RQ1: Are FinTech adoption and intellectual capital positively associated with the quality of ESG disclosure?

RQ2: Is ESG disclosure positively associated with the dividend policy of DSE-listed financial institutions?

RQ3: Does financial performance condition the association between ESG disclosure and dividend policy?

RQ4: Is dividend policy positively associated with firm value after controlling for firm characteristics and common year effects?

Agency Theory:

Agency theory explains the dividend-policy stage of the framework. When managers control substantial free cash flow, they may retain resources or invest in projects that do not maximise shareholder value. Dividend payments reduce discretionary cash and expose firms to external capital-market scrutiny when additional financing is required. In this study, agency theory does not independently explain FinTech adoption or ESG disclosure. It complements the signalling mechanism by clarifying why a disciplined payout policy may be valued by investors, particularly when ESG expenditure creates concern about managerial overinvestment or symbolic sustainability spending.

Signalling Theory:

Signalling theory provides the central explanation for the disclosure and valuation stages. Managers possess information about earnings quality, liquidity, digital capability, and sustainability commitments that outside investors cannot observe directly. Credible ESG disclosure can reduce this gap when it is supported by verifiable data and consistent operating performance. Dividend payments provide an additional costly signal because they require distributable cash and cannot be sustained indefinitely without financial capacity. The framework therefore predicts that ESG disclosure and dividend policy jointly communicate firm quality, although the strength of the signal depends on financial performance and disclosure credibility.

Resource-Based View (RBV):

The resource-based view explains the antecedent stage. FinTech infrastructure, analytics capability, employee expertise, organisational routines, and stakeholder relationships represent intangible resources that can improve data capture, internal control, service innovation, and reporting quality. Intellectual capital and FinTech adoption are therefore modelled as capabilities that enable ESG disclosure rather than as unrelated valuation variables. Their value arises when firms combine these resources in ways that are difficult to replicate and direct them toward credible sustainability reporting and operational decision-making. Rababah, Molavi, and Farhangdoust (2022) similarly show that R&D, marketing, and financial decisions contribute to value creation, while Rababah, Javed, and Malik (2022) emphasise the performance role of social capital and managerial intangible capability.

Stakeholder Theory:

Stakeholder theory explains why firms use these resources to produce ESG disclosure. Financial institutions face demands from shareholders, depositors, borrowers, employees, regulators, and communities for information on environmental exposure, social conduct, and governance quality. Disclosure responds to these demands and can preserve legitimacy and access to resources. Within the integrated model, stakeholder pressure determines the relevance of ESG information, while the resource-based view explains the firm’s capacity to produce it and signalling theory explains how investors interpret it.

Information Asymmetry:

Information asymmetry links the theoretical stages. Annual reports and sustainability disclosures reduce the difference between managerial and investor information, but only when the underlying data are sufficiently complete and credible. FinTech tools can support traceability, timeliness, and consistency, while intellectual capital determines whether the firm can interpret and govern these systems. Reduced information asymmetry may lower uncertainty and improve investor assessment of dividend sustainability and firm value. This argument supports associations among the constructs; it does not by itself establish causal identification.

Technology Adoption (Diffusion):

Technology-adoption theory provides a boundary condition rather than a separate causal pathway. Financial institutions adopt FinTech when expected efficiency, regulatory compliance, customer demand, and competitive benefits exceed implementation costs and risks. Adoption intensity may therefore reflect organisational readiness and external pressure. The study uses this perspective to interpret variation in the FinTech index, while the resource-based view remains the principal theory connecting adoption to ESG disclosure.

A. The adoption of FinTech and its implications for ESG Disclosure

The introduction of FinTech into corporate structures is gaining greater recognition for its ability to drive change in Environmental, Social, and Governance (ESG) development (Du et al., 2022). Such a disruptive effect can be attributed to FinTech’s ability to increase information openness, improve reporting procedures, and enable more effective capital allocation to sustainable projects (Wang, 2025). In particular, FinTech can enable small- and medium-sized ventures to more closely monitor their environmental footprints and align their performance with sustainability objectives, thereby enhancing transparency (Campanella et al., 2025). Furthermore, the use of FinTech solutions can provide substantial support for a firm’s ESG by enabling real-time tracking of environmental indicators and promoting the creation of more effective governance frameworks. This integration plays a vital role in achieving a comprehensive understanding of the effects a firm has on stakeholders and the environment, which will eventually lead to improvements in overall ESG performance.

Moreover, the introduction of artificial intelligence on FinTech platforms will directly and indirectly improve ESG performance, especially in green finance practices (Sohail et al., 2025). This is done through reduced credit risks, enhanced supervision efficiency, increased product innovation, and enhanced information sharing, all of which are critical to developing green finance. All of these technological advances will allow making more accurate measurements of investments based on sustainability and building new financial solutions that can encourage companies to act in a responsible way towards the environment (Du et al., 2022; Sohail et al., 2025) FinTech adoption, consequently, would greatly contribute to corporate sustainability through resource optimization, reduced carbon footprint and general environment sustainability performance of the companies (Tian et al., 2023).

Path 2: How Intellectual Capital affects ESG Disclosure

Intellectual Capital is essential in defining a firm’s ability to fully express ESG disclosure, as it involves the knowledge, skills, and organisational processes that make a firm innovative and strong (Demiraj, 2025). In particular, intellectual capital, encompassing human, structural, and relational aspects, provides the foundation for the resources needed to develop and implement more advanced ESG strategies and reporting systems. Proper utilisation of intellectual capital will enable companies not only to identify salient ESG issues but also to develop solutions and convey them transparently to stakeholders (Sun and Guo, 2025). The increase in intellectual capital is reputed to have been associated with more investments in employee training, innovation and knowledge systems, which are a part of ESG strategies and contribute to greater development of human and structural capital, resulting in increased financial performance (Demiraj, 2025).

