Does corporate digital transformation restrain ESG decoupling? Evidence from China

The study investigates whether and how corporate digital transformation restrains ESG decoupling—the gap between firms’ external ESG reporting and their actual ESG activities. Motivated by growing concerns over symbolic ESG disclosure and information asymmetry, the authors argue that digital technologies enhance firms’ information collection, processing, and communication, improving internal control and increasing external monitoring, thereby curbing managerial discretion in ESG reporting. China provides a pertinent context due to rapid, policy-driven digitalization and rising ESG reporting among A-share listed firms. The paper’s purpose is to test the relationship between digital transformation and ESG decoupling, explore underlying mechanisms (information processing capability and reduced information asymmetry), examine regional and reporting-standard heterogeneity, and assess economic consequences for high-quality development.

The literature review covers three streams. (1) Digital transformation: defined as organization-wide reengineering leveraging connectivity, computing, and information technologies to improve efficiency and performance. Prior work links digitalization to operational transformation, enhanced customer value, productivity gains, performance improvements, financing benefits, and innovation, while noting potential ethical issues. (2) ESG decoupling: rooted in organizational decoupling theory and related to CSR decoupling and greenwashing, but broader due to inclusion of governance. Theoretical foundations include legitimacy theory (symbolic disclosure to gain acceptance) and information asymmetry (managers possess superior ESG information, enabling impression management). (3) Determinants of ESG decoupling: prior studies emphasize regulatory/monitoring contexts, market pressures, firm characteristics, governance, and managerial psychology. Gaps remain on technological drivers, measurement interdependencies across ESG databases, and the possibility that decoupling reflects information-processing limitations, not only impression management. The authors posit that digital transformation can mitigate decoupling by enhancing internal control (input side of reporting) and reducing information asymmetry (output side). They develop hypothesis H1: corporate digital transformation significantly mitigates ESG decoupling.

Data and sample: Chinese A-share listed firms (Shanghai and Shenzhen) that disclosed ESG-related reports from 2010–2019. Exclusions: financial/insurance industries, ST/*ST/PT firms, and observations with missing variables. Final sample comprises 6,185 firm-year observations. Continuous variables are winsorized at the 1% and 99% levels. Main data sources: CSMAR (firm data and MD&A texts), CNRDS (ESG performance ratings), Diebold (internal control index), and manually collected ESG reports for textual analysis of optimistic tone. Measurement of ESG decoupling: Defined as the standardized difference between the optimistic tone of ESG reports and actual ESG performance. Optimistic tone is computed via textual analysis using the Loughran–McDonald dictionary: (positive word count − negative word count)/(positive + negative) × 100; alternative specification scales by total words. ESG performance is proxied by CNRDS ESG score; in robustness, RKS ratings are used. Both components are converted to z-scores before differencing. Higher values indicate greater decoupling. Measurement of digital transformation: Construct a digital terminology dictionary (AI, big data, cloud computing, blockchain, digital applications) using national policy texts and media from 2010–2019, Python tokenization, and manual vetting (terms with frequency ≥5). Apply machine learning-based textual analysis to firms’ MD&A sections to count dictionary keyword frequencies; the natural log of total frequency is the Digital variable. Robustness uses an alternative measure: the ratio of digital economy-related intangible assets to total assets (keywords such as intelligent platform, management system, client, network, software, and related patents). Controls: Financial leverage (Lev), return on equity (ROE), firm size (Size, ln employees), firm age (Age), growth (sales growth), ownership concentration (First), CEO-chair duality (Dual), board size (DirNum), institutional ownership (Inst), and state ownership (SOE). Year and industry fixed effects are included in all baseline regressions. Empirical strategy: Baseline OLS regressions of Decoupling on Digital with controls and fixed effects. Endogeneity addressed through multiple approaches: (1) Instrumental variables (IV) using historical tele-density (IV_1984) à la Nunn and Qian (2014) to predict Digital; (2) Heckman two-stage model to correct sample selection into ESG reporting; (3) Propensity score matching (1:3 nearest neighbor with caliper) to balance covariates; (4) Extensive robustness tests including alternative measures, lagged Digital (1- and 2-year lags), exclusion of potential strategic disclosure cases (high-tech firms, abnormal digital word disclosures, firms punished for disclosure, and only good/excellent disclosure-assessed firms), and inclusion of province × year fixed effects to absorb regional-time shocks. Mechanism tests: Mediation via (a) internal control quality (Icw, Diebold composite index) representing improved information inputs; and (b) information asymmetry proxied by real earnings management (EM) representing reduced manipulation at output. Two-step regressions per Baron and Kenny (1986) assess partial mediation. Heterogeneity: Subsamples by region (eastern vs central-western China) and by GRI adherence (GRI-following vs non-GRI). Economic consequences: Stepwise regressions test whether digital transformation enhances firms’ total factor productivity (TFP; Levinsohn–Petrin method) directly and indirectly via reduced ESG decoupling.

