When Do Green Initiatives Translate into Financial Performance? ESG Transmission and Corporate Life-Cycle Contingencies in China’s Heavily Polluting Industries

1. Introduction

Corporate environmental transformation has become a central financial and strategic issue because firms are increasingly evaluated not only by realized earnings but also by their capacity to manage environmental liabilities, transition risks, stakeholder expectations, and long-horizon resource constraints. The accumulated literature generally finds that environmental, social, and governance (ESG) performance is non-negatively related to financial performance, yet the magnitude and even the sign of the relationship depend on the outcome, institutional setting, measurement system, and research design (Friede et al., 2015; Gillan et al., 2021). Recent panel evidence also shows that ESG, green innovation, and financial outcomes can be connected through multiple channels rather than a single reduced-form relationship (Tripopsakul, 2025; Zheng et al., 2022).

Two gaps remain particularly important. First, the term “green strategy” often aggregates economically different activities. Environmental capital expenditure commits internal resources; green patenting reflects technological capability; environmental subsidies provide externally supplied policy resources; and ISO 14001 certification formalizes an environmental management system. These activities may all improve external sustainability assessment, but they differ sharply in cash-flow consequences, implementation costs, timing, and informational content. Existing studies usually examine one activity at a time—for example, green innovation (Liu et al., 2024; Zheng et al., 2022) or green investment (Hou et al., 2025)—which makes cross-initiative comparisons difficult because samples, measures, and estimators vary across studies.

Second, the mechanism by which green action is translated into financial performance is incompletely specified. ESG ratings can operate as an external evaluation channel: observable environmental actions enter sustainability assessments, and those assessments can affect stakeholder confidence, financing conditions, supply-chain access, regulatory exposure, and reputation. However, a positive ESG-mediated pathway does not imply that the total contemporaneous financial effect of a green initiative must be positive. Investment, research, compliance, matching expenditure, and designated-use constraints can coexist with stakeholder benefits. The distinction between a positive indirect pathway and a weak or negative remaining direct component is therefore economically meaningful rather than statistically anomalous.

A further issue is whether this transmission mechanism is equally strong across stages of corporate development. Firm life cycle captures systematic variation in financing needs, investment opportunities, organizational routines, and resource-integration capacity. Prior accounting and finance research documents life-cycle differences in financial resources, governance, and corporate social-responsibility behavior (Habib & Hasan, 2019; Hasan & Habib, 2017). Recent evidence from China also shows that life-cycle conditions can alter relationships among green innovation, financial performance, and ESG reporting (Rauf et al., 2026). Yet the moderator is often used as a broad subgroup device rather than located at a specific point in the mechanism. This study specifies life cycle on the first stage—green initiatives to ESG performance—because resource deployment and implementation capacity are expected to determine how effectively a green action becomes externally visible sustainability performance.

The empirical setting is Chinese A-share firms in heavily polluting industries from 2015 to 2024. This population is particularly informative because pollution-control investment, green innovation, environmental subsidies, certification, and ESG assessment are simultaneously salient under strong regulatory pressure and rapid development of green-finance and sustainability-disclosure institutions. The final panel contains 6,579 firm-year observations from 1,114 firms. We estimate firm and year fixed-effects models with firm-clustered standard errors, use product-of-coefficients mediation with 5,000 firm-cluster bootstrap replications, apply Benjamini-Hochberg false-discovery-rate adjustment to prespecified hypothesis families, and evaluate life-cycle differences through conditional indirect-effect contrasts.

The results show a layered pattern. All four initiatives are positively associated with ESG ratings. ESG performance is positively associated with the accounting outcomes ROA and ROE but not with the market-based outcome Tobin’s Q. All four initiatives exhibit positive statistical indirect associations with ROA through ESG; however, the total contemporaneous association of green subsidy with ROA is negative, and the total associations for investment, innovation, and certification are not statistically distinguishable from zero. Life-cycle moderation is selective: the conditional indirect associations for green investment and green subsidy are significantly stronger in growth-stage firms than in mature and declining firms, while corresponding contrasts are not established for innovation or certification.

This study contributes in four ways. First, it compares four economically distinct green initiatives within the same firms, period, control structure, and estimator, thereby separating initiative-specific differences from cross-study design differences. Second, it positions ESG as an intermediate external-evaluation mechanism and explicitly decomposes total associations into ESG-mediated and remaining direct components. Third, it locates corporate life cycle at a theoretically specified first stage and tests conditional indirect-effect contrasts rather than inferring moderation from subgroup significance. Fourth, it distinguishes accounting profitability from market valuation and retains qualified and null findings, which helps define the boundary of claims about ESG-related value creation. The study therefore advances a mechanism-based interpretation: green action can be recognized by ESG evaluators without generating an immediate or uniform financial premium.

The remainder of the paper is organized as follows. Section 2 develops the theoretical framework and hypotheses. Section 3 describes the sample, variables, models, and inference procedures. Section 4 presents the empirical results and robustness evidence. Section 5 discusses the findings and their theoretical implications. Section 6 outlines practical implications, and Section 7 concludes with limitations and directions for future research.

2. Theoretical Framework and Hypothesis Development

2.1. Green Initiatives and ESG Performance

Stakeholder theory argues that firms depend on relationships with shareholders, creditors, employees, customers, suppliers, governments, communities, and other groups whose claims extend beyond short-run profit (Freeman, 1984). In this setting, ESG ratings are a compact external representation of how firms manage environmental, social, and governance responsibilities. A firm’s green actions can therefore affect financial outcomes indirectly when they change the information and signals available to stakeholders.

The resource-based view complements this logic by emphasizing heterogeneity in resources and in firms’ ability to organize and deploy them (Barney, 1991; Wernerfelt, 1984). Environmental investment, technological knowledge, policy-supported resources, and formalized management systems are different resources or capabilities. Their value depends not only on acquisition but also on effective integration into operations. A dynamic resource perspective further suggests that the economic usefulness of a resource can change with organizational development and capability evolution (Helfat & Peteraf, 2003).

Green investment consists of expenditures on pollution control, cleaner production, wastewater treatment, desulfurization, denitrification, and related projects. Such investment is observable evidence that environmental objectives are backed by resource commitment. Prior evidence links green investment to pollution reduction and environmental outcomes (Ren et al., 2022), while recent firm-level research in China shows that environmental investment can be reflected in ESG ratings (Hou et al., 2025). Because rating agencies evaluate environmental governance and disclosed environmental actions, increases in green investment should be associated with stronger ESG assessment.