Furthermore, companies with high intellectual capital are better placed to navigate the intricacies of ESG reporting, including integrating diverse data sets and providing a statement of long-term sustainability (Chouaibi and Chouaibi, 2020). This will allow them not only to comply with regulatory disclosure requirements but also to engage voluntarily in disclosure activities more indicative of a stronger commitment to sustainability. This active involvement can enable companies to improve their reputation, attract not only able investors but also encourage the creation of value over the long-term horizon, as this can be achieved with a highly developed knowledge base and skilled human resources.

The growing interest of stakeholders in corporate social responsibility and clear environmental stewardship has only heightened the significance of ESG disclosure, thereby affecting key corporate financial judgments, such as dividend payments (Pham et al., 2024). Firms with strong ESG disclosures tend to be viewed as more stable and controlled, which, in turn, can be reflected in a more stable, predictable dividend policy, indicating financial soundness and sustainability to investors in the long term (Jorgji et al., 2024). Moreover, well-established ESG functioning and reporting can ensure the attraction of a larger segment of sustainably-oriented investors who can consider the payment of dividends as evidence of the company’s shaping its shareholder returns and its sustainable interests (Duong and Nguyen, 2025).

On the other hand, companies with less robust ESG reporting can be scrutinised and doubted more by investors, which can lead to changes in dividend policies as they strive to appease stakeholders or to control the impression that a company is a risky undertaking. This interaction implies that strong ESG reporting can normalise and even increase dividend payments because these explicit forms of reporting are usually viewed by investors as indicating reduced risk and long-term corporate value (Zahid et al., 2022). It means that comprehensive ESG disclosures may also be a tool for promoting the sustainability activities of a firm and complying with the requirements of corporate responsibility, which eventually leads to a constructive correlation with dividend policy due to the inflow of sustainability-sensitive investors (Jorgji et al., 2024).

B. Intellectual Capital and Dividend Policy

As greater focus is developing on the importance of intellectual capital, it is becoming evident as a crucial factor in the dividend policy of a particular firm in influencing its ability to produce sustainable returns as well as allocating them to the shareholders (Odat and Bsoul, 2024). In particular, it has a set of elements of intellectual capital, including human capital (superiority of employees), structural capital (organisational procedures and patents), and relational capital (friends and enemies). Having a high human capital base may drive operational performance and innovation, leading to higher earnings and sustainable dividend payouts (Duong and Nguyen, 2025). In the same vein, healthy structural capital, or structures that cover the organisation’s processes and intellectual property, can elevate operational efficiency and minimise expenses, thereby liberating capital to be passed on to shareholders. Further, strong relational capital, driven by stakeholder involvement and brand recognition, can boost market and consumer trust, which will be translated into stable revenue, thereby improving a firm’s capacity to support or raise dividend payments.

C. Financial Performance as a Moderator

The averting effect arises from the fact that high financial performance gives firms the space and capacity to invest in ESG activities and maintain dividend disbursement levels, even in the face of economic uncertainty (Wang, 2025). In this way, financially sound companies can more easily absorb the costs of extensive ESG disclosures and even adopt sustainable practices without undermining their loyalty to shareholders by paying them dividends (Almulhim et al., 2024). In turn, companies with weaker financial positions can have difficulty allocating resources to comprehensive ESG reporting and stable dividend distributions, making it a trade-off between these two strategic goals. This implies that financial strength can be seen as an important facilitator that enables businesses to address the perceived dilemma between investing in sustainability and giving dividends to shareholders (Jain and Malhotra, 2025).

D. Dividend Policy and Firm Value

As a key financial choice, the dividend policy has a direct impact on the valuation of a firm, on the one hand, by its signalling effect on future profitability and risk, and, on the other hand, on the preferences of investors and capital structure of the firm (Malik and Kashiramka, 2024). In particular, a properly designed dividend policy may signal financial stability and good corporate governance, which, in turn, would boost shareholder value (Zhou and Bu, 2025). Dividend infections can be used to attract and maintain long-term investors who often consider regular payments as a sign of good financial status and willingness of the company to pay its shareholders their capital back (Almulhim et al., 2024; Jain and Malhotra, 2025). Additionally, dividend policy may create information asymmetry between management and investors, thereby minimising the cost of equity and increasing the firm’s total market value (Joshi and Joshi, 2024). In its turn, an unpredictable or missing dividend policy can be viewed as a negative indicator of financial insecurity or mismanagement, which may incidentally result in a loss of investor trust and a reduction in the company’s value (Kusumawati and Hadiyanto, 2024). It aligns with the idea that regular, large dividend payments are commonly viewed as signs of firm stability and administrative effectiveness and thus have a positive impact on the firm’s value (El-Deeb and Allam, 2024). (Hypothesis 6) The influence of dividend policy, specifically, high dividend payout ratios, is assumed to have a positive effect on the market value of a company as it is a positive indicator of financial performance and may attract a group of investors who need additional returns in the long term (Moolkham, 2024).

E. The Mediation of ESG Disclosure

Such a mediation effect means that ESG disclosure will convert intangible assets and operational decisions into tangible signals that will shape internal policy-making and perceptions among external stakeholders (Jain and Malhotra, 2025). In particular, ESG disclosure mediates the relationship between proactive tax planning and firm value, and between intellectual capital and dividend policy as transparency and accountability improve (Adthajak et al., 2025). This disclosure, in its turn, will aid in eliminating the problems of information asymmetry as it will lead to investor trust, which may become a value generator for the firm’s value. In addition, sustainability reporting, especially in active tax planning, has been shown to enhance ESG performance, which, in turn, drives overall business value, particularly for companies with strong financial capacity. This increased transparency, stemming from full ESG disclosure, provides another monitoring mechanism for stakeholders, demonstrating the quality of the company’s operations and strengthening its image (Matuszewska-Pierzynka et al., 2023).

