis) are multiplied by 10⁴ to account for the scale difference between News Spotlight and the dependent variables as seen in Table 1. Refer to Tables D2 and C2 (Supplemental materials) for the full table and variable definitions.

Statistical significance is denoted as ***p < 0.01, **p < 0.05, and *p < 0.1.

To ensure the validity of our findings, we carried out a series of robustness checks. First, we validated our primary results using alternative measures of firm financial performance and different constructs of the News Spotlight variable. For financial performance metrics, we extended our analysis beyond the variables used in Table 2 to include ROI, M–B, and an alternative measure of ROA defined as (SALE − COGS − SG&A)/Total Assets. As shown in Table C6 in the Supplemental materials, our results remain robust, corroborating hypothesis 1. Furthermore, we devised alternative measures of the News Spotlight variable by calculating the weighted ⁴ sum of the number of total and unique climate action terms in a firm's news coverage within a year. We also redefined the News Spotlight variable by scoring news articles with Mistral Large 2 instead of our vocabulary-based approach, where articles were rated from 0 to 100 based on relevance to firm climate actions using one-shot learning. The regression estimates remain consistent under the alternative News Spotlight measures, showing the validity of the vocabulary-based News Spotlight measures and the robustness of our findings (Tables B5, B6, B7, C7, and C8 in the Supplemental materials).

To further validate our findings, we employed an alternative IV for News Spotlight, termed “Peer Green Spotlight,” representing the aggregated news spotlight of peer firms within the same industry. Based on the first-stage Cragg-Donald Wald statistic, this instrument also passes the weak IV test. This result is expected, given the substantial role of firms’ industry peers in attracting media attention. Moreover, with Firm-fixed effects, it is unlikely that firm performance is influenced by industry-wide News Spotlight, except through its impact on the firm's News Spotlight, thereby fulfilling the exclusion restriction condition. The regression estimates with this alternative instrument are consistent with our initial results (Table C9 in the Supplemental materials). Additionally, we recreated the “Democratic Voting” ratio IV using the 2004 US presidential election results (Bush vs. Kerry) and found consistent estimates (Table C10 in the Supplemental materials). To minimize potential correlation with our target variable (firm financial performance) during the study period (2005–2017), we avoided using more recent election data as an IV.

As shown in Table 1, a significant portion of our treatment variable, News Spotlight, is clustered at zero. This systematic clustering indicates that many firms do not receive any news spotlight on climate action, which is expected. To address this censoring issue, we employ the Tobit regression model, which is well-suited for estimating relationships in datasets where zero values of the treatment variable are systematic and carry meaningful information. The reliability of the Tobit model results is further enhanced by incorporating the “Democratic Voting” ratio as an IV. This approach provides a robust estimation of the causal effects of News Spotlight on firm financial performance while considering the zero-inflated nature of our treatment variable. Consistent with our initial findings, the estimates from the Tobit regression indicate that an increased news spotlight on firm climate action negatively impacts their financial performance and positively affects operational costs (hypotheses 1 and 2; Table C11 in the Supplemental materials). Figure B4 (Supplemental materials) reveals a nonstationary trend in News Spotlight over the study period (2005–2017). Since firm financial performance is also nonstationary due to macroeconomic factors, we re-estimated our main analysis using first-differenced variables instead of within fixed effects. All results remained robust (Tables C16 and 17 in the Supplemental materials). We also respecified the model by introducing lags of up to 3 years for all independent variables to study the longer-term impacts of News Spotlight on firms’ financial performance. Although the magnitude of the News Spotlight coefficient slightly decreases over time, the negative effect remains consistent and statistically significant (Tables C18 and 19 in the Supplemental materials).

6.3. Mediation Analysis: COGS and SG&A

Next, we conduct a mediation analysis to investigate how increased News Spotlight leads to reduced firm financial performance. In this model, we treat firm operational costs (i.e., COGS and SG&A expenses) as parallel mediators between News Spotlight and various firm financial performance metrics. Given the potential endogeneity concerns with News Spotlight, we conduct a causal mediation analysis with parallel mediators following the guidelines of Dippel et al. (2020) and Hayes’ (2017) PROCESS Model 4, using a single instrument (i.e., “Democratic Voting” ratio). The mediation analysis also includes the control variables (Var. 2 to 12 in Table 1), State, Industry, and Year-fixed effects. Our results, presented in Table 4 and Figure 4, indicate that operational costs (COGS and SG&A expenses) significantly mediate the relationship between News Spotlight and financial performance, corroborating hypothesis 3. Specifically, COGS and SG&A mediate 25% and 75% of the impact of News Spotlight on ROA, respectively.