Green innovation is measured through green patent applications and captures technological activity directed toward cleaner production, resource efficiency, and environmental protection. Prior studies show that green innovation is associated with environmental performance (Rehman et al., 2021; Tang et al., 2018) and, in Chinese listed firms, with ESG performance (Liu et al., 2024; Zheng et al., 2022). Patent applications do not guarantee commercialization or physical emission reduction, but they provide an externally observable signal of a firm’s technological orientation.

Green subsidies are targeted government resources for environmental protection, energy conservation, emission reduction, cleaner production, or green technology projects. Subsidies can relax project financing constraints and signal policy recognition. Evidence from China links government environmental or R&D support to green innovation and environmental performance (Bai et al., 2019; Du et al., 2023; Han et al., 2024; Luo et al., 2024). Thus, even when subsidized projects impose complementary expenditures, the receipt and deployment of targeted support can be positively reflected in ESG assessment.

Green certification is represented by valid ISO 14001 environmental management system certification. ISO 14001 formalizes procedures for environmental policy, monitoring, review, corrective action, and continual improvement (International Organization for Standardization [ISO], 2015). Certification provides third-party verification and can reduce information asymmetry about environmental-management processes. Prior work connects ISO 14001 with environmental and supply-chain practices (Arimura et al., 2011; Erauskin-Tolosa et al., 2020; Heras-Saizarbitoria et al., 2011; Potoski & Prakash, 2005), although certification status does not guarantee identical implementation quality across firms.

H1a. 

Green investment is positively associated with corporate ESG performance.

H1b. 

Green innovation is positively associated with corporate ESG performance.

H1c. 

Green subsidy is positively associated with corporate ESG performance.

H1d. 

Green certification is positively associated with corporate ESG performance.

2.2. ESG Performance and Corporate Financial Performance

ESG performance can influence financial outcomes through several stakeholder channels. Stronger sustainability assessment can reduce perceived compliance and reputational risk, improve access to creditors and supply chains, support customer and employee relationships, and signal management quality. Broad reviews find predominantly non-negative but heterogeneous ESG–financial-performance relationships (Endrikat et al., 2014; Friede et al., 2015; Gillan et al., 2021). The heterogeneity is important because accounting profitability and market valuation capture different economic objects.

ROA and ROE measure realized accounting performance and can reflect changes in operating continuity, financing costs, risk management, and asset utilization over a relatively short horizon. Tobin’s Q, by contrast, reflects market expectations about future cash flows and is influenced by growth opportunities, industry prospects, discount rates, liquidity, and investor sentiment. ESG information may therefore improve accounting performance without being immediately capitalized into market valuation, particularly in pollution-intensive sectors where better sustainability assessment can coexist with expectations of future transition expenditure (Eccles et al., 2014; Fatemi et al., 2018).

H2. 

Corporate ESG performance is positively associated with corporate financial performance, with outcome-specific evidence assessed for ROA, ROE, and Tobin’s Q.

2.3. ESG as a Statistical Transmission Channel

Combining the first two pathways yields an indirect mechanism in which green initiatives improve externally assessed ESG performance and ESG performance is subsequently associated with financial outcomes. This mechanism differs from a claim that green activity directly raises production efficiency. It instead captures an external-evaluation channel through which stakeholders recognize and respond to the firm’s sustainability profile.

The remaining direct component can differ in sign across initiatives. Environmental capital expenditure can generate operating savings but also requires current resource commitment; green innovation can have commercialization lags; subsidy-supported projects can impose matching expenditure, designated-use restrictions, or administrative costs; and certification entails system design, audits, training, and ongoing compliance. Accordingly, the study predicts the direction of the indirect effect but does not impose a directional hypothesis on the remaining direct component. This distinction is consistent with evidence that sustainability-related benefits and resource costs can coexist (Endrikat et al., 2014; Heras-Saizarbitoria et al., 2011; Tang et al., 2018).

H3a. 

Green investment has a positive statistical indirect association with financial performance through ESG performance.

H3b. 

Green innovation has a positive statistical indirect association with financial performance through ESG performance.

H3c. 

Green subsidy has a positive statistical indirect association with financial performance through ESG performance.

H3d. 

Green certification has a positive statistical indirect association with financial performance through ESG performance.

2.4. Corporate Life Cycle as a First-Stage Boundary Condition

Corporate life cycle captures differences in investment opportunity, financing demand, asset structure, strategic orientation, and organizational routines. The cash-flow-pattern framework developed by Dickinson (2011) provides an observable classification that can be applied to listed firms, while the broader literature documents systematic stage differences in financial resources and corporate behavior (Habib & Hasan, 2019; Hasan & Habib, 2017). A resource-based interpretation predicts that the same increment in a green resource need not be converted into ESG performance with equal effectiveness across stages.

Growth-stage firms are expanding capacity and establishing operating systems. Environmental facilities and cleaner production technologies can therefore be embedded in new capacity, while policy-supported projects can complement expansion opportunities. Mature firms typically have more established environmental assets and systems, so additional environmental spending may involve retrofit or replacement and produce smaller marginal changes in externally observable ESG indicators. Declining firms face tighter cash-flow constraints and strategic contraction, making green expenditure more likely to focus on minimum compliance rather than broad capability building. These differences imply a stronger initiative-to-ESG conversion in growth-stage firms, particularly for initiatives that require current resource deployment.

By contrast, the second stage from ESG assessment to financial performance primarily reflects external stakeholder responses to ESG signals. Those responses are shaped by capital markets, creditors, customers, and supply-chain institutions and are not assumed ex ante to differ systematically with the firm’s life-cycle stage. The moderator is therefore located on the first stage. This specification is more precise than treating life cycle as an all-purpose subgroup variable and differs from research that examines broader life-cycle moderation of financial performance or ESG reporting (Rauf et al., 2026).

H4a. 

The ESG-mediated association of green investment with ROA is stronger in growth-stage firms than in mature firms.

H4b. 

The ESG-mediated association of green innovation with ROA is stronger in growth-stage firms than in mature firms.

H4c. 

The ESG-mediated association of green subsidy with ROA is stronger in growth-stage firms than in mature firms.

H4d. 

The ESG-mediated association of green certification with ROA is stronger in growth-stage firms than in mature firms.

H5a. 

The ESG-mediated association of green investment with ROA is stronger in growth-stage firms than in declining firms.

H5b. 

The ESG-mediated association of green innovation with ROA is stronger in growth-stage firms than in declining firms.