F. The Mediation of Dividend Policy

Huge dividend payout, as an example, may indicate high cash flows and high profitability, hence, positively affecting the investor perceptions and thus, the firm value (Yuniningsih et al., 2022). It is also supported by the fact that the dividend policy can be viewed as an indicator of a corporation’s future, which positively affects a firm’s value and helps attract investment (Handayani and Ibrani, 2023). This signalling capability is especially applicable to sustainable capitalism: the value of firms is no longer defined by financial performance, but by strategic responses to environmental challenges and the incorporation of green innovation capacity (Widagdo and Rahmawati, 2025). Additionally, healthy ESG performance is further reinforced by media coverage, which enhances transparency and strengthens a company’s image, thereby attracting investment and helping firms stand out in the competitive market (Xue et al., 2024). This interrelatedness highlights the problem of how clear disclosure of ESG activities, as well as a steady dividend policy, are used in the long-term sustainability and value of a firm in the market through trust and indicators of a healthy operational and financial activity (Metwally et al., 2025; Tamasiga et al., 2024).

The conceptual framework follows a sequential Digital–ESG–Dividend–Value pathway. FinTech adoption and intellectual capital are treated as enabling capabilities. They improve the firm’s capacity to capture data, maintain reporting controls, interpret stakeholder requirements, and prepare credible ESG disclosure. ESG disclosure is the central informational mechanism. It reduces uncertainty concerning sustainability practices and can influence dividend decisions by clarifying whether the firm possesses the financial and governance capacity to balance stakeholder investment with shareholder distributions. Dividend policy is the final signalling mechanism through which internally generated information is reflected in market valuation. Financial performance conditions the ESG–dividend association because profitable and liquid institutions are better able to sustain both ESG commitments and distributions. This structure reduces conceptual overload by assigning each construct a distinct position rather than modelling all variables as simultaneous and interchangeable causes of firm value.

The framework therefore tests four connected propositions: digital and intellectual capabilities are associated with stronger ESG disclosure; ESG disclosure is associated with dividend policy; financial performance moderates the ESG–dividend relationship; and dividend policy is associated with firm value. Direct paths from FinTech adoption, intellectual capital, and ESG disclosure to firm value are retained as controls for partial mediation, but they are not interpreted as separate theoretical mechanisms. The empirical models assess conditional associations within observational panel data. They do not claim experimental or structural causality.

The hypotheses follow the sequential conceptual framework.

To test H1–H5, you specify a fixed-effects panel model:

FVit=αi+δt+β1FINTECHit+β2ESGit+β3ICit+β4DIVit+γ′Xit+uit,

where FVit is firm value (Tobin’s Q, market-to-book ratio, or ROA); αi are firm effects; δt are year effects; Xit includes size, leverage, growth, profitability, and sector/country dummies; and uit is the error term. Study tests the hypotheses by examining the signs and significance of β1 through β4 . A positive β1 supports H1, and so on. If the study finds that FinTech adoption, ESG disclosure, intellectual capital, and dividend policy all have positive and significant coefficients, then accept H5.

Testing moderation (H6)

To test whether ESG disclosure moderates the link between dividend policy and firm value, include an interaction term:

FVit=αi+δt+β1FINTECHit+β2ESGit+β3ICit+β4DIVit+β5(ESGit×DIVit)+γ′Xit+uit.

A positive and significant β5 indicates that ESG disclosure strengthens the dividend policy effect, thus supporting H6.

Dynamic specification and endogeneity correction

Firm value often exhibits persistence. To account for dynamics and address potential endogeneity, we extend the model:

FVit=αi+δt+ϕFVi,t−1+β1FINTECHit+β2ESGit+β3ICit+β4DIVit+β5(ESGit×DIVit)+γ′Xit+εit.

Lagging firm value introduces a correlation between FVi,t−1. Moreover, firm fixed effects creates bias. You apply the system GMM estimator to resolve this. The estimator uses differenced equations to remove firm effects and instruments lagged dependent variables with further lags. The moment conditions are E[FVi,t−sεit]=0 for s≥2 . Study test instrument validity using Hansen’s J-statistic and examine first- and second-order serial correlation using the Arellano–Bond tests. Study accepts H1–H6 if the estimated parameters remain consistent in the GMM framework.

The sampling frame comprises bank-based financial institutions listed on the Dhaka Stock Exchange during 2015–2023. Firms are retained when annual reports, audited financial statements, market information, and the disclosures required to construct the study variables are available. Bangladesh is appropriate for the research question because its financial system is bank-centred, digital banking expanded rapidly during the sample period, and ESG reporting developed within an evolving regulatory environment. These features create variation in digital capability and disclosure quality while holding the national institutional setting constant. The design improves internal comparability but limits external generalisation. The results should therefore be interpreted as evidence from listed Bangladeshi financial institutions rather than as estimates that automatically apply to all emerging economies.

Dependent variable. The three complementary measures used to determine firm value include the Tobin Q, the market-to-book ratio, and return on assets (ROA). Tobin’s Q is computed as Tobin Q = market capitalisation/book debt/ total assets. The market-to-book ratio is a measure of the relationship between market equity and book equity. ROA- Net income/total assets. These indicators show market perceptions, opportunities for growth and performance of the operations.

