How Does Corporate Information Environment Influence CSR?

Abstract

This study investigates the impact of outsiders’ demand for more information (or transparency) on corporate social responsibility (CSR) initiatives. Drawing on a dataset of U.S. companies from 2010 to 2023, CSR performance is measured using ASSET4 ratings, while CSR disclosure levels are captured through the number of words and sentences in reports. Utilizing within-industry and -firm OLS regressions, our analyses reveal a positive relationship between the demand for more information and future CSR investments, showing that firms with higher demand for information not only enhance their CSR performance but also expand the length of their CSR reports. These results suggest that increased pressures for information encourage organizations to engage more deeply with social responsibility, resulting in more robust CSR activities and more comprehensive reporting practices. This study contributes to the existing literature by highlighting the strong predictive role of outsiders’ demand for more information in promoting CSR investment and disclosure, and by offering important insights for policymakers and practitioners on fostering corporate responsibility through enhanced transparency.

1. Introduction

Investor decisions in capital markets heavily rely on transparency. Its opposite, opacity, signifies an information gap. This occurs when a select group possesses confidential company knowledge unavailable to others limited to public market data (L. Lei & Luo, 2023; Thakor & Merton, 2023). Additionally, opacity arises when certain market participants have superior information about trading conditions or possess distinct skills for processing it (Cortes, 2021; Ertan et al., 2025). This situation fosters agency problems, increasing associated costs for the firm (Jensen & Meckling, 1976). Such problems emerge when managers (agents) and investors (principals) have conflicting goals, making oversight difficult. Managers often exploit their private information advantage for personal gain (Healy et al., 1999), leading to information asymmetry. This asymmetry creates a “lemon” problem due to misaligned incentives between principals and agents (Akerlof, 1970).

Mitigating information asymmetry and lowering agency costs necessitates a channel requiring managerial effort. Prior research proposes that managers can utilize corporate social responsibility (CSR) initiatives to enhance transparency. Analyzing U.S. firms from 1991 to 2010, Cui et al. (2018) demonstrate an inverse relationship between CSR engagement and information asymmetry. Their findings further indicate this negative link is stronger for high-risk companies, implying CSR is employed to reduce investor adverse selection. Supporting the conclusion that CSR performance diminishes information asymmetry, studies by Siew et al. (2016), Choudhary et al. (2019), Naqvi et al. (2021), and Hamrouni et al. (2022) yield consistent results. Although substantial research explores how CSR affects transparency, the reverse impact—how information asymmetry influences CSR performance—has received less attention.

Reviewing the literature, we found opposing implications on the influence of transparency on CSR performance. One stream suggests that transparency might decrease CSR investment. According to the literature, investing in CSR yields long-term benefits (e.g., Shirasu & Kawakita, 2021; Hamrouni et al., 2022; Garel & Petit-Romec, 2021; Ma & Yasir, 2023; Tsai & Wu, 2022). However, the challenge lies in evaluating the true value and cost of CSR efforts, which can deter managers from making such investments. Prior literature emphasizes the long-term benefits of responsible behavior, contrasting them with short-term costs (Adhikari, 2016; Qian et al., 2019; Z. Lei et al., 2022). They document that pressure to meet short-term performance expectations may lead managers to prioritize immediate gains over CSR improvement, especially when under scrutiny from financial analysts. Managers, thus, may feel compelled to limit CSR investment to satisfy short-term demands, particularly when facing pressure from a high number of financial analysts. Conversely, a reduction in analyst coverage may grant managers more leeway to focus on long-term CSR enhancement (Adhikari, 2016; Qian et al., 2019; Z. Lei et al., 2022). This suggests a negative influence of transparency on CSR investments. On the other hand, increased transparency may not necessarily limit a company’s investment in socially responsible activities; instead, it could potentially boost the company’s CSR investments. Research indicates a shift among analysts toward more favorable views of companies demonstrating strong CSR performance (Ioannou & Serafeim, 2015; Luo et al., 2015). Supporting this, Gao et al. (2016) find that firms providing high-quality CSR disclosure attract a larger analyst following, particularly when their actual CSR performance is also high. This evolving trend indicates that heightened analyst attention might encourage firms to prioritize socially responsible initiatives, recognizing the positive impact on their reputation and performance. That is, by aligning with societal expectations and demonstrating commitment to responsible practices, companies may attract more favorable assessments from analysts, leading to increased support and potential financial benefits. Thus, more transparency might increase CSR investments. These conflicting perspectives leave a critical empirical question unresolved: Does external transparency enhance or constrain CSR investments?

The primary objective of this research is to resolve the existing ambiguity by exploring the impact of market scrutiny—comprising analysts and investors—on CSR performance and reporting. While earlier studies predominantly concentrate on financial analysts (Adhikari, 2016; Qian et al., 2019; Z. Lei et al., 2022; Tsang et al., 2024), we extend the analysis to incorporate the overall pressure from external stakeholders, as theorized by Anderson et al. (2009).

Employing a dataset consisting of U.S. companies spanning from 2010 to 2023, our study delves into the effect of external transparency on CSR performance and disclosure. We gauge CSR performance using ASSET4 ratings and quantify disclosure through word and sentence counts, consistent with Nazari et al. (2017) and Mahmoudian et al. (2025). Our results indicate a positive correlation between external transparency and forthcoming CSR investments in social and environmental domains. Additionally, we observe that this influence prompts an expansion in the length of CSR reports.

Our research significantly contributes to the existing literature in multiple dimensions. Firstly, we enhance the understanding of CSR determinants by demonstrating that external transparency serves as a robust predictor of CSR investment. Furthermore, we extend the discourse on transparency’s impact on CSR (García-Sánchez et al., 2021; Qian et al., 2019) by revealing its influence on the volume of CSR reports, prompting managers to produce more comprehensive reports. Importantly, our study also elucidates the policy and practical implications, highlighting how heightened external pressures for transparency drive organizations towards greater social responsibility.

The subsequent sections of this study are organized as follows. Section 2 establishes the motivation and explores the rationale for the connection between external transparency and forthcoming CSR investments. Section 3 details the research design and sample characteristics. Section 4 presents the empirical findings, and Section 5 concludes by discussing the implications derived from these results.

2. Motivation and Hypotheses

Recently, there has been a global rise in CSR initiatives, focusing more on considering the interests of various stakeholders rather than solely on generating profits for shareholders. This shift underscores the importance of having reliable information about a company’s CSR practices when making investment decisions. Early scholars like Bowen (1953) defined CSR as businesses’ obligations to align their policies and decisions with societal objectives and values. CSR entails a company’s response to societal issues beyond just economic, technical, and legal concerns (Davis, 1973). It acknowledges that firms have responsibilities to a wider range of stakeholders beyond shareholders. This perspective, rooted in stakeholder theory (Freeman, 1984), suggests that companies performing well in the long term consider the needs of various stakeholders. Consequently, CSR practices facilitate meaningful dialogue between companies and stakeholders, enhancing stakeholder support and ensuring long-term viability (Branco & Rodrigues, 2008; Gray et al., 1995). However, existing research often oversimplifies CSR’s relationship with shareholder value, lacking clarity on when they intersect (Awaysheh et al., 2020). Moreover, studies fail to distinguish between different types of CSR actions among firms (Hawn & Ioannou, 2016), potentially leading to misinterpretation of their behaviors under conflicting stakeholder and investor pressures (Cho et al., 2015; Luo et al., 2015). In this study, we fill the gap by exploring how companies invest in CSR activities when they are under pressure of market scrutiny.