H5c. 

The ESG-mediated association of green subsidy with ROA is stronger in growth-stage firms than in declining firms.

H5d. 

The ESG-mediated association of green certification with ROA is stronger in growth-stage firms than in declining firms.

3. Materials and Methods

3.1. Data, Sample, and Institutional Setting

The study covers ten complete fiscal years from 2015 to 2024. The period begins with a phase of more systematic ecological-civilization reform in China and spans the subsequent development of green finance, the carbon-peaking and carbon-neutrality agenda, and progressively stronger sustainability-disclosure expectations. The ten-year window also provides within-firm variation for fixed-effects identification and sufficient temporal coverage for lagged and staggered-adoption sensitivity analyses.

The population consists of Shanghai and Shenzhen A-share firms operating in heavily polluting industries. Industry scope is identified using three-digit China Securities Regulatory Commission (CSRC) codes together with the scope of heavily polluting industries specified by environmental authorities. The final sample covers 16 three-digit industry categories, including chemicals, rubber and plastics, non-metallic mineral products, metal processing, electricity and heat production, textiles, paper, petroleum processing, chemical fiber, and mining-related industries.

Firm-year observations are screened to exclude financial and insurance firms, ST and *ST/delisting-risk observations, observations outside the firm’s listed period, and observations missing variables required for the primary ESG and ROA models. The resulting unbalanced panel contains 6,579 firm-year observations from 1,114 firms. Tobin’s Q is available for 6,577 observations. The life-cycle composition is 2,638 growth-stage observations (40.10%), 2,960 mature-stage observations (44.99%), and 981 decline-stage observations (14.91%).

Data are integrated by stock code and year. CSMAR provides firm characteristics, accounting data, governance variables, market valuation, industry information, and detailed subsidy records. Huazheng provides the primary ESG rating, continuous ESG score, and component scores. Green patent applications are drawn from a patent research database and classified using relevant International Patent Classification codes; the underlying registrations are publicly verifiable through the China National Intellectual Property Administration. Environmental investment is constructed from annual-report note items and cross-checked against a corporate environmental-responsibility data source. ISO 14001 records are obtained from publicly accessible certification information and supplemented with annual reports and corporate social-responsibility reports. City-level macroeconomic controls are obtained from regional statistical sources. Refinitiv ESG data are used only for cross-provider sensitivity analysis because coverage is materially smaller.

Table 1. Analytical sample composition.

Dimension	Category	Firm-years	Share
Life cycle	Growth	2,638	40.10%
Life cycle	Mature	2,960	44.99%
Life cycle	Decline	981	14.91%
Panel	Total	6,579	100.00%
Firms	Distinct listed firms	1,114	—
Period	2015–2024	10 years	—

Note. The panel is unbalanced. Life-cycle stage is assigned from the sign pattern of operating, investing, and financing cash flows following Dickinson (2011).

Table 1. Analytical sample composition.

Dimension	Category	Firm-years	Share
Life cycle	Growth	2,638	40.10%
Life cycle	Mature	2,960	44.99%
Life cycle	Decline	981	14.91%
Panel	Total	6,579	100.00%
Firms	Distinct listed firms	1,114	—
Period	2015–2024	10 years	—

Note. The panel is unbalanced. Life-cycle stage is assigned from the sign pattern of operating, investing, and financing cash flows following Dickinson (2011).

3.2. Variable Measurement

The primary financial outcome is return on assets (ROA), calculated as net profit divided by average total assets. Return on equity (ROE) is the secondary accounting outcome, and Tobin’s Q is the market-based outcome. ROA is used for the total-association, mediation, and moderated-mediation decompositions because it provides a common asset-scaled profitability measure across the sample.

The four green initiatives are measured separately rather than aggregated into a composite index because their scales and economic meanings differ substantially. Green investment is environmental investment divided by operating revenue. Green innovation is the natural logarithm of one plus annual green patent applications. Green subsidy is the value of environmental government subsidies identified under a strict project-name rule divided by operating revenue. Green certification equals one when the firm holds a valid ISO 14001 environmental management system certificate in a given year and zero otherwise.

The primary mediator is the Huazheng ESG rating coded from 9 for AAA to 1 for C. The rating is used because it represents the provider’s final categorical assessment observed by external stakeholders. In the linear fixed-effects models, the nine ordered categories are entered numerically; the resulting coefficients are therefore interpreted as within-firm movements on the ordered rating scale rather than as evidence that every adjacent grade has identical economic distance. The continuous Huazheng ESG score is used as an alternative measure in second-stage sensitivity analysis to assess dependence on discretization. Refinitiv combined ESG and environmental-pillar scores provide a separate cross-provider check for the smaller 2019–2024 subsample.

Corporate life cycle is classified from the signs of operating, investing, and financing cash flows following Dickinson (2011). The original introduction and growth patterns are combined as growth, the original maturity pattern remains maturity, and shake-out and decline patterns are combined as decline. This consolidation reflects the limited number of listed firms with true start-up patterns and the asset-disposal characteristics of shake-out observations in this population.

Controls include firm size (log total assets), leverage, firm age, operating cash flow scaled by assets, revenue growth, ownership concentration, CEO duality, board size (log number of directors), city-level GDP (log), and industrial structure upgrading (tertiary-industry value added divided by secondary-industry value added). Continuous variables are winsorized at the 1st and 99th percentiles on duplicated analysis variables; indicator and ordinal variables remain untransformed. All main models are repeated with raw continuous variables as a sensitivity check.

Table 2. Variable definitions.

Variable	Role	Measurement
ROA	Outcome	Net profit/average total assets
ROE	Outcome	Net profit/average shareholders’ equity
Tobin’s Q	Outcome	Market value/replacement cost of capital
Green investment	Independent	Environmental investment/operating revenue
Green innovation	Independent	ln(green patent applications + 1)
Green subsidy	Independent	Strictly identified green subsidy/operating revenue
Green certification	Independent	Valid ISO 14001 certificate: 1=yes, 0=no
ESG rating	Mediator	Huazheng rating coded AAA=9 to C=1
Life-cycle stage	Moderator	Growth, maturity, decline based on cash-flow sign patterns
Controls	Covariates	Firm characteristics, governance, and regional macro controls

Note. Green investment and green subsidy are revenue-scaled ratios; coefficient magnitudes across the four initiative variables are therefore not directly comparable.

Table 2. Variable definitions.