Key predictors. FinTech adoption is measured with a firm-year text index constructed exclusively from annual reports. The dictionary contains terms grouped into six domains: digital banking and mobile financial services; electronic payments and transaction platforms; artificial intelligence, machine learning, and data analytics; blockchain and distributed-ledger applications; cloud computing, application programming interfaces, and open banking; and cybersecurity and digital identity. The complete dictionary, stemming rules, exclusion terms, and firm-year scores are reported in the online supplement. Each report is converted to machine-readable text, lower-cased, stripped of tables and boilerplate where possible, and tokenised. Term counts are divided by the total number of words and multiplied by 10,000. Two researchers independently reviewed the dictionary and a stratified sample of reports. Inter-coder agreement is reported using Cohen’s kappa, and construct validity is assessed by comparing the index with disclosed digital-service launches, IT expenditure where available, and year trends. Alternative indices based on binary term presence and TF-IDF weighting are used in robustness tests. ESG disclosure is measured consistently through a single self-constructed disclosure index because commercial Bloomberg, MSCI, and Refinitiv coverage is incomplete for the DSE sample. The index applies the same checklist, coding protocol, and weights to every firm-year observation. Items cover environmental, social, and governance disclosure, and the supplement reports the checklist, scoring rules, inter-coder reliability, and sensitivity estimates using unweighted and pillar-specific scores. Mixing commercial ratings with manually coded observations is avoided.

Control variables. The size is the natural logarithm of total assets. Leverage is the ratio of total debt to total assets. Growth is analytically calculated as the percentage change in annual sales. The return on equity (ROE) captures profitability. Sector and country dummies capture unobservable heterogeneity within industries and across regulatory jurisdictions. The year-fixed effects control for macroeconomic shocks and regulatory changes. A detailed description of variables and measurements is displayed in Table 1.

The empirical strategy assigns a distinct purpose to each method and avoids treating methodological variety as evidence of causality. Firm and year fixed effects provide the primary hypothesis tests because they control for time-invariant institutional characteristics and common annual shocks. System GMM is used as an endogeneity sensitivity analysis for persistence, reverse association, and time-varying omitted variables. Quantile regression assesses whether coefficients differ across the conditional distribution of firm value. CS-ARDL is reported only as supplementary evidence on temporal adjustment and is interpreted cautiously because the panel has a relatively short time dimension. The deep neural network is separated from explanatory inference and evaluates out-of-sample prediction, nonlinearity, and feature importance. Agreement across methods is interpreted as robustness of association, not proof of causal mechanisms.

1. Baseline fixed-effects model

To test the core hypotheses (H1 through H5), the study employs a linear panel regression with firm and year fixed effects:

FVit=αi+δt+β1FINTECHit+β2ESGit+β3ICit+β4DIVit+γ′Xit+εit,

where i indexes firms and t indexes years. FVit is firm value (Tobin’s Q, market-to-book ratio, or ROA). αi captures unobserved firm-specific effects; δt captures year-specific shocks; FINTECHit , ESGit , ICit and DIVit represent FinTech adoption, ESG disclosure, intellectual capital and dividend policy; Xit contains controls (size, leverage, growth, profitability, sector and country dummies); and εit is the idiosyncratic error. The fixed-effects estimator removes time-invariant heterogeneity. The hypotheses are tested by the signs and statistical significance of β1 through β4 . A positive β1 suggests FinTech adoption increases firm value (H1); a positive β2 suggests ESG disclosure increases firm value (H2); a positive β3 suggests that intellectual capital increases firm value (H3), and a positive β4 supports the dividend policy effect (H4). Joint significance of these coefficients supports H5.

2. Moderation tests

To test whether ESG disclosure moderates the effect of dividend policy on firm value (H6), an interaction term is added:

FVit=αi+δt+β1FINTECHit+β2ESGit+β3ICit+β4DIVit+β5(ESGit×DIVit)+γ′Xit+εit.

A positive and significant β5 indicates that firms with higher ESG disclosure benefit more from dividend payouts, aligning with evidence that ESG practices enhance stakeholder trust. For the role of financial performance in the ESG–value relationship (H7), a second interaction between ESG disclosure and ROE or ROA is included:

FVit=αi+δt+β1FINTECHit+β2ESGit+β3ICit+β4DIVit+β5(ESGit×DIVit)+β6(ESGit×ROEit)+γ′Xit+εit.

A significant β6 indicates that financial performance moderates the ESG–value link, consistent with findings that ESG effectiveness depends on economic strength.

3. Dynamic panel model and endogeneity correction

Firm value may exhibit persistence, and reverse causality could bias estimates. To address these issues, a dynamic specification includes the lagged dependent variable:

FVit=αi+δt+ϕFVi,t−1+β1FINTECHit+β2ESGit+β3ICit+β4DIVit+β5(ESGit×DIVit)+γ′Xit+εit.

Here, ϕ measures persistence. In the two-step system GMM specification, the lagged dependent variable and the principal strategic variables are treated as endogenous or predetermined according to theoretical timing. Instruments are restricted to collapsed lag sets, normally lags two and three, to keep the instrument count below the number of firms and reduce overfitting. Year indicators enter as standard instruments. The analysis reports the number of instruments, number of groups, Arellano–Bond AR(1) and AR(2) tests, Hansen’s J test, and difference-in-Hansen tests for instrument subsets. A satisfactory specification requires expected first-order correlation, no second-order serial correlation, and Hansen statistics that do not indicate invalid instruments or implausibly perfect fit. Estimates are described as endogeneity-adjusted associations rather than causal effects.

4. Cross-sectional dependence, slope heterogeneity and CS-ARDL

Because banks operate in a common economic environment, cross-sectional dependence may arise. Pesaran’s cross-sectional dependence (CD) test evaluates the null of independence:

CD=2N(N−1)∑i<jρ̂ij,

where ρ̂ij is the pairwise residual correlation. Significant CD suggests common shocks. The Swamy test tests for slope homogeneity across firms; rejecting the null indicates heteroscedasticity. When cross-sectional dependence is present, the Cross-Sectionally Augmented Autoregressive Distributed Lag (CS-ARDL) model is used to capture long-run relationships while accounting for common factors. Its error-correction form is:

ΔFVit=ψi(FVi,t−1−λi0−λi1FINTECHi,t−1−λi2ESGi,t−1−λi3ICi,t−1−λi4DIVi,t−1−λi5′Xi,t−1)+∑p=0P−1θipΔFVi,t−1−p+∑q=0Q−1ϕiqΔZi,t−1−q+δ0FV¯t−1+δ1′Z¯t−1+eit,

where Zit=(FINTECHit,ESGit,ICit,DIVit,Xit) , ·¯ denotes cross-section averages, ψi is the speed of adjustment, and λij are long-run coefficients. A significant negative ψi confirms cointegration.

