Credit risk reflects the possibility that borrowers may default on their obligations and constitutes a central risk in banking. Traditional credit assessments rely on historical statements and standardized scoring systems, which are often backward-looking, constrained by limited data, and susceptible to information asymmetry (Berger & Udell, 2002). FinTech has the potential to fundamentally transform this landscape by enhancing the accuracy, timeliness, and scope of assessments. The core logic of FinTech in mitigating credit risk lies in reducing information asymmetry in lending (Stiglitz & Weiss, 1981). FinTech adoption enables banks to leverage big data analytics and machine learning algorithms to improve risk assessment models. In the pre-lending stage, these technologies enhance client screening, especially for borrowers with limited traditional credit histories, thereby providing more precise credit decisions and reducing the likelihood of non-performing assets from the outset (Cheng & Qu, 2020). During post-lending monitoring, real-time data processing allow banks to dynamically track borrowers’ operational conditions and repayment status, enabling early identification of potential risk through intelligent warning systems (Sebastian, 2025). This continuous monitoring mechanism not only strengthens the timeliness of risk control, but also enhances the proactivity of risk response, effectively preventing credit risk from becoming more severe. Thus, FinTech contributes to lowering default probabilities and reducing credit risk, which leads to the following hypothesis: Hypothesis 2b. FinTech adoption is negatively associated with the credit risk of banks. FinTech Adoption and Liquidity Risk Liquidity risk arises when banks fail to meet short-term funding obligations. Ineffective liquidity management may force banks to rely on costly external funding, thereby increasing vulnerability to funding shocks (Chen et al., 2018), FinTech adoption enhances the liquidity risk management of banks by improving data integration, real-time information processing, and forecasting accuracy of cash flows. Specifically, digital technologies enable banks to monitor deposit flows, loan drawdowns, and payment activities more frequently and timely (Hasan et al., 2023). Advanced analytics and automated systems facilitate more accurate liquidity demand forecasting and dynamic adjustment of funding strategies (Harle et al., 2016). Moreover, FinTech applications improve internal coordination across treasury, risk management, and business units, allowing banks to respond more promptly to liquidity pressures. As a result, FinTech adoption strengthens the ability of banks to manage funding liquidity and reduces exposure to liquidity risk. Accordingly, we propose Hypothesis 2c: Hypothesis 2c. FinTech adoption is negatively associated with banks’ liquidity risk. Bank FinTech, Bank Risk, and Profitability The impact of FinTech adoption on bank profitability may operate indirectly through its impact on bank risk. Consistent with the risk–return framework, lower risk levels contribute to more stable earnings and sustainable profitability by reducing losses, funding costs, and capital constraints (Saunders et al., 2026). By strengthening information acquisition, processing efficiency, and real-time monitoring, FinTech improves the risk assessment and management capabilities of banks, leading to reduced risk across multiple dimensions and ultimately translating into higher profitability (Thakor, 2020). Specifically, reductions in overall bank risk can enhance operational stability and resilience, leading to more sustainable financial performance. Lower credit risk reduces loan losses and provisioning expenses, thereby directly improving profitability. Similarly, lower liquidity risk can decrease reliance on costly external funding and mitigate the likelihood of liquidity shortfalls, contributing to improved earnings performance. In addition, lower risk exposure stabilizes earnings and capital positions, further supporting sustainable profitability (Berger & Bouwman, 2013). These arguments suggest that bank risk serves as a key transmission channel linking FinTech adoption to bank profitability. Accordingly, we propose the following mediation hypotheses: Hypothesis 3a: FinTech adoption enhances bank profitability through reducing overall risk. Hypothesis 3b: FinTech adoption enhances bank profitability through reducing credit risk. Hypothesis 3c: FinTech adoption enhances bank profitability through reducing liquidity risk.