Variable	Role	Measurement
ROA	Outcome	Net profit/average total assets
ROE	Outcome	Net profit/average shareholders’ equity
Tobin’s Q	Outcome	Market value/replacement cost of capital
Green investment	Independent	Environmental investment/operating revenue
Green innovation	Independent	ln(green patent applications + 1)
Green subsidy	Independent	Strictly identified green subsidy/operating revenue
Green certification	Independent	Valid ISO 14001 certificate: 1=yes, 0=no
ESG rating	Mediator	Huazheng rating coded AAA=9 to C=1
Life-cycle stage	Moderator	Growth, maturity, decline based on cash-flow sign patterns
Controls	Covariates	Firm characteristics, governance, and regional macro controls

Note. Green investment and green subsidy are revenue-scaled ratios; coefficient magnitudes across the four initiative variables are therefore not directly comparable.

3.3. Econometric Specification

The baseline specification uses firm and year fixed effects. Firm effects absorb time-invariant heterogeneity in business models, managerial capability, and industry niches, while year effects absorb common macroeconomic, regulatory, and capital-market shocks. Standard errors are clustered at the firm level because residuals may be serially correlated and heteroskedastic within firms. This approach follows standard panel-data inference practice in corporate finance (Petersen, 2009).

ROA_it = α + βGreen_it + γ′X_it + μ_i + λ_t + ε_it

(1)

Equation (1) estimates the total association of each green initiative with ROA. The first-stage model for H1 replaces the outcome with the ESG rating:

ESG_it = α₁ + aGreen_it + γ₁′X_it + μ_i + λ_t + ε_it

(2)

H2 is evaluated by estimating ESG separately against ROA, ROE, and Tobin’s Q. For mediation, the initiative-specific second-stage model includes both ESG and the focal initiative:

ROA_it = α_2m + b_mESG_it + c′Green_it + γ_2m′X_it + μ_i + λ_t + ε_it

(3)

The statistical indirect effect is a × b_m, the remaining direct component is c’, and the total association is a × b_m + c’. Because the sampling distribution of the product term is generally asymmetric, inference is based on 5,000 bootstrap replications resampled at the firm-cluster level rather than on a normal approximation (Efron & Tibshirani, 1993; Preacher & Hayes, 2008).

The moderated-mediation model places corporate life cycle on the first stage. Growth is the reference group, and Mature and Decline are indicator variables:

ESG_it = α₃ + a₁Green_it + a₂(Green_it × Mature_it) + a₃(Green_it × Decline_it) + a₄Mature_it + a₅Decline_it + γ₃′X_it + μ_i + λ_t + ε_it

(4)

Conditional indirect effects are computed as a₁ × b_m for growth, (a₁ + a₂) × b_m for maturity, and (a₁ + a₃) × b_m for decline. H4 and H5 are tested using growth-minus-maturity and growth-minus-decline contrasts of these conditional indirect effects. The contrasts and their confidence intervals are obtained from the same firm-cluster bootstrap samples, ensuring internally consistent inference.

Multiple related tests raise the probability of false positives. The four H1 tests form one family, the four H3 indirect effects form a second family, and the eight H4/H5 life-cycle contrasts form a third family. Benjamini-Hochberg false-discovery-rate adjustment is applied within these prespecified families (Benjamini & Hochberg, 1995). H2 is retained as one substantive proposition with outcome-specific reporting.

3.4. Model Diagnostics and Sensitivity Design

Model-selection diagnostics support the two-way fixed-effects specification. The firm-effects F test and Breusch-Pagan LM test reject pooled OLS, the Hausman test rejects random-effects consistency, and year effects are jointly significant. The Wooldridge test indicates first-order serial correlation and the modified Wald test indicates groupwise heteroskedasticity, reinforcing the use of firm-clustered standard errors. Multicollinearity is limited: the mean variance-inflation factor is 1.318 and the maximum is 2.344.

The robustness programme varies measurement, outcome, control set, fixed-effects structure, sample, timing, and treatment of extreme values. It includes a broader subsidy rule, an alternative source investment measure, continuous Huazheng ESG, Refinitiv ESG measures, alternative profitability measures, industry-by-year effects, exclusion of 2020, alternative heavy-pollution flags, one- and two-year lags, and raw variables. First ISO 14001 certification is additionally analyzed through a cohort-robust staggered difference-in-differences design because certification has an identifiable adoption date (Callaway & Sant’Anna, 2021). System GMM is reported only as a supplementary diagnostic for persistence and potential reverse causality (Blundell & Bond, 1998; Windmeijer, 2005). Generative AI was used only in manuscript preparation as disclosed in the Acknowledgments; it was not used to generate or alter the underlying data or statistical estimates.

All data cleaning, variable construction, diagnostics, estimation, bootstrap inference, multiple-testing adjustment, and robustness analyses were implemented programmatically in Stata 18, with scripts and execution outputs retained for reproducibility. During conversion of the dissertation materials into this journal manuscript, ChatGPT (OpenAI, GPT-5.6 Sol) was used to assist with article organization, drafting, and language refinement. The empirical design, variable definitions, source data, statistical models, estimates, and tables were supplied from the first author’s pre-existing dissertation research workflow; the tool was not used to generate, modify, or independently analyze the research data. All AI-assisted text was reviewed and verified by the authors.

4. Results

4.1. Descriptive Statistics and Within-Firm Variation

The main variables display sufficient within-firm variation for fixed-effects identification. The within-firm standard deviation of the ESG rating is 0.7704, while those of ROA, green investment, green innovation, green subsidy, and certification are 0.0410, 0.02838, 0.6395, 0.000618, and 0.3083, respectively. The four green initiatives are only weakly correlated with one another, supporting their treatment as distinct dimensions rather than a single composite index. The strongest pairwise correlation among explanatory variables is between firm size and green innovation (Pearson r = 0.624), but the corresponding VIF values remain well below conventional concern thresholds.

Table 3. Selected descriptive statistics.

Variable	N	Mean	SD	Min	Max
ROA	6,579	0.0406	0.0581	-0.1606	0.2135
ROE	6,579	0.0612	0.1203	-0.5144	0.3882
Tobin’s Q	6,577	1.7396	0.9591	0.7887	6.1635
Green investment	6,579	0.01608	0.04358	0.0000	0.2960
Green innovation	6,579	1.0000	1.1546	0.0000	4.6052
Green subsidy	6,579	0.000323	0.000926	0.0000	0.00670
Green certification	6,579	0.3909	0.4880	0.0000	1.0000
ESG rating	6,579	4.0702	1.1057	1.0000	9.0000

Note. Continuous variables are winsorized at the 1st and 99th percentiles; indicator and ordinal variables are shown in raw form.