5. Quantile regression and machine-learning models

Quantile regression explores heterogeneity across the distribution of firm value. For quantile τ, estimate:

Qτ(FVit|Zit)=ατ+βτ1FINTECHit+βτ2ESGit+βτ3ICit+βτ4DIVit+βτ5(ESGit×DIVit)+γτ′Xit.

Estimating this model at, say, the 10th, 25th, 50th, 75th, and 90th percentiles reveals whether the effects differ between firms with low and high market valuations.

Finally, the deep neural network is used as a predictive complement rather than as an additional hypothesis-testing estimator. Normalised FinTech adoption, ESG disclosure, intellectual capital, dividend policy, and control variables enter the network as predictors of each firm-value proxy. The data are divided by time and firm to prevent information leakage, with earlier observations used for training and later observations reserved for testing. Network depth, hidden units, learning rate, regularisation, and dropout are selected through nested cross-validation. Performance is compared with fixed-effects predictions, elastic net, random forest, and gradient boosting using out-of-sample mean absolute error, root mean squared error, and R-squared. SHAP values identify nonlinear feature contributions and interaction patterns. These outputs are used to determine whether predictive relationships overlooked by the linear models exist; they are not interpreted as causal coefficients or substitutes for the theoretical model.

6. Estimation procedure and diagnostics

The estimation sequence begins with descriptive statistics, missing-data checks, pairwise correlations, and variance inflation factors. The discussion explicitly considers the moderate dispersion of ROA and FinTech adoption. Limited within-firm variation can widen standard errors and reduce the power of fixed-effects estimates because identification relies on changes within the same institution over time. The analysis therefore reports within and between variation, uses alternative FinTech constructions, and avoids interpreting statistical insignificance as evidence of no economic relationship. The Hausman test informs the fixed-effects choice, while firm-clustered standard errors address heteroskedasticity and within-firm serial correlation. The primary fixed-effects model is followed by moderation and sequential-transmission tests. System GMM then evaluates sensitivity to persistence and endogeneity, with full instrument diagnostics. Quantile regressions assess distributional heterogeneity. CS-ARDL results are retained as supplementary temporal evidence, and the DNN analysis is evaluated only on held-out observations. Robustness checks use alternative firm-value proxies, dividend yield, alternative text indices, alternative ESG scores, winsorisation, and pre- and post-2019 subsamples.

Descriptive statistics in Table 2 show moderate dispersion across the principal variables. The relatively narrow range of ROA is expected for regulated financial institutions but implies that profitability-based models may have less within-firm variation than market-based valuation models. FinTech adoption also changes gradually because annual-report terminology reflects cumulative digital capability rather than discrete investment shocks. This limited dispersion does not invalidate the analysis, but it may reduce coefficient precision and makes fixed-effects estimates dependent on relatively small changes within each institution. For this reason, the study reports within and between variation, applies alternative text-index constructions, and treats differences in statistical significance across specifications cautiously.

Table 3 shows that there are positive relationships between the proxy for firm value and the constructs of FinTech adoption, ESG disclosure, and intellectual capital. The highest pair-wise correlation is between VAIC and Market-to-Book, where the two are very closely aligned in knowledge-based efficiency and equity value. The dividend payout ratios also show a moderate positive relationship with firm value, consistent with signalling theory. Notably, the correlations between independent variables do not exceed the threshold for traditional multicollinearity; the Variance Inflation Factor ranges from 1.42 to 2.76, which is considerably lower than the critical value of 10. All these results indicate the absence of severe multicollinearity, thus confirming the legitimacy of all the explanatory variables in the regression formulas. Both predictors are shown to provide unique information about the change in firm value.

According to the CD test conducted by Pesaran (2004), the dependence across the panel sample across the cross-section is statistically significant, implying that common shocks that arise are common among the banks and may be a result of regulatory policies, monetary conditions, or other macro conditions in Bangladesh, see output in Table 4. Such dependence can lead to incorrect standard errors and estimates of coefficients that are ignored. The Swamy slope homogeneity test rejects the null hypothesis of homogeneous slopes, indicating heterogeneity in the effects of the explanatory variable on bank valuation. As a result, these results support the use of second-generation panel methods, i.e., CS-ARDL and CCEMG, which can account for cross-sectional correlation and heterogeneous dynamics. The findings highlight the importance of sophisticated panel techniques rather than simple, pooled estimators.

Table 5 presents the baseline fixed-effects estimates for three firm-value proxies. These are the firm-fixed effects and year-fixed effects models. The specification captures unobserved time-varying differences across banks, including their governance cultures, ownership structures, and risk appetites. It is also calculated to eliminate the impact of typical annual shocks, e.g., macroeconomic conditions, regulatory adjustments and industry-wide liquidity events. The findings directly support H1-H5 across both market-based and accounting-based values.

H1 estimates that the adoption of FinTech augments firm value. All three models have a positive, statistically significant coefficient for FINTECH. For Tobin’s Q, the estimated value is 0.183, with a significant p-value. It suggests a relationship between the higher level of FinTech engagement, as measured by the text-based index, and a higher market valuation relative to assets (AlQudah et al., 2025; Najaf et al., 2023). The findings of the analysed materials assume that the adoption of FinTech will notably improve the financial stability of a bank, its financial performance, and the efficiency of its operations, which will help increase the overall valuation of the firm and promote its sustainable growth (Handayani and Ibrani, 2023; Kayed et al., 2024). This impact is also in line with the resource-based view, which states that increased operational efficiency, better customer service, and the novelisation of offered products may result from a strategic investment in advanced IT resources, including those related to FinTech (Tarawneh et al., 2024).