Through an extensive literature review, it becomes apparent that the visibility demanded by external stakeholders plays a pivotal role in cultivating trust among creditors and investors (Thakor & Merton, 2023; Liao & Zhang, 2024; Sun et al., 2024). Studies indicate that companies exhibit a greater propensity to prioritize their CSR initiatives when subjected to public and societal scrutiny (E. H. Kim & Lyon, 2015). This external scrutiny heightens the risk of exposing greenwashing practices, thereby eroding public trust. Consequently, public oversight serves as a deterrent against greenwashing behaviors (Berrone et al., 2017), guiding managers away from self-serving motives and fostering authentic CSR endeavors. Pressure for transparency has been identified as crucial in ensuring the accuracy of CSR data, which can significantly impact the precision of financial forecasts. High-quality CSR data help financial analysts assess the reliability of company reports, reducing forecasting errors. A larger analyst following increases the scrutiny placed on firms, helping to align disclosed data with actual performance and preventing misleading CSR disclosures from distorting financial predictions (García-Sánchez et al., 2021). This enhanced scrutiny not only promotes accurate CSR reporting but also supports the long-term financial benefits of ethical business practices. In alignment with this view, Marquis et al. (2016) demonstrate that organizations with greater visibility are more likely to comply with institutional requirements, thus minimizing selective disclosure. This tendency arises from the heightened pressure exerted by external stakeholders (King, 2008; Bansal & Roth, 2000). Gong et al. (2021) also emphasize the role of transparency in environments with weak institutional frameworks, where creditors face difficulties in assessing a company’s genuine CSR commitment. Transparent CSR activities are therefore perceived as more trustworthy, underscoring the importance of transparency in fostering stakeholder trust.

Therefore, external pressures for transparency can serve as a potent alternative to regulation, diminishing managers’ decision-making autonomy. Consequently, organizations with heightened external visibility and transparency face increased public scrutiny, leading to a greater emphasis on stakeholder interests. This fosters enhanced CSR investments, improving both CSR performance and disclosure.

Having said that, Qian et al. (2019) suggest that in the absence of significant transparency pressures, firms may resort to adopting aggressive disclosure strategies, leading to extensive disclosure of CSR-related information. Qian et al. (2019) build upon the concept that investing in CSR often yields long-term benefits but may not result in immediate profits, as activities like environmental protection or corporate philanthropy typically have delayed financial returns (Aguinis & Glavas, 2012; Margolis & Walsh, 2003; Orlitzky et al., 2017). These investments involve uncertainty and require time to translate into financial gains (Wang & Choi, 2013), with studies indicating that environmentally proactive firms may initially experience negative financial performance (Cordeiro & Sarkis, 1997; Gull et al., 2022). The challenge of assessing the value of such investments can deter managers from prioritizing them (Berchicci & King, 2007; Z. Liu et al., 2021). While the benefits of responsible behavior, such as risk reduction, are long-term, costs are immediate (Fieseler, 2011). Qian et al. (2019) then hypothesize that short-term performance pressure, often driven by the expectations of investors and analysts, may constrain managers from investing in CSP, a pressure that diminishes when fewer analysts cover the firm. Thus, this stream suggests that external transparency lowers CSR investment.

The collective impact of external transparency on CSR investments poses an empirical inquiry. Consequently, we formulate the following hypothesis to elucidate its implications:

H1: 

External transparency increases CSR investment.

3. Research Design

3.1. Sample and Data

The sample in this research includes all publicly-listed companies in the U.S. that were part of the ASSET4 database from 2009 to 2023, following the 2008–2009 global financial crisis. This post-crisis period was deliberately chosen due to the heightened pressure on U.S. companies during the crisis, which may have influenced their CSR behaviors and disclosures as they sought attention from critical stakeholders (Lins et al., 2017; Poursoleyman et al., 2024a). Financial data for these companies were retrieved from the Worldscope database, while CSR disclosure was measured by identifying firms that released standalone sustainability reports during the respective fiscal years. We first checked an indicator on the ASSET4 database that specifies whether a company issued such a report. From there, we visited each company’s website to download PDF versions of the reports they had published. To analyze the reports, we sought to calculate basic metrics like the number of words and sentences in each one. For this, we used the online readability analysis tool at Readable.com. The website provides descriptive reports on uploaded PDF files that extract details like word and sentence counts. After uploading the sustainability reports, we collected from company websites; some reports could not be properly analyzed by the Readable tool. Those reports had to be excluded from our sample. In total, after gathering financial data, identifying sustainability reports, downloading the reports, and analyzing them using Readable.com’s text analysis features, we were left with a data set of 2613 sustainability reports covering 7426 firm-year observations from publicly-listed U.S. companies in the post-financial crisis period. This formed the basis for our analyses.

In summary, our sample is representative of larger, publicly listed U.S. companies with active CSR disclosure practices, as these firms are systematically tracked by ASSET4 and more likely to produce machine-readable reports. However, smaller firms, those without sustainability reports, or companies with non-standard report formats are underrepresented. Despite this limitation, our sample aligns with studies focusing on CSR disclosure trends among prominent market participants, which are critical for understanding investor- and stakeholder-facing behaviors in the post-crisis era.

3.2. Model

To explore the influence of external transparency on CSR investments, we follow Lys et al.’s (2015) model which captures the determinants of CSR expenditures:

$$\begin{array}{l}{\mathrm{C}\mathrm{S}\mathrm{R}\mathrm{P}\mathrm{E}\mathrm{R}\mathrm{F} \mathrm{o}\mathrm{r} \mathrm{C}\mathrm{S}\mathrm{R}\mathrm{D}\mathrm{I}\mathrm{S}}_{\mathrm{i},\mathrm{t}+1}\\ ={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{E}\mathrm{X}\mathrm{T}\mathrm{R}\mathrm{A}\mathrm{N}\mathrm{S}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\beta }}_2{\mathrm{A}\mathrm{D}\mathrm{V}\mathrm{E}\mathrm{R}\mathrm{T}\mathrm{I}\mathrm{S}\mathrm{I}\mathrm{N}\mathrm{G}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\beta }}_3{\mathrm{S}\mathrm{A}\mathrm{L}\mathrm{E}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\beta }}_4{\mathrm{C}\mathrm{A}\mathrm{S}\mathrm{H}}_{\mathrm{i},\mathrm{t}}\\ +{\mathsf{\beta }}_5{\mathrm{C}\mathrm{F}\mathrm{O}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\beta }}_6{\mathrm{C}\mathrm{O}\mathrm{R}\mathrm{P}\mathrm{G}\mathrm{O}\mathrm{V}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\beta }}_7{\mathrm{L}\mathrm{E}\mathrm{V}\mathrm{E}\mathrm{R}\mathrm{A}\mathrm{G}\mathrm{E}\mathrm{G}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\beta }}_8{\mathrm{L}\mathrm{I}\mathrm{T}\mathrm{I}\mathrm{G}\mathrm{A}\mathrm{T}\mathrm{I}\mathrm{O}\mathrm{N}}_{\mathrm{i},\mathrm{t}}\\ +{\mathsf{\beta }}_9{\mathrm{M}\mathrm{T}\mathrm{B}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\beta }}_{10}{\mathrm{R}\mathrm{O}\mathrm{A}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\beta }}_{11}{\mathrm{R}\mathrm{N}\mathrm{D}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\beta }}_{12}{\mathrm{S}\mathrm{I}\mathrm{Z}\mathrm{E}}_{\mathrm{i},\mathrm{t}}+{\displaystyle \sum _{\mathrm{j}}^{\mathrm{a}}}{\mathsf{\beta }}_{\mathrm{j}}{\mathrm{I}\mathrm{N}\mathrm{D}\mathrm{U}\mathrm{S}\mathrm{T}\mathrm{R}\mathrm{Y}\mathrm{F}\mathrm{E}}_{\mathrm{i}}\\ +{\displaystyle \sum _{\mathrm{j}+\mathrm{a}+1}^{\mathrm{b}}}{\mathsf{\beta }}_{\mathrm{j}+\mathrm{a}+1}{\mathrm{Y}\mathrm{E}\mathrm{A}\mathrm{R}\mathrm{F}\mathrm{E}}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}+1}\end{array}$$

(1)

where CSRPERF represents the company’s CSR performance which is an arithmetic average of social and environmental pillars. CSRDIS denotes the volume of CSR disclosure. We use two proxies including CSRDIS_WORDS and CSRDIS_SENTS. These three variables are our main dependent variables. The intendent variable of interest is EXTRANS, which shows external transparency. A positive coefficient on EXTRANS supports our main hypothesis (H1). Other variables are our control variables. We provide a short summary of the variables in Appendix A.

