Fintech and Commercial Banking

Results and Discussion

Descriptive Statics

The descriptive statistics are given in Table 2. It is found that the Z value of the explained variable (RISK) is distributed between -0.47 and 2.27, with a mean of 1.58 and a standard deviation of 0.87. It shows that during the statistical period, partial risks found in some years, but the overall Chinese banks, are robust with less change. Similar results were found in previous studies. The average value of FIN is 0.53, with a standard deviation of 0.37 and a distribution interval between 0 and 0.78. It shows that the financial technology of Chinese banks has been steadily improved with less fluctuation. Although overall changes in other variables are stable, they fluctuate slightly in the past 10 years.

Table 2. Descriptive statistics of variables.

Impact of Fintech on Commercial Banks' Risk-Taking

We used a fixed-effect model to analyze the impact of fintech on commercial banks' risk-taking (Table 3). It is found that both the full sample and different types of banks, fintech are significantly positively correlated with bank risk. The explained variable of bank risk is the z-value, which indicates the stability of the bank. Therefore, fintech can effectively improve the operational stability of the bank and reduce bank risk.

The results verified hypothesis 1. The result of heterogeneity analysis revealed that state-owned banks are the most effective, and rural commercial banks are the worst effects in the risk prevention effect of financial technology on banks. State-owned banks have strong funds, obvious talent reserves, and technical advantages. Meanwhile, state-owned banks have strong research and development capabilities for key technologies related to financial technology, start technology earlier, and invest on a larger scale with more mature applications and devolvement. Moreover, state-owned banks' business strategies are more conservative, with relatively rich experience in risk management and more prudent decision-making behavior.

Therefore, when using financial technology, state-owned banks have the most obvious risk prevention effect and higher robustness. For rural commercial banks, due to their relatively weak R&D and innovation capabilities, financial technology research and development relies on third-party technology companies. Although the application of financial technology is conducive to reducing bank risks, the risk prevention effect is the least obvious.

Table 3. The impact of fintech on commercial banks' risk-taking.

Furthermore, we examined the impact of financial technology indicators on commercial banks' risk-taking (Table 4). For the full sample data, the three dimensions, namely, payment and settlement, business development, and investment management, have a significant positive impact on the stability of commercial banks. It indicates that for the full sample, financial technology plays an important role in the field of payment and settlement, business development, and investment management applications. Fintech will significantly improve the operational stability of commercial banks and reduce bank risks. In terms of different banks, the two dimensions of payment settlement and investment management have significant positive effects on state-owned banks, joint-stock banks, city commercial banks, and rural commercial banks.

The dimension of business development has a significant positive effect on state-owned banks, joint-stock banks, and rural commercial banks. The dimension of resource allocation has a significant positive effect on state-owned banks and joint-stock banks. For state-owned and joint-stock banks, the application of financial technology in the field of payment and settlement, business development, investment management, and resource allocation can help to improve their stability and reduce bank risk-taking.

For urban commercial banks, the application of fintech only in the field of payment and settlement and investment management will significantly improve the stability of their banks and reduce bank risk-taking. Similarly, for rural commercial banks, the application of fintech in the field of payment and settlement, business development, and investment management reduces bank risk-taking.

Table 4. The impact of various decomposition indicators of fintech on the banks' risk-taking.

Analysis of Mediating Effect

We used the mediation effect model to analyze the impact of fintech on banks' risk-taking. It should be pointed out that the mediation effect analysis model is for the full sample data. Column 1 of Table 5 showed the results of basic regression for the total effect of financial technology on the risk-taking of commercial banks. While columns 2, 3, and 4 represent the regression results for financial technology (FIN) on the cost-income ratio (CIR), non-interest income ratio (NII), and loan impairment to total loan ratio (ILR), it represents the intermediary effect or indirect effect of financial technology on commercial banks' risk-taking through operational efficiency, financial innovation, and risk management.

Column 5 is the regression result of financial technology, cost-to-income ratio, noninterest income ratio, loan impairment to total loan ratio combined on the risk-taking of commercial banks, representing the direct effect of financial technology on commercial banks' risk-taking, where the coefficient of financial technology represents the direct effect of the impact of financial technology on banks' risk-taking.