The effect size of the Market-to-Book model is even greater, 0.271, with extremely high significance. Such a trend suggests that equity investors reward banks with a stronger digital orientation, presumably because digital capabilities enhance service access, reduce transaction costs, and improve information processing (Bueno et al., 2024; Chen and Srinivasan, 2023). Research results in support of the position that at the stage of FinTech development, the profitability of banks is boosted many times over and the level of risk-taking is negatively influenced, which means that there is an excellent and statistically significant effect on financial activity and the level of financial stability (Kayed et al., 2024). Moreover, ESG-related issues, especially in combination with FinTech innovation, can further improve performance, as excessive quantification of the role of sustainable strategies in the financial sector (AlQudah et al., 2025). These improvements are eventually followed by increased financial stability and sustainability, as evidenced by sustained positive changes across different model forms and financial stability indicators (AlHares et al., 2022). A positive FINTECH effect, with a coefficient of 0.0082, which is significant at traditional levels, is also indicated in the ROA model. This inconsistency between proxies and differences in scale, but consistency across proxies, supports H1. The most fundamental assumption is that FinTech adoption is linked to market expectations and operating performance. Based on research findings it can be assumed that FinTech innovations help to add more automated processes and achieve better customer experience due to the use of such tools as chatbots and mobile apps and ensure better fraud detection opportunities in the FinTech industry with the help of machine learning, which in turn lead to a higher profitability and efficiency of a bank (Alsmadi et al., 2023).

H2 hypothesis a positive relationship between the ESG disclosure and the firm value. In all three models, ESG has a positive value and is statistically significant. In the Tobin Q model, the ESG coefficient is 0.0041, and in the Market-to-Book model, it is 0.0068. These estimates suggest that the wider the ESG disclosure, the higher the market value. The coefficient of ROA, in turn, indicates a smaller yet significant relationship between ESG reporting and improved profitability; this relationship is evident in the coefficient of 0.00018, though the channels through which it operates likely involve investor confidence, funding conditions, and stakeholder relationships. The fact that the stronger of the market-based measures indicates a greater influence of ESG disclosure suggests that it will have a primary impact on valuation through perceptions, risk reassessment, and the anticipated predictability of cash flows. This qualifies H2 and aligns with the idea that transparency minimises information asymmetry and reputational risk. Our results are consistent with the current literature, including studies suggesting that the positivity of ESG performance, which FinTech frequently supports, has a positive effect on banks’ performance and overall financial stability (Hamdouni, 2025; Mokhtar and Alam, 2023; Yuan, 2025). Also, the collaboration between fintech and green finance can lead to a more sustainable future for the global banking industry by incorporating environmental performance into the financial practices of banking organisations and making them more goal-oriented in terms of sustainability (Kassetty et al., 2024).

H3 is that firm value increases due to intellectual capital. The VAIC is both good and meaningful across all specifications. The Q coefficient of the Tobin is 0.057, and the Market-to-Book is 0.091. The two are emphatically important. The coefficient of ROA is also significant at 0.0019. This trend signifies that more efficient banks in terms of using human resources, structure, and capital invested resources have better market value and high profitability (Alkababji and Mushtaha, 2023). The findings indicate that intellectual capital is a value-creation resource in the banking industry, where the quality of services, risk analytics, product design, and process discipline depend on knowledge resources. According to the market-derived coefficients, which were found to be relatively higher than accounting-derived coefficients, this implies that investors valued intangible capability more than current profitability. This proposition is aligned with the literature of Lev and Radhakrishnan (2005); William et al. (2019).

In the context of H4, it is assumed that dividend policy increases a firm’s value. The dividend payout ratio shows positive, significant coefficients in all three models. This is less in the Tobin Q, with a coefficient of 0.112 significant at the five per cent level, but more in the Market-to-book, with a coefficient of 0.186 significant at the five per cent level. The DIV in the ROA model is also positive and considerable, with a coefficient of 0.0065. This evidence suggests that dividend payout is a valuation indicator for listed financial institutions. Thin and non-tinier dividend distributions may be interpreted as a sign of the quality of earnings, the strength of the liquidity situation, and a disciplined managerial team (Kılınçarslan, 2018; Nkn, 2018). Hypothesis-4 (H4) is supported by the positive correlation that is repeated in the Q, Market-to-Book and ROA of Tobin. It also indicates that the dividend decisions are associated with both the market confidence and operating returns.

H5 implies that the joint influence of FinTech adoption, ESG disclosure, intellectual capital, and dividend policy explains firm value when the impact of common firm-specific factors is mitigated. Table 4 supports H5 in two ways. Once, none of the four focal variables is found to have negative signs or statistical significance when controlling for and including fixed effects. Second, the measures of model fit suggest significant within-firm explanatory power. In the range of R² between 0.389 and 0.436, which is large with the fixed-effects design and the banking environment in which most determinants are time-invariant, there is a value. The model F-statistics are highly significant across all specifications (see Figure 2), indicating that a combination of regressors is significant in explaining the time-varying value of firms in the banking industry. This combined importance aligns with the structure that underpins digital capability, sustainability reporting, intangible efficiency, and payout strategy working in unison.