- Descriptive patterns: ESG decoupling mean is 0.048 (SD 1.509), indicating, on average, more optimistic ESG disclosure relative to performance. Digital transformation varies widely (Digital mean 1.159; min 0; max 4.990). - Univariate comparison: Firms with digital transformation exhibit significantly lower ESG decoupling (mean −0.024) than those without (0.189), p < 0.01. - Baseline regressions: Digital is significantly negatively associated with ESG decoupling (coefficients −0.187 and −0.117 across specifications with controls and fixed effects; both p < 0.01). Several controls are significant (e.g., Size negative; SOE positive; larger boards associated with lower decoupling). - Endogeneity checks: • IV approach: The instrument (IV_1984) strongly predicts Digital (first-stage significant; Kleibergen-Paap LM = 14.532, p < 0.01). The second stage shows Predicted_Digital significantly reduces decoupling (coefficient ≈ −2.394, p < 0.01). • Heckman selection: After correcting for selection into ESG reporting, Digital remains negatively related to decoupling (−0.113, p < 0.01). • PSM: In the matched sample, Digital continues to significantly reduce decoupling (−0.108, p < 0.01). - Robustness: Findings hold using an alternative Digital measure based on digital intangible assets (−0.252, p < 0.05), alternative decoupling constructions (both using alternative tone scaling and RKS ESG performance; negative and significant at p < 0.01), lagged Digital (1- and 2-year lags negative and significant), exclusion of potential strategic disclosure effects (all subsamples show negative, p < 0.01), and with province × year fixed effects (Digital remains negative at p < 0.05). - Mechanisms: Digital transformation improves internal control quality (positive and significant), and better internal control is associated with lower decoupling; Digital reduces earnings management (negative and significant), and higher earnings management is associated with higher decoupling. Both channels show partial mediation, supporting that digitalization enhances information inputs and reduces opportunistic outputs, thereby narrowing ESG talk–walk gaps. - Heterogeneity: The negative effect of Digital on decoupling is stronger and significant in eastern China (−0.139, p < 0.01) but insignificant in central-western regions; Chow tests confirm coefficient differences (p < 0.0001). The effect is significant among firms not following GRI (−0.090, p < 0.01) but insignificant among GRI-followers; differences are statistically significant (p < 0.0001). - Economic consequences: Digital transformation significantly increases TFP (0.043, p < 0.01). ESG decoupling is negatively associated with TFP (−0.0106, p < 0.05). Including decoupling slightly attenuates Digital’s coefficient on TFP, consistent with a pathway whereby digitalization enhances high-quality development by reducing ESG decoupling.

The findings support H1: corporate digital transformation mitigates ESG decoupling. By strengthening internal control systems (better information acquisition, processing, and reliability) and reducing information asymmetry (greater transparency, enhanced stakeholder monitoring, lower supervision costs, and reduced managerial discretion), digitalization aligns ESG narratives with actual practices. This resolution of the talk–walk gap improves the credibility and usefulness of ESG disclosures, informing stakeholders and capital markets more effectively. The study advances the ESG decoupling literature by highlighting technological drivers and by using text-based measures that directly compare report tone to performance, addressing prior measurement interdependencies. It also broadens digital transformation research by evidencing social disclosure quality benefits and by documenting that digitalization’s governance and empowering roles extend to sustainability reporting. Regional and reporting-standard heterogeneity underscore the importance of digital infrastructure and institutional frameworks in shaping digitalization’s effectiveness. Finally, demonstrating that reduced decoupling contributes to higher productivity links ESG reporting quality to firms’ high-quality development.

Using 6,185 firm-year observations of Chinese A-share listed companies (2010–2019), the study shows that corporate digital transformation significantly reduces ESG decoupling. Results remain robust across IV, Heckman, PSM, alternative measures, lag structures, strategic disclosure exclusions, and fixed-effect enrichments. Mechanism tests reveal that improved internal control and reduced information asymmetry partially mediate the effect. The effect is stronger in eastern regions and among non-GRI reporters. Moreover, digital transformation enhances firms’ TFP and part of this benefit operates through lower ESG decoupling. Contributions include identifying technological forces governing ESG decoupling, employing text analytics to measure decoupling directly from ESG reports, and demonstrating economic consequences for productivity. Policy and practice implications encourage support for digitalization, refinement of ESG governance in digital contexts, and managerial use of digital tools to strengthen internal control and truthful disclosure. Future research can build more granular, firm-specific digitalization measures, develop comprehensive Chinese sentiment dictionaries for ESG texts, and test generalizability across international contexts and institutional environments.

- Measurement of digital transformation relies on keyword-based textual analysis and digital intangible asset ratios, which may not fully capture multi-dimensional changes in technology, talent, organization, and strategy or potential over/under-disclosure. - ESG report tone is measured using the Loughran–McDonald dictionary; lack of an authoritative, comprehensive Chinese sentiment dictionary may lead to measurement noise in Chinese ESG texts. - Generalizability beyond China should be made cautiously due to institutional differences; replication in international settings is needed. - Despite extensive controls and robustness checks, residual endogeneity cannot be entirely ruled out.