Table 4.

Coefficient estimates of the effects of News Spotlight, COGS, and SG&A expenses on firm financial performance in mediation analysis.

	${\text{News Spotlight}}_t$	News Spotlight through COGS		News Spotlight through SG&A		Total
Dependent variable	Direct effect	Indirect effect	Prop. mediated	Indirect effect	Prop. mediated	Indirect effect	Prop. mediated
$\text{RO}{\text{A}}_{t+1}$	−0.83	−0.21**	0.25	−0.65***	0.75	−0.87***	1.0
$\text{RO}{\text{E}}_{t+1}$	−1.15	−0.10*	0.19	−0.44**	0.81	−0.54***	1.0
$\text{RO}{\text{I}}_{t+1}$	−0.99	−0.23**	0.20	−0.94***	0.80	−1.17***	1.0
$\text{RO}{\text{S}}_{t+1}$	−3.01\ast \ast \ast	−0.59**	0.12	−1.26***	0.26	−1.85***	0.38
$\text{NP}{\text{M}}_{t+1}$	−3.26\ast \ast \ast	−0.68**	0.13	−1.48***	0.27	−2.16***	0.40

Note. COGS = cost of goods sold; SG&A = selling, general, and administrative; ROA = return on assets; ROE = return on equity; ROI = return on investment; ROS = return on sales; NPM = net profit margin.

Controls: Var. 2 to 12 (Table 1); State, Year, and Industry-fixed effects. The coefficients are multiplied by 10⁴ to account for the scale difference between News Spotlight and the dependent variables, as seen in Table 1. Refer to Tables D3 and C2 (Supplemental materials) for the full table and variable definitions.

Statistical significance is denoted as ***p < 0.01, **p < 0.05, and *p < 0.1.

Figure 4.

Structure of the mediation analysis. Coefficient values shown in the arrows may appear small due to the scale difference between News Spotlight and the other variables as seen in Table 1.

6.4. Media Spotlight Sentiment and Financial Performance

Positive and negative news spotlights could impact firms’ financial performance and value through distinct channels. Using FinBERT, a deep learning method fine-tuned on 1.8 million financial news articles from Reuters TRC2 (Araci, 2019), we classified climate action news articles into three categories: positive (24%), negative (42%), and neutral (34%). Our regression analysis separately examines the effects of positive and negative news spotlights on various financial performance measures, controlling for Industry, Year, and State-fixed effects (Table 5). This approach ensures that our findings reflect average within-industry changes in financial performance attributable to variations in positive and negative news spotlights after accounting for macroeconomic, temporal, and geographical factors. Using the “Democratic Voting” ratio as an IV, we address potential endogeneity concerns. Our results reveal that both positive and negative news spotlights on climate actions have negative impacts on firms’ financial performance and value, corroborating hypotheses 4 and 5. While a negative effect from positive news coverage might seem counterintuitive, such outcomes are not unexpected in the literature. Krüger (2015) found negative returns on positive ESG events, and others (Capelle-Blancard and Petit, 2019; Cornell and Damodaran, 2020) have indicated that positive ESG announcements often fail to yield shareholder returns. The positive news spotlight increases COGS, SG&A expenses, and Debt-to-EBITDA (Table 5), likely due to heightened environmental expectations and scrutiny that prompt firms to expedite sustainability initiatives. These efforts typically lead to immediate cost escalations and elevated operational expenses, driven by activities such as supplier vetting, intensified marketing, and enhanced reporting and auditing processes (Di Giuli and Kostovetsky, 2014; Marinovic and Varas, 2016). These findings align with recent “greenhushing” trends, where firms opt not to publicize their positive climate initiative to avoid costs and potential backlash. A study featured by Financial Times reports that “a quarter of the 1,200 firms surveyed in 12 countries chose not to promote their science-based net zero targets” (Speed, 2022). According to South Pole, a global carbon offsets developer, firms perceive it as “too risky to publicize efforts to improve climate records,” leading to strategic silence. AlphaSense has revealed a 31% year-over-year decline in mentions of sustainability initiatives on US earnings calls, even as mentions in proxy statements have increased, suggesting sustained operational environmental commitments despite reduced public discourse to avoid the spotlight (Brue, 2023). Similarly, a recent Wall Street Journal article attributes this silence in publicizing climate goals to increased regulatory scrutiny and stringent industry expectations (Uribe, 2024). The negative climate action spotlight also increases firms’ operational costs, possibly due to heightened regulatory scrutiny and reputational risks, necessitating public relations efforts and potential investments in sustainable transitions to mitigate reputational damage. Firms may also incur costs related to litigation and penalties (Foerderer and Schuetz, 2022; Kölbel et al., 2017; Krüger, 2015). Mediation analyses show that COGS and SG&A expenses mediate the effects of both positive and negative spotlights on financial performance (Tables C12 and C13 in the Supplemental materials).