3.3. Variable Measurement

3.3.1. CSR Investment

Our approach to measuring CSR investment encompasses both performance and disclosure aspects. To evaluate CSR performance, we use the ASSET4 database, which applies a four-step rating system: (1) collecting comprehensive raw data points, (2) aggregating them into specific indicators, (3) compiling these indicators into thematic categories (e.g., 18 categories in 2014), and (4) grouping these categories into broad pillars. Before 2017, ASSET4 included four pillars: environmental, social, corporate governance, and economic. However, after the removal of the economic pillar in 2017, the framework was revised to focus on three pillars: environmental, social, and corporate governance, with the number of categories reduced to ten. Since corporate governance is closely related to transparency, incorporating the governance pillar could obscure the true effect of transparency on CSR performance. Therefore, our main variable (CSRPERF) is the average of the social and environmental pillar scores, while the individual pillars are used as alternative metrics.

In line with the methodological precedents set by prior research (e.g., Nazari et al., 2017; Mahmoudian et al., 2025), the present study measures the volume of CSR disclosure through two primary quantitative indicators: the count of words and the count of sentences contained in corporate social responsibility reports. This choice is methodologically sound; word count has been a longstanding metric for disclosure volume (Chiang et al., 2019; Neu et al., 1998), and sentence count similarly serves this purpose effectively (Clarkson et al., 2020; Hackston & Milne, 1996). The core strength of utilizing the absolute number of words and sentences lies in their capacity to serve as unambiguous, standardized, and readily calculable proxies for the total informational output in CSR communications. It is noteworthy that accounting scholars frequently leverage these exact counts to derive measures of informational volume in diverse disclosure contexts (e.g., Li, 2008; Callen et al., 2013). For our investigation, focusing on the disclosure dimension of CSR investment, we specifically operationalize this concept by employing the natural logarithm of the total word count (variable: CSRDIS_WORDS) and the natural logarithm of the total sentence count (variable: CSRDIS_SENTS), a transformation commonly applied to address skewness and normalize distributions in such count data.

3.3.2. External Transparency

Transparency reflects either insiders’ attempts to control information or insufficient attention from market participants or intermediaries. Consequently, scholarly proxies capture both internal and external dimensions (Anderson et al., 2009). This means proxies based on external factors are still influenced by internal transparency. We therefore employ Anderson et al.’s (2009) framework to differentiate between measures of external and internal transparency. This involves regressing total transparency measures against internal ones and utilizing the resulting residuals to isolate external factors from the analysis. That is, to derive the external component, we conduct regression analysis using the average index of transparency and disclose quality or internal opacity measures and the resulting residuals ($\widehat{\epsilon }$) represent external transparency (EXTRANS) (Anderson et al., 2009):

$${TRANS}_{i,t}={\beta }_0+{\beta }_1{GOVERNANCE}_{i,t}+{\beta }_2{ACCRUALS}_{i,t}+{\beta }_3{∆EPS}_{i,t}+{\epsilon }_{i,t}$$

(2)

where TRANS represents the firm’s comprehensive transparency measure. Following Anderson et al. (2009), we construct this measure using a framework distinguishing variables based on market trades and analyst coverage. First, we utilize market-based indicators: trading volume (VOLUME) and bid-ask spread (SPREAD). VOLUME, calculated as the natural logarithm of the average daily dollar trading volume during the fiscal year, acts as an indicator of investor disagreement and uncertainty (Barth et al., 2020). SPREAD, reflecting information asymmetry, represents the compensation required by market makers for trading against better-informed investors. A smaller SPREAD indicates reduced adverse selection problems (O. Kim & Verrecchia, 1994), computed as the annual average of the daily (ask price − bid price)/[(ask price + bid price)/2]. Second, we employ analyst-based proxies: forecast errors (ERROR) and coverage (COVERAGE). ERROR measures the absolute difference between mean analyst earnings forecasts and actual earnings, scaled by stock price, indicating market prediction accuracy. COVERAGE, the natural logarithm of one plus the number of analysts following the firm, captures analyst scrutiny intensity. We rank all proxies into country-year deciles relative to the sample. SPREAD and ERROR are inverse transparency indicators (higher values = lower transparency), ranked from 1 (most transparent) to 10 (least transparent). VOLUME and COVERAGE are direct indicators (higher values = higher transparency), ranked from 1 (least transparent) to 10 (most transparent). For consistency, we rescale SPREAD and ERROR deciles to align directionally with transparency (1 = least transparent, 10 = most transparent). TRANS is then the average of these four standardized decile rankings.

Our measurement of disclosure quality includes three proxies: corporate governance (GOVERNANCE), earnings quality (ACCRUALS), and the year-on-year change in earnings per share (∆EPS). The GOVERNANCE proxy is derived from the ASSET4 governance score. Earnings quality is measured through the accruals model developed by Dechow and Dichev (2002) and refined by Francis et al. (2005). This involves performing a regression of total current accruals (CACCRUALS) against five independent variables. CACCRUALS is calculated as CACCRUALS = ΔCA − ΔCL − ΔCASH + ΔDEBT, where ΔCA refers to the change in current assets (scaled by total assets), ΔCL refers to the change in current liabilities (scaled by total assets), ΔCASH represents the change in cash and cash equivalents (scaled by total assets), and ΔDEBT represents the change in current liabilities due to debt (scaled by total assets). The regression model is then estimated as follows:

$${CACCRUALS}_{i,t}={\beta }_0+{\beta }_1{CFO}_{i,t-1}+{\beta }_2{CFO}_{i,t}+{\beta }_3{CFO}_{i,t+1}+{\beta }_4{∆SALE}_{i,t}+{\beta }_5{PPE}_{i,t}+{\epsilon }_{i,t}$$

(3)

where CFO signifies cash flow from operations calculated indirectly. It is obtained by taking net income before extraordinary items and subtracting total accruals, where total accruals equal total current accruals minus depreciation and amortization. CFO, ∆SALE (the annual change in sales), and PPE (the gross value of property, plant, and equipment) are all normalized by total assets. The model is estimated cross-sectionally for each country-year combination. Accrual quality is proxied by the standard deviation of a firm’s residuals from this model over a five-year window, with larger standard deviations signifying lower accrual quality. The inclusion of ∆EPS, measured as the annual rate of change in earnings per share, reflects the rationale that management’s disclosure behavior is influenced by EPS fluctuations and the motivation to prevent negative EPS results.