Table 3. Selected descriptive statistics.

Variable	N	Mean	SD	Min	Max
ROA	6,579	0.0406	0.0581	-0.1606	0.2135
ROE	6,579	0.0612	0.1203	-0.5144	0.3882
Tobin’s Q	6,577	1.7396	0.9591	0.7887	6.1635
Green investment	6,579	0.01608	0.04358	0.0000	0.2960
Green innovation	6,579	1.0000	1.1546	0.0000	4.6052
Green subsidy	6,579	0.000323	0.000926	0.0000	0.00670
Green certification	6,579	0.3909	0.4880	0.0000	1.0000
ESG rating	6,579	4.0702	1.1057	1.0000	9.0000

Note. Continuous variables are winsorized at the 1st and 99th percentiles; indicator and ordinal variables are shown in raw form.

4.2. Green Initiatives and ESG Performance

Table 4, Panel A, reports the first-stage fixed-effects estimates. Green investment, green innovation, green subsidy, and green certification are each positively associated with the ESG rating at the 1% level. The separate-model estimates are 1.8021, 0.0942, 68.7565, and 0.2385, respectively. Because the four variables use different measurement scales, the raw coefficient magnitudes should not be interpreted as a ranking of substantive importance. In an additional joint specification, all four coefficients remain positive and statistically significant (1.7757, 0.0962, 63.3822, and 0.2463, respectively), indicating that the first-stage findings are not driven by pairwise overlap among the initiative measures. All four H1 statements remain significant after within-family false-discovery-rate adjustment.

4.3. ESG and Financial Performance

Panel B shows that the ESG rating is positively associated with both accounting outcomes. A one-point increase in the ESG rating is associated with a 0.00180 increase in ROA (p = 0.007) and a 0.00398 increase in ROE (p = 0.012), conditional on the controls and fixed effects. The Tobin’s Q coefficient is -0.01698 (p = 0.103), so a positive market-valuation association is not established. Replacing the ordinal rating with the continuous Huazheng ESG score produces the same qualitative pattern: the score is positive for ROA and ROE and statistically insignificant for Tobin’s Q. H2 is therefore supported for accounting-based performance but not for the market-based outcome.

4.4. Total Associations of Green Initiatives with ROA

Panel C reports total associations before including the ESG mediator. Green investment (0.00299), green innovation (0.00064), and green certification (-0.00156) are not statistically significant in their separate total-association models. Green subsidy is negatively associated with ROA (-2.15510, p < 0.05). The negative subsidy relationship is not a mechanical artifact of grant income entering profit: when recognized green grants are removed from pre-tax profit, the subsidy coefficient becomes more negative (-3.4031), and it remains negative across operating ROA, pre-tax ROA, EBIT ROA, and operating margin specifications.

Table 4. Core two-way fixed-effects estimates.

Path/variable	Coefficient	Clustered SE	p-value	Within R2
Panel A: Initiative -> ESG rating
Green investment	1.8021***	0.3307	<0.001	0.0954
Green innovation	0.0942***	0.0186	<0.001	0.0961
Green subsidy	68.7565***	15.5205	<0.001	0.0941
Green certification	0.2385***	0.0417	<0.001	0.0998
Panel B: ESG rating -> financial performance
ROA	0.00180***	0.00067	0.007	0.4054
ROE	0.00398**	0.00158	0.012	0.3402
Tobin’s Q	-0.01698	0.01040	0.103	0.3032
Panel C: Total association -> ROA
Green investment	0.00299	0.01733	n.s.	0.4043
Green innovation	0.00064	0.00087	n.s.	0.4044
Green subsidy	-2.15510**	0.83835	<0.05	0.4053
Green certification	-0.00156	0.00165	n.s.	0.4045

Note. All models include the full control set, firm fixed effects, and year fixed effects. Panel A uses the initiative-specific models. Panel C reports the total-association models used in the mediation decomposition. *** p < 0.01, ** p < 0.05. “n.s.” denotes not statistically significant at 10%.

Table 4. Core two-way fixed-effects estimates.

Path/variable	Coefficient	Clustered SE	p-value	Within R2
Panel A: Initiative -> ESG rating
Green investment	1.8021***	0.3307	<0.001	0.0954
Green innovation	0.0942***	0.0186	<0.001	0.0961
Green subsidy	68.7565***	15.5205	<0.001	0.0941
Green certification	0.2385***	0.0417	<0.001	0.0998
Panel B: ESG rating -> financial performance
ROA	0.00180***	0.00067	0.007	0.4054
ROE	0.00398**	0.00158	0.012	0.3402
Tobin’s Q	-0.01698	0.01040	0.103	0.3032
Panel C: Total association -> ROA
Green investment	0.00299	0.01733	n.s.	0.4043
Green innovation	0.00064	0.00087	n.s.	0.4044
Green subsidy	-2.15510**	0.83835	<0.05	0.4053
Green certification	-0.00156	0.00165	n.s.	0.4045

Note. All models include the full control set, firm fixed effects, and year fixed effects. Panel A uses the initiative-specific models. Panel C reports the total-association models used in the mediation decomposition. *** p < 0.01, ** p < 0.05. “n.s.” denotes not statistically significant at 10%.

4.5. ESG-Mediated Indirect Associations

The mediation decomposition provides the central mechanism evidence. All four product terms are positive, their 95% percentile bootstrap intervals exclude zero, and their false-discovery-rate-adjusted p-values are approximately 0.010. The indirect association is 0.003251 for green investment, 0.000167 for green innovation, 0.131430 for green subsidy, and 0.000449 for green certification. H3a–H3d are therefore supported.

The decomposition also shows why total-effect regressions alone are incomplete. For green investment, the positive indirect association (0.003251) is larger than the small negative remaining direct estimate (-0.00026), leaving a small and imprecise positive total association. For green innovation, both the indirect and remaining direct point estimates are positive, but only the indirect component is statistically precise. For green subsidy, the positive ESG-mediated association (0.131430) is substantially smaller than the negative remaining direct component (-2.28653), yielding a negative total association (-2.15510). For certification, the positive indirect component (0.000449) partially offsets a negative but imprecise direct component (-0.00201). Thus, external ESG recognition and contemporaneous profitability can move in different directions.

Table 5. Decomposition of initiative-ROA associations through ESG.