Table 5.

IV regression coefficients: positive and negative News Spotlight impact on firm financial performance and cost and debt metrics.

	$RO{A}_{t+1}$	$RO{S}_{t+1}$	$RO{E}_{t+1}$	$\text{Tobin’s }{Q}_{t+1}$ (Ln)	$NP{M}_{t+1}$	$COG{S}_{t+1}$	$SG\mathrm{\&}{A}_{t+1}$	${\text{Debt-to-EBITDA}}_{t+1}$
${\text{Indep. variables}}_t$	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Instrumented positive spotlight	−4.21***	−9.55***	−5.01***	−12.35**	−9.19***	14.53**	6.26***	103.18***
	(1.09)	(2.40)	(1.80)	(6.08)	(2.16)	(5.91)	(2.26)	(39.02)
Instrumented negative spotlight	−2.27***	−5.15***	−2.70***	−6.74**	−4.96***	7.83**	3.15***	55.68***
	(0.56)	(1.22)	(0.91)	(3.30)	(1.10)	(3.13)	(0.98)	(20.50)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Industry-, State-, and Year-FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
No. of observations	2,182	2,182	2,182	2,130	2,176	2,182	929	2,172
Cragg-Donald Wald statistic	29.93	29.93	29.93	26.69	29.96	29.93	30.66	30.20
	31.23	31.23	31.23	26.99	31.21	31.23	36.55	31.41

Note. IV = instrumental variable; ROA = return on assets; ROS = return on sales; ROE = return on equity; NPM = net profit margin; COGS = cost of goods sold; SG&A = selling, general, and administrative; FE = fixed effect.

Controls: Var. 2 to 12 (Table 1). The coefficients and robust standard errors (in parenthesis) are multiplied by 10⁴ to account for the scale difference between News Spotlight and the dependent variables, as seen in Table 1. Refer to Tables D4.1, D4.2, and C2 (Supplemental materials) for the full table and variable definitions.

Statistical significance is denoted as ***p < 0.01, **p < 0.05, and *p < 0.1.