3.3.3. Control Variables

To maximize the explanatory power of our firm and industry factors regarding CSR expenditures, we include several factors outlined in the existing literature that significantly determine CSR expenditures. These variables also influence information asymmetry and transparency (e.g., Poursoleyman et al., 2024b). In our analysis, we incorporate advertising (ADVERTISING) and research and development (RND) expenditures. This decision stems from the observation that companies allocating substantial resources to advertising and RND tend to engage more extensively in CSR-related initiatives. By integrating these factors into our study, we aim to capture the nuanced relationship between corporate investment patterns and CSR endeavors (McWilliams & Siegel, 2001). We incorporate litigation expenses (LITIGATION) into CSR expenditures, recognizing their role as a form of reputation insurance. By allocating resources to mitigate legal risks, the company proactively safeguards its reputation, demonstrating the commitment to responsible corporate behavior and enhancing stakeholder trust (Peloza, 2006). In our analysis, we incorporate the natural logarithm of total assets (SIZE) as a proxy for firm size. This choice is based on the premise that larger firms tend to possess greater resources, which can facilitate increased CSR expenditures. Consequently, these firms may face heightened expectations and pressures to participate in CSR-related initiatives due to their perceived capacity to effect meaningful change and their potential impact on societal well-being (Teoh et al., 1999; Whited & Wu, 2006). Incorporating a metric for the firm’s comprehensive corporate governance score (CORPGOV) is imperative due to indications suggesting a correlation between corporate governance practices and the breadth and efficacy of CSR expenditures. This addition allows for a more nuanced analysis, acknowledging the intricate interplay between governance structures and CSR initiatives (Johnson et al., 2020). In our analysis, we incorporate metrics such as book leverage (LEVERAGE) and market-to-book ratio (MTB), as empirical evidence suggests that stable firms with lower risk profiles tend to exhibit a greater propensity to engage in CSR expenditures (Cochran & Wood, 1984). By integrating these indicators into our evaluation, we aim to capture the nuanced relationship between financial stability, risk management, and CSR investment decisions, thereby enriching our understanding of the factors influencing CSR expenditure patterns within firms. Next, we incorporate metrics such as cash reserves (CASH), operating cash flow (CFO), total revenues (SALE), and return on assets (ROA) as indicators of firm performance. Campbell (2007) and Preston and O’Bannon (1997) posit that these financial metrics may stimulate or elicit external expectations for CSR expenditures. This is predicated on the notion that firms demonstrating robust financial performance may be perceived as having greater capacity and obligation to engage in socially responsible activities. By including these variables in our model, we aim to explore the potential influence of firm performance on the demand for CSR expenditures, thereby contributing to the ongoing discourse on the drivers of corporate social responsibility. Finally, we incorporate industry fixed effects to account for the diverse range of environmental impacts, growth potentials, disclosure mandates, and regulatory frameworks across various sectors. These factors are anticipated to exert influence on the magnitude of CSR expenditures. By including industry fixed effects, we aim to control for industry-specific characteristics that may confound the relationship between firm-level factors and CSR expenditures. This approach enhances the robustness of our analysis by mitigating potential biases arising from industry disparities and enables a more accurate assessment of the impact of firm-specific determinants on CSR spending (Karpoff et al., 2005; Naughton et al., 2019).

4. Results

4.1. Descriptive Statistics

Presented in

Table 1. Descriptive statistics. The table displays descriptive statistics for the study’s variables. EXTRANS measures external transparency, constructed following Anderson et al. (2009). CSRPERF is the average corporate social responsibility performance, derived from ASSET4’s social and environmental pillar scores. CSRPERF_SOC (ASSET4) reflects social performance, while CSRPERF_ENV (ASSET4) indicates environmental performance. Disclosure extent is captured by CSRDIS_WORDS (ln[total words in CSR report]) and CSRDIS_SENTS (ln[total sentences in CSR report]). ADVERTISING is advertising expense divided by net sales. SALE represents total revenue scaled by total assets. CASH is cash and equivalents divided by total assets. CFO denotes operating cash flow scaled by total assets. CORPGOV is the governance index sourced from ASSET4. LEVERAGE is total debt divided by total assets. LITIGATION is litigation expense divided by net sales. MTB (market-to-book) signifies growth opportunities. ROA measures return on assets. RND is R&D expenditure scaled by total assets. SIZE, firm size, is ln(total assets).

	N	Mean	Std. Dev.	1st Perc.	p25	Median	p75	99th Perc.
EXTRANS	7426	0	0.089	−0.239	−0.056	0.002	0.062	0.174
CSRPERF	7426	0.412	0.236	0.05	0.201	0.379	0.608	0.896
CSRPERF_SOC	7426	0.48	0.219	0.092	0.303	0.461	0.651	0.937
CSRPERF_ENV	7426	0.344	0.283	0	0.067	0.31	0.584	0.914
CSRDIS_WORDS	2613	9.526	0.834	7.107	9.073	9.625	10.105	11.083
CSRDIS_SENTS	2613	7.288	0.849	4.913	6.786	7.4	7.889	8.888
ADVERTISING	7426	0.283	0.639	0.015	0.118	0.225	0.349	0.862
SALE	7426	0.928	0.636	0.111	0.48	0.757	1.205	3.168
CASH	7426	0.149	0.156	0.001	0.033	0.094	0.21	0.668
CFO	7426	0.11	0.077	−0.097	0.066	0.101	0.148	0.342
CORPGOV	7426	0.523	0.219	0.052	0.357	0.539	0.7	0.918
LEVERAGE	7426	0.266	0.199	0	0.11	0.254	0.385	0.877
LITIGATION	7426	0.308	1.373	−0.048	0	0	0	8.244
MTB	7426	4.184	5.735	−9.72	1.7	2.81	4.97	29.43
ROA	7426	0.058	0.087	−0.23	0.024	0.054	0.094	0.295
RND	7426	0.049	0.248	0	0	0	0.041	0.366
SIZE	7426	15.131	1.676	11.432	13.956	14.996	16.236	19.148

are detailed descriptive statistics for all continuous variables incorporated in our empirical analysis. Consistent with methodological best practices for financial data, each continuous variable underwent winsorization at both the 1st and 99th percentiles to minimize distortion from extreme outliers. The variable EXTRANS, derived from residual values calculated per Anderson et al.’s (2009) methodological framework, logically demonstrates a mean value of zero—an expected outcome given its computational basis in regression residuals.

Analysis of CSR performance metrics reveals meaningful patterns. The composite CSRPERF variable shows a mean score of 0.412 (SD = 0.236), indicating moderately positive CSR engagement overall across the sample firms, with the standard deviation reflecting substantial cross-sectional variation. When disaggregating dimensions, notable disparities emerge: The social performance metric (CSRPERF_SOC) displays relatively normal distribution characteristics (mean = 0.480, SD = 0.219), while environmental performance (CSRPERF_ENV) exhibits both lower average performance (mean = 0.344, SD = 0.283) and greater distributional skewness. This divergence is further evidenced in quartile analysis, where the first quartile for environmental performance registers at just 0.067 versus 0.303 for social performance. These differentials collectively suggest U.S. firms systematically achieve stronger social responsibility outcomes than environmental stewardship outcomes.

Corporate governance effectiveness (CORPGOV) presents a mean score of 0.523 (SD = 0.219), with the first quartile at 0.357. This distribution profile indicates generally robust governance structures among sampled firms, aligning with established literature documenting governance standards in U.S. markets.

Turning to disclosure metrics, the transformed word count variable (CSRDIS_WORDS, natural log of CSR word volume) shows a mean of 9.526. When reverse-transformed, this corresponds to an average original document length of approximately 13,711 words. Parallel sentence-level analysis indicates an average of 1463 sentences per report. Within our full sample of 7426 firm-years, 2613 observations (35%) contained formal CSR disclosures suitable for textual analysis. The interquartile range for CSRDIS_WORDS (25th percentile = 9.073; 75th percentile = 10.105) demonstrates concentrated central tendency while accommodating substantial variation in report length, including significantly longer documents at upper distribution tails.

Collectively, these distributional characteristics—including central tendencies, dispersion metrics, and comparative performance patterns—conform to established expectations within the CSR literature and provide a statistically robust foundation for subsequent multivariate analysis.