Switching to controls, leverage is negative and significant in all proxies. This is consistent with the fact that increased leverage increases risk and valuation, while also squeezing profitability. The market-based models hurt firm size, which may indicate the effect of maturity and reduced growth in large institutions. The growth in sales is positive and substantial, indicating that expansion is advancing in value and profit. ROE, one of the controls used to determine profitability, is positive and significant, and it acts as anticipated. Among the focal variables, FinTech adoption and VAIC have the strongest and most consistent drivers in market-based valuation models, whereas ROA shows lesser but notable impacts. This distinction is educative. Market valuations are expectations of future efficiency and growth, whereby they respond more intensively to FinTech and intellectual capital. Accounting profitability reflects the current returns and possibly slower changes. The general evidence in Table 4 confirms H1 5 and provides a weighted foundation for the subsequent models that address endogeneity, long-term dynamics, and nonlinear effects using system GMM, CS ARDL, and deep learning.

Interaction terms are presented in Table 6 to test the moderating roles spelt out in H6 and H7. Adding ESG x Dividend and ESG x Financial Performance to the explanatory power greatly enhances the value, as shown by the higher within-R2 values across all three firm value proxies compared to the baseline model.

The relationship between ESG disclosure and dividend payout is positive and significant for the Q, Market-to-Book, and Tobin’s Q. This finding shows that dividend policy is also value-relevant in the cases when ESG disclosure is more effective in that investors turn dividend payments more reliable and sustainable when firms are represented by clear evidence of environmental, social, and governance practices at the same time. The observed higher magnitude in the Market-to-Book model indicates that equity markets are more sensitive to the combined signalling effect of sustainability disclosure and payout commitment. In the case of ROA, the correlation is positive but weaker, and operating effectiveness is improved through ESG engagement in dividend decisions. Based on the existing literature, high ESG performance has the potential to improve a firm’s profitability and, in turn, its ability to pay future dividends. Also, empirical evidence shows that strong ESG practices can greatly increase shareholder value, and dividend policy is a credible signal of financial sustainability and proper governance.

The correlation between ESG disclosures and financial performance is also strong and highly statistically significant across all three specifications. This result shows that the market greatly compensates for ESG disclosure when companies have strong financial fundamentals, as greater disclosure of ESG strengthens the positive link between ESG and financial performance. Arguably, this indicates that ESG undertakings can only result in increased valuation based on profitability and earnings stability (Demirgüneş, 2015). The moderation effect is not only statistically but also economically significant in market-based proxies (Q and Market-to-Book) and in the accounting-based model (ROA). Moreover, this symbiotic interaction highlights that whereas the importance of ESG engagement is essential, the financial effect is greater when combined with the company’s financial health, which will help investors more readily identify the real value of sustainable value creation and distinguish greenwashing from genuine sustainability (Malhotra, 2025). These results support existing studies that show that ESG disclosure has a positive impact on firm value across most performance indicators, such as Q and Return on Assets (Hamdouni, 2025).

Table 7 reports the two-step system GMM estimates as an endogeneity and persistence check rather than as proof of causal effects. The specification limits and collapses the lag instruments so that the instrument count remains below the number of firms. The table reports the number of groups and instruments, Hansen’s J statistic, difference-in-Hansen tests, and Arellano–Bond AR(1) and AR(2) diagnostics. The expected AR(1) result and the absence of statistically significant AR(2) support the serial-correlation assumptions. Hansen and difference-in-Hansen probabilities are interpreted jointly; very high values are not treated automatically as evidence of validity because they may indicate weak diagnostics caused by instrument proliferation. The similarity between fixed-effects and GMM coefficient signs is therefore interpreted as robustness to alternative assumptions, while the magnitude and significance are discussed as conditional associations.

To crucially test whether long-run equilibrium relations and to investigate short-run dynamics, it is estimated in error-correction form using the CS-ARDL model, see Table 8. The error-correction measure is large and significant across all specifications (p<0.01), which is strong evidence of cointegrated movement and a consistent mechanism for eliminating such deviations (replenishing the shocks to the firm value) in the long-run: errors are fixed at a slow pace (of not more than 20-30% per period), making the long-run equilibrium persistent auto-correlation in the banking market plausible. Overall, the use of FinTech has a strong, positive, statistically significant impact in the long run, with an average beta of 0.15-0.25 and a p-value of less than 0.01, which proves that long-term digital transformation not only boosts but also has a long-lasting strengthening impact on market value due to the increase in operational performance and competitiveness. ESG disclosure also delivers a positive, significant long-run coefficient ($\beta = 0.10-0.20, p=0.05), confirming that sustained sustainability reporting enhances firm credibility, reduces perceptions of risk, and yields premiums in investor portfolios. The potency of intellectual capital lies in its long-run effects, as underscored by its role as a key structural driver of long-term competitiveness through knowledge advantages and innovation rents. Dividend policy also has high importance, as it reflects stable financial health and a strong long-run valuation. Smaller in size yet positive are the short-run dynamics of FinTech and ESG variables, which exhibit initial adaptation frictions that yield strategic dividends over time. The CS-ARDL model that incorporates cross-sectional averages is a powerful means to eliminate the prevalent macroeconomic shocks and enhance the reliability of the estimates, as well as causality. Twelve solidly supported by these findings are assertions of stable cointegrated relationships, which empirically confirm theoretical hypotheses that digital prowess, sustainability transparency, and intellectual assets are sterling foundations for long-run value creation in financial institutions. In particular, the adoption of AI-based FinTech-associated technologies has been proven to have a beneficial effect on the financial performance, sustainability, and stability of the banking sector (Hamdouni, 2025).

The quantile regression findings (see Table 9) clearly indicate a high level of heterogeneity in the firm value distribution, i.e., the coefficients increase in magnitude as the quantile is shifted toward the lower or upper ends, with the largest being for market-based measures. This validates a nonlinearity, skewed and upslope influence in which the most effective performers enjoy disproportionate rewards.

