Impact of Fintech on Bank Risk-Taking: Evidence from China
Data, Variables, and Models
Yüklə 1,3 Mb. Pdf görüntüsü
|
- Bu səhifədəki naviqasiya:
- 4. Data, Variables, and Models
| 4. Data, Variables, and Models
4.1. Samples and Data Sources Considering that 2013 is the first year of internet finance in China and the availability of data, we select 155 local commercial banks (including 101 urban commercial banks and 54 rural commercial banks) as the research objects from 2011 to 2016. Bank-level financial statement data comes from the Wind economic database (Wind economic database is a financial database, the database covers the stock, fund, bond, foreign exchange, insurance, futures, derivatives, spot trading, macroeconomic, financial news, and some other fields.). The Fintech index is derived from the municipal digital financial index compiled by the Peking University Digital Finance Research Center ( Guo et al. 2019 ) and the development level of urban Fintech measured by keyword Baidu retrieval method (Using the Baidu search engine to search for relevant hot words in important policy documents, news and conferences and using the resulting number of pages to reflect a certain indicator) ( Li et al. 2020 ). The city economic data corresponding to the bank registration place comes from the CSMAR (China Stock Market Accounting Research) database, the China Statistical Yearbook, and the statistical yearbooks of each city. And this article chooses Stata measurement software for empirical analysis, the reason please see Appendix D .
4.2.1. Core Variables (1) Explained variable: bank risk assumption (Risk). Since it is difficult to directly observe the bank’s risk-taking behavior and the degree of Fintech development level, in view of the availability of data, this paper selects proxy variables that are closely related to bank risk for regression. Banks’ risk-taking can be measured by several indicators, but there are some drawbacks: (i) the non-performing loan ratio focuses on the risk and quality of loan assets and has been systematically distorted under the pressure of supervision ( Zhang and He 2012 ); (ii) missing data on indicators such as expected default rate, fran- chise value, and risk-to-asset ratio, the accuracy of which is questioned ( Fang et al. 2012 ). Comprehensive consideration, we choose Z as the benchmark proxy variable for bank risk exposure, the calculation formula as follows: Z -Score = SDROA
ROA + CAR where ROA is the return on total assets, CAR is the ratio of capital to assets, SDROA is the standard deviation of the return on total assets (calculated over a 3-year rolling window). The banks’ risk-taking increases with the increase of Z-value. Taking into
Risks 2021 , 9 , 99 10 of 27
account the characteristics of the Z-value spike, thick tail, and small value, and carrying out the logarithmic treatment of ln (Z + 1). In addition, we choose SDROA, the standard deviation of the return on total assets, as the auxiliary proxy variable for bank risk, and after logarithmic processing ln ( SDROA
+ 1 ) testing its robustness. (2) The core explanatory variable: Fintech development. We select the city-level China Digital Finance Index compiled by Peking University Digital Finance Research Center (
Guo et al. 2019 ) as the proxy variable for the degree of Fintech development. The index uses the underlying data of Ant Financial’s trading account (Ant Financial is an internet financial company, the Alipay and the Ye’Bao mentioned above are its sub business segments) to show the level of Fintech development in various regions of China from multiple dimensions. In the early stages of Fintech development, the growth of the index was mainly reflected in the breadth of coverage. In recent years, the important driver of the index has been depth of use. Therefore, based on the practice of Qiu et al. ( 2018 ), the coverage breadth of digital finance index is selected as the proxy variable (Fintech 1) to measure the degree of regional financial technology development. In the robustness test, this paper selects city-level Fintech index ( Li et al. 2020 ) as the auxiliary proxy variable (Fintech 2). We choose this index to measure Fintech for the following reasons. On the one hand, the Baidu retrieval method in the domestic search engine market is an absolute dominant position, and Fintech-related issues cover a comprehensive range. Therefore, searching relevant hot terms on Baidu can accurately reflect the development level of Fintech in relevant regions. On the other hand, according to the relevant studies in China’s top journals, Baidu search method is widely used in the existing literature ( Shi and Jin 2019 ; Sheng and Fan 2020 ; Song et al. 2021 ). In specific processing, take the logarithm after removing the zero, as there are 21 zero values in this index. Obviously, this indicator is not directly affected by the choice of banks, which helps to alleviate endogenous problems. 4.2.2. Control Variable (1) Bank level. With reference to the existing literature, we choose the size of bank to control the difference in individual bank size ( Zhang et al. 2019 ). Adopt debt to asset ratio (DAR) to reflect banks’ financial leverage ( Xu and Chen 2012 ). The use of capital adequacy ratio (CAR) to reflect the bank’s capital replenishment capacity, and the net operating margin (Netprf) reflects the bank’s profitability. The cost-to-income ratio (CIR) is used to indicate the bank’s operating efficiency, and non-interest income ratio (NIRR) is used to control the difference of bank’s income structure. Learning from Ma and Li ( 2019
), and the bank’s liquidity is represented by the deposit asset ratio (SAR). Detailed bank-level data is described in Appendix A . (2) City level. With reference to the existing literature, this article selects GDP per capita (PGDP) to control the degree of regional economic development, and selects the total wage-deposit ratio (FinDev) to indicate the degree of regional financial development. (3) Macro level. At the macroeconomic level, this article selects the broad money (M2) growth rate to control the impact of monetary policy trends on the entire banking industry. 4.2.3. Mediation Variable Referring to the existing literature, we select net interest margin (NIM) and Gover- nance as the proxy variables for the internal function channels of Banks, and choose the bank competition intensity (is tested by HHI) and per capita savings (PSaving) as the proxy variables for the external function channels of Banks. Considering the availability of data, the Herfindahl-Hirschman Index (HHI) is calculated from the sum of the squares of the ratio of each bank’s network points to the total number of local bank networks during the same period. 4.2.4. Instrument Variable In the endogeneity test, internet penetration rate is usually selected as the instrumental variable. The paper argues that internet technology can not reflect the change of new Risks 2021 , 9 , 99 11 of 27
technology well. 1 To this end, we select the Urban Innovation Index (Innovation) compiled by the Industrial Development Research Center of Fudan University as an instrumental variable ( Kou and Liu 2017 ). Please see Table 1 for the details of the variables. Yüklə 1,3 Mb. Dostları ilə paylaş:
|