Table 2. Correlation matrix. The correlation matrix of the variables is presented in the table, with their descriptions provided in Table 1. Significance levels are marked as follows: *** for p < 0.01 and ** for p < 0.05.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
(1) EXTRANS	1.000
(2) CSRPERF	0.158 ***	1.000
(3) CSRPERF_SOC	0.156 ***	0.920 ***	1.000
(4) CSRPERF_ENV	0.142 ***	0.953 ***	0.758 ***	1.000
(5) CSRDIS_WORDS	0.045 **	0.430 ***	0.410 ***	0.380 ***	1.000
(6) CSRDIS_SENTS	0.045 **	0.444 ***	0.421 ***	0.393 ***	0.962 ***	1.000
(7) ADVERTISING	−0.042 ***	−0.054 ***	−0.016	−0.077 ***	−0.063 ***	−0.069 ***	1.000
(8) SALE	0.044 ***	−0.094 ***	−0.076 ***	−0.098 ***	−0.102 ***	−0.101 ***	−0.115 ***	1.000
(9) CASH	0.052 ***	−0.094 ***	−0.037 ***	−0.128 ***	−0.042 **	−0.031	0.224 ***	−0.024 **	1.000
(10) CFO	0.244 ***	0.057 ***	0.073 ***	0.038 ***	0.010	0.011	−0.143 ***	0.249 ***	0.214 ***
(11) CORPGOV	0.011	0.422 ***	0.376 ***	0.412 ***	0.187 ***	0.192 ***	−0.063 ***	−0.004	−0.114 ***
(12) LEVERAGE	−0.057 ***	0.112 ***	0.079 ***	0.125 ***	0.050 **	0.048 **	−0.084 ***	−0.203 ***	−0.351 ***
(13) LITIGATION	−0.035 ***	0.057 ***	0.064 ***	0.045 ***	0.059 ***	0.056 ***	0.010	−0.077 ***	0.029 **
(14) MTB	0.159 ***	0.082 ***	0.101 ***	0.058 ***	−0.025	−0.025	0.037 ***	0.048 ***	0.178 ***
(15) ROA	0.265 ***	0.091 ***	0.101 ***	0.074 ***	−0.014	−0.007	−0.164 ***	0.231 ***	0.182 ***
(16) RND	−0.055 ***	−0.015	0.023 **	−0.043 ***	0.003	0.007	0.801 ***	−0.130 ***	0.292 ***
(17) SIZE	0.145 ***	0.632 ***	0.545 ***	0.631 ***	0.346 ***	0.347 ***	−0.130 ***	−0.250 ***	−0.245 ***
Variables	(10)	(11)	(12)	(13)	(14)	(15)	(16)		(17)
(10) CFO	1.000
(11) CORPGOV	0.017	1.000
(12) LEVERAGE	−0.124 ***	0.056 ***	1.000
(13) LITIGATION	−0.018	0.024 **	0.003	1.000
(14) MTB	0.210 ***	−0.018	−0.045 ***	0.002	1.000
(15) ROA	0.581 ***	0.043 ***	−0.164 ***	−0.029 **	0.204 ***	1.000
(16) RND	−0.191 ***	−0.054 ***	−0.112 ***	0.014	0.044 ***	−0.164 ***	1.000
(17) SIZE	−0.054 ***	0.302 ***	0.280 ***	0.046 ***	0.000	0.010	−0.099 ***		1.000

details the Pearson correlation coefficients among all variables central to our empirical investigation. A thorough analysis of these relationships reveals particularly instructive patterns regarding external transparency (EXTRANS). This variable demonstrates statistically robust and economically meaningful positive associations with multiple dimensions of corporate social responsibility. Most notably, EXTRANS exhibits a correlation coefficient of 0.158 (p-value < 0.01) with the composite CSR performance measure (CSRPERF), indicating that a single-unit elevation in external transparency corresponds to a 15.8% increase in overall CSR performance. This foundational relationship extends consistently to both constituent domains: social performance (CSRPERF_SOC: r = 0.156, p < 0.01) and environmental performance (CSRPERF_ENV: r = 0.146, p < 0.01). Collectively, these highly significant coefficients provide preliminary multivariate validation for our core hypothesis, suggesting that firms exhibiting greater external transparency subsequently demonstrate superior CSR performance outcomes.

The relationship between transparency and CSR disclosure volume proves equally noteworthy. Both logarithmic disclosure metrics, CSRDIS_WORDS and CSRDIS_SENTS, maintain positive correlations with EXTRANS of approximately 0.045. While more modest in magnitude than performance linkages, this consistent directional relationship implies that enhanced external transparency may precede expansions in the comprehensiveness of CSR reporting. Furthermore, we observe a pronounced interdependence between CSR performance and disclosure practices. The strong positive correlation between CSRPERF and CSRDIS_WORDS (r = 0.430, p < 0.01) substantiates that firms with advanced CSR engagement tend to produce more voluminous reports—a finding aligned with voluntary disclosure theory.

Additional correlation patterns merit attention for their theoretical consistency. Firm size (SIZE) maintains significant positive relationships with both CSR performance dimensions (ranging from r = 0.12 to r = 0.28) and disclosure metrics (r ≈ 0.35), corroborating the well-established literature indicating greater resource availability and stakeholder pressures among large corporations. Control variables including leverage, profitability, and market-to-book ratios exhibit correlation patterns largely congruent with prior studies in governance and sustainability research (e.g., Core et al., 2006; Dhaliwal et al., 2011), confirming our sample’s alignment with established empirical regularities. Crucially, variance inflation factors (VIF) remained below conventional thresholds (max VIF < 3.2), mitigating multicollinearity concerns for subsequent regression modeling.

4.2. Baseline Model

The multivariate regression analysis presented in

Table 3. Regression of CSR performance and disclosure on external transparency: baseline analysis. This table reports the main regression estimates. All explanatory and control variables are defined in Table 1. The reported t-statistics are based on robust standard errors adjusted for firm-level clustering and heteroscedasticity. Statistical significance is denoted using conventional asterisk notation: *** coefficient significant at α = 1% (p < 0.01), ** coefficient significant at α = 5% (p < 0.05), * coefficient significant at α = 10% (p < 0.10).

	(1)	(2)	(3)	(4)	(5)
Dependent Variable=	CSRPERFt+1	CSRPERF_SOCt+1	CSRPERF_ENVt+1	CSRDIS_WORDSt+1	CSRDIS_SENTSt+1
	b/t	b/t	b/t	b/t	b/t
EXTRANS	0.121 ***	0.139 ***	0.103 ***	0.473 **	0.449 **
	(5.16)	(5.95)	(3.63)	(2.17)	(2.00)
ADVERTISING	0.004	0.006	0.001	−0.105	−0.187
	(0.66)	(1.25)	(0.21)	(−0.73)	(−1.26)
SALE	0.020 ***	0.019 ***	0.021 ***	−0.020	−0.037
	(5.11)	(4.57)	(4.56)	(−0.57)	(−1.00)
CASH	0.005	−0.023	0.034 *	−0.008	0.071
	(0.34)	(−1.50)	(1.81)	(−0.06)	(0.49)
CFO	0.233 ***	0.216 ***	0.250 ***	1.005 ***	0.938 ***
	(7.04)	(6.46)	(6.35)	(3.66)	(3.29)
CORPGOV	0.240 ***	0.214 ***	0.266 ***	0.356 ***	0.380 ***
	(25.40)	(22.69)	(23.33)	(4.16)	(4.28)
LEVERAGE	−0.041 ***	−0.027 **	−0.055 ***	−0.146	−0.157
	(−3.85)	(−2.52)	(−4.18)	(−1.50)	(−1.55)
LITIGATION	0.002	0.002	0.003 *	0.011	0.009
	(1.62)	(1.18)	(1.72)	(1.27)	(0.96)
MTB	0.001 ***	0.001 **	0.001 ***	−0.002	−0.002
	(2.89)	(1.99)	(3.13)	(−0.87)	(−0.93)
ROA	−0.022	−0.016	−0.027	−0.247	−0.223
	(−0.78)	(−0.57)	(−0.82)	(−1.07)	(−0.95)
RND	0.032 **	0.025	0.039 **	0.222	0.325
	(2.21)	(1.54)	(2.47)	(0.71)	(0.97)
SIZE	0.090 ***	0.074 ***	0.106 ***	0.216 ***	0.223 ***
	(65.83)	(52.10)	(65.48)	(17.05)	(17.22)
_cons	−1.166 ***	−0.817 ***	−1.514 ***	5.670 ***	3.292 ***
	(−34.09)	(−23.41)	(−35.80)	(14.02)	(7.89)
Year Dummies	Yes	Yes	Yes	Yes	Yes
Industry Dummies	Yes	Yes	Yes	Yes	Yes
N	7426	7426	7426	2613	2613
R2	0.543	0.471	0.524	0.204	0.206
Adj. R2	0.540	0.467	0.521	0.188	0.190
F	235.435	166.367	220.582	13.116	13.253