Initiative	Path a	Path b	Direct	Indirect	95% CI	Total
Green investment	1.8021***	0.00180**	-0.00026	0.003251**	[0.00080, 0.00608]	0.00299
Green innovation	0.0942***	0.00178***	0.00047	0.000167***	[0.000039, 0.000316]	0.00064
Green subsidy	68.7565***	0.00191***	-2.28653***	0.131430***	[0.03477, 0.25210]	-2.15510**
Green certification	0.2385***	0.00188***	-0.00201	0.000449***	[0.000126, 0.000811]	-0.00156

Note. Percentile confidence intervals and bootstrap standard errors are based on 5,000 replications resampled over the 1,114 firm clusters. The indirect effect is a × b. The quantities represent statistical mediation and do not identify a causal mechanism. *** p < 0.01, ** p < 0.05.

Table 5. Decomposition of initiative-ROA associations through ESG.

Initiative	Path a	Path b	Direct	Indirect	95% CI	Total
Green investment	1.8021***	0.00180**	-0.00026	0.003251**	[0.00080, 0.00608]	0.00299
Green innovation	0.0942***	0.00178***	0.00047	0.000167***	[0.000039, 0.000316]	0.00064
Green subsidy	68.7565***	0.00191***	-2.28653***	0.131430***	[0.03477, 0.25210]	-2.15510**
Green certification	0.2385***	0.00188***	-0.00201	0.000449***	[0.000126, 0.000811]	-0.00156

Note. Percentile confidence intervals and bootstrap standard errors are based on 5,000 replications resampled over the 1,114 firm clusters. The indirect effect is a × b. The quantities represent statistical mediation and do not identify a causal mechanism. *** p < 0.01, ** p < 0.05.

4.6. Corporate Life Cycle and Conditional Indirect Associations

Life-cycle differences are concentrated in green investment and green subsidy. For green investment, the conditional indirect association is 0.00489 in growth, 0.00151 in maturity, and 0.00007 in decline. The growth-minus-maturity contrast is 0.00338 with a 95% CI of [0.00064, 0.00686], and the growth-minus-decline contrast is 0.00482 with a 95% CI of [0.00094, 0.00974]. After adjustment across the eight life-cycle contrasts, both differences remain significant (adjusted p = 0.026). H4a and H5a are supported.

The subsidy pattern is similar but larger on its own scale. Conditional indirect associations are 0.26825 in growth, 0.08182 in maturity, and 0.08110 in decline. The growth-minus-maturity and growth-minus-decline contrasts are 0.18643 and 0.18715, respectively; both remain significant after adjustment (adjusted p = 0.026). H4c and H5c are supported. In the first-stage interaction model, the growth-stage subsidy-to-ESG slope is 150.0953, while the interactions with maturity and decline are approximately -104, consistent with a much weaker conversion of subsidy intensity into ESG assessment outside the growth stage.

By contrast, the innovation and certification contrasts are not statistically established. Green innovation has conditional indirect associations of 0.000178, 0.000161, and 0.000074 in growth, maturity, and decline, but the adjusted p-values for the growth-stage contrasts are 0.765 and 0.107. Certification has positive conditional indirect associations in all three stages (0.000449, 0.000438, and 0.000277), but the growth-stage contrasts have adjusted p-values of 0.897 and 0.240. H4b, H4d, H5b, and H5d are not supported. The life cycle is therefore a selective, rather than universal, boundary condition.

Table 6. Conditional indirect associations by corporate life-cycle stage.

Initiative	Growth	Mature	Decline	Growth-Mature [95% CI]	Adj. p	Growth-Decline [95% CI]	Adj. p
Green investment	0.00489**	0.00151*	0.00007	0.00338 [0.00064, 0.00686]	0.026	0.00482 [0.00094, 0.00974]	0.026
Green innovation	0.000178**	0.000161**	0.000074	0.000017 [-0.000072, 0.000105]	0.765	0.000104 [-0.000005, 0.000256]	0.107
Green subsidy	0.26825***	0.08182**	0.08110**	0.18643 [0.04237, 0.38195]	0.026	0.18715 [0.04307, 0.38458]	0.026
Green certification	0.000449**	0.000438**	0.000277**	0.000011 [-0.000185, 0.000220]	0.897	0.000173 [-0.000064, 0.000514]	0.240

Note. Growth is the reference stage. Confidence intervals use 5,000 firm-cluster bootstrap replications. Adjusted p-values use the Benjamini-Hochberg procedure across the eight growth-stage contrasts. Stars refer to the unadjusted conditional indirect-effect tests; formal H4/H5 decisions are based on the adjusted contrast p-values.

Table 6. Conditional indirect associations by corporate life-cycle stage.

Initiative	Growth	Mature	Decline	Growth-Mature [95% CI]	Adj. p	Growth-Decline [95% CI]	Adj. p
Green investment	0.00489**	0.00151*	0.00007	0.00338 [0.00064, 0.00686]	0.026	0.00482 [0.00094, 0.00974]	0.026
Green innovation	0.000178**	0.000161**	0.000074	0.000017 [-0.000072, 0.000105]	0.765	0.000104 [-0.000005, 0.000256]	0.107
Green subsidy	0.26825***	0.08182**	0.08110**	0.18643 [0.04237, 0.38195]	0.026	0.18715 [0.04307, 0.38458]	0.026
Green certification	0.000449**	0.000438**	0.000277**	0.000011 [-0.000185, 0.000220]	0.897	0.000173 [-0.000064, 0.000514]	0.240

Note. Growth is the reference stage. Confidence intervals use 5,000 firm-cluster bootstrap replications. Adjusted p-values use the Benjamini-Hochberg procedure across the eight growth-stage contrasts. Stars refer to the unadjusted conditional indirect-effect tests; formal H4/H5 decisions are based on the adjusted contrast p-values.

4.7. Supplementary Identification and Robustness Evidence

The robustness evidence indicates a clear hierarchy rather than uniform invariance. The contemporaneous first-stage initiative-to-ESG relationships are the most stable. Results remain positive under the broader subsidy definition and alternative investment measure and are largely preserved across control-set changes, exclusion of 2020, and alternative fixed-effect structures. The one-year-lag first-stage coefficient remains positive and significant for innovation, subsidy, and certification, but not for investment. Cross-provider evidence is more qualified: in the smaller Refinitiv sample (908 firm-years from 237 firms), investment and certification remain positively associated with the combined ESG measure, whereas innovation and subsidy do not.