METHODOLOGY Sample Selection and Data Source This study used panel data for 82 Chinese commercial banks over the period 2013–2023. The sample included five (5) state-owned commercial banks, 11 joint-stock commercial banks, 48 city commercial banks, and 18 rural commercial banks. In total, they represented over 90% of the total banking assets in China. The data for this study came from several sources. Bank-level variables were compiled from the annual reports of banks and the CSMAR database. FinTech adoption indicators were constructed based on textual analysis of the annual reports of banks, combined with the entropy weight approach. Macroeconomic variables were obtained from official statistics released by the National Bureau of Statistics of China. The resulting panel dataset has enabled a rigorous empirical analysis of the links between FinTech adoption, bank risk, and profitability within the Chinese banking industry. Variable Measurements Dependent Variables: Bank Profitability Bank profitability is proxied by Return on Assets (ROA), defined as the ratio of net income to total assets. ROA is commonly used in banking studies to evaluate the efficiency of banks in generating earnings from their asset base (Zheng et al., 2023). It provides a standardized view of performance, facilitating cross-bank comparability. In robustness checks, we alternatively used Return on Equity (ROE) to ensure our findings were not sensitive to the choice of profitability measure. Independent Variable: FinTech Adoption Bank-level FinTech adoption is measured using an index derived from textual disclosures in the annual reports of banks. The construction proceeded in the following three stages. First, following Huang et al. (2023) and Wang (2024), we developed a FinTech lexicon encompassing 118 keywords, which were categorized into six core dimensions: artificial intelligence, big data, blockchain, cloud computing, online-based FinTech applications and mobile-based FinTech applications. Secondly, this study employed Python-based web crawlers and text mining techniques to programmatically collect and analyze the full-text annual reports. For each bank-year observation, our algorithm scanned the reports to identify and count keywords occurrences, and then aggregating the totals for each of the six predefined dimensions. Third, these raw dimension-level term frequencies were synthesized into a single composite index using the entropy weight method, which objectively assigned weights based on the informational variation each dimension provided across the sample. This approach goes beyond simple text analysis by integrating the entropy weight method, thereby yielding a more objective FinTech adoption index. The full list of keywords is provided in Appendix A. Mediating Variables: Bank Risk To examine the channels through which FinTech adoption affected profitability, we considered three distinct dimensions of bank risk as mediating variables as follows: overall risk, credit risk, and liquidity risk. Overall bank risk is measured by Z-score (ZSCORE =

ROA+CAR ). A higher Z-score reflects greater bank σ(ROA)

stability and, correspondingly, lower overall risk (Conlon et al., 2020; Demirgüç-Kunt & Huizinga, 2010). Credit risk is captured by the non-performing loan ratio (NPL). This ratio serves as a direct and widely accepted measure of asset quality in banking. A higher NPL signals poorer loan portfolio quality and implies higher expected credit losses (Cheng & Qu, 2020; Wu et al., 2023). Liquidity risk is captured by the liquidity ratio (LR), measured as liquid assets divided by total assets. A higher LR indicates a stronger buffer of liquid assets and a greater ability to withstand funding shocks, and is, thus associated with lower liquidity risk (Yahaya et al., 2022). Control Variables Following prior studies (e.g., Khan et al., 2023; Zhao et al., 2023), we incorporated a set of control variables at both the bank and macroeconomic level. Bank-level controls encompass size (SIZE), measured as the natural logarithm of total assets; the cost-to-income ratio (CIR); the capital adequacy ratio (CAR); the debt-to-asset ratio (DAR); and the loan-to-deposit ratio (LDR). At the macro level, economic development is controlled by using provincial per capita GDP (PGDP) to account for crossregional heterogeneity. A detailed definition of variables is as shown in Table 1.

To empirically examine the relationships proposed in Hypotheses 1–3, we employed panel regression models. Equation (1) examines the direct effect of FinTech adoption on bank profitability： ROA it = αo + α1 FINit + α2 Controlit +μi + δt + εit

To examine Hypotheses 2a–2c, we then regressed each risk dimension on FinTech adoption: ZSCORE it = αo + α1 FINit + α2 Controlit +μi + δt + εit

NPL it = αo + α1 FINit + α2 Controlit +μi + δt + εit

LR it = αo + α1 FINit + α2 Controlit +μi + δt + εit

Finally, to test whether bank risk mediated the link between FinTech and profitability (Hypotheses 3a– 3c), we estimated the following regressions: ROA it = αo + α1 FINit + α2 ZSCOREit + α3 Controlit +μi + δt + εit