6.5. Moderation Effects of Firms’ Environmental Performance and Industry “Dirtiness”

This section examines how heterogeneous factors, including firms’ environmental performance (proxied by CO₂ equivalent emissions) and their industries’ pollution status (“dirtiness”), influence the relationship between news spotlight and firm performance. To capture these effects, we introduce their interaction with the News Spotlight to our regression model. Column 2 in Table 6 includes the interaction term CO₂ × News Spotlight, and column 4 includes Ind. Dirt. × News Spotlight. In the first stage of the analysis, we employ the “Democratic Voting” ratio from the 2000 US presidential election as the IV, along with control variables, Industry, State, and Year-fixed effects, to estimate News Spotlight (columns 1 and 3). The coefficient of “Democratic Voting” in the first-stage 2SLS regression is positive and statistically significant, indicating that firms headquartered in regions with higher democratic voting ratios receive greater news spotlight. In the second stage, we use fitted values of the News Spotlight variable and interaction terms to examine their effects on ROA (columns 2 and 4). The significant positive coefficients of the interaction terms suggest that environmental performance and industry pollution levels moderate the relationship between News Spotlight and ROA, corroborating hypotheses 6 and 7. Since the coefficients of the interaction terms exceed the negative coefficient of News Spotlight, this suggests that increased media attention may financially benefit high-polluting firms. With a lower baseline of environmental reputation, these firms have more opportunities for cost-effective ecological improvements (Sadovnikova and Pujari, 2017). High-polluting firms may also experience regulatory cost savings and operational efficiencies from timely responses to climate attention, driven by heightened stakeholder scrutiny on these firms (Cordeiro and Tewari, 2015; Flammer, 2013). Conversely, low-polluting firms, which are already operating with relatively more efficient and eco-friendly practices, face higher relative costs for additional environmental actions due to a smaller scale of benefits. For them, spotlight-driven environmental investments may yield marginal improvements at substantial operational costs, providing limited financial benefits.

Table 6.

IV regression coefficients: heterogeneous effects of News Spotlight on ROA with interactions.

	${\text{News Spotlight}}_t$	$RO{A}_{t+1}$	${\text{News Spotlight}}_t$	$RO{A}_{t+1}$
	1st stage	2nd stage	1st stage	2nd stage
${\text{Independent variables}}_t$	(1)	(2)	(3)	(4)
Democratic Voting ratio (i.e., instrumental variable)	0.265***		0.367***
	(0.038)		(0.041)
Spotlight_IV (i.e., fitted value of News Spotlight)		−0.036***		−0.042***
		(0.007)		(0.008)
CO2 Eq. emissions	0.409***	0.290	0.349***	−0.019
	(0.050)	(0.241)	(0.042)	(0.014)
Ind. Dirt. a			−0.590***	−0.024***
			(0.069)	(0.005)
Spotlight_IV × CO2 Eq. emissions		0.097***
		(0.031)
Spotlight_IV × Ind. Dirt.				0.054**
				(0.024)
Control variables	Yes	Yes	Yes	Yes
Industry FE	Yes	Yes	No	No
State and Year FE	Yes	Yes	Yes	Yes
No. of observations	2,453	2,453	2,453	2,453
Cragg-Donald Wald statistic	36.90		81.06

Note. IV = instrumental variable; ROA = return on assets; FE = fixed effect.

Controls: Var. 3 to 12 (Table 1). All variables are standardized for enhanced interpretability. Refer to Tables D5, D6, and C2 (Supplemental materials) for the full results and variable definitions.

^a

Ind. Dirt. (=1: environmentally “dirty” industries based on two-digit SIC codes: (10) Metal mining, (13) Oil and gas extraction, (26) Paper and allied products, (28) Chemicals and allied products, (29) Petroleum refining, (33) Primary metal industries, (49) Electric, gas, and sanitary services; =0: otherwise) (Hsu and Wang, 2013). Statistical significance is denoted as ***p < 0.01, **p < 0.05, and *p < 0.1.

Similarly, the results reveal that firms within “dirty” industries may financially benefit from the increased climate action spotlight, likely due to unique market dynamics within these firms. In industries where environmental action is becoming a competitive imperative, firms can strengthen their market position by pursuing sustainability initiatives (Cordeiro and Tewari, 2015).