In the case of Tobin Q, the FinTech coefficient shot up from 0.094 to 0.257 across the 10th to the 85th-90th percentile, or a 173 per cent increase, which is a solid indicator that digital transformation delivers considerably better value creation for well-performing banks. The fact that this incremental value is 0.163 throughout the distribution is a fabricated testament to the market leaders having a disproportionate share in integrating FinTech. Similarly, VAIC shows a monotonic increase from 0.031 at the lower quantiles to 0.083 at the higher quantiles, indicating that intellectual capital exhibits increasing marginal returns, most pronounced when banks are valued at a premium. The results show that innovation in FinTech is a highly significant contributor to Total Factor Productivity, and the quantile differences are statistically significant at the 10th percentile. This indicates that the effectiveness of FinTech use in improving bank performance is not evenly distributed, and the strongest performers have a greater capacity to convert technological expenditures into clear financial performance (Hassan et al., 2025; Li et al., 2024).

The ESG disclosure also increases, ranging from 0.0021 to 0.0061, in the Q model of the Tobin, definitively suggesting that sustainability transparency is an effective enhancer of the valuation of strong market firms. This tendency is reflected in dividend policy, which increased its values from 0.052 to 0.169, indicating the greater power of payout signalling in promoting high-value entities. According to the research results, the noted positive correlation between FinTech and the performance of the bank is especially high in the higher quantiles, which proves that banks with more favourable initial performance metrics are in a position to use the FinTech investments to improve their financial results (Hassan et al., 2025). This shows that FinTech leads to greater performance improvements in banks with lower performance, as they can access financial services previously unavailable (Hassan et al., 2025).

The Market-to-Book version further widens this heterogeneity: FinTech impacts increase fourfold, to 0.364, highlighting the extreme sensitivity of equity markets to the adoption of digital technology in the elite enterprise. VAIC soars up by 0.046 up to 0.136, as the key position of knowledge efficiency is sealed. ESG coefficients almost tripled, justifying the predisposed investor reward for stability in well-valued companies. Even within the ROA model, where all coefficients are moderately scaled, the trend is persuasive. FinTech increases consistently by 0.0031 to 0.0118, VAIC increases similarly by 0.0009 to 0.0033, and dividends increase accordingly. This is a clear indication of how operational efficiencies from digitalisation, sustainability, and intellectual capital are focused within the ranks of the best banks. The results show an intent to conclude that, whereas FinTech can improve financial inclusion by expanding services to underserved populations, the advantages are more practically achieved by better-performing organisations, which can obtain economies of scale and strategically integrate technologies.

The deep neural network results in Table 10 address a different question from the econometric estimates. They assess whether the study variables improve out-of-sample prediction of firm value and whether nonlinear patterns are present. The DNN is compared with linear and tree-based benchmarks using a holdout period and firm-aware cross-validation. Improvements in test-set error indicate predictive information, not theoretical confirmation or causality. SHAP values are used to examine whether FinTech adoption, ESG disclosure, intellectual capital, and dividend policy contribute consistently to predictions and whether their marginal contributions vary across observations. When the SHAP ranking broadly accords with the panel estimates, the result strengthens confidence that the explanatory variables contain valuation-relevant information. Divergence between SHAP patterns and linear coefficients is interpreted as evidence of nonlinearity or interaction that warrants further research, rather than as a reason to replace the econometric findings.

The deep neural network results in Table 10 address a different question from the econometric estimates. They assess whether the study variables improve out-of-sample prediction of firm value and whether nonlinear patterns are present. The DNN is compared with linear and tree-based benchmarks using a holdout period and firm-aware cross-validation. Improvements in test-set error indicate predictive information, not theoretical confirmation or causality. SHAP values are used to examine whether FinTech adoption, ESG disclosure, intellectual capital, and dividend policy contribute consistently to predictions and whether their marginal contributions vary across observations. When the SHAP ranking broadly accords with the panel estimates, the result strengthens confidence that the explanatory variables contain valuation-relevant information. Divergence between SHAP patterns and linear coefficients is interpreted as evidence of nonlinearity or interaction that warrants further research, rather than as a reason to replace the econometric findings.

This study has several limitations. The sample contains 24 DSE-listed bank-based financial institutions and therefore offers limited statistical power and external generalisability. The nine-year period also constrains long-run estimation, so CS-ARDL results are treated as supplementary. The FinTech index measures disclosed adoption rather than audited technology expenditure or system usage. Although the keyword dictionary, validation checks, and alternative constructions improve reliability, annual-report language may contain strategic emphasis. The ESG index measures disclosure and may not fully represent underlying ESG performance. Moderate within-firm dispersion in ROA and FinTech adoption can reduce fixed-effects precision. Finally, fixed effects, lag structures, system GMM, and placebo tests reduce selected endogeneity concerns but do not eliminate all omitted-variable bias or establish causal effects. Future research should use regulatory shocks, digital-policy discontinuities, matched cross-country samples, audited technology expenditure, and external ESG assurance to strengthen identification and measurement.

Second, the FinTech adoption index is based on annual-report text. This approach captures disclosed digital activity, not necessarily actual technology intensity. Future studies can combine annual-report text with FinTech investment data, mobile banking transaction data, digital branch indicators, or IT expenditure.

Third, ESG disclosure is measured from public reporting. Disclosure may differ from actual ESG performance. Future studies can compare disclosure scores with third-party ESG ratings, regulatory penalties, board-level sustainability committees, or environmental and social performance indicators.

Fourth, the study uses observational panel data. Although lagged variables, fixed effects, and system GMM reduce endogeneity concerns, causal inference remains limited without an external shock. Future research can use regulatory reforms, FinTech adoption mandates, or ESG reporting rule changes as quasi-natural experiments.

Fifth, the study focuses on dividend policy as the main transmission channel from ESG disclosure to firm value. Future studies can examine other channels, such as risk reduction, cost of capital, analyst coverage, institutional ownership, and market liquidity.