rigorously examines the complex interrelationships among CSR performance and disclosure practices, and external transparency (EXTRANS). This table systematically reports five distinct model specifications: Models (1) through (3) employ CSR performance metrics as the dependent variables—specifically, the composite CSRPERF score (Column 1), social performance (CSRPERF_SOC, Column 2), and environmental performance (CSRPERF_ENV, Column 3). Columns (4) and (5) shift the analytical focus to CSR disclosure volume, utilizing the natural logarithm of word count (CSRDIS_WORDS) and sentence count (CSRDIS_SENTS) as respective dependent variables. Methodological rigor is ensured through the inclusion of both year and industry fixed effects across all specifications, effectively controlling for temporal trends and sectoral heterogeneity. Furthermore, all reported standard errors incorporate adjustments for potential heteroskedasticity, enhancing the reliability of statistical inferences.

Model (1) yields a statistically and economically significant result: EXTRANS demonstrates a positive coefficient of 0.121 (t-statistic = 5.16, p-value < 0.01). This indicates a robust positive association between current external transparency and subsequent CSR performance. Quantifying the economic magnitude reveals that a one-standard-deviation increase in EXTRANS corresponds to a 1.08 basis point enhancement in future CSRPERF. This foundational relationship persists when examining performance dimensions separately. Model (2) shows EXTRANS significantly predicts social performance (CSRPERF_SOC: β = 0.139, t = 5.95, p < 0.01), while Model (3) confirms its positive association with environmental performance (CSRPERF_ENV: β = 0.103, t = 3.63, p < 0.01). The differential magnitude between social (β = 0.139) and environmental (β = 0.103) coefficients suggests that external transparency exerts a stronger influence on social responsibility outcomes. This observed asymmetry may stem from the inherent resource intensity of environmental initiatives, which often require substantial capital investment (e.g., carbon abatement technologies), potentially making them less immediately responsive to transparency pressures compared to social programs.

The disclosure analysis in Columns (4) and (5) provides compelling corroborative evidence. EXTRANS exhibits significant positive coefficients of 0.473 (t = 2.17, p < 0.05) for CSRDIS_WORDS and 0.449 (t = 2.00, p < 0.05) for CSRDIS_SENTS. This indicates that firms with higher external transparency subsequently produce more extensive CSR reports. The economic translation demonstrates that a one-standard-deviation increase in EXTRANS leads to a 4.2 basis point increase in report length (words) and a 3.99 basis point increase in sentence count. These findings collectively substantiate our core hypothesis that external transparency acts as a catalyst for more comprehensive CSR disclosure practices.

Analysis of control variables reveals theoretically consistent patterns within the CSR performance models (Columns 1–3). In terms of the positive associations, we reveal the following coefficients: SALE (total revenue: β > 0), CFO (cash flow: β > 0), CORPGOV (governance quality: β > 0), MTB (market-to-book ratio: β > 0), RND (R&D intensity: β > 0), and SIZE (firm size: β > 0). These align with our conceptual framework: larger firms (SALE, SIZE) face amplified stakeholder expectations; financially robust firms (CFO, MTB) possess greater resource capacity; innovation-focused firms (RND) exhibit aligned strategic priorities; and strong governance (CORPGOV) enables effective CSR implementation. Additionally, for the negative associations, we find that LEVERAGE (debt ratio: β < 0) shows a significant negative coefficient, consistent with prior literature (e.g., Benlemlih, 2017). This supports the contention that financial constraints impede CSR investment, whereas stability facilitates it.

5. Robustness Checks

5.1. CSR Reporters and Non-Reporters

This section presents robustness tests for our primary findings on the CSR performance-external transparency relationship. Initial analysis of the correlation matrix revealed moderately strong associations between CSR performance and disclosure metrics (0.430 for word count and 0.444 for sentence count). These correlations suggest the possibility that our main results could be influenced by the quantitative aspects of CSR reporting rather than substantive performance. To investigate this concern, we divided our sample into firms that issue CSR reports and those that do not, then separately analyzed the CSR performance-transparency relationship within each subgroup.

Table 4. Regression of CSR performance on external transparency: robust to sub-samples of reporters and non-reporters. This table presents comparative regression analyses examining the relationship between corporate social responsibility performance (CSRPERF) and external transparency (EXTRANS) across strategically defined subsamples. The analytical framework partitions firms into two distinct cohorts: (1) entities publishing formal standalone CSR reports (Columns 1–3) and (2) entities without such dedicated disclosures (Columns 4–6). All variables incorporated within these specifications—including dependent, independent, and control variables—are comprehensively defined in Table 1, ensuring consistent operationalization throughout the study. Reported t-statistics derive from robust standard errors adjusted for heteroscedasticity (non-constant error variance) via the Huber-White estimator. Statistical significance is denoted using conventional asterisk notation: *** coefficient significant at α = 1% (p < 0.01), ** coefficient significant at α = 5% (p < 0.05), * coefficient significant at α = 10% (p < 0.10).