6.6. Moderation Effects of Market Orientation, Durability Classification, Index Constituency, and Greenwashing

The literature has shown that firms with higher market visibility and direct consumer interaction are more sensitive to consumer perceptions and public scrutiny, potentially influencing their response to climate-related media spotlight (Chiu and Sharfman, 2011; Servaes and Tamayo, 2013). To investigate this, we interact News Spotlight with binary variables for B2B and B2C firms. Our findings indicate that B2C firms, which engage directly with customers, observe a stronger financial impact from the climate news spotlight than B2B firms, which exhibit a less negative reaction due to their lower consumer proximity (Tables D7 and D8 in the Supplemental materials). Additionally, firms in industries with complex manufacturing processes and capital-intensive structures, such as those in the durable goods sector, often face challenges in adapting swiftly to external pressures, including media attention (Leung and Sun, 2021). To account for this, we interact News Spotlight with a binary indicator for firms in durable goods industries, assessing how organizational rigidity in these capital-intensive sectors may moderate the impact of the media spotlight on financial performance. We found that firms in durable goods industries experience a less negative impact from News Spotlight, likely due to their long product development cycles and significant fixed asset investments, which hinder rapid strategic adjustments to the spotlight (Table D9 in the Supplemental materials). Also, larger, more visible, established firms with dedicated public relations (PR) departments and substantial resources to manage public perception may respond differently to the media spotlight (Bednar et al., 2013; Kölbel et al., 2017). To capture this, we included S&P 500 status as an interaction term with News Spotlight and found that these firms are impacted more negatively by increased news spotlight on their climate actions, likely due to the higher costs and reputational risks associated with managing public relations of S&P 500 firms (Table D10 in the Supplemental materials). The literature also documents the negative financial impacts of greenwashing on firm performance (Darendeli et al., 2022; Walker and Wan, 2012). To examine this, we created a greenwashing measure based on the positive media spotlight a firm received for climate action relative to its actual environmental performance (refer to EC.D2 in the Supplemental materials for more details). Our analysis reveals that firms with higher greenwashing scores experience more negative financial performance from the spotlight, underscoring the financial risks associated with deceptive climate communications.

7. Conclusion

News coverage of firms plays a critical role in shaping their public perception and strategic positioning. However, whether and how the news spotlight on firms’ climate actions affects their operations, financial performance, and market value remains primarily unanswered. With rising concerns over climate change, firms’ climate actions have become a central focus of the media spotlight and stakeholder scrutiny. This increased attention may prompt firms to take operational actions to address the spotlight, often through public relations efforts to validate or counter the disclosed information. Previous literature has focused on studying the effects of isolated environmental events on stock prices, primarily examining polluting industries where these incidents are prevalent, without providing a comprehensive cross-sectoral view. Past studies have used news coverage of environmental incidents as a proxy for their actual environmental performance, but they do not isolate the effects of the news spotlight itself. Furthermore, these studies lack a systematic framework for detecting events from text and thus rely on manual news reading and coding or proprietary datasets. In this study, we addressed these gaps by systematically measuring the climate action news spotlight on firms and investigating its impact on their financial performance and market value while controlling for their actual environmental performance.

We contribute to the current body of literature in several notable ways.

First, we developed an automated machine learning-based text mining framework and used it to construct a climate action vocabulary. This vocabulary, created using natural language processing and information retrieval techniques, provides an effective, systematic, and interpretable measure of the language related to climate actions in news articles. Using this vocabulary, we analyzed news content related to firms and proposed quantitative metrics to measure the intensity of climate-related media attention over specific periods. We evaluated the semantic validity of our vocabulary by creating ground truth labels for a sample of news data using Amazon Mechanical Turk ratings and calculated the accuracy of our vocabulary-based metrics in detecting climate-related news (AUC = 0.97, F1 score = 0.93). Also, we benchmarked our vocabulary-based method against Mistral Large 2, a state-of-the-art large language model, and found comparable performance (AUC = 0.96, F1 score = 0.89). This comparison showed that our method performs on par with state-of-the-art LLMs while offering distinct advantages in transparency, interpretability, and replicability. We aggregated the climate action-based news spotlight metrics across industries and observed that environmentally “dirty” sectors (e.g., mining) exhibited the highest values of the news spotlight metric. We further developed machine learning models to assess the predictive utility of our metrics in forecasting firms’ environmental scores. These validation tests confirm the semantic validity of our vocabulary in capturing climate action language from text documents.