The second-stage ESG-to-financial-performance relationship is more specification-sensitive. Continuous Huazheng ESG scores remain positively associated with ROA and ROE, and several alternative contemporaneous profitability measures yield positive ESG coefficients. However, one- and two-year lagged ESG coefficients are not significant, the raw-variable ROA model weakens, and the smaller Refinitiv sample does not reproduce the same second-stage result. These variations support a cautious interpretation of the contemporaneous accounting association rather than a claim of a persistent causal ESG premium.

Certification allows a stronger timing-based design because first adoption is observable. The cohort-robust staggered-adoption sample contains 547 usable firms, including 244 treated firms and 303 never-certified firms. The average treatment effect on the treated is +0.26990 for ESG (p = 0.004), -0.00957 for ROA (p = 0.036), and -0.00521 for Tobin’s Q (p = 0.943). The ESG result reinforces the baseline certification-to-ESG pattern, while the financial results again separate external sustainability recognition from contemporaneous profitability. In this supplementary design, treatment is absorbing after the first observed certification year; it therefore captures post-first-adoption timing rather than the effect of current-year certificate validity used in the baseline status regression. Pre-treatment evidence is not uniformly clean, so the design is treated as stronger temporal evidence rather than definitive causal identification (Callaway & Sant’Anna, 2021).

System GMM does not provide confirmatory evidence. Although the AR(2) tests are acceptable, Hansen tests reject the joint validity of the overidentifying restrictions for both ROA and ROE. The GMM coefficients are therefore not used to claim robustness. Reporting this failure is important because estimator validity should govern interpretation rather than the sign of a coefficient alone.

5. Discussion

5.1. A Common ESG-Recognition Stage Across Distinct Green Actions

The most stable empirical finding is that four economically different green initiatives are positively associated with ESG performance. This common first stage is theoretically important because it shows that external sustainability assessment can recognize capital commitment, technological activity, policy-supported projects, and institutionalized environmental management without implying that these actions are interchangeable. The result is consistent with recent evidence on green innovation and ESG in Chinese firms (Liu et al., 2024; Zheng et al., 2022) and with firm-level evidence that environmental investment is incorporated into ESG ratings (Hou et al., 2025).

The comparative design adds information that single-initiative studies cannot provide. All four relationships are estimated within the same institutional period, population, controls, and fixed-effects framework, which reduces the risk that apparent differences across green actions are merely artifacts of different samples or estimators. The limited pairwise correlations among the initiatives further indicate that they capture non-redundant dimensions of corporate environmental behavior.

At the same time, ESG ratings are external assessments rather than physical measures of environmental quality. A patent application may never be commercialized, an ISO 14001 system may differ in implementation depth, an investment may be inefficient, and a subsidy may be awarded to firms facing unusually high environmental burdens. The positive first-stage results should therefore be interpreted as recognition relationships, not proof that each measured initiative produces a specific quantity of emission reduction.

5.2. Why Accounting and Market Outcomes Diverge

The positive ESG coefficients for ROA and ROE are consistent with stakeholder mechanisms involving financing conditions, compliance risk, operating continuity, customer relationships, and supply-chain access. These channels can influence realized profitability relatively quickly. The absence of a significant positive Tobin’s Q coefficient indicates that market valuation does not mirror the accounting relationship one-for-one. This divergence is consistent with prior work emphasizing the role of disclosure, expectations, and measurement in the ESG-value relationship (Endrikat et al., 2014; Fatemi et al., 2018).

In heavily polluting industries, investors may interpret stronger ESG assessment as favorable information about governance, compliance, and transition capability while simultaneously anticipating continued capital expenditure or regulatory costs. The net market response can therefore be weak even when current accounting performance is positively associated with ESG. The result is not evidence that ESG reduces firm value; rather, it shows that a positive market-based relationship is not established in this sample and specification.

This distinction also explains why broad claims that “ESG improves financial performance” are too coarse. Financial performance is multidimensional, and the timing of realized profitability can differ from the timing of market revaluation. The sensitivity of the second-stage relationship to lags and rating provider further reinforces the need to specify the outcome, horizon, and measurement system when interpreting ESG-finance evidence.

5.3. Positive Indirect Associations Can Coexist with Weak or Negative Total Associations

The mediation results show that ESG provides a common positive statistical transmission channel, but the remaining direct and total components differ across initiatives. This pattern is particularly informative for green subsidy. Targeted environmental support is positively associated with ESG, yet the contemporaneous total ROA relationship is negative and remains negative after removing the direct accounting contribution of grant income. This is consistent with, but does not identify, mechanisms such as matching expenditure, designated-use restrictions, remediation needs, or selection of environmentally constrained firms into subsidy receipt. Prior research likewise emphasizes that government support can stimulate green innovation or environmental outcomes without implying an immediate private profitability gain (Bai et al., 2019; Du et al., 2023; Han et al., 2024; Luo et al., 2024).

For investment and innovation, the positive indirect pathway is statistically clearer than the total contemporaneous profitability relationship. Environmental capital projects can be visible to ESG raters before cost savings are realized, while patenting may precede implementation and commercialization. This sequencing is consistent with evidence that the performance effects of green innovation and environmental investment depend on type, organizational conditions, and time horizon (Eccles et al., 2014; Rehman et al., 2021; Tang et al., 2018).

Certification provides a similar separation between recognition and short-run return. ISO 14001 can strengthen credibility and formalize routines, which is reflected in the positive ESG relationship, but implementation and audit costs can offset immediate financial benefits. Selection into certification is also a long-standing concern in the literature (Heras-Saizarbitoria et al., 2011). The staggered-adoption evidence, which shows a post-certification ESG increase but a negative average ROA estimate, reinforces the need to analyze informational recognition and financial pay-off separately.

5.4. Life-Cycle Moderation Is Selective, Not Universal

The life-cycle results refine the resource-based explanation. Significant stage differences are concentrated in green investment and green subsidy, the two initiatives most directly dependent on current resource deployment and complementary implementation capacity. Growth firms can embed pollution-control facilities in new capacity and allocate targeted policy resources to expanding project portfolios. Mature firms are more likely to retrofit established assets, while declining firms face liquidity constraints and strategic contraction. Consequently, the marginal conversion of new investment or subsidy intensity into observable ESG improvement is stronger in growth.