Table 5. Impact of fintech on intermediary variables.

The results of Table 5 confirmed that the estimated coefficients of the financial technology level of commercial banks (FIN) to the cost-to-income ratio (CIR), non-interest income ratio (NII), and loan impairment to total loan ratio (ILR) are all significantly positive. It indicates that the improvement of the level of financial technology will significantly improve their financial efficiency, financial innovation level, and risk management and control ability of commercial banks.

When main explanatory variables are added to the intermediary effect model, the estimated coefficient of commercial bank fintech remains significant and positive. Moreover, the intermediary variables' cost-income ratio (CIR), non-interest income ratio (NII), and loan impairment to total loan ratio (ILR) are also significant and positive.

The results showed that financial efficiency, financial innovation, and risk management play an intermediary role in the impact of commercial banks' financial technology level on their risk reduction. In the contribution of financial technology level to commercial banks' risk-taking, the total effect is 2.187, the direct effect is 1.262, and the proportion of direct effect is 57.70%. The intermediary effects of financial technology in reducing its risk-taking through financial efficiency, financial innovation, and risk management account for 8.51, 7.18, and 5.77% of the total effect, respectively.

It reveals that the three channels of improving financial efficiency, financial innovation, and risk management contribute 8.51, 7.18, and 5.77% to explain the improvement of banks' fintech level and thus reduce their risk-taking. The empirical results effectively identified the transmission mechanism of financial technology through three paths of financial efficiency, financial innovation, and risk management, and then affected banks' risk-taking. These findings satisfy hypotheses 2, 3, and 4.

Test of Robustness

Following Zhang et al., we used the method of replacing the explained variables to test the robustness, specifically replacing the Z value with the nonperforming loan ratio (NPL) as a proxy for the risk-taking results of commercial banks. The robustness test results are shown in Table 6. It can be seen that the impact of fintech on the risk-taking results of commercial banks is still significant because the nonperforming loan ratio is a negative indicator, and the fintech coefficient is significantly negative. It indicates that fintech has reduced the risk-taking level of commercial banks. Moreover, it is confirmed that the basic test is robust.

Table 6. Robustness test.

Analysis of Early Warning Mechanism

We used the KLR model to analyze the risk warning of commercial banks. The early warning indicators were constructed at the macro-level and microlevel. In particular, 10 early warning indicators were selected, and weights were given to the indicators. The KLR model was first proposed by Kaminsky and was mainly used for currency crisis research. Then, many people put it into early warning analysis of the financial crisis, early warning analysis of futures crisis, and early warning analysis of stock market systemic risk.

Construction of Risk-Warning Indicator System for Commercial Banks

Microlevel Indicators

At the microlevel, bank risk is closely related to the bank itself. The higher the risk of an individual business, the greater the vulnerability and the higher the probability of systemic risk. At the same time, the empirical results in the third part of the article showed that the application of financial technology can effectively reduce the risk level of banks. Therefore, we also incorporated financial technology into the risk-warning indicator system of commercial banks. Finally, in the selection of microlevel indicators, we determined five secondary indicators such as capital adequacy ratio, provision ratio, liquidity ratio, nonperforming loan ratio, and financial technology level.

Macro-Level Indicators

At the macro level, the level of economic development will affect the operation of banks. Whether it is pro-cyclical or countercyclical, the macroeconomy will affect the development of banks. The deterioration of the macroeconomy will be transmitted to the banks in various ways, such as the deterioration of the real economy, the failure of many business customers, and the inability to repay bank loans, resulting in credit risks. Therefore, in the selection of macro indicators, five secondary indicators are determined, such as the CPI growth rate, GDP growth rate, interest rate, exchange rate, and banking prosperity index (Table 7).

Table 7. Indicator system of commercial bank’s risk-warning.

Analysis of a Risk-Warning Model of Commercial Banks

In this section, we selected the sample of the fourth part, the quarterly data from the first quarter of 2011 to the fourth quarter of 2020, a total of 40 time periods. In this study, we used ten indicators, particularly capital adequacy ratio, provision ratio, liquidity ratio, nonperforming loan ratio, financial technology level, CPI growth rate, GDP growth rate, interest rate, exchange rate, and banking prosperity index. We substituted the data into equation 4 to calculate the risk index of each bank.