Second, we examined whether firms benefit or lose from the media spotlight on their climate actions. Our findings revealed that increased media attention on climate actions, regardless of sentiment, has an adverse effect on firms’ financial performance, contradicting the assumption that positive news is always beneficial. We found that heightened media spotlight significantly raises firms’ operational costs and debt levels, explaining the observed negative financial impact. Although the adverse financial impact of positive climate spotlight may initially seem counterintuitive, it is not unprecedented. Relevant literature has shown that firms often experience negligible gains or even losses following positive ESG disclosures. A heightened positive spotlight on climate action raises the bar for firms regarding environmental responsibility, pressuring firms to accelerate costly sustainability efforts. Such costs can arise from activities like scrutinizing suppliers, intensifying marketing initiatives, and implementing more frequent reporting and audits. Additionally, we uncovered the mechanisms through which the media spotlight impacts financial performance. Our findings could explain why some firms have recently adopted a “silence is golden” approach, opting for “greenhushing” to minimize public disclosure of their sustainability efforts and avoid escalating costs associated with heightened scrutiny and stakeholder expectations. Our analysis reveals that increased operational costs mediate the relationship between news spotlight and financial performance. Essentially, the media spotlight on a firm's climate actions tends to result in higher operational costs, which in turn negatively affect financial performance. Such costs may arise from heightened visibility from stakeholders, regulatory pressures, reactive investments in sustainable technologies, and public relations efforts aimed at maintaining a green corporate image.

We also analyzed the heterogeneous impacts of firms’ environmental performance and industry “dirtiness” on the studied relationship. Our findings show that for firms with high pollution levels and those from “dirty” industries, the financial effect of the spotlight is “flipped” compared to low-polluting firms, indicating a significant positive impact. High-polluting firms, with a baseline of poor environmental reputation, have greater potential for improvement, allowing them to capitalize more cost-effectively on operational changes driven by media scrutiny. These changes can lead to significant savings in reputational and regulatory costs, driven by intense stakeholder pressure on these firms. In contrast, firms with already low pollution levels may face relatively higher costs for limited environmental gains, given their already low emission levels. We also examined the interaction between firms’ climate action spotlight and their market orientation, durability classification, index constituency, and greenwashing status. We found these factors to be significant moderators of the relationship.

In conclusion, our study uncovers the financial implications of the media spotlight on firms’ climate actions. We emphasize the importance of firms managing the spotlight, considering its potential to impact their financial performance by incurring operational costs. Our findings revealed disparate financial impacts of the climate action spotlight across firms, based on their pollution levels and the environmental impact of their industries. Cleaner firms and those from low-polluting industries should be cautious about attracting such media attention, considering its associated costs. In contrast, high-polluting firms and those from traditionally polluting industries may derive financial benefits from the spotlight. While firms cannot control the spotlight they receive, they can strategically influence it through public relations efforts and agenda-setting strategies. For high-polluting firms, starting from a low baseline of environmental reputation, media attention can serve as a driving force for change, yielding both financial gains and sustainable growth. One size does not fit all; firms should tailor their strategy to attract or avoid the spotlight based on their environmental performance and needs. Our study also highlights the societal benefits of directing the media spotlight on high-polluting rather than low-polluting firms. Low-polluting firms, with limited capacity for environmental improvements, may seek media attention due to external pressures or for virtue signaling, public relations purposes, or perceived benefits to the brand. Such pursuits, which offer no financial benefits to the firms, as shown by our results, can clutter the media with superficial claims and divert society's focus from the critical sustainability issues facing high-polluting firms. The media's role is pivotal in informing the public about firms’ actions and capturing their attention; thus, undue focus on low-polluting firms can detract from the attention needed for significant climate actions in more impactful sectors. From a societal perspective, our findings advocate for a judicious media strategy that prioritizes coverage of firms with the greatest potential for environmental strides, steering societal attention to where it matters the most for sustainability.

As a limitation, we note that our findings primarily address the financial performance of firms. However, media spotlight on climate actions may also yield intangible, long-term benefits that are not captured by these financial measures. Future research may further explore the complex dynamics between the media spotlight and financial performance in light of evolving global sustainability priorities to better understand managerial and operational responses to media scrutiny. Another research direction could be to consider the climate action-related spotlight in social media and its effect on firm performance. We believe that our current study should prompt investigations along these lines to further our understanding of the role of media in the context of firms’ climate actions.

Supplemental Material

sj-pdf-1-pao-10.1177_10591478251387803 - Supplemental material for No News About Climate Action is Good News for Low-Polluting Firms

Supplemental material, sj-pdf-1-pao-10.1177_10591478251387803 for No News About Climate Action is Good News for Low-Polluting Firms by Nima Safaei and Gautam Pant in Production and Operations Management