	(1)	(2)	(3)	(4)	(5)	(6)
Sample=	Reporters			Non-Reporters
Dependent Variable=	CSRPERFt+1	CSRPERF_SOCt+1	CSRPERF_ENVt+1	CSRPERFt+1	CSRPERF_SOCt+1	CSRPERF_ENVt+1
	b/t	b/t	b/t	b/t	b/t	b/t
EXTRANS	0.106 **	0.091 **	0.121 **	0.104 ***	0.132 ***	0.077 **
	(2.48)	(2.07)	(2.33)	(3.97)	(4.98)	(2.42)
ADVERTISING	−0.018	0.011	−0.047	0.003	0.004	0.002
	(−0.61)	(0.36)	(−1.32)	(0.84)	(1.20)	(0.39)
SALE	0.024 ***	0.016 *	0.032 ***	0.016 ***	0.017 ***	0.014 ***
	(3.06)	(1.85)	(3.51)	(3.70)	(3.77)	(2.82)
CASH	−0.055 *	−0.060*	−0.049	0.018	−0.016	0.052 **
	(−1.80)	(−1.87)	(−1.35)	(1.07)	(−0.90)	(2.54)
CFO	0.383 ***	0.341 ***	0.425 ***	0.109 ***	0.116 ***	0.102 **
	(4.82)	(4.39)	(4.62)	(3.24)	(3.29)	(2.51)
CORPGOV	0.155 ***	0.144 ***	0.165 ***	0.206 ***	0.186 ***	0.227 ***
	(9.16)	(8.24)	(8.08)	(19.32)	(17.12)	(17.52)
LEVERAGE	−0.029	−0.032	−0.025	−0.024 **	−0.010	−0.037 **
	(−1.52)	(−1.62)	(−1.07)	(−1.97)	(−0.82)	(−2.53)
LITIGATION	0.005 ***	0.007 ***	0.004 **	0.000	−0.001	0.002
	(3.06)	(3.39)	(2.03)	(0.22)	(−0.85)	(1.14)
MTB	−0.000	−0.001	−0.000	0.000	0.000	−0.000
	(−0.79)	(−1.53)	(−0.03)	(0.17)	(0.56)	(−0.15)
ROA	−0.065	−0.055	−0.075	−0.031	−0.019	−0.043
	(−0.89)	(−0.79)	(−0.89)	(−1.05)	(−0.58)	(−1.25)
RND	0.442 ***	0.233 **	0.652 ***	0.016	0.014	0.017
	(4.68)	(2.48)	(5.82)	(1.13)	(0.84)	(1.23)
SIZE	0.076 ***	0.065 ***	0.087 ***	0.073 ***	0.059 ***	0.086 ***
	(29.37)	(24.05)	(28.37)	(39.52)	(30.85)	(38.86)
_cons	−0.669 ***	−0.477 ***	−0.860 ***	−0.884 ***	−0.577 ***	−1.190 ***
	(−6.33)	(−5.73)	(−6.13)	(−23.12)	(−14.04)	(−25.44)
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
Industry FE	Yes	Yes	Yes	Yes	Yes	Yes
N	2613	2613	2613	4813	4813	4813
R2	0.435	0.385	0.393	0.413	0.352	0.409
Adj. R2	0.423	0.373	0.381	0.407	0.345	0.402
F	37.159	30.303	31.363	79.318	65.514	77.502

reports these findings.

The subsample regression analyses presented in Table 4 delineate distinct findings for firms based on their CSR reporting status. Columns (1) through (3) present results exclusively for firms that publish standalone CSR reports, whereas Columns (4) through (6) correspond to firms without formal CSR disclosures. Critically, the relationship between external transparency (EXTRANS) and aggregate CSR performance demonstrates remarkable consistency across reporting status. For reporting firms (Column 1), EXTRANS yields a coefficient of 0.106, statistically significant at the 5% level (t-stat ≈ 2.3). Similarly, among non-reporting firms (Column 4), the coefficient is 0.104, achieving significance at the stricter 1% threshold (t-stat ≈ 2.6). This parity in coefficient magnitude and direction strongly indicates that the fundamental positive association between external transparency and subsequent CSR performance holds irrespective of a firm’s current disclosure practices.

Disaggregating performance into its constituent dimensions, however, reveals nuanced differential effects. Concerning social performance (CSRPERF_SOC), reporting firms exhibit a significant positive coefficient of 0.091 (p < 0.05, Column 2). Contrastingly, non-reporting firms demonstrate a substantially larger coefficient of 0.132 (p < 0.01, Column 5). This divergence suggests that external transparency exerts a comparatively stronger influence on improving social responsibility outcomes within firms that currently abstain from formal CSR reporting. The pattern reverses for environmental performance (CSRPERF_ENV). Reporting firms display a robust coefficient of 0.121, significant at the 2.5% level (p < 0.025, Column 3). Meanwhile, non-reporting firms show a markedly lower coefficient of 0.077, significant only at the 5% level (p < 0.05, Column 6). This asymmetry implies that firms already engaged in CSR disclosure experience a disproportionately greater positive effect of external transparency on their environmental initiatives relative to their non-reporting counterparts. Collectively, these dimensional analyses reveal that while the core transparency-CSR performance link transcends reporting status, the relative sensitivity of social versus environmental dimensions is meaningfully moderated by a firm’s existing commitment to formal disclosure.

In summary, partitioning the sample and analyzing the sub-samples independently helps address potential concerns around the volume of CSR disclosure driving the overall results. Both reporting and non-reporting firms exhibit a significant positive relationship between CSR performance and external transparency. However, the dimension-level analysis provides some nuanced differences, with external transparency more influential on social performance for non-reporters and environmental performance for reporters. Overall, these robustness checks reinforce the conclusions from our main analysis.

5.2. Within-Firm Estimate

While our primary empirical specifications incorporated year and industry fixed effects to mitigate temporal fluctuations and sectoral heterogeneity, these controls cannot fully account for persistent, unobservable firm-specific factors—such as organizational culture, strategic orientation, or managerial philosophy—that may systematically influence the nexus between external transparency, CSR performance, and disclosure practices. To address this inherent limitation and rigorously isolate within-firm dynamics, we implement firm fixed effects regression models. The results of this advanced specification are systematically presented in

Table 5. Regression of CSR performance and disclosure on external transparency: robust to within-firm estimate. This table presents firm fixed effects regression estimates examining the relationship between CSR dimensions (performance and disclosure) and external transparency (EXTRANS). This advanced specification addresses unobserved time-invariant firm heterogeneity by absorbing all cross-sectional differences through entity-specific intercepts, thereby isolating the within-firm relationship between transparency dynamics and CSR outcomes over time. The model structure parallels our main analysis: Columns (1)–(3) feature CSR performance metrics (aggregate, social, environmental) as dependent variables, while Columns (4) and (5) utilize logarithmic transformations of disclosure volume measures. Statistical significance thresholds adhere to conventional econometric standards: *** denotes significance at α = 1% (p < 0.01), ** denotes significance at α = 5% (p < 0.05), * denotes significance at α = 10% (p < 0.10).

	(1)	(2)	(3)	(4)	(5)
Dependent Variable=	CSRPERFt+1	CSRPERF_SOCt+1	CSRPERF_ENVt+1	CSRDIS_WORDSt+1	CSRDIS_SENTSt+1
	b/t	b/t	b/t	b/t	b/t
EXTRANS	0.126 ***	0.119 ***	0.133 ***	0.994 ***	1.023 ***
	(5.23)	(4.92)	(4.32)	(3.67)	(3.80)
ADVERTISING	−0.024	−0.026	−0.021	−0.047	−0.416
	(−1.14)	(−1.22)	(−0.85)	(−0.07)	(−0.62)
SALE	−0.007	0.003	−0.017	−0.049	−0.087
	(−0.56)	(0.24)	(−1.09)	(−0.40)	(−0.71)
CASH	0.083 ***	0.071 **	0.095 ***	0.103	0.235
	(3.17)	(2.44)	(3.18)	(0.33)	(0.71)
CFO	0.080 **	0.074 **	0.086 *	0.924 *	0.631
	(2.20)	(2.04)	(1.93)	(1.82)	(1.24)
CORPGOV	0.166 ***	0.157 ***	0.175 ***	0.495 ***	0.535 ***
	(11.03)	(10.07)	(9.41)	(3.11)	(3.33)
LEVERAGE	0.009	0.033	−0.014	−0.235	−0.271
	(0.36)	(1.21)	(−0.47)	(−0.85)	(−0.98)
LITIGATION	0.000	−0.000	0.001	0.002	0.001
	(0.34)	(−0.00)	(0.58)	(0.23)	(0.09)
MTB	0.001 ***	0.001 **	0.001 **	−0.000	0.001
	(2.59)	(2.31)	(2.27)	(−0.02)	(0.18)
ROA	−0.072 ***	−0.048 *	−0.096 ***	−0.229	−0.317
	(−2.74)	(−1.82)	(−3.03)	(−0.56)	(−0.80)
RND	0.095	0.125	0.065	0.298	0.642
	(1.10)	(1.36)	(0.64)	(0.21)	(0.48)
SIZE	0.131 ***	0.110 ***	0.153 ***	0.625 ***	0.643 ***
	(15.62)	(13.96)	(13.73)	(6.26)	(6.63)
_cons	−1.678 ***	−1.300 ***	−2.055 ***	−0.872	−3.283 **
	(−12.59)	(−10.33)	(−11.68)	(−0.51)	(−1.96)
Year FE	No	No	No	No	No
Industry FE	No	No	No	No	No
Firm FE	Yes	Yes	Yes	Yes	Yes
N	7426	7426	7426	2613	2613
R2	0.261	0.193	0.230	0.096	0.108
Adj. R2	0.259	0.192	0.229	0.092	0.104
F	45.936	37.121	36.011	10.150	12.687

.