Green innovation and certification do not show statistically established stage contrasts. This is equally informative. Knowledge-based innovation can draw on accumulated technological capability that spans annual life-cycle categories, and ISO 14001 has standardized, externally verifiable management-system requirements that can be implemented across stages. The absence of significant contrasts does not prove exact equality; rather, it indicates that the data do not establish systematic stage dependence for these two pathways after multiple-testing correction.

This selective pattern advances the life-cycle literature by locating the boundary condition within the mechanism. Rauf et al. (2026) show that firm life cycle can moderate relationships involving financial performance, green innovation, and ESG reporting in China. The present design asks a different question: whether the slope that converts an incremental green initiative into ESG assessment differs by stage and whether that first-stage difference produces a different conditional indirect association with ROA. The results suggest that “life cycle matters” is too general; it matters most for green activities whose effectiveness depends on immediate resource deployment.

5.5. Theoretical Contributions

First, the study contributes construct differentiation. It places four distinct environmental actions in one framework while preserving their economic meanings. The common ESG relationship demonstrates a shared external-recognition stage, whereas divergent direct, total, and life-cycle patterns show why aggregation into a single green index could conceal meaningful heterogeneity.

Second, the study contributes mechanism specification. ESG is modeled as an intermediate stakeholder-evaluation variable rather than merely an outcome, explanatory variable, or control. The positive statistical indirect associations identify a common transmission pathway without requiring a positive total effect.

Third, the study contributes boundary-condition precision by specifying corporate life cycle on the initiative-to-ESG first stage and testing conditional indirect-effect contrasts. The selective moderation of investment and subsidy, but not innovation and certification, refines the dynamic resource-based view by showing that stage dependence varies with the type of green resource.

Fourth, the study separates accounting and market performance and calibrates evidential strength. The strongest evidence concerns the first-stage initiative-to-ESG relationships; the second-stage ESG-to-financial relationship is positive in contemporaneous accounting models but more specification-sensitive. This hierarchy avoids treating all statistically significant coefficients as equally robust.

6. Practical Implications

For policymakers, the findings support differentiated evaluation of environmental policy. Green investment, innovation, subsidy, and certification should not be judged by a single short-run profitability criterion. Policy dashboards should distinguish (i) the amount or adoption of green action, (ii) changes in externally assessed ESG performance, and (iii) realized environmental and financial outcomes. A negative contemporaneous subsidy-profitability association does not necessarily imply policy failure when the objective is pollution control or technological transition; conversely, a higher ESG rating is not proof that physical environmental objectives have been achieved.

For corporate managers, green strategy should be managed as a portfolio of distinct activities. Capital-intensive investment and subsidy-supported projects should be aligned with life-cycle-specific implementation capacity. Growth firms can integrate environmental systems into expansion projects, mature firms should focus on retrofit efficiency and asset-replacement cycles, and declining firms should distinguish strategic transition projects from minimum-compliance spending under liquidity pressure. The absence of significant stage contrasts for innovation and certification means that these activities should not be assumed to lose relevance outside the growth stage.

For investors and lenders, ESG information should complement rather than replace conventional financial analysis. The source of ESG improvement matters. A rating increase linked to verified management systems can have a different cost and risk profile from one driven by pollution-control capital expenditure or subsidy-supported remediation. The lack of a significant Tobin’s Q relationship and the cross-provider sensitivity imply that due diligence should look behind aggregate scores to the underlying action, investment horizon, financing need, and evidence of realized environmental outcomes.

For ESG rating agencies and certification bodies, greater methodological transparency would improve interpretability. Users should be able to distinguish whether a rating change reflects investment inputs, technological outputs, disclosure, formal certification, governance changes, or realized environmental performance. Clearer certification histories—including validity, suspension, scope, and recertification—would also help separate nominal certification from persistent implementation.

7. Conclusions, Limitations, and Future Research

This study examines how four corporate green initiatives are associated with financial performance through ESG assessment and how the first stage of this mechanism varies across the corporate life cycle. In an unbalanced panel of 6,579 firm-year observations from 1,114 Chinese A-share firms in heavily polluting industries over 2015–2024, all four initiatives are positively associated with ESG performance. ESG is positively associated with ROA and ROE but not with Tobin’s Q. Green investment, innovation, subsidy, and certification each exhibit positive statistical indirect associations with ROA through ESG. However, their total contemporaneous financial associations differ, and green subsidy shows a negative total ROA association. Life-cycle moderation is significant for the ESG-mediated pathways of investment and subsidy, with stronger associations in growth-stage firms, but is not established for innovation or certification.

The central conclusion is therefore not that every green initiative produces an immediate financial premium. Rather, corporate green actions can pass through a common external-recognition stage while retaining distinct resource costs, timing profiles, and boundary conditions. ESG improvement and profitability are related but not equivalent outcomes. This distinction is especially important in pollution-intensive industries where environmental transformation often requires substantial current expenditure.

Several limitations constrain causal interpretation. First, the observational design cannot remove all time-varying confounding or reverse causality. The mediation estimates are statistical indirect effects, not causal mediation. The certification staggered-adoption analysis strengthens temporal evidence but has imperfect pre-treatment patterns, and system GMM fails its Hansen specification tests. Second, the primary Huazheng ESG grade is ordinal. Entering its nine ordered categories numerically in the fixed-effects framework is a modeling approximation; the continuous Huazheng score reduces this concern for the second-stage sensitivity analysis but does not eliminate the measurement limitation in the full moderated-mediation framework. Third, ESG ratings are provider-specific external assessments and may diverge because of data, weighting, and aggregation choices (Berg et al., 2022). The smaller Refinitiv sample does not reproduce every baseline relationship, and provider differences are confounded with sample differences. Fourth, the four initiative measures are proxies: expenditure does not measure project effectiveness, patent counts do not measure quality or commercialization, subsidy classification depends on project descriptions, and certification records formal status rather than implementation depth. Fifth, the decline-stage sample is smaller, reducing power for life-cycle contrasts. Finally, the findings are specific to listed Chinese firms in heavily polluting industries and should not be treated as universal parameters.

Future research can strengthen identification by exploiting policy shocks, eligibility thresholds, regional regulatory changes, or other quasi-natural experiments. Richer project-level data could distinguish pollution-control investment from energy-efficiency or clean-production projects; patent quality and commercialization could refine innovation measures; subsidy objectives and matching requirements could clarify the negative total association; and certification histories could be linked to audit, penalty, emissions, and energy-intensity data. Longer horizons and explicit modeling of life-cycle transitions would also help determine whether weak or negative contemporaneous financial associations represent temporary transition costs or persistent burdens.