ROA it = αo + α1 FINit + α2 NPLit + α3 Controlit +μi + δt + εit

ROA it = αo + α1 FINit + α2 LR it + α3 Controlit +μi + δt + εit

In these models, ROA it denoted return on assets, FINit measured FinTech adoption, ZSCORE it , NPL it , and LR it represented overall risk, credit, and liquidity risk, respectively. Controlit included bankspecific and macroeconomic control variables. Bank and year fixed effects were captured by μi and δt , while εit denoted the error term. RESULTS AND DISCUSSIONS Descriptive Statistics Analysis As shown in Table 2, the sample consists of 902 bank-year observations. The average ROA was 0.846, with a relatively low dispersion across banks. FIN showed a mean of 0.089, accompanied by a standard deviation of 0.108, reflecting substantial variation in FinTech adoption among banks. Regarding bank risk indicators, the sample banks exhibited a high average ZSCORE (3.438) and low dispersion (SD = 0.520), implying a generally stable banking system. The non-performing loan ratio (NPL) averaged 1.455, but ranged widely from 0.030 to 4.690, reflecting divergence in credit risk levels. Meanwhile, liquidity risk (LR) showed a mean value of 0.620, pointing to generally contained liquidity positions across the sample. Control variables also exhibited reasonable dispersion. Bank size (SIZE) varied substantially across institutions, while CIR, CAR, DAR, LDR and PGDP remained within economically plausible ranges.

Table 3 presents the baseline regression results. Column (1) reports the specification without control variables, while Columns (2) and (3) progressively include bank-specific controls and macroeconomic factors. Across all specifications, the estimated coefficient on FIN was consistently positive and significant at the 1% level. In the full model (Column 3), a one-unit increased in FinTech adoption was associated with a 0.3420 rise in ROA. This finding is consistent with existing empirical evidence showing that FinTech adoption enhances bank profitability through improvements in operational efficiency, information processing, and service innovation (e.g., Singh et al., 2021; Tang & Wu, 2024). The adoption of mobile banking platforms, AI-driven credit assessment and blockchain-based payment systems improves operational efficiency by reducing reliance on physical branches and labor-intensive processes. Data-driven analytics enable more precise personalized services, expanding revenue streams beyond traditional interest income. Digital technologies also enhance process transparency and regulatory compliance, helping banks lower operational risk and compliance costs. Specifically in the Chinese context, these effects are reinforced by supportive regulatory frameworks, such as the FinTech Development Plan2 (The People’s Bank of China, 2022), and close collaboration between banks and technology giants. This enables Chinese banks to more effectively leverage FinTech to reduce costs, enhance efficiency, and diversify revenue sources, thereby improving profitability. Regarding control variables, bank size exhibited a positive effect on profitability once additional controls were included, suggesting scale advantages for larger banks. CIR exhibited a strong negative relationship with ROA, indicating that operational inefficiency reduced profitability. CAR showed a positive but statistically insignificant effect, whereas DAR and LDR were negatively related to The People’s Bank of China, PBC Issues Fintech Development Plan (2022 to 2025), available at: https://www.pbc.gov.cn/en/3688110/3688172/4437084/2025080817504626255/index.html profitability, suggesting that excessive leverage and aggressive lending might undermine the financial performance of banks. Moreover, regional economic development (PGDP) exerted a significantly positive relationship with ROA, highlighting the importance of macroeconomic conditions for bank profitability. Overall, the baseline results support Hypothesis 1, which is in line with prior findings that FinTech adoption significantly enhances bank profitability.