Table 5 maintains a parallel structure to our core analysis: Columns (1) through (3) examine CSR performance dimensions (aggregate, social, and environmental), while Columns (4) and (5) focus on disclosure volume metrics (log-transformed word and sentence counts). Crucially, when accounting for firm-level invariant heterogeneity, the relationship between external transparency (EXTRANS) and aggregate CSR performance (CSRPERF, Column 1) remains robustly positive and statistically significant (β = 0.126, p < 0.01), aligning closely with our baseline estimates. This consistency extends to the social (CSRPERF_SOC, Column 2) and environmental (CSRPERF_ENV, Column 3) dimensions, reaffirming the fundamental stability of the transparency–performance linkage.

However, the disclosure analysis reveals a substantively enriched narrative. For CSR report word count (CSRDIS_WORDS, Column 4), EXTRANS exhibits a markedly stronger coefficient of 0.994 (p < 0.01) relative to our main specification. Similarly, sentence count (CSRDIS_SENTS, Column 5) shows an amplified coefficient of 1.023 (p < 0.01). This systematic intensification of effect magnitudes under firm fixed effects is methodologically coherent. Cross-sectional heterogeneity—stemming from deeply embedded differences in reporting norms, resource endowments, or sustainability cultures across firms—likely attenuated the disclosure–transparency relationship in pooled specifications. By contrast, the firm fixed effects estimator identifies effects exclusively from longitudinal variation within each entity, effectively differencing out time-invariant confounders. Consequently, these results capture the “purified” marginal effect of changes in transparency on changes in disclosure intensity at the individual firm level.

This methodological refinement yields a critical theoretical insight: The substantive impact of external transparency on CSR reporting extensiveness is significantly more pronounced when evaluated through the lens of intra-firm evolution than through cross-sectional comparisons. While the absolute levels of disclosure vary considerably across firms due to fixed attributes, the firm fixed effects estimates demonstrate that transparency enhancements dynamically catalyze report expansion within the same organization over time. This reinforces our core proposition that transparency operates as a dynamic governance mechanism, exerting its strongest disclosure influence through within-firm temporal channels rather than static cross-sectional associations.

6. Discussion, Conclusion, and Limitations

This research demonstrates a robust and positive correlation between a firm’s external transparency and its performance in CSR, as well as the comprehensiveness of its CSR disclosures. This outcome strongly supports the proposition that scrutiny from financial markets serves as a pivotal catalyst for genuine CSR engagement. Consequently, our findings offer significant empirical evidence relevant to the persistent scholarly discussion concerning whether transparency acts as a driver or a potential constraint on CSR investment.

The literature presents conflicting viewpoints on this relationship. One stream of research, represented by scholars like Fieseler (2011) and Qian et al. (2019), posits that the pressure for transparency might paradoxically hinder long-term CSR commitments. Their argument centers on the notion that heightened visibility exposes firms to intense short-term performance demands from the market, potentially leading them to deprioritize substantive, long-horizon CSR investments. However, our results align with an alternative perspective championed by García-Sánchez et al. (2021) and E. H. Kim and Lyon (2015). This view emphasizes that increased transparency fosters greater accountability, thereby encouraging authentic CSR efforts rather than superficial reporting. Specifically, García-Sánchez et al. (2021) contend that external scrutiny effectively reduces “CSR decoupling”—the gap between a company’s stated CSR commitments and its actual practices—by compelling alignment between disclosure and performance. Our study empirically corroborates this accountability mechanism.

Notably, our conclusions diverge from those of Qian et al. (2019), who suggested transparency pressures could limit CSR investment. This divergence may be explained by a key methodological distinction: our conceptualization of “market scrutiny” is intentionally broader. While prior studies, including Qian et al. (2019), often focused narrowly on the influence of analyst coverage, we incorporate the combined pressures emanating from both financial analysts and investors. This more comprehensive view captures a wider spectrum of external oversight.

Furthermore, to achieve greater analytical clarity, we employed the framework established by Anderson et al. (2009). This allowed us to rigorously disentangle the distinct effect of external transparency (information available to outsiders) from the separate concept of internal disclosure quality (internal reporting and management systems). By isolating the influence of external market pressures in this way, our analysis provides a more refined and nuanced understanding of how these specific forces independently shape CSR outcomes, moving beyond the limitations of prior conflated approaches.

Moreover, our findings reinforce recent evidence from Caputo et al. (2021), who demonstrate that regulatory mandates on non-financial reporting substantially enhance the transparency of firms’ environmental disclosures, and Y. Liu et al. (2023), who show that performance transparency reduces consumer skepticism and increases the effectiveness of CSR initiatives. By situating our results alongside these recent works, it becomes clear that external transparency not only aligns disclosed information with actual CSR performance but also bolsters stakeholder trust and engagement.

This research offers several significant contributions to the existing body of knowledge. First, we extend the discourse on CSR determinants by demonstrating that external transparency is a robust predictor of CSR investment, addressing a gap identified in studies that oversimplify the CSR-shareholder value relationship (Awaysheh et al., 2020). Second, we advance the transparency-CSR literature by revealing that external scrutiny not only enhances CSR performance but also increases the comprehensiveness of disclosure, a dimension underexplored in prior work. This aligns with stakeholder theory (Freeman, 1984), as firms under scrutiny prioritize stakeholder interests to mitigate reputational risks and build trust. Third, our methodological approach—using residuals to isolate external transparency from internal governance factors—offers a novel lens for future research on transparency’s multifaceted impacts.

The mechanisms driving our findings can be contextualized through agency theory and institutional pressures. Heightened external transparency reduces information asymmetry, curbing managers’ ability to prioritize short-term gains over long-term CSR commitments (Jensen & Meckling, 1976). Simultaneously, institutional pressures compel firms to align with societal expectations, as visible firms face higher costs for greenwashing (Berrone et al., 2017). This dual dynamic explains why externally transparent firms invest more substantively in both social and environmental initiatives, albeit with variations across dimensions. For example, the stronger impact of transparency on social performance for non-reporting firms suggests that public scrutiny compels laggards to address immediate social concerns, while environmental investments—often requiring capital-intensive, long-term commitments—are prioritized by firms already engaged in reporting.

The practical takeaway from our research is the clear necessity of robust regulatory oversight and functioning market signals to encourage substantive CSR initiatives. Policymakers could incentivize transparency through mandates or ESG reporting frameworks, while investors might leverage transparency metrics to identify firms with authentic CSR commitments. For managers, proactively addressing stakeholder demands through robust CSR disclosure can enhance reputational capital and mitigate litigation risks, as evidenced by the positive association between litigation expenses and CSR performance in our robustness checks.

Despite its contributions, this study faces certain limitations that suggest the need for further investigation. First, while we address endogeneity through firm fixed effects, experimental designs could further isolate causality. Second, our focus on U.S. firms limits generalizability; cross-country analyses exploring institutional heterogeneity (Bhattacharya et al., 2003) could yield richer insights. Finally, qualitative investigations into how firms balance transparency pressures with CSR strategy formulation would deepen theoretical understanding. These avenues align with our call for research on the interplay between transparency, cultural contexts, and CSR outcomes.