Abstract This thesis examines the relationship between digital monitoring, sustainability disclosure, and the cost of capital among listed energy and environmental firms. The dependent variable is firm-level weighted average cost of capital (WACC). Three disclosure-based measures are examined: ESG reporting scope, a digital monitoring disclosure index, and disclosed evidence of AI and predictive risk modelling. Only the latter two are technology-related; ESG reporting scope is a general sustainability-disclosure coverage measure, not a measure of FinTech-based transparency. The empirical analysis draws on a panel of 50 publicly listed firms from Germany, Italy, China, India, and Vietnam over 2020–2024: a non-probability purposive country-quota sample (ten firms per country) that is not statistically representative of each country's population of eligible firms. The full descriptive sample includes 250 firm-year observations. Because the three explanatory variables are lagged by one year, the regression period covers 2021–2024; after listwise deletion of missing values, mainly missing ESGReportingScope data, the main regression uses 131 firm-year observations from 39 firms. Missing values are left blank, not imputed. Vietnam has very limited representation in the main regression (2 of 131 observations) because most Vietnamese firm-years lack LSEG ESG Reporting Scope coverage. The methodology uses panel regression with firm and year fixed effects, controlling for stable firm differences and common yearly shocks. The main specification includes the three lagged explanatory variables, firm-level controls for investment scale (LnCAPEX) and leverage, and country-year controls for the interest-rate environment and the country-level ESGIndex, with standard errors clustered at the firm level. Five supplementary specifications (M1–M5) are also reported, and the sensitivity of the main result to the standard-error method is reported transparently. The results provide suggestive but not conclusive evidence of a negative within-firm association between ESG reporting scope and WACC in the available LSEG-covered sample: the coefficient is negative and significant at the 5% level under conventional standard errors and at the 10% level under standard errors clustered at the firm level. This is the clearest association among the three disclosure measures, though it is sensitive to the choice of standard-error method and should be interpreted cautiously. The DigitalMonitoringDisclosureIndex coefficient is negligibly small and not statistically significant. AI_PredictiveRiskDisclosure — recoded against a stricter standard requiring explicit evidence of artificial intelligence, machine learning, algorithmic risk assessment, or predictive risk modelling — carries a negative coefficient in every fixed-effects specification but does not reach statistical significance. H2 and H3 are therefore not supported by the empirical results. All results are interpreted as conditional statistical associations rather than causal effects.
Digital Monitoring, Sustainability Disclosure, and the Cost of Capital: Evidence from Listed Energy and Environmental Firms in Germany, Italy, China, India, and Vietnam
PANAHKHAHBAMEHRKARPASAND, PARMIDASADAT
2025/2026
Abstract
Abstract This thesis examines the relationship between digital monitoring, sustainability disclosure, and the cost of capital among listed energy and environmental firms. The dependent variable is firm-level weighted average cost of capital (WACC). Three disclosure-based measures are examined: ESG reporting scope, a digital monitoring disclosure index, and disclosed evidence of AI and predictive risk modelling. Only the latter two are technology-related; ESG reporting scope is a general sustainability-disclosure coverage measure, not a measure of FinTech-based transparency. The empirical analysis draws on a panel of 50 publicly listed firms from Germany, Italy, China, India, and Vietnam over 2020–2024: a non-probability purposive country-quota sample (ten firms per country) that is not statistically representative of each country's population of eligible firms. The full descriptive sample includes 250 firm-year observations. Because the three explanatory variables are lagged by one year, the regression period covers 2021–2024; after listwise deletion of missing values, mainly missing ESGReportingScope data, the main regression uses 131 firm-year observations from 39 firms. Missing values are left blank, not imputed. Vietnam has very limited representation in the main regression (2 of 131 observations) because most Vietnamese firm-years lack LSEG ESG Reporting Scope coverage. The methodology uses panel regression with firm and year fixed effects, controlling for stable firm differences and common yearly shocks. The main specification includes the three lagged explanatory variables, firm-level controls for investment scale (LnCAPEX) and leverage, and country-year controls for the interest-rate environment and the country-level ESGIndex, with standard errors clustered at the firm level. Five supplementary specifications (M1–M5) are also reported, and the sensitivity of the main result to the standard-error method is reported transparently. The results provide suggestive but not conclusive evidence of a negative within-firm association between ESG reporting scope and WACC in the available LSEG-covered sample: the coefficient is negative and significant at the 5% level under conventional standard errors and at the 10% level under standard errors clustered at the firm level. This is the clearest association among the three disclosure measures, though it is sensitive to the choice of standard-error method and should be interpreted cautiously. The DigitalMonitoringDisclosureIndex coefficient is negligibly small and not statistically significant. AI_PredictiveRiskDisclosure — recoded against a stricter standard requiring explicit evidence of artificial intelligence, machine learning, algorithmic risk assessment, or predictive risk modelling — carries a negative coefficient in every fixed-effects specification but does not reach statistical significance. H2 and H3 are therefore not supported by the empirical results. All results are interpreted as conditional statistical associations rather than causal effects.| File | Dimensione | Formato | |
|---|---|---|---|
|
Panahkhahbamehrkarpasand_Parmidasadat.pdf
accesso aperto
Dimensione
3.64 MB
Formato
Adobe PDF
|
3.64 MB | Adobe PDF | Visualizza/Apri |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/118983