Notes. t-statistics in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

Table 4 examines the association between FinTech adoption and the various dimensions of bank risk, including overall risk (ZSCORE), credit risk (NPL), and liquidity risk (LR). The findings indicate that FinTech adoption was systematically associated with lower bank risk across multiple dimensions. Specifically, FIN was positively correlated with ZSCORE (β = 0.2125, p < 0.05), suggesting enhanced bank stability and reduced overall risk. FIN was negatively related to NPL at the 1% significance level (β = -0.6633), consistent with improved credit risk control and lower nonperforming loans. Furthermore, FIN exhibited a positive and significant relationship with LR (β = 0.2951, p < 0.01), implying stronger liquidity positions among banks with higher levels of FinTech adoption. These risk-mitigating effects can be attributed to several underlying mechanisms. First, AI-driven credit assessment and real-time monitoring systems enhance banks’ ability to identify and manage credit risks before they materialize, reducing ex-post non-performing loans. Second, big data analytics enable more accurate stress testing and liquidity forecasting, allowing banks to maintain optimal liquidity buffers and respond proactively to market shocks. Third, digital platforms facilitate diversified funding sources and improve deposit stability, thereby reducing liquidity risk. These technological advancements collectively strengthen the risk management capacity of banks and contribute to greater financial stability, aligning with prior empirical evidence (He et al., 2023; Liu & Wang, 2022). In China, these effects are amplified through proactive prudential regulation and institutional mechanisms, such as the regulatory sandbox, which allows banks to pilot AI-driven risk tools under close supervisory oversight, ensuring that innovation enhances rather than compromises financial stability. Taken together, the results support Hypotheses 2a–2c and indicate that FinTech adoption not only enhanced bank profitability, but also contributed to risk mitigation by improving overall stability, reducing credit risk, and strengthening liquidity positions. This dual effect of promoting returns and reducing risk provides a direct motivation to examine whether risk reduction serves as a key channel through which FinTech influences profitability.

FinTech, Risk Profiles, and Bank Profitability: A Mediation Analysis To further examine the mechanism through which FinTech adoption affects bank profitability, this section investigates whether bank risk profiles serve as mediating channels. Building on the baseline model, we sequentially incorporated measures of ZSCORE, NPL, and LR to assess their mediating roles. Table 5 presents the mediation analysis results. Column (1) reports the baseline specification linking FinTech adoption to bank profitability. Columns (2)–(4) introduce ZSCORE, NPL, and LR as mediating variables, respectively. Across all specifications, the coefficient on FIN remained positive and statistically significant, and its magnitude decreased after the inclusion of risk variables. This pattern provides initial evidence that bank risk profiles partially mediate the relationship between FinTech adoption and profitability. Specifically, ZSCORE carries a positive and highly significant coefficient, indicating that lower overall risk (higher stability) is associated with higher profitability. The coefficient of FIN decreased from 0.3420 to 0.2783, suggesting that improved overall stability constituted an important channel through which FinTech enhanced profitability. NPL exhibited a significantly negative effect on ROA, implying that lower credit risk contributed to higher profitability. The FIN coefficient further declined to 0.2625, indicating that credit risk reduction constituted a key transmission channel. Similarly, LR was also positively related to profitability, while the coefficient on FIN decreased relative to the baseline, suggesting that reduced liquidity risk (improved liquidity conditions) also mediated the FinTech– profitability relationship. To formally test the significance of the indirect effects, we conducted Sobel tests and bootstrap mediation tests with 1,000 replications. The results, as reported in the Table 6, confirm that the indirect effects through ZSCORE, NPL, and LR are statistically significant. Notably, the credit risk channel accounted for the largest share (23.23%) of the total effect, followed by overall risk (18.62%) and liquidity risk (10.17%). This ordering of mediation effects suggests that FinTech adoption influences bank profitability through multiple risk dimensions, with the credit risk channel exhibiting a relatively more pronounced contribution, while improvements in overall stability and liquidity conditions play complementary roles. The distinct magnitudes of these mediating channels provide insights into how FinTech adoption reshapes bank profitability. The pronounced role of the credit risk channel can be explained by the structural characteristics of China’s banking sector. Unlike their counterparts in developed markets where non-interest income constitutes a substantial share of revenues, Chinese commercial banks remain predominantly creditdriven, with interest income accounting for approximately 70-80% of total revenues (China Banking Association, 2023). Consequently, FinTech-driven improvements in credit assessment, risk identification and monitoring directly enhance asset quality and reduce non-performing loans, yielding the strongest positive effect on profitability. In contrast, improvements in overall stability (ZSCORE) and liquidity conditions (LR) mainly enhance operational resilience and resource management, providing complementary but relatively smaller contributions to bank profitability. These results are broadly consistent with prior empirical studies emphasizing not only the critical importance of credit risk management in banking profitability (Adekunle, 2025; Duffie & Singleton, 2003), but also that they highlight the particular significance of the credit risk channel in the Chinese context, where regulatory guidance and digital lending platforms enable banks to leverage FinTech effectively for loan quality control. These findings further underscore that the risk-mitigating effects of FinTech should be interpreted in light of the specific institutional and business model characteristics of the Chinese banking system. By investigating these multiple risk-based transmission channels within a unified analytical framework, this study fills an important empirical gap regarding the risk-mediated pathways linking FinTech adoption to bank profitability, and offers a clearer mechanism-based interpretation of how FinTech reshapes the risk structures of banks to enhance profitability. Thus, these findings support Hypotheses 3a–3c and demonstrate that bank risk is not merely an outcome of FinTech adoption, but a central mechanism through which FinTech affects bank profitability.