Estimation of Indicator Warning Threshold

Correlation Analysis Between Each Index and Bank Risk Index

If an indicator is positively correlated with the BRI, the larger the indicator value, the higher the probability of bank systemic risk. If the indicator is negatively correlated with the BRI, the smaller the indicator value, the higher the probability of bank systemic risk. The capital adequacy ratio is negatively correlated with the BRI. The provision ratio is the ratio of provision for bad and doubtful debts, which is positively related to the bank's risk index. The liquidity ratio reflects the asset liquidity of commercial banks and is negatively correlated with the BRI. The nonperforming loan ratio reflects the bank's nonperforming loan situation and is positively correlated with the bank's risk index. The level of fintech is negatively correlated with the BRI. The more severe the inflation, the greater the possibility of a financial crisis, so the CPI growth rate is positively correlated with the BRI. The economic growth is stable and the probability of a financial crisis is small; therefore, the GDP growth rate is negatively correlated with the BRI. Both interest rates and exchange rates are positively correlated with the BRI. The banking sentiment index is negatively correlated with the banking risk index. Previous studies have shown similar findings.

Threshold and Optimal Noise Signal Ratio

Using the KLR method for crisis early warning analysis, it is necessary to clarify the thresholds of various indicators. We have taken the noise-to-signal ratio, which is the minimum NSR. The threshold and optimal NSR of each index are given in Table 8.

Table 8. Thresholds of each indicator and optimal noise signal ratio.

If the indicator calculation result is within the threshold range, the system will not issue a risk warning signal. If the indicator calculation result exceeds the threshold range, the system will issue a risk warning signal. At the same time, it is also necessary to pay attention to the correlation analysis of indicators. If there is a positive relationship between the indicators and the crisis, an early warning signal will be issued when the indicators are higher than the threshold range. If there is a negative relationship between the indicator and the crisis, there will be no warning signal below the threshold range. It is also necessary to pay attention to the optimal noise signal ratio.

The smaller the calculation result, the stronger the early warning ability of the corresponding indicator. Generally, the indicators with the optimal noise signal ratio higher than 1 can be directly eliminated, their reference value is not obvious, and the early warning ability is limited. For example, the optimal noise signal of the two indicators of banking climate index and nonperforming loan ratio is relatively large, indicating that the two indicators have low early warning ability. While the optimal noise signal of the CPI growth rate is relatively small, the indicator has a high warning ability. The contribution of the level of early warning ability to the final comprehensive early warning index of systemic risk is reflected in the weight of each index.

Test of Early Warning Effect

According to the determined threshold range, the release of crisis signals is determined by the indicators, and the relevant values are substituted into equation 7 to calculate the bank's risk early warning index (RWI). According to the sample and time selection, we calculated the BRI and bank risk-warning index (RWI) of 37 listed commercial banks from the first quarter of 2011 to the fourth quarter of 2020. By comparing the BRI_t and RWI_t of 37 banks, the risk-warning ability of each bank can be observed. Our early warning effect test focused on measuring the overall risk situation and early warning capability of the banks.

The calculation results are shown in Figure 1. Comparing the trend of the BRI and the bank risk-warning index, it can be seen that if the signal month is set to 12 months, the bank risk-warning index can play a warning effect in this cycle. Observing the changing trend of the indicators BRIt and RWIt, the risk situation can be predicted. In fact, in addition to the strong predictive ability of the overall risk early warning system, the individual risk early warning index for each bank can also well predict the risk situation of each bank.

Figure 1. Bank's overall risk index and risk-warning index.

In the process of building the bank's risk-warning index, it is found that the early warning capabilities of each index are different. Indicator correlation analysis can clarify the direction of macro-prudential management. Particularly, it effectively reduces the nonperforming loan ratio and CPI growth rate, comprehensively increases the relative proportion of liquidity, forms a sound early warning system  structure, controls the capital periodicity of the adequacy ratio, ensures the stability of bank deposit interest rates, reduces the bank provision rate, and improves banks' fintech degree. Through these concerns, the occurrence of systemic risk can be reduced.