Robustness checks were conducted by modifying key variable definitions and sample compositions, with the corresponding results reported in Table 7. Alternative Variable Definitions First, ROE was used as a substitute for ROA. Column (1) shows that FIN remained positive and statistically significant, confirming that FinTech adoption still enhanced bank profitability. Second, we adopted an alternative proxy for FinTech adoption; FIN_ratio, which was defined as the proportion of FinTech-related keywords in the annual report of banks. The results show that FIN_ratio was also positively and significantly related to ROA, confirming that our findings were insensitive to the specific construction of FinTech adoption. Alternative Sample Compositions Third, to account for potential structural breaks caused by the 2015 stock market turmoil, we excluded observations from that period. The results in Column (3) remain consistent with the baseline findings, with FIN remaining positive and statistically significant. Finally, we removed observations from the COVID-19 affected years 2020 and 2021 to mitigate the influence of pandemic-related shocks. Column (4) shows that FIN remained positively related to profitability, supporting the robustness of our conclusions. Taken together, the robustness checks suggest that the positive association between FinTech adoption and bank profitability was stable across alternative profitability measures, different FinTech proxies, and various sample restrictions.

Endogeneity may arise from reverse causality and omitted variable bias. To address this concern, we conducted both lagged-variable and instrumental-variable tests. Lagged FinTech Adoption First, we replaced FinTech adoption with its one-period and two-period lags. Table 8 shows that both lagged FinTech variables remained positive and statistically significant in Columns (1) and (2), alleviating concerns about short-term reverse causality. Instrumental Variable Approach Second, we adopted a two-stage least squares (2SLS) approach, using the one-period lag of the regional Digital Inclusive Finance Index as an instrument for bank FinTech. This instrument is theoretically justified: regional FinTech development is likely correlated with an individual bank’s adoption (relevance condition), but is unlikely to directly affect a specific bank’s profitability after controlling for economic conditions (PGDP, CPI) and bank-level factors (exclusion restriction). The first-stage results (Column 3) confirm a strong positive predictive power of the instrument for FIN (β= 0.3121, p < 0.01). The Anderson LM statistic and the Cragg-Donald Wald F statistic both rejected the null hypothesis of a weak instrument. As shown in Column (4), the second-stage results reveal a statistically positive influence of FinTech adoption on bank profitability (β= 0.8710, p < 0.05), which is in line with the baseline evidence.

We additionally explored whether the effect of FinTech adoption on bank profitability varied across bank types and regional digital infrastructure levels. Heterogeneity by Bank Type Table 9 presents the results for national and regional banks in Column (1) and Column (2), respectively. FinTech adoption significantly enhanced profitability for regional banks (β = 0.5233, p < 0.01), whereas the effect was insignificant for national banks. A plausible explanation is that regional banks rely more heavily on technological innovation to overcome geographical constraints, optimize services, and enhance competitiveness. In contrast, the pre-existing scale and business advantages of large national banks may reduce the marginal benefits of FinTech adoption.

Column (3) and Column (4) report heterogeneity across regions with different levels of digital infrastructure. The coefficient of FinTech adoption was significantly positive only in regions with less developed digital infrastructure (β= 0.3952, p < 0.01), whereas it was insignificant in high-level regions. This finding aligns with the “catching-up” effect: in regions with weaker digital foundations, investments in FinTech can more effectively fill service gaps and reach new customers, yielding higher marginal returns. In well-developed regions, however, competition in technology application is already intense, leaving limited room for additional gains. The heterogeneity analyses indicate that the profitability-enhancing impact of FinTech was not uniform but was significantly moderated by the strategic positioning of banks and the external digital